PROCEEDINGS - PT-CONF · Romania (INM) is mainly responsible for developing, maintaining and...

Oradea, Romania

10th – 13th September 2019

PROCEEDINGS

OF THE 7TH INTERNATIONAL PROFICIENCY TESTING

CONFERENCE

ORGANIZED BY:

CEPROCIM SA, Bucharest, Romania -Research, Consulting and Process Development in the binders’ field

AFSIC Association of Proficiency Testing

Providers Agora University of Oradea and Research & Developme nt Agora Ltd Oradea

www.pt-conf.org

Editor: Dr. Eng. Virginia - Graziela GUSLICOV

PROCEEDINGS OF THE 7TH IN TERNATIONAL

PROFICIENCY TESTING CONFERENCE

ISSN 2066 – 737 X

Publisher: Smart Publishing

Bucharest

With the support of:

The Romanian Ministry of Research and Innovation

Smart Publishing Bucharest, 2019

FORWARD Motto: "There is no science without measurements, no quality without testing and no global market without standards." The Commission of the European Union

Dear Colleagues, The seventh International Proficiency Testing Conference - www.pt-conf.org was been organized by CEPROCIM SA, Bucharest (Research, Consulting and Process Development in the binders field) - www.ceprocim.ro and Association of Proficiency Testing Providers (AFSIC) – www.afsic.ro, in co-operation with Agora University of Oradea (http://univagora.ro/) and Research & Development Agora Ltd Oradea under the auspices of the Romania's Ministry of Research and Innovation - http://www.research.gov.ro/.

PT –Conf 2019 was held on 10th to 13th September, in Oradea (Romania).

The main objective of this event have been to provide the framework for debates, for interchange of knowledge and expertise, of ideas and initiatives amongst academia, scientists, engineers, managers from supplier bodies of PT schemes, managers from laboratories of analyses and tests, from all over the world and from all ranges interested in Proficiency Testing (PT) and the connected areas with PT.

The themes of the conference have included: Proficiency Testing Schemes (PTS); Participation at Proficiency Testing and Accreditation of Quality Management (PPT); Reference Materials and Proficiency Testing (RM); Validation of Testing Methods and Proficiency Testing (VTM); Uncertainty of Measurement and Proficiency Testing (UM); Internal Quality Control of Analyses and Proficiency Testing (IQC); Metrology, Traceability and Proficiency Testing (MT); Development of the Analyses Methods in the Testing Laboratories (DAM).

A number of 105 authors and co-authors have prepared the manuscripts of 32 papers which have been presented during the conference as oral, or as poster presentations.

At the conference workings a number of ~ 80 specialists from 20 countries (Argentina, Australia, Belgium, Brazil, China, Czech Republic, France, Germany, Italy, Malaysia, Norway, Poland, Portugal, Turkey, Romania, Serbia, Slovakia, UE, United Kingdom, and United States) has involved.

A training course namely: “The evaluation of the measurement uncertainty in the PT schemes from the point of view of suppliers and participants”, has been held during the conference session.

The working group discussions named "ISO/IEC 17043 WILL BE REVISED STARTING IN 2020. WHAT SHOULD BE INCLUDED IN THE REVISION?" has been held during the conference session.

Exhibitors from Germany, United Kingdom and USA have presented their products at the conference exhibition.

The conference discussions have greatly contributed to disseminate knowledge and innovative ideas with a focus on Proficiency Testing.

Virginia - Graziela GUSLICOV

7th Proficiency Testing Conference President

PRESIDENT OF PT-CONF 2019 - Virginia Graziela GUSLICOV INTERNATIONAL ADVISORY COMMITTEE

Ricardo BETTENCOURT DA SILVA

Centro de Química Estrutural of the University of Lisbon (https://cqe.tecnico.ulisboa.pt/) and the Faculty of Sciences of the University of Lisbon (https://www.ulisboa.pt/en/unidade-organica/faculty-sciences), Portugal

Philip BRIGGS

Proficiency Testing Australia, Australia (www.pta.asn.au)

Mirella BUZOIANU

National Institute of Metrology (INM), Romania (http://www.inm.ro/)

Fernando CORDEIRO RAPOSO

European Commission, Directorate Gen Joint Research Centre, Belgium (https://ec.europa.eu/jrc)

Ioan DZIȚAC Agora University of Oradea (http://univagora.ro/en/university/)

Fanel IACOBESCU Romanian Legal Metrology Office (BRML), Accreditation Association of Romania RENAR, Romania ( www.brml.ro ; www.renar.ro )

Michael KOCH

Institute for Sanitary Engineering, Water Quality and Solid Waste Management, University Stuttgart, Germany (www.iswa.uni-stuttgart.de)

Mișu-Jan MANOLESCU Agora University of Oradea (http://univagora.ro/en/university/)

Hariton George PREDESCU

Research, Consulting and Process Development in the Binders Field CEPROCIM SA, Bucharest, Romania (http://www.ceprocim.ro/en/)

Doru Vladimir PUŞCAŞU Research, Consulting and Process Development in the Binders Field CEPROCIM SA, Bucharest, Romania (http://www.ceprocim.ro/en/)

Maria SAHAGIA IFIN-HH, Horia Hulubei National Institute of R&D for Physics and Nuclear Engineering, Romania (www.nipne.ro)

Dan THOLEN

A2LA ILAC, USA (www.a2la.org)

STEERING COMMITTEE

Virginia-Graziela GUSLICOV CEPROCIM SA, Bucharest, Romania

Simona DZITAC University of Oradea and Research & Development Agora Ltd, Romania

Nicoleta VLAD CEPROCIM SA, Bucharest, Romania

Rares BENEA Blue Star Design, Arad, Romania

http://www.ceprocim.ro/en/

http://univagora.ro/

http://www.ceprocim.ro/en/

http://www.bluestardesign.ro/

Cristina DUMITRACHE CEPROCIM SA, Bucharest, Romaniahttp://www.ceprocim.ro/en/

CONTENTS

Invited lectures Mirella Maria Buzoianu WORK AT THE INM TO DEVELOP MEASUREMENT CAPABILITIES TO ASSIGN

REFERENCE VALUES IN PROFICIECY TESTING SCHEMES..................................

1 Eric Ziegler, Anne Tirard, Abdelkader Boubetra FEEDBACK FROM A CUSTOMIZED PT-TEST WITH AN ANTI-COLLUSION

DESIGN.........................................................................................................................

10 Steffen Uhlig, Bertrand Colson THE CASE OF LOW PARTICIPANT NUMBERS: TAKING LABORATORY

UNCERTAINTIES INTO ACCOUNT IN THE HAMPEL ESTIMATOR..........................

16 Proficiency Testing schemes (PTS) Leïla Ali Mandjee, Eloïse Riouall, Jean-Christophe Augustin and Vincent Carlier DEVELOPMENT OF A PROFICIENCY TESTING SCHEME IN FOOD

MICROBIOLOGY ON CAMPYLOBACTER...................................................................

24 Marleen Abdel Massih, Viviane Planchon, Elena Pitchugina, Florence Ferber and Jacques Mahillon A VIRTUAL PROFICIENCY TESTING IN FOOD MICROBIOLOGY............................ 32 Danica Boljević, Milenko Maričić, Nikola Maričić APPROACH AND OVERVIEW OF THE PT SCHEME IN ACOUSTIC FIELD.............. 41 Boris Constantin, Eric Ziegler, Anne Tirard, Abdelkader Boubetra, Azziz Assoumani, Hugues Biaudet PRESERVATION OF PARABENS IN WATER SAMPLES........................................... 49 Marie DANGERVILLE, Abdelkader BOUBETRA, Romain LE NEVE, Sandrine NGUYEN, Anne TIRARD MICROBIOLOGY PROFICIENCY-TESTING SCHEME IN COSMETICS.................... 57 Caterina MAZZONI, Anne TIRARD, Abdelkader BOUBETRA, François ROUSSEY, Eric ZIEGLER PROFICIENCY-TESTING SCHEME FOR HALOANISOLES AND HALOPHENOLS

IN OAK WOOD..............................................................................................................

62 Virginia-Graziela GUSLICOV 30 YEARS OF INTER-LABORATORY TESTS FOR CEMENT IN ROMANIA.............. 71 Ke-xiao Yu, Sha-sha Wu, Hui-chao Liang, Ying Xu, Xiao-ling Zhu, Zhong-bao Guo, Jian-bo Zhou, Ruo-lin Li RESEARCH ON THE PROFICIENCY TESTING PROGRAM OF DETERMINATION

OF SOLUBLE LEAD IN SYNTHETIC MATERIALS TRACK SURFACES....................

79 Sha-sha Wu, Ke-xiao Yu, Hui-chao Liang, Ying Xu, Jian-bo Zhou, Ruo-lin Li OVERVIEW OF PROFICIENCY TESTING IMPLEMENTATION OF STYRENE

BUTADIENE STYRENE MODIFIED BITUMIOUS SHEET MATERIALS......................

87

Ying Xu, Kexiao Yu, Shasha Wu, Xiaoling Zhu, Huichao Liang ESTABLISH A NATIONWIDE CEMENT TESTING COMPARISON PLATFORM TO

REALIZE INFORMATION-BASED MANAGEMENT.....................................................

94 Cristina Stancu THE 10TH EDITION OF INTERLABORATORY TESTS FOR ADHESIVES FOR

CERAMIC TILES - AN ANNIVERSARY EDITION.........................................................

99 Martin Vana, Eva Cizmarova SCOPE OF PROFICIENCY TESTING PROGRAMMES PROVIDED BY UKZUZ FOR

AGRICULTURAL MATRICES........................................................................................

107 Thibaut Cugnon1, Florence Ferber ², Jacques Mahillon3, Richard Lambert1 CONCLUSIONS OF A BLIND PROFICIENCY TESTING SCHEME IN MANURE

ANALYSIS.......................................................................................................................

111 José Ricardo Bardellini da Silva, Adelcio Rena Lemos, Carla Thereza Coelho, Paulo Roberto da Fonseca Santos BRAZILIAN’S NMI PROFICIENCY TESTING PROGRAMS AN IMPORTANT TOOL

FOR COMPETITIVENESS, INNOVATION AND INDUSTRY DEVELOPMENT............. 116

Participation at Proficiency Testing and Accreditation of Quality Management (PPT) Huichao Liang, Kexiao Yu , Shasha Wu, Jianbo Zhou, Ying Xu, Xiaoling Zhu, Ruolin Li，Ying Xu IMPLEMENTATION MODE FOR GOVERNMENT ORDERED PROFICIENCY

TESTING PROGRAM...................................................................................................

123 Mohammed Z. Alrabeh, Mohammed A. Almutairi, Mostafa A. Alsamti

TROUBLESHOOTING OF CHLORITE OVERLAPPING OVER BROMATE PEAK IN DRINKING WATER.......................................................................................................

127

Validation of Testing Methods and Proficiency Testing (VTM) Majda Alenezi, Adnan S. Almussallam, Abdullah T. Bawazir, Ahmed I. Al-Ghusn, Nawaf F. Alsowidan and Ahmad Hanafi IMPLEMENTATION AND VALIDATION OF FORMALDEHYDE ANALYSIS AT SFDA

LABORATORIES...........................................................................................................

135 Peter Kruspan, Markus Arndt, Klaus Lipus VALIDATION OF A METHODOLOGY FOR ASSESSING THE LEVEL OF

RESPIRABLE CRYSTALLINE SILICA (RCS) IN BULK POWDERS RELEVANT TO THE CEMENT INDUSTRY............................................................................................

143 Irina-Mihaela Ciortan

REVISITING A ROUND ROBIN TEST TOWARDS THE EVALUATION OF IMAGING SPECTROSCOPY SYSTEMS FOR CULTURAL HERITAGE APPLICATIONS.............

151

Adnan S. Al-Mussallam, Fahad S. Aldawsari and Nawaf F. Alsowidan PHTHALATES DETERMINATION IN COSMETICS AT SFDA LABORATORIES.......... 159

Uncertainty of Measurement and Proficiency Testing (UM)

Carla Palma, Vanessa Morgado, Ricardo J. N. Bettencourt da Silva MODELLING THE VARIATION OF MEASUREMENT UNCERTAINTY WITH THE

MEASURED VALUE FROM PROFICIENCY TEST RESULTS................................

165 Internal Quality Control of Analyses and Proficiency Testing (IQC) Gherghina Ciortan , Marina Martin STATISTIC CONTROL OF THE RESULTS FOR THE DETERMINATIONS OF

MAJOR OXIDES FROM A PORTLAND CEMENT, BY USING OF SHEWHART DIAGRAMS...................................................................................................................

173

Metrology, Traceability and Proficiency Testing (MT) Maria Sahagia, Aurelian Luca ABSOLUTE STANDARDIZATION OF THE RADIONUCLIDE 54Mn AND

PARTICIPATION AT INTERNATIONAL COMPARISONS..............................................

187 Others Jaroslaw Rachubik COMPARISON OF PT PERFORMANCE IN RADIOACTIVITY MEASUREMENT

LABORATORIES USING SCINTILLATION AND SEMICONDUCTOR DETECTORS....

194 Hui Ling Li, Khairul Anuar Abdul Aziz DEPARTMENT OF CHEMISTRY MALAYSIA PROFICIENCY TESTING PROVIDER

(JKM PTP): CURRENT PRACTICES AND FUTURE CHALLENGES.............................

195 Hulusi Demircioğlu, İlgi Karapınar INTER-LABORATORY COMPARISON TEST FOR ELECTROCHEMICAL

ANALYTICAL METHODS IN REAL WASTEWATER SAMPLE: Z-SCORE MODEL AND PARTICIPANTS EFFECT ABSTRACT..................................................................

196 Javier Luiso, Mariano Palacios DEVELOPMENT OF PROFICIENCY TESTING PROGRAM IN SOFTWARE SAFETY

VERIFICATION THROUGH NUMERICAL EQUIVALENCE TEST................................. 197

Javier Luiso, Mariano Palacios DEVELOPMENT OF PROFICIENCY TESTING PROGRAM IN RANDOM NUMBER

GENERATOR VERIFICATION ACCORDING TO REMOTE GAMBLING AND SOFTWARE TECHNICAL STANDARD..........................................................................

198 Mariano Palacios, Javier Luiso NEW APPROACH PT-PROGRAMS FOR SOFTWARE TESTING LABORATORIES

IN INNOVATIVE TECHNOLOGIES.................................................................................

199 Index of authors

The Seventh

International Proficiency Testing Conference

Oradea, Romania 10th − 13th September 2019

1

WORK AT THE INM TO DEVELOP MEASUREMENT

CAPABILITIES TO ASSIGN REFERENCE VALUES IN PROFICIECY TESTING SCHEMES

Mirella Buzoianu

National Institute of Metrology, Sos.Vitan-Bârzesti 11, Bucharest, Romania

[email protected]

Abstract Proficiency tests are increasingly used both by the accreditation bodies as a tool to ensure the credibility of their accreditation process and by the accredited laboratories and/or laboratories seeking for accreditation. Most of the proficiency testing schemes (PTS) available at national and European levels use evaluation criteria based on a consensus value from the participants results. The use of consensus value has already demonstrated its advantages and its limitations. Although the ISO/CEI 17043 standard allows the proficiency testing providers to externally subcontract the reference value used in a PTS, quite few of them apply this facility. This practice is however peculiar since any national institute of metrology is widely recognised a provider of a reliable reference value, especially when there is a capability recognised for this service, chemical field being well known for the reference measurements services. In this frame, the paper present recent scientific developments at the INM aiming at confirmation measurement capabilities to provide reference values in food matrices. Since 2009, different European metrology research programs enabled the National Institute of Metrology (INM) from Romania to join several projects in chemical field with the main outcome to developing and implementing different kinds of measurement capabilities. The European Metrology Programme for Innovation and Research (EMPIR) facilitated expanding the existing metrology capabilities regarding inductive couple plasma mass spectrometry (ICP-MS) with isotope dilution (ID) within the frame of EMPIR RPOT1601 contract. Aspects are presented regarding developing ID ICP-MS for measuring mass fractions of elements in a natural matrix. Key words Reference value, Calibration and Measurement Capability (CMC), metrological traceability.

Mirella Buzoianu: Work at the INM to Develop Measurement Capabilities to Assign Reference Values in PTS

2

1 INTRODUCTION National Metrology Institutes (NMIs) develop and maintain national measurement standards, based on the definitions of the quantities and units of the International System of Units (SI) or, where this is not yet possible, to other internationally recognized standards. The NMIs are the foundation of metrological traceability and disseminate it to their customers. Accordingly, the National Metrology Institute from Romania (INM) is mainly responsible for developing, maintaining and operating the 21 national measurement standards. The INM ensures the metrological traceability to the SI for its own standards and provides a wide range of services with the competence demonstrated and mutual recognised in the frame of CIPM MRA. Where intrinsic primary standards are involved INM ensures the metrological comparability. The Mutual Recognition of national measurement standards and of calibration and measurement certificates issued by national metrology institutes, CIPM-MRA, requires the INM to participate in key and supplementary comparisons providing the technical confidence in its measurement capabilities. Many Calibration and Measurement Capabilities (CMCs) claimed by INM, further evaluated, approved and published, cover also services related to providing reference values. Consequently, within the established international frame, the activity of the INM can cover also the proficiency testing (PT) needs. Therefore, the advantages of using metrological traceable reference values in PTS is ensured. Latest developments in the scientific activity of the INM related to measurement capabilities to assign reference values of mass fractions of metals in PT food samples are presented. This competence would be quite useful in certain proficiency testing schemes (PTS) in food area where the added value of using reference values (instead of the consensus values) for a more objective evaluation of laboratory performance has been demonstrated. 2 METROLOGICAL BACKGROUNDS AND CONCEPTS IN PTS The first purpose of an interlaboratory comparison quoted in the Introduction to [1], adopted also as national standard, is the ‘evaluation of the performance of laboratories for specific tests or measurements and monitoring laboratories' continuing performance’. When measurements are involved, it is of utmost importance to consider in such comparisons the interrelations between the test item (material under investigation), the applied measurement/testing procedure and the participating laboratory. The ‘evaluation of participant performance against pre-established criteria by means of interlaboratory comparisons’ defines the proficiency testing [1]. Since any interlaboratory comparison relies on tests/measurements, understood in a very broad scope, many concepts specific to measurement science apply in PTS. Note that main metrological concepts related to the technical requirements for PT schemes (PTS) have been presented in [2]. Briefly, validation, metrological traceability, measurement uncertainty, and accuracy (trueness and fidelity) are required in [1] at different stages of planning, designing, performing and reporting. Thus, metrological traceability and measurement uncertainty associated with the assigned value shall be included in the PT plan (clause 4.4.1.3 (q)), they need to be carefully considered in designing the statistical analyses (clause 4.4.4.3 (a)), they shall be taken into account in the documented procedure for establishing the assigned value in order to


3

demonstrate that the scheme is fit for purpose (clause 4.4.5.1) and they need to be reported to participants (clause 4.8.2 (m)). In addition, measurement accuracy need to be carefully considered in designing the statistical analyses (clause 4.4.4.3 (a)). 2.1 Metrology related concepts in PTS Measurement uncertainty and metrological traceability are intrinsic characteristics of any measurement result. Measurement results reported by any participant in a PT make no exception to this rule. The two above mentioned intrinsic characteristics are conditions for the metrological comparability and compatibility of such results. Consequently, metrological traceability, comparability and compatibility have a special place in any PTS. Note that the definitions of metrological traceability and measurement uncertainty are to be found in the chapter 3 ‘Terms and definitions’ of [1]. According to the definition of ‘metrological traceability’ also given in [3] only results with an appropriate uncertainty statement can be traceable. As often is outlined, the metrological traceability of a measurement result does not ensure that the measurement uncertainty is adequate for a given purpose or that there is an absence of mistakes during the measurement process. Metrological comparability of measurement results, or simply metrological comparability, is defined in [3] as comparability of measurement results, for quantities of a given kind that are metrologically traceable to the same reference. Note that metrological comparability of measurement results does not necessitate that the measured quantity values and associated measurement uncertainties compared are of the same order of magnitude. Metrological compatibility is defined in [3] as property of a set of measurement results for a specified measurand, such that the absolute value of the difference of any pair of measured values from two different measurement results is smaller than some chosen multiple of the standard uncertainty of that difference. Metrological compatibility represents the criterion for deciding whether two measurement results refer to the same measurand or not. If in a set of measurements of a certain measurand, thought to be constant, a measurement result is not compatible with the others, either the measurement was not correct (e.g. its measurement uncertainty was assessed as being too small) or the measured quantity changed between measurements. The ultimate outcome of a PTS is the evidence regarding the laboratory’s measurement capability defined in [4] as the ‘ability to measure a quantity of a given kind, in a specified interval of quantity values, embodied in a specified material, as demonstrated by a measurement uncertainty’ with the note: ‘A comparison of the measurement uncertainty in the measurement result obtained by one participant to that of the measurement result obtained by another laboratory for the same quantity in the same material, compares their respective measurement capabilities.’ 2.2 Reference value in a PTS A major factor of risk in evaluating the performance of a participant in any PTS is the way the PT provider chooses to assign the value, further used as reference value, with its standard deviation. Five ways of determining the assigned value in a PTS are described in [5] based on: - calculation from the masses of properties used/added (formulation); - the certified value of a CRM used as the test item; - an assigned value determined and given by a single laboratory using a reference (primary) method of measurements;


4

- a consensus value given by expert laboratories; - a consensus value from the participant results. Each of the above mentioned way has its advantages and its limitation described in [5]. From a metrologist point of view the safest way is to obtain the value of a known sample (such as a CRM) or given by a reliable reference laboratory (such as a National Institute of Metrology with a CMC published in the BIPM data base [6]). On many occasions, even the participants expect an independent and as traceable as possible reference value; in such situations, their evaluation would be performed and documented in an undisputable and objective way. However, the common approach in practice is the use of consensus value from the participant results by classical or robust statistics. In this respect, a study performed during 2005-2012 [7] on PT providers from different European countries and USA revealed interesting conclusions. One of these referred to the impact of data to the probability distribution when performing the assessment of laboratories in PTS with consensus values from the participants in the PT. According to the definition given in [5]: ‘consensus value is the value derived from a collection of results in an interlaboratory comparison’ with o note to entry stating that ‘the phrase ‘consensus value’ is typically used to describe estimates of location and dispersion derived from participant results in a proficiency test round, but may also be used to refer to values derived from results of a specified subset of such results or, for example, from a number of expert laboratories’. From metrology perspective a first question raised [8] was whether a ‘consensus value’ is actually a ‘measured value’ since it appears to be the ‘consequence of a decision supposedly taken after the ‘absence of a sustained opposition’. Also, it is used in connection with pooling measurement results that are more or less independent. In metrological terms consensus value was interpreted as ‘not inconsistent with the relevant set of measurement results’ [9]. In many PTS the participants are not requested to declare their measurement uncertainty, de facto thus preventing any evaluation. Further, the consensus value does not take into account the measurement uncertainty associated with the measurement result reported by the participants. Therefore, there are still some arguments in debating the quality of a consensus value [8,9]. As indicated in [5], an assigned value of a proficiency test item may be determined by a single laboratory using a suitable reference measurement method, preferable primary, completely described, understood and appropriate for the PTS or from a calibration against the reference values of a closely matched certified reference material. Reference values provided by highly skilled, with proved competence applying measurement methods of appropriate metrological level (i.e. with an associated measurement uncertainty smaller than a target measurement uncertainty set in the planning step), clearly allow a fair and scientifically thorough interlaboratory comparison evaluation. In addition, proficiency data with stated uncertainties represent the best opportunity for also testing the consistency of the uncertainty claims with [10]. This case where reference values are provided by highly competent laboratory applies mostly to the NMIs that disseminate the traceability to PT providers via CMCs available in the BIPM KCDB. This goal is looked upon in metrology in chemistry area at the INM, as illustrate in BIPM KCDB [6]. Some results described below demonstrate the interest to also expand the range of matrices in near future. Note that frequently one may observe that the quality of laboratory evaluation is better when using a reference value. Successful PT schemes require well characterised test materials, sufficiently homogenous and stable within the period of the PT. In practical terms these kind of PT


5

materials should qualified as reference materials. Consequently the SR EN ISO/CEI 17043:2010 includes specific requirements related to the homogeneity and stability of the test materials starting with the planning up to the final reports (4.8.2 (i)). When CRMs are used (portions of a CRM unknown to participants), although more expensive, the PTS make use of reference values traceable assigned. The reported results are comparable independent of number of participants. An obvious limitation is related to the commutability of the CRM. Assigned values can be determined using an interlaboratory comparison with expert laboratories using a protocol that specifies the numbers of proficiency test items and replicates and any other relevant conditions. The results provided by the expert laboratory with measurement uncertainty statement, correspond to the definition given in [3] as well as to the GUM approach [10], i.e. ‘in general, the result of a measurement is only an approximation or estimate of the value of the measurand and thus is complete only when accompanied by a statement of the uncertainty of that estimate’. Depending on the number of results reported by the expert laboratories, it is then the PT provider task to establish/estimate the reference value of the PTS to be further used in assessments. Working with reference values provided outside the laboratories participating in a PTS has some clear advantages synthetized in [11]. They mainly refer to: - metrological sound evaluation by comparing the reported results with measurement uncertainty associated against a reference value with associated measurement uncertainty not influenced at all by any participant; - solving the issue of whether eliminating the outliers; each measurement result reflects the measurement capability and gives indications on further corrective actions; - elimination of a variable reference value given by consensus value/average; - wider application both for small and large population of participants, especially since consensus value showed its limits in PTS with a small number of participants; - better monitor the performance of a participant over a long period of time. 2.3 Metrology services in support to PTS The national measurement standards maintained and operated at INM cover main measurement fields of mass, length, angle, pressure, force, electrical quantities, time and frequency, acoustics, temperature and optical quantities. The reference measurement standards cover in addition flow, ionization radiation and some physico-chemical quantities areas. The established measurement standards (the national and reference ones) are disseminated by providing a wide range of services: calibration, reference measurements, CRMs etc. In support to PTS, the INM is capable to provide reference values in almost every area of activity, chemistry included as indicated in the 197 CMCs published on the BIPM data base [6]. With its main objective to establish the degree of equivalence of the national measurement standards to insure world-wide uniformity of measurement and to provide for the mutual recognition of calibration and measurement certificates issued by the NMIs the CIPM MRA allowed INM to fully integrate in the international system of measurements and to provide mutual recognised metrology services. Aspects of the active participation of the INM in the key comparisons organized by the Consultative Committees created by the CIPM and/or in key and supplementary regional comparisons organised by Regional Metrology Organisations (RMOs) to establish INM’s technical confidence, are described in [12].


6

3 RECENT ACTIVITES IN THE INM TO DEVELOP MEASUREMT CAPABILITIES TO ASSIGN REFERENCE VALUES TO PTS The EMPIR 16RPT01 ChemMet-Cap project [13], entitled ‘Development of scientific and technical capabilities in the field of chemical analysis’, started at 1st of June 2017, to addresses several scientific and technical objectives: - developing traceable measurement capabilities for the analysis of heavy metals (for concentrations at ng/kg and mg/kg levels, with uncertainties <10 % by developing isotope dilution mass spectrometry (ID-ICPMS) methodology as a primary procedure for elemental determination; - applying the ID-ICPMS methods developed to environmental and food samples to determine the heavy metals content in representative matrices, such as potable and natural waters, sediments, and different types of fish/biota samples; - validating the developed methods (ID-ICPMS) by participation in suitable international comparisons and hence to underpin the development of appropriate CMCs (Calibration and Measurement Capabilities) for submission to the BIPM Key Comparison Database. The EMPIR 16RPT01 project is carried out by four NMIs from France (LNE), also acting as project coordinator, Romania (INM), Bulgaria (BIM), Turkey (TUBITAK UME), one Designated Institute (IAPR/EXHM/GCSL-EIM) from Greece and one research institute Tunisia During very first activities of the EMPIR 16RPT01 ChemMet-Cap project INM developed the ICP-MS techniques with direct and reverse isotope dilution (ID) due to the fact that the IDMS was recognized the most accurate analytical tool capable of determining the mass fraction of an analyte (element) with lowest associated uncertainty and highest technical quality. In the frame of the project an exercise was planned and performed where the partners had to assign mass fraction values of Hg, Zn, Cu and Se to two samples of seabass material provided by the TUBITAK in order to verify and validate the implemented direct ID-ICPMS method for food matrix. Mass fraction of copper, zinc and selenium in the seabass material were measured at the INM partner. The applied measurement procedure included moisture determination, digestion step of the samples spiked with the isotopes of interest, measurement of isotope ratios, data treatment and measurement uncertainty evaluation and reporting the measurement result. The ERM BB442 Elements in Fish Muscle was treated in the same way as the seabass samples for validation purposes. The material moisture was determined in the range of (3.05 … 3.23) % for the two received bottles, in accordance with the recommendation of TUBITAK. The moisture of the ERM was estimated as indicated in the accompanying documentation provided by the producer. The microwave digestion program: 3 min-250 W-100 °C; 5 min-1000 W-180 °C; 10 min-1000 W-180 °C; 5 min-0 W-50 °C, was applied on amounts of samples/ERM BB442 lying in the range of (0.2 … 0.8) g. Nitric acid TraceSELECT 69.0 % Fluka Analytical was used. Copper, Zink and Selenium Isotope standards produced by VHG-LGC, were used to spike the samples and ERM in the digestion step. The isotope ratios of 63Cu/65Cu, 64Zn/68Zn and 78Se/82Se were measured with the optimized instrument ELAN DRCe, Perkin Elmer type.


7

NIST SRM 3168 and INM Cu primary standard solution (9.93 ± 0,026) g/kg were used as calibration standards. A preliminary step related to the determination of the isotopic composition on nonspiked samples was included in the measurement procedure. For that purpose, the sample and the ERM were digested without spike using the same digestion program. The results reported by the INM are summarized in table 1.

Table 1 . Mass fraction of Zn, Cu and Se reported by the INM

Bottle No. Run No. Sample mass Mass fractions (mg/kg)

(g) Zn Cu Se

56

1 0.45912 15.38 1.194 0.990 2 0.52215 16.83 1.086 0.981 3 0.47231 16.08 1.109 1.072 4 0.44682 15.50 1.142 1.008 5 0.71865 16.44 1.116 1.013

30

1 0.36891 15.91 1.158 1.045 2 0.49356 16.42 1.281 1.025 3 0.46439 15.71 1.147 1.025 4 0.81121 17.32 1.111 0.980 5 0.51563 18.12 1.105 0.970

Mean value 16.37 1.145 1.011 Combined standard uncertainty (uc) 1.00 0.059 0.050

Coverage factor (k) 2 2 2 Expanded uncertainty (U) 2.00 0.118 0.100

Range of accepted mass fraction, mg/kg 16.4 …19.4 1.10 … 1.30 0.87 … 1.13

As one may note all the measurement results reported by the INM overlapped well with the range of accepted mass fraction values for Zn, Cu and Se. To measure the mass fraction of Se in sample, the following measurement equation was considered:

(1) where: -ω82Se_spike is the mass fraction of Se the spike, mg/kg; - mspike is the mass of spike, g; - msoln is the mass of sample, g; - MSe_smp is the Molar mass of Se in sample, g/mol; - MSe_spike is the Molar mass of Se in spike, g/mol; - A78Se

spike is the abundance of 78Se in spike; - A82Se

spike is the abundance of 82Se in spike; - A82Se

smp is the abundance of 82Se in sample; - A78Se

smp is the abundance of 78Se in sample; - Rblend is the isotope Ratio 78Se /82Se in the blend; - ωblk is the mass fraction of Se in the digested blank, mg/kg; - fdil, dilution factor; - - is the correction factor for moisture.


8

Accordingly, the uncertainty budget for selenium is presented in table 2.

Table 2 . Uncertainty budget for selenium measurements

Quantity Value Standard

uncertainty Probability distribution

Relative standard

uncertainty mass fraction of Se the spike, ωb, mg/kg 10.009018 0.0991 normal 0.009901 Molar mass of Se in sample, Mx, g/mol 78.971 0.008 rectangular 0.0001013 Molar mass of Se in spike, Mb, g/mol 81.9882 0.008 rectangular 0.0000975 mass of sample, mx, g 35.56439 0.00020 normal 0.0000056 mass of spike, msp, g 0.029528 0.00028 normal 0.0094827 abundance of 82Se in sample 0.08825 0.00040 rectangular 0.0004533 abundance of 82Se in spike 0.9972 0.001 rectangular 0.0010028 abundance of 78Se in spike 0.0003 0.00007 rectangular 0.2333333 abundance of 78Se in sample 0.23685 0.00007 rectangular 0.0002955 Ratio 78Se/82Se in blend, Rblend 0.463615 0.0655 normal 0.0141372 dilution factor, fdil 81.54752 0.0005 normal 0.0000061 mass fraction of the blank, ωblk, mg/kg 7.400 0.008 normal 0.0010811 mass fraction of Se the sample, ωx, mg/kg

0.970 0.050 0.0515464

Next step planned for the project include a PTS organised by the Greek partner in October-December 2019, on a lyophilised tuna fish spiked with heavy metals. The reference value of the PTS will be provided by the group of expert laboratories (LNE, TUBITAK UME, INM, BIM and IAPR). The intended mass fraction ranges are (0,005 … 0,5) mg/kg for Cd; (0,05 … 5,0) mg/kg for Pb, Hg, Mn, Cr, As and Cu; and (5 … 500) mg/kg for Zn. At the end of the project in 2020, INM expects to be able to meet some of the requirements needed to claim new CMCs in food area for providing reference values with the main application in PT schemes. 4 CONCLUSIONS The paper presented aspects regarding the advantages of using reference value in PTS and selected outcomes of the participation of the INM in the EMPIR16RPT01 project. The added value of the research activities performed in European metrology projects is reflected in developing and implementing new measurement methods (ID ICP-MS) of inorganic species consistent with the European practice. The INM contributes directly to the comparability and traceability of measurement results performed in Romania. Traceability is disseminated by means of calibrations, reference measurements, as well as some certified reference materials in line with the state-of-art at the European level. The available proficiency testing schemes in Romania partly covers the needs of accredited laboratories. In addition, the lack of reliable reference values in such schemes limits the possibility to improve the analytical quality.


9

REFERENCES

[1] ISO/IEC 17043:2010 ‘Conformity assessment — General requirements for proficiency testing’

[2] M.Buzoianu “Key Measurement Concepts Required in PT Schemes” Proceedings of the Fifth International Proficiency Testing Conference, Timisoara, Romania, (15)16th − 30th September, 2015

[3] JCGM 200:2012 International vocabulary of metrology — Basic and general concepts and associated terms (VIM); VIM Definitions with Informative Annotations 2017

[4] P.De Bièvre, R.Dybkaer, A.Fajgelj, D.Brynn Hibbert, Metrological traceability of measurement results in chemistry: Concepts and implementation (IUPAC Technical Report), Pure Appl. Chem., Vol. 83, No.10, (2011):1871–1933

[5] ISO 13528:2015 ‘Statistical methods for use in proficiency testing by interlaboratory comparison’

[6] www.bipm.kcdb.org [7] F.Medeiros de Albano, C.Schwengber ten Caten, Proficiency tests for

laboratories: a systematic review, Accreditation and Quality Assurance Volume 19(4) (2014):245-257

[8] P.De Bièvre, Is ‘consensus value’ a correct term for the product of pooling measurement results?, Accreditation and Quality Assurance Volume 17 (2012):639:640

[9] K.Heydorn, The quality of consensus value, Accreditation and Quality Assurance Volume 18 (2013):243:245

[10] JCGM 100:2008 Evaluation of measurement data — Guide to the expression of uncertainty in measurement

[11] P.De Bièvre, Measurement results in a Proficiency Testing Programme need to be evaluated against independent criteria, Accreditation and Quality Assurance Volume 21 (2016):167:169

[12] M.Buzoianu “Role of the IINM in Providing Reference Values in Proficiency Testing Schemes” Proceedings of the Sixth International Proficiency Testing Conference, Cluj - Napoca, Romania, (03)04th − 06th October 2017

[13] Publishable Summary for 16RPT01 ChemMet-Cap Development of scientific and technical capabilities in the field of chemical analysis, April 2018, https://www.euramet.org/research-innovation/search-research-projects/details/project/development-of-scientific-and-technical-capabilities-in-the-field-of-chemical-analysis/?

The Seventh



10

FEEDBACK FROM A CUSTOMIZED PT-TEST

WITH AN ANTI-COLLUSION DESIGN

Eric Ziegler , Anne T irard , Abdelkader B oubetra

BIPEA, 189 rue d’Aubervilliers, 75018 PARIS – FRANCE

[email protected]

Abstract Interlaboratory comparisons are a key step for the recognition of laboratories skills and performance, whether for accreditation purpose, supplier/ customer referencing and so on. Therefore, the consequences of poor results obtained in proficiency tests can be very important and so participants can try to avoid such unsatisfactory results with more that only their analytical abilities. Laboratories can especially communicate with other participants in order to compare their results to at least check their results before submitting. The ISO 17043 standard requires to take reasonable precautions to prevent collusion between participants or falsification of results. It remains nevertheless quite hard for a proficiency tests provider to detect such issues and avoid some possible biases. However, in the framework of customized proficiency test ordered by a customer for a specific purpose, things are a little different than for regular tests with volunteer participants. Such customers are now aware of such issues and ask for avoiding collusion in order that the requested test is efficient. Following such a request, two rounds were carried out. The design especially included the productions of two different levels for some of the defined series, not indicating to the participants which one the two levels they were sent. The tests provided some valuable information even it was not always that easy to conclude. The results of these tests, the advantages and limitations of such a design are presented in this paper. Key words Proficiency-testing schemes, collusion, laboratory performance

Eric Ziegler, Anne Tirard, Abdelkader Boubetra: Feedback from a customized PT-test with an anti-collusion design

11

1 INTRODUCTION BIPEA provides a wide range of proficiency testing schemes in feed, food, beverages water, soil, fertilizers, cosmetics, etc. In addition to proficiency tests on regular basis, customized interlaboratory comparison can be implemented on request of a specific customer. In such test, the supplier can request all characteristics are fit for purpose regarding its needs, in terms of matrix, parameters, content and concentrations, and so on. More and more, avoiding collusion is part of the request. For the renewal of a collaboration presented in this paper, the customer even listed this collusion issue as the main point, indicating that the tests would be stopped if no satisfaction was given on this subject. In consequence, a specific design was implemented to meet this need. 2 DESIGN 2.1 Distribution of the samples

Each round consists of three series of samples (S1 to S3), with different parameters to be analysed for each series. Each participant is enrolled for a defined set of parameters according to its activity, and receives from one to three samples to cover its enrolment. For the first round, it was decided to produce two batches of samples for each of the three series (Batch A and Batch B). A design was built to distribute efficiently the three series of samples between the participants. The test being worldwide, the main issue of collusion is between participants from a same geographical area (suspicions actually emerged from past tests in some areas). Consequently, the distribution was organised as presented in the table below.

Table 1. Design of the distribution of the samples between the participants

Laboratory Country Sample S1 Sample S2 Sample S3 Laboratory 1 Atlantide B - - Laboratory 2 Atlantide A - - Laboratory 3 Atlantide B A - Laboratory 4 Atlantide A B A Laboratory 5 Atlantide B B B Laboratory 6 Atlantide B - - Laboratory 7 Atlantide A - - Laboratory 8 Atlantide B A Laboratory 9 Atlantide B A B

Laboratory 10 Eldorado B - B Laboratory 11 Eldorado A A A Laboratory 12 Eldorado A - B Laboratory 13 Eldorado B B A Laboratory 14 Lilliput A B B Laboratory 15 Lilliput B A A


12

Two participants of a same country can have the same set of samples, a partly different set, or a fully different set. The samples were labelled exactly the same way between the two batches of a same series, with for example written Sample S1 and the name of the sample. Only the bar code differs from one batch to another, which does not allow the participant to see any difference. Once the participants registered according the defined pattern, the bar code allow not to mix-up any sample, the appropriate bar code requiring to be scanned for all the samples of a dedicated parcel. 2.2 Content of the samples The difference of concentrations/contents of the parameters to analyse between the two batches is an important factor to define. The goal is to have not too different levels, otherwise it could be too obvious, but different enough to be able to conclude on the results regarding for example the uncertainty of the concerned determination. An example of concentrations used in the test is presented below.

Table 2. Concentration of the parameters in the two different batches of a round

Parameter Concentration batch 1

Concentration batch 2

Parameter 1 593 737 Parameter 2 300 600 Parameter 3 92 115 Parameter 4 1.30 1.62 Parameter 5 2.64 3.30 Parameter 6 159 199 Parameter 7 9 13 Parameter 8 15 28

Parameter 11 180 80 Parameter 12 160 70 Parameter 13 1.8 2.25 Parameter 14 180 80 Parameter 15 160 70

2.3 Communication As for previous rounds, the test was announced to the participants, with especially the shipment of a start-up information containing the needed information for the test (dates, description of the samples, submission of the samples and so on), but of course without mentioning that they will not have the same samples.


13

RESULTS AND DISCUSSIONS For each parameter, it was especially focus on unsatisfactory results that were close to the assigned value of the sample from the other batch. The table below shows the results for one parameter.

Table 3. Set of results for one of the parameters

Assigned values: 767 - SDPA: 8 Result z-score Result z-score Result z-score Result z-score (A) 765 -0,25 (A) 770 0,38 (A) 760 -0,88 (A) 774 0,88 (A) 772 0,63 (A) 611 -19,50 (A) 791 3,00 (A) 770 0,38 (A) 772 0,63 (A) 790 2,88 (A) 765 -0,25 (A) 785 2,25 (A) 768 0,13 (Z) 770 0,38 (A) 771 0,50 (A) 775 1,00 (A) 813 5,75 (A) 765 -0,25 (A) 774 0,88 (A) 761 -0,75 (A) 697 -8,75 (A) 767 0,00 (A) 765 -0,25 (A) 767 0,00 (A) 770 0,38 (Z) 751 -2,00 (A) 750 -2,13 (A) 620 -18,38 (A) 720 -5,88 (A) 774 0,88 (A) 766 -0,13 (A) 772 0,63 (A) 770 0,38 (A) 770 0,38 (Z) 763 -0,50 (A) 766 -0,13 (A) 776 1,13 (A) 760 -0,88 (A) 758 -1,13 (A) 760 -0,88 (A) 773 0,75 (A) 771 0,50 (A) 760 -0,88 (A) 770 0,38 (A) 766 -0,13 (Z) 775 1,00 (A) 770 0,38 (A) 768 0,13 (A) 761 -0,75 (A) 771 0,50 (A) 755 -1,50 (A) 776 1,13 (A) 759 -1,00 (A) 769 0,25 (A) 775 1,00 (A) 770 0,38 (A) 757 -1,25 (A) 771 0,50

For this parameter, two results appear especially to be suspicious. First, they show the biggest z-score and most importantly they are very close to the assigned value of 610 met for the other batch. Both participants did not determine many parameters in this test and were out of range only for this parameter. One of the two laboratories had only one other fellow countryman in the test which had a sample of the second batch. The second of the two laboratories had three fellow countrymen in the test, one having the same sample and two having the sample of the other batch. For the first round with this design, about fifteen of such occurrences were met in the first round, with especially three for a same participant. Some other suspicious data were observed but being more in-between. It could correspond to either laboratories that try to average some results or simply far away results. Even for results that appear to be quite obviously related to the other batch, it remains however possible to face other explanations. Nevertheless, a specific tangible case come out. Full of honesty and innocence, one of the participants contact us to indicate that there was homogeneity issue with our samples. At a first step two laboratories from a same country compare their results and obtain quite different results. They therefore carry out the analyses again and confirm their results. They then decided to send each other their samples to performed crossed analyses. Of course, they confirm on each sample what the other


14

laboratory have found and conclude they were an issue with at least one of the samples. This example shows how far can participant go to check whether the results they are ready to submit in an interlaboratory comparison are consistent. It also highlights how important is the difference between the two levels. For the mentioned participants, in this case, it was obviously too different to lead them to submit their results as they were obtained. On the other hand, for other parameters, assigned values appear to be not different enough to appreciate potential collusion. The tuning of the level’s difference requires experience and cannot be easily or necessarily met for all the parameters. An issue met with this design concerns parameter with fewer results. The population being split, it can remain rather few results for some more specific parameters. As the assigned value is estimated from the robust mean of the participants, the reliability of the assigned value can be impacted. In this case collusion, can first hardly be seen, but moreover proficiency assessment which remains the first goal of the test can be less efficient. For example, it can be needed to widen the standard deviation for proficiency assessment to counterbalance the uncertainty on the assigned value due to a reduce number of results. The sending of the results also appeared to rise some difficulties. First of all, there are actually two reports for each series. It was not mentioned in the report that there were two different batches and so on, but of course only half of the participants are in each report which can be a clue for regular and observant participant. It is then needed that each participant receives the reports corresponding to the received which can be managed with a careful registration and organization. Nevertheless, in this specific case, there were some regional supervisor, and it was not easy for them to understand how it works. It especially happened that a regional supervisor forwards some reports to participants for which it was not intended, and there were consequently some questions from participants not finding their results in the report. The subjects mentioned above lead especially the participants (at least some of them even if probably not all) to be aware that they could receive different samples than another participant. In consequence such a design can become less efficient trough time and actually less obvious crossing were met during the second round. However, even if known, it gives less opportunity to the participants to check their results between them. It can therefore encourage them to rely more on theirselves and provide their results as they obtain them, which match the goal of being able to assess individually the proficiency of the participants. Nevertheless, in order not establish a new routine, it is decided to change the design the test from one round to another. Double batch are now produced form some of the series at each round but not necessarily for all. It first allows to reduce the number of batches to produce, as it represents a significant amount of work, and then to have for some tests bigger populations of results.


15

3 CONCLUSIONS The issues related to collusion become increasingly important. It is now requested by some customers or even for call for public tenders, to include some series for which all the participants have not the same samples. The use of two series of samples in a same proficiency test can actually allow to detect such specific collusion issues, at least for the most obvious ones, as it is in fact not easy to conclude in most of the case. Too much should consequently not be waited from such a design, but at least, even if clear evidences are not seen, it encourages the participants to deal with the test only by their own. The difference between the level/content of the two series is an important factor to think of, as the difference should not be too big, appearing as too obvious, or on the opposite too similar and therefore not allowing to conclude. The first use of such design is probably the most efficient as being unexpected by participants. If the tests are repeated, it appears to be important not to keep systematically the same design, to maintain some uncertainty about what is provided. The number and frequency of such design shall also be defined according to the purpose, as of course such design requires more works for the proficiency tests provider. These kinds of tests shall finally be dedicated to tests with a sufficient number of participants. Indeed, for these tests, the number of provided results is split between the two batches. If the assigned values are estimated independently from the data of the test, there is no issue, but as most of the time the assigned values are estimated from the robust mean of the participants, a low number of results can impact the reliability of assigned values and therefore be more damaging than useful. REFERENCES

[1] ISO 17043:2010 - Conformity assessment - General requirements for

proficiency testing

The Seventh



16

THE CASE OF LOW PARTICIPANT NUMBERS: TAKING LABORATORY UNCERTAINTIES INTO ACCOUNT IN THE HAMPEL

ESTIMATOR

Steffen Uhlig , Bertrand Colson

QuoData GmbH, Prellerstr. 14, 01309 Dresden, Germany

[email protected], [email protected]

Abstract According to ISO/IEC 17025:2017, testing and calibration laboratories should apply procedures for estimating measurement uncertainty. It would thus make sense to include these uncertainty estimates in the evaluation of laboratory performance in proficiency testing. According to ISO 13528:2015, submitted laboratory-specific uncertainties can be taken into account in the assessment of the performance of PT participants by means of ζ (or En) scores. It is important to note, however, that the laboratory uncertainties are not taken into account in the actual estimation of consensus means. In this paper, a proposal for a modification of the Hampel estimator is presented. In the original Hampel estimator, the laboratory weights are determined on the basis of a single reference quantity (e.g. ). In the modified estimator, this single reference quantity is replaced by the laboratory-specific uncertainties. Thus, all available information is taken into consideration in the estimation of the consensus mean, ensuring a greater level of reliability. Since is no longer needed in the computation of the consensus mean, the requirements regarding minimum participant numbers can thus be relaxed. Key words Measurement uncertainty ,Consensus mean, Hampel estimator, laboratory assessment

Steffen Uhlig, Bertrand Colson: The case of low participant numbers: Taking laboratory uncertainties into account in the Hampel estimator

17

1 INTRODUCTION The past few years have witnessed renewed interest in the application of robust algorithms for the computation of consensus values in connection with laboratory performance assessment in proficiency testing (PT) rounds. In particular, in the ISO 13528:2015 standard [1], not only are several different robust estimators presented (for the computation both of the consensus mean and of the standard deviation for proficiency assessment ) – but their respective merits and shortcomings are discussed in the light of objective criteria such as efficiency and breakdown point. In the normative Annexes C and D, it is shown that the Q/Hampel method – though computationally intensive – offers a higher degree of reliability. Simultaneously, the question how to meaningfully take reported measurement uncertainty estimates into account in the evaluation of PT data is currently receiving increased attention. Indeed, according to ISO/IEC 17025:2017 [2], laboratories shall estimate the uncertainty of test results and calibrations. Strictly speaking, a lone test result (i.e. a test result without uncertainty interval) is meaningless; uncertainty estimation constitutes an integral part of every measurement procedure. It would thus make sense to include these uncertainty estimates in the assessment of laboratory performance. According to ISO 13528:2015, submitted laboratory uncertainties can be taken into account in the assessment of the performance of PT participants by means of (or ) scores. As far as taking the laboratory uncertainties into account in the actual estimation of the consensus mean is concerned, however, no detailed procedures are provided in ISO 13528:2015 – though an overview of possible approaches and relevant considerations is provided in Section 7.6.3, including weighed averages and making allowing for over dispersion. Combining the focus on robust algorithms and the need to incorporate uncertainty estimates in the assessment of laboratory performance, it would make perfect sense to take the uncertainties into account in the computation of the consensus mean by means of a suitable modification of a robust estimator. Given the reliability of the Q/Hampel algorithms – as measured in terms of breakdown point and efficiency – this paper proposes a modified Hampel estimator in which reported measurement uncertainty estimates determine the size of the individual contributions to the consensus mean. This modified Hampel estimator is called the measurement uncertainty-weighted Hampel estimator – abbreviated as MU-weighted Hampel. Since the use of the MU-weighted Hampel estimator circumvents the need to compute a estimate, this algorithm ensures a high degree of reliability even in cases where the number of participants is very low (say, 6 participants) – as long as reported measurement uncertainty estimates are sufficiently accurate. Indeed, the approach described here may even be applicable in cases where the number of participants is as low as 3 or 4 participants. 2 THE MU-WEIGHTED HAMPEL ESTIMATOR In order to identify a satisfactory approach for including reported standard uncertainties in the computation of the consensus mean, it is reasonable to propose the following two requirements


18

a) test results will be downweighted if the corresponding uncertainty estimates are high (i.e. a similar effect to the weighted arithmetic mean);

b) inconsistencies between deviations and standard uncertainties will be penalized (i.e. the computational procedure includes a built-in assessment of the uncertainty estimate).

The proposal presented in this paper consists in a modification of the Hampel estimator described in ISO 13528:2015 designed to meet the two above requirements. In Annex C.5.3.3 of ISO 13528:2015, the Hampel estimate is implicitly defined as the solution of the following equation

Equation 1

where denotes the mean value across test results for laboratory , denotes the number of laboratories and (Psi) denotes the piecewise defined function displayed in Fig. 1.

1.5-1.5 3 4.5-4.5 -3

1.5

-1.5

Fig. 1. Ψ (Psi) function of the Hampel estimator There is no explicit computational rule for the Hampel estimate based on this function. For a discussion of different approaches to solving Equation 1, the reader is therefore referred to the theory of robust M estimators for which this function plays a central role [3]. In the MU-weighted Hampel estimator, the Hampel function is applied not to the

but to the values, where denotes the reported standard uncertainty

estimate for laboratory . This means that the influence of the individual laboratory mean values is determined on the basis of a separate criterion for each laboratory. The larger the reported value, the larger the tolerance towards deviations before downweighting takes place. Conversely, if inconsistencies between observed deviations and reported standard uncertainties are present – i.e., if large deviations are observed while the reported uncertainties are low – the influence of the corresponding results on the estimate will be reduced. This ensures that requirement (2) set forth at the outset of the present section is fulfilled.


19

Turning now to requirement (1), we must address another aspect of the proposed modifications to the Hampel estimator. Indeed, in the MU-weighted Hampel estimator, the function itself is weighted in terms of the reported standard uncertainty estimates: the higher the reported uncertainty, the lower the influence. This is illustrated in Fig. 2 for three different values of the standard uncertainty . Note that the value corresponds to the original Hampel function. It is clear that this modification ensures that the requirement (1) is also met; namely, that test results will be downweighted if the corresponding uncertainty estimates are high – in accordance with the basic principle of the weighted arithmetic mean (where the weights are defined as the squared inverse uncertainty estimates). Indeed, the weighted function behaves identically to the function of the weighted arithmetic mean, on the interval - to .

Fig. 2. The weighted Hampel (Psi) function for three standard uncertainty ( ) values. The x axis represents the standardized deviation (i.e. the difference between laboratory mean and Hampel estimate, divided by the laboratory

uncertainty), while the y axis represents corresponding Psi values. The weighting mechanism functions as follows: the higher the laboratory

uncertainty estimate, the lower the weighting. Note that the value corresponds to the original (i.e. unweighted) Hampel function.

3 WHEN CAN THE APPROACH BE APPLIED? It should be ensured that the uncertainty estimates reported by the participant laboratories were obtained on the basis of a reliable and realistic approach. Failing this, the MU-Hampel cannot be seen as a more reliable estimator than the Hampel estimator described in ISO 13528:2015. Ideally, it would be possible for PT providers to apply a procedure which allowed them to check the reliability of the laboratory uncertainty estimates. Guidance is provided in ISO 13528:2015, Section 9.8. One


20

possible approach is the Naji plot, a graphic tool simultaneously providing information about laboratory deviations compared to ( scores), laboratory deviations compared to the uncertainty of the consensus mean and laboratory uncertainties ( scores) and the ratio between standard laboratory uncertainties and . As long as

the latter lies below a certain threshold (e.g. ), satisfactory scores can be

seen as evidence that the corresponding reported uncertainty estimates are accurate. Note that if scores are unsatisfactory – i.e. the corresponding uncertainty estimates are too low – then the present approach can be expanded to take this into consideration: namely, an extra source of variability called “dark noise” is computed in such a way as to inflate all the uncertainty values. It is important to include the dark noise in the computation of the MU-weighted Hampel mean when there is a large sample-heterogeneity component, and/or the laboratory uncertainties are implausibly low. Further information regarding the applicability of this approach can be found in [4]. 4 APPLYING THE NEW APPROACH IN THE CASE OF LOW PART ICIPANT NUMBERS The following 3 examples will illustrate the reliability of the proposed estimator in the case of low participant numbers. Each of the three examples shows test results (diamonds) and expanded uncertainty intervals (boxes) for 6 laboratories. Horizontal lines represent the consensus mean estimates computed using 4 different estimators; arithmetic mean, Q/Hampel, MU-weighted Hampel and MU-weighted Hampel with dark noise. From one example to the other, slight modifications are observed, resulting in instructive changes in the consensus means.

Fig. 3. Comparison Q/Hampel, MU-weighted Hampel and MU-weighted Hampel with dark noise – Example 1. Two laboratories are outliers with respect to level and

uncertainty. The MU-weighted Hampel estimate lies clearly above the other three estimates.

LaboratoryL16 L8 L13 L11 L9 L24

mg/

kg

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1


21

In the first example (Fig. 3), two of the laboratories (L16 and L8) have both lower recovery rates and higher uncertainty values. The values of the 4 remaining laboratories lie close to each other, and the uncertainties are lower. The result is that the MU-weighted Hampel lies higher than both the arithmetic mean and the Q/Hampel estimate. The dark noise is nonzero, meaning that the reported uncertainty estimates do not account for all observed variability – that is say, the deviations of laboratories L16 and L8 are too large in comparison to the reported uncertainties. Including the dark noise component results in a lower estimate – indeed, not only does the dotted lie lower than the dashed line, it even lies lower than the bold solid line (Q/Hampel). This is due to the fact that the added dark noise component homogenizes the uncertainty values of all 6 laboratories, leading to higher contributions from the 2 outlying laboratories. In the second example (Fig. 4), the test results of laboratories L13, L24, L11 and L9 (the “non-outliers”) are now near-identical – and the Q/Hampel estimate now practically coincides with the MU-weighted Hampel estimate. This is due to the fact the Q estimate is now much lower, so that the contributions to the Q/Hampel estimate from L16 and L8 are nullified. It should be noted that the MU-weighted Hampel with dark noise estimate is the same as in Example 1 (Fig. 3).

Fig. 4. Comparison Q/Hampel, MU-weighted Hampel and MU-weighted Hampel with dark noise – Example 2. Two laboratories are outliers with respect to level and

uncertainty. The results of the remaining 4 laboratories are now near-identical. The Q/Hampel and the MU-weighted Hampel estimates lie clearly above the other 2

estimates. In the third example (Fig. 5), the standard uncertainty estimates reported by the two outlying laboratories (L16 and L8) are now comparable to those of the remaining 4 laboratories. As a result, the degree of inconsistency between reported uncertainties and observed deviations is such that these laboratories no longer play a role in the computation of the dark noise. Accordingly, the dark noise is now zero and the resulting MU-weighted Hampel with and without dark noise estimates coincide.


mg/

kg

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1


22

Fig. 5. Comparison Q/Hampel, MU-weighted Hampel and MU-weighted Hampel with dark noise – Example 3. The uncertainty estimates of the 2 outlier laboratories (L16 and L8) is now comparable to those of the remaining 4 laboratories. As a result, the

dark noise is zero, and the corresponding estimate now coincides with the MU-weighted Hampel estimate.

Finally, the question arises what the minimum number of laboratories is for the application of the MU-weighted Hampel estimator. While difficult to answer this question without criteria derived from a particular context, the following considerations are provided in order to shed some light on this issue. Let us return to the second example (Fig. 4). In this example, the test results from the “good” laboratories (i.e. the laboratories having reported lower measurement uncertainty values) lie very close together, which can be seen as a confirmation that these four laboratories do perform relatively well. Now what happens if the number of laboratories is reduced to three, instead of six? In particular, what happens if only one of the four “good” laboratories participates? This means that the data itself now completely misrepresents the actual state of affairs: the majority of the labs (two against one) display a considerable negative bias. Accordingly, it can hardly be expected that any algorithm will perform very well. Nevertheless, it is informative to have a look at this extreme scenario. Table 1 provides a comparison of the results obtained with the MU-Hampel estimator (both with and without dark noise correction) for the original data (with 6 participating labs) and with only 3 labs, whereby the four possible cases are considered regarding the selection of one of the four labs. Table 1. Comparison of the results obtained with the MU-Hampel estimator (both with and without dark noise correction) for the original data from the second example (i.e.

with 6 participating labs, see Fig. 4) and with only 3 labs

MU-weighted Hampel

MU-weighted Hampel

with dark noise Original data (6 labs) 0.625 0.552

3 labs

L16, L8 and L13 0.433 0.457 L16, L8 and L24 0.372 0.461 L16, L8 and L11 0.372 0.465 L16, L8 and L9 0.372 0.465


mg/

kg

0.75

0.70

0.65

0.60

0.55

0.50

0.45

0.40

0.35

0.30

0.25


23

As can be seen, while the MU-weighted Hampel is led astray in three of the four cases (discarding the “good” laboratory altogether), the MU-weighted Hampel with dark noise is consistently able to retain the “good” laboratory in the calculation. Accordingly, though more work is required in order to fully characterize the uncertainty of the MU-Hampel in cases where very few laboratories participate, it can be said that the MU-weighted Hampel with dark noise provides an acceptable level of reliability in this type of scenario. 5 CONCLUSION The proposed estimator includes reported uncertainties in the computation of the consensus mean. In doing so, it is ensured that test results are downweighted if the corresponding uncertainty estimates are high and that inconsistencies between deviations and uncertainties are penalized. Moreover, a dark noise component makes up for unaccounted variability. This proposal has wide and far-reaching applicability. In particular, it allows a reliable assessment of laboratory performance even if the number of participants is very low. REFERENCES

[1] ISO 13528:2015, Statistical methods for use in proficiency testing by interlaboratory comparison.

[2] ISO/IEC 17025:2017, General requirements for the competence of testing and calibration laboratories.

[3] Huber, PJ et. al. “Robust Statistics (2nd ed.).” Hoboken, NJ: John Wiley & Sons, Inc., 2009.

[4] Uhlig S, Colson B, Gowik P. “Taking laboratory uncertainties into account in the Hampel estimator.” Accred Qual Assur (2018) 16: 483. https://doi.org/10.1007/s00769-018-1332-x.

The Seventh



24

DEVELOPMENT OF A PROFICIENCY TESTING SCHEME IN FOOD

MICROBIOLOGY ON CAMPYLOBACTER

Leïla Ali Mandjee 1, Eloïse Riouall 1, Jean-Christophe Augustin 1,2 and Vincent Carlier 1,2

1 Animal Société Aliment – 7 avenue du Général de Gaulle – Bât. Jean Girard –

94700 Maisons-Alfort – France 2 Ecole Nationale Vétérinaire d’Alfort – 7 avenue du Général de Gaulle – 94700

Maisons-Alfort – France [email protected], [email protected],

[email protected] Abstract Since 2010, Animal Société Aliment (ASA) has developed a proficiency testing scheme in food microbiology named Gel RAEMA. The following parameters are proposed on a gelified matrix : enumeration of Bacillus cereus, Pseudomonas, Lactic acid bacteria and Yeast/Moulds. ASA is accredited by Cofrac according to ISO 17043 since 2015 for Bacillus cereus and lactic acid bacteria, and since 2019 for Pseudomonas, Yeast/Moulds. ASA ensures that requirements of laboratories are met; for that, surveys are regularly submitted to the participants. Needs for detection and enumeration of Campylobacter have been highlighted. The new version of the standards ISO 10272-1 / 10272-2 (2017), may also increase the interest of laboratories for these parameters. So, development works have been carried out by ASA, considering the strict growth and survival conditions of Campylobacter, as a microaerophilic microorganism; and considering homogeneity and stability requirements. Two research areas have been investigated. The first one concerned the composition of the matrix and the second one focused on the preparation of the strain, before adding it in the matrix. These research studies led to the survival of Campylobacter; it has been translated by its enumeration at the expected concentration of 4 log CFU/g. This concentration remains stable during 10 days before a limited decrease to 3 log CFU/g after 15 days of storage at 4°C. These promising results allowed ASA to expand these works to a large-scale and propose an experimental testing scheme. Campylobacter was then included in the Gel RAEMA scheme that took place in May 2018. 33 laboratories participated to the enumeration test and 17 participated to the detection one.

Leïla Ali Mandjee, Eloïse Riouall, Jean-Christophe Augustin, Vincent Carlier : Development of a proficiency testing schemes in food microbiology on Campylobacter

25

Key words Proficiency testing scheme, food microbiology, gelified matrix, Campylobacter, homogeneity, stability 1 INTRODUCTION In 2017, in the European Union, Campylobacter was the most commonly reported gastrointestinal bacterial pathogen in humans. It has been so since 2005 [1]. 246 158 cases of campylobacteriosis were reported confirmed in 2017. The number of cases was high, but their severity in reported case fatality was low (0.04%). Poultry meat is considered to be the main source of human campylobacteriosis; the occurrence of Campylobacter is around 30% in broilers and turkeys. Undercooking and cross contamination are likely to be the main routes of infection. Other sources are also described such as milk, water, other meat and contact with pet or farm animals [8]. On the basis of these scientific and epidemiologic data, European Commission decided to review its Regulation (EC) N° 2073/2005 by the Regulation (EC) N°2017/1495. A limit of <1 000 CFU/g has been fixed as an hygiene criterion for broiler carcasses [2]. This criterion could help to keep Campylobacter in broiler carcasses under control and to reduce the risk for public health related to poultry meat consumption. Standards for analyses of Campylobacter have also been reviewed, ISO 10272-1 [3] and ISO 10272-2 [4]. According to these elements, food analytical laboratories have to assess themselves on the analysis of this bacterium by participating in a proficiency testing scheme. ASA has taken into account this requirement and has carried out development works. The goal of this work is to propose a proficiency testing scheme on this parameter, in agreement with the standard ISO 17043 [5]. 2 EXPERIMENTAL DESIGN 2.1 Strain of Campylobacter Development works have been carried out with a strain of Campylobacter jejuni subsp. jejuni (ATCC® 33291). It is described as a non-spore-forming Gram negative bacterium, highly motile with a spiral shape. This strain belongs to the group of thermophilic Campylobacter species that grow between 37°C and 42°C [6]. It resides as a commensal in a wide range of diverse animal hosts. For its survival, it requires an atmosphere with reduced oxygen and elevated carbon dioxide concentrations. A mixture of 5% oxygen, 10% carbon dioxide and 85% nitrogen provides optimal growth conditions. The main stress encountered by Campylobacter is this oxygen concentration in the ambient atmosphere [7]. It has been highlighted that Campylobacter jejuni had an incapability to use glucose and other carbohydrates as growth substrates. This property distinguishes it from other gastrointestinal pathogens like Salmonella, E. coli and Listeria monocytogenes. All these physiological characteristics must be considered in the experiments conducted. Following development works have been carried out with this in mind.


26

2.2 Preparation of the strain The strain of Campylobacter , used for these works, is stored at the laboratory at -20°C in a medium composed by Bolton broth and 20% of glycerol. In the first time, as Campylobacter is a microaerophilic strain, we started to prepare it in microaerobic conditions. 100 µl of the initial broth is incubated in 10ml of Bolton broth at 41.5°C during 48h. The subculture is then diluted 1/100th in Bolton broth, and incubated at 41.5°C during 48h. The preparation continues by the centrifugation of the strain three times (10 min 5000 xg) and recovering of the pelet in physiological water. In the second time, the ability of Campylobacter to survive in an ambient atmosphere has been studied in order to enhance it. The goal of this work is to supply a better resistance of the strain after its inclusion in the matrix. This ability is implemented during the strain’s preparation [7]. For this, 100 µl of the strain is incubated in 10ml of Bolton broth at 41.5°C during 48h under microaerobic conditions. Then the subculture is diluted 1/100th in Bolton broth, and incubated at 41.5°C during 48h under aerobic conditions. Preparation continues like the first one by centrifugations. This protocol is named PA. To increase resistance of the strain against stress conditions, we left it under aerobic conditions before its storage as a reserve [8]. This was made by consecutive cultures of the strain on horse blood agar at 37°C during 48h under aerobic conditions. After these first 48h, a reserve of Campylobacter is prepared; one colony is isolated on another horse blood agar plate and incubated at 37°C during 48h. Reserve of Campylobacter are then prepared after 48h, 96h and 168h. This protocol is named protocol PB. 2.3 Choice of the matrix and its compounds As other parameters proposed in Gel RAEMA schemes, the developed matrix for the parameter Campylobacter is a gelified one. In order to ensure the stability, the composition of the matrix must be adapted to the requirements of Campylobacter. One concentration of the gelifying agent has been fixed. It allows gel forming and its satisfactory homogenization during its analysis. The matrix must also allow the strain to maintain itself at the same concentration during the period of the scheme. For that, composition of Campylobacter’s selective culture media has been studied: Karmali, Bolton, Preston (Thermo Fisher Scientific). Some nutritive compounds, source of metabolites, are part of their composition, they have been tested: yeast extract, meat extract and serums. It is also well established that amino acids plays an important role in fueling the central metabolism of Campylobacter [6]. Pyruvate and glutamate have then been tested. 2.4 Preparation and monitoring of experiments The gelified matrix is produced in an autopreparator on permanent stirring. A sterilization cycle is led on the matrix (15 min at 121°C) before its artificial inoculation, at 46°C, by a liquid suspension of bacterium. The whole gel is stirred during 40 min and dispatch in bottles using a peristaltic pump. In this way, it allows to get


27

homogeneous samples, one requirement for the setting up of Proficiency Testing, according to ISO 17043 [5]. Bottles are then stored at refrigerated temperature, 4°C. Some of them undergo a change in temperature. In order to take into account sending conditions, they are transferred at room temperature during 1 and 2 days. The concentration is controlled at the beginning and followed over time during approximately twenty days. The controls are performed with specific culture medium: Campyfood agar (Biomerieux) for enumeration and detection analysis. Four samples stored at 4°C are analyzed, each time, in duplicate. Two samples with change of temperature are analyzed each time in duplicate. These samples are only analyzed after changes have been made. In this way, the stability is controlled. The homogeneity is verified for the most convincing tests on ten samples. Each sample is then analyzed in duplicate on the same media used for the stability controls. 3 RESULTS AND DISCUSSION 3.1 Effect of the matrix composition In a basic gel matrix, without additional compounds, some difficulties are encountered to enumerate and detect Campylobacter. Thus, nutritive compounds have been added in order to enhance its survival. Several tested compounds did not lead to conclusive results (data not shown). Fig. 1 presents the behaviour of Campylobacter in two matrices. The basic composition is the same one for the two matrices. The only difference is the addition of one specific compound, named C1, for the curve with the blue markers. Without this compound (green markers), the concentration tends to decrease; it cannot be enumerated after 17 days of storage. It is specified that the testing period by participants is between day 5 and day 12. The stability must be verified during it. The presence of this compound (blue markers) improved the stability of Campylobacter during 10 days when samples are stored at 4°C. Then the concentration tends to decrease. The concentration is not as stable for samples when transfered at 25°C (square and triangle markers). To this day, results obtained are the most promising one; the compound C1 has then been retained for next experiments in order to confirm the behaviour of the strain.


28

Fig. 1. Monitoring of the Campylobacter concentration in gelified matrix and in gelified matrix + compound C1

3.2 Effect of the strain’s preparation In parallel, with the same purpose of enhancing the stability of Campylobacter in the matrix, we tried to improve its resistance to environmental conditions that are not optimum for it. For this, experiments were made on the strain’s preparation method. First experiments, carried out with the strain kept under microaerobic conditions, did not produce satisfactory results (data not shown). Thus, two other protocols have been set up. These protocols included one or many aerobical steps along the preparation method. The initial concentration of the strain after incubation has not changed; it is always around 8 log CFU/ml. While the strain’s preparation in microaerobic conditions did not allow its enumeration at the beginning of the test, the two others protocols gave the ability back to Campylobacter to be enumerated during the first days of experiments. The initial targeted rate of the strain in the matrix being 4 log CFU/g. Fig. 2, presented below, shows the behaviour of the strain in the gelified matrix stored at 4°C (circle) and with change of temperature for 1 day (square) and 2 days (triangle). Green markers correspond to a strain’s preparation including just one step under aerobic condition (protocol PA); the blue ones matches with consecutive situations of the strain under aerobic conditions (protocol PB).


29

Fig. 2. Monitoring of the Campylobacter concentration in gelified matrix after strain

preparation protocol PA and PB There is not a significant difference between the blue and green curves stored at 4°C. At the beginning, the population looks stable, a decrease is then observed from 15 days. Data of transferred samples look closed between 1 and 2 days of transfer. Overall, transferred data are lower than those of samples stored at 4°C throughout the experiment. A decrease of 1 log CFU/g is observed. Data from the protocol PA are more scattered than those obtained with the protocol PB. This could be due to the preparation protocol of the strain that promotes a better stability of the concentration rate along the experiment. Nevertheless, results obtained are interesting and promising. 3.3 Organization of a Proficiency Testing scheme In order to have more information about the behaviour of the strain on a large scale, we propose this parameter as an experimental test during the scheme of May 2018. 33 laboratories participated for the enumeration and 17 for the detection of Campylobacter. Samples were constituted of the gelified matrix containing the compound C1. The strain was prepared according to the protocol PA. As all RAEMA schemes, laboratories applied the method they want and which they usually used. Several methods have thus been used by them such as NF EN ISO 10272-1, NF EN ISO 10272-2, AFNOR BIO 12/30-05/10, AFNOR BRD 07/25-01/14. For the detection test, only one laboratory did not detect the sample as positive (data not shown). For the enumeration test, the distribution of the laboratories results is presented in fig. 3.


30

Fig. 3. Distribution of laboratories results

For this test, the assigned value was 3.61 log CFU/g and the standard deviation for proficiency assessment was 0.596 log CFU/g. The assigned value was conformed to the attended value. However, the standard deviation appeared to be higher than values obtained for other parameters (around 0.3 log CFU/g). This first experimental scheme gave important information, essential for improvement of future schemes. 4 CONCLUSIONS According to new ISO standards and the new European regulation, requirements of laboratories were taken into account by ASA. Consequently, ASA set up development works to answer them. Considering that Campylobacter survives under strict environmental conditions, experimental works have been carried out on two sides: matrix composition and strain preparation method. Promising results have been obtained. Experiments must be carried on. Other large-scale tests will be implemented. This will allow us to assess the homogeneity and to compare it to the first one. This will permit to adjust the technical protocol followed for the preparation of the schemes. REFERENCES

[1] European Commission. “The European Union summary report on trends and sources of zoonoses, zoonotic agents and food-borne outbreaks in 2017”. EFSA Journal - Vol 16 - 2018: 5500.

[2] European Commission. “Commission Regulation (EU) 2017/1495 of 23 August 2017 amending Regulation (EC) N° 2073/2005 as regards Campylobacter in broiler carcases”. Official journal L218 – 24.8.2017 – p. 1-6.

[3] ISO 10272-1: 2017, Microbiology of the food chain – Horizontal method for detection and enumeration of Campylobacter spp. – Part 1: Detection method


31

[4] ISO 10272-2: 2017, Microbiology of the food chain – Horizontal method for

detection and enumeration of Campylobacter spp. – Part 1: Colony count technique

[5] NF EN ISO/IEC 17043: 2010 Conformity assessment – General requirements for proficiency testing

[6] Dirk Hofreuter. “Defining the metabolic requirements for the growth and colonization capacity of Campylobacter jejuni”. Frontiers in Cellular and Infection Microbiology – September 2014 – Volume 4 – Article 137

[7] Rodrigues et al. “Description of Campylobacter jejuni Bf, an atypical aerotolerant strain”. Gut Pathogens (2015) 7:30

[8] P. M. O’Kane and I. F. Connerton. “Characterisation of aerotolerant forms of a robust chicken colonizing Campylobacter coli”. Frontiers of Microbiology March 2017 – Volume 8 – Article 513

The Seventh



32

A VIRTUAL PROFICIENCY TESTING IN FOOD MICROBIOLOGY

Marleen Abdel Massih 1, Viviane Planchon 2, Elena Pitchugina 3, Florence Ferber 3 and Jacques Mahillon 1

1 Laboratory of Food and Environmental Microbiology, Earth and Life Institute, UCLouvain, Croix du Sud L7.05.12, B-1348 Louvain-la-Neuve, Belgium

2 Walloon Agricultural Research Centre (CRA-W), Agricultural Production Department, Rue de Liroux, 9, B-5030 Gembloux, Belgium

3 Walloon Agricultural Research Centre (CRA-W), Coordination Unit of REQUASUD, Rue de Liroux, 9, B-5030 Gembloux, Belgium

[email protected]

Abstract REQUASUD has been organizing food microbiology proficiency testing (PT) since 1989. Traditional PT provide few information on the causes of unsatisfactory results, as the analytical process in the laboratories is a “black box” whose details are not transmitted to the PT provider. To identify more precisely the causes of false analytical results, an in silico PT scheme was organized in February 2019, involving 16 laboratories from Belgium and Luxembourg. In this virtual PT, pictures of Petri dishes and confirmation tubes, obtained from real analyses, were sent to the laboratories to assess the post-analytical steps. For this virtual PT, REQUASUD selected analytical methods used by most participants in routine: enumeration of total aerobic flora, lactic bacteria and Clostridium perfringens. The laboratories were asked to return the following results:

- Number of colonies counted on each Petri dish picture; - Interpretation of the confirmation test, based on the pictures of five tubes; - Calculation of the final result; - Assessment of the conformity of the sample with respect to legislation.

In each of the 16 laboratories, one to three operators took part to the exercise, which enabled the PT organizer to assess intra- and inter-laboratory variability. The results enabled to assess the contribution of four post-analytical steps (enumeration, confirmation, calculation and conclusions on conformity) to the global measurement uncertainty of the food microbiological analyses. For these steps, some

Marleen Abdel Massih, Viviane Planchon, Elena Pitchugina, Florence Ferber and Jacques Mahillon: A virtual proficiency testing in food microbiology.

33

differences in practices between laboratories were highlighted, which can result in additionnal variability of the final analytical results. This study also showed the usefulness of virtual PT schemes, which represent a privileged observation post to highlight and quantify individual sources of errors during the analytical and post-analytical process. Key words Virtual Proficiency Testing, Food Microbiology, Sources of errors, Enumeration, Confirmation, Calculation 1 INTRODUCTION REQUASUD has been organizing proficiency testing (PT) since 1989, in five analytical fields: minerals in soils or foodstuffs, near infra-red spectroscopy, nitrates in soils and food microbiology. It is a policy of REQUASUD to distribute PT samples as close as possible to real-life samples (e.g. naturally contaminated food samples). In these PT, the samples are sent to 5 to 26 laboratories, which return their results to REQUASUD for the evaluation of their analytical performances. However, these traditional PT provide limited information on the potential causes of unsatisfactory results, as the details of the analytical processes are not transmitted to the PT provider. A virtual PT consists in sending in silico items (e.g. pictures, numbers) to the participant laboratories, instead of real samples. Although not applicable to any field of activity, virtual PT display numerous advantages: i) no transportation of the samples, as the virtual PT items are generally sent online; ii) no stability or homogeneity issue; iii) reduced costs; and iv) no limitations of the number of participants. Another advantage of virtual PT is the possibility to focus on a specific step of the analysis, e.g. assess only the counting or the calculation step. Virtual PT have successfully been organized in the clinical field, for instance to assess the ability of laboratories to differentiate blood cells on the basis of images of microscope slides [1] but, to our knowledge, no true and complete virtual PT has been described in the field of food microbiology. 2 OBJECTIVES Since 1989, the organisation of inter-laboratory comparisons enabled REQUASUD to highlight several issues linked to food microbiological analyses [2]. However, the information about the precise origin of these errors remained inaccessible with the traditional design of PT. The purpose of this virtual PT was to quantify the contribution of four post-analytical steps (colony counting, confirmation test, calculation and interpretation of the results) on the total error of analytical results. The influence of all the other steps of the analyses (e.g. sample preparation, dilution, incubation) on the results was discarded.


34

3 EXPERIMENTAL 3.1 Methods A virtual PT scheme was organized by REQUASUD in February 2019, involving 16 laboratories from Belgium and Luxembourg. For this in silico PT, pictures of Petri dishes and confirmation tubes, obtained from real analyses, were sent to the laboratories to assess the post-analytical steps. Analytical methods used by most participants in routine were selected, to ensure that the laboratories were familiar with the culture media and the characteristics of the expected colonies. Three enumeration analyses were proposed in this first virtual PT: total aerobic flora on Plate Count Agar (PCA), lactic acid bacteria on Man-Rogosa-Sharpe agar (MRS) and Clostridium perfringens on Tryptose Sulphite Cycloserine agar (TSC) with confirmation in Lactose-Sulphite tubes (LS). 3.2 Materials Good-quality pictures of Petri dishes and tubes, obtained from real analyses, were taken. The optical conditions (light intensity and white, grey or black backgrounds) were optimized to reduce light reflection and enable appropriate contrast of the bacterial colonies.

An Excel form, containing real-size pictures, was sent to the participants. The laboratories were asked to submit this file to one to three operators and to return, within the 2 weeks, the following results for each operator: - The number of characteristic colonies counted on each Petri dish picture; - The interpretation of the confirmation test, based on the pictures of the tubes, for the presence of gas and precipitate; - The calculation of the final result; - An assessment of the sample conformity with respect to legislation.

The participants were asked to enumerate the colonies directly on the screen, to avoid any additional variability linked to printing quality in the different laboratories. They were also asked to analyse the pictures at their 100 % size, without zooming, to remain close to real-life conditions. 4 RESULTS Some technical hurdles were linked with the use of pictures on a computer instead of real Petri dishes. Indeed, the operators are not used to enumerate the colonies on a screen and they do not have the possibility to move the Petri dishes, to better count the colonies on the edges, as they do in routine.

The results of the enumeration, confirmation, calculation and conformity assessment steps are presented separately below. 4.1 Enumeration of the colonies The variability of the enumeration results among operators was assessed for each analytical parameter. A representative result is illustrated in Fig. 1.


35

Fig. 1. A picture sent to the participants: Petri dish for the enumeration of total flora at the 10-5 dilution.

On average, the operators enumerated 11 colonies on this picture. The results of enumeration returned by each operator are presented in Fig. 2.

Fig. 2. Enumeration of colonies on the Petri dish displayed in Fig.1 by 16 laboratories

and from one to three operators per laboratory. We observed variability in the colony counts between laboratories (inter-laboratory variability) but also among operators (intra-laboratory variability). In the example presented above, the counts for the same picture ranged from 9 to 17 colonies. The standard deviations (SD) associated with the enumeration step, for the three different analyses, are summarized in Table 1.

Table 1. Average relative SD observed in the enumeration step for the three

analyses.

Total flora Lactic bacteria Cl. perfringens Intra -laboratory s R (%) 7,98 11,76 4,39 Inter -laboratory s R (%) 10,53 13,06 4,39


36

For two of the three analyses, the variability of the enumeration results among operators turned out to be surprisingly high. The exception is the enumeration of Cl. perfringens, as the characteristic colonies form large black spots easy to distinguish. This basic step of colony counting turned out to contribute greatly to the total error of the analytical result. This observation lends support to the postulate of Augustin and Carlier [2] that the step of colony counting contributes significantly to the measurement uncertainty in food microbiology. 4.2 Interpretation of the confirmation test The participants received five pictures of lactose-sulphite (LS) tubes, for the confirmation of Cl. perfringens (Fig. 3).

Fig. 3. The pictures of the five LS tubes, sent to the participants. Each tube was interpreted as positive or negative by the operators, on the basis of the amount of gas (the minimum to be deemed positive is a quarter of the bell) and the presence of a black precipitate in each tube. The results are presented in Fig. 4.

Fig. 4. Interpretation of the five confirmation tubes (LS).


37

The standard deviations linked to the confirmation step are summarized in Table 2. Globally, inside a same laboratory, the operators follow coherent decision rules. However, between laboratories, the instructions seem to differ greatly, as the number of tubes (out of five tubes) deemed positive varies from two to five, leading to an inter-laboratory SD of 21,82 % for the reading of the confirmation test.

Table 2. Relative standard deviations obtained for the LS confirmation.

Confirmation (LS) Intra -laboratory s R (%) 8,12 Inter -laboratory s R (%) 21,82

4.3 Calculation and final result The final results were calculated by the participants, based on the enumeration result of each Petri dish and, for Cl. perfringens, also on the number of positive tubes for the confirmation. An example of the variability of the final results is shown in Fig. 5.

Fig. 5. Enumeration results of Cl. perfringens, expressed in Colony-Forming Units

(CFU) per gram of sample. The standard deviations of the final results are summarized in Table 3. The global variability of the final results can be considered as the consequence of the differences in counting, test interpretation and calculation.

Table 3. Average relative standard deviations of the final results for the 3 analyses

Total flora Lactic bacteria Cl. perfringens Intra -laboratory s R (%) 22,86 8,05 14,26 Inter -laboratory s R (%) 38,17 8,34 27,99


38

4.4 Conformity assessment Based on their analytical results, the laboratories were asked to conclude on the conformity or non-conformity of the sample, with respect to the regulatory limits (Table 4). The sample was described by the PT organizer as a “rice and shrimp salad”.

Table 4. Range of results returned by the participants; legal criteria used for

interpretation; and conclusion on the sample conformity for the three analyses.

Range (CFU/g) Criteria (CFU/g) Conformity Total flora 220 000 - 580 000 1 000 000 95% Lactic bacteria 240 000 - 380 000 1 000 000 100% Cl. perfringens 900 - 1700 1000 50% The conformity assessment was very coherent among the laboratories. The only exception was for the Cl. perfringens parameter, which could be interpreted as conform or not-conform in function of the result calculated by the operator. 5 DISCUSSION 5.1 Differences in practices The results of this first virtual PT organized by REQUASUD were discussed with the participants, for educational purposes. A first observation is that the practices for colony-enumeration and interpretation of confirmation tests vary between the laboratories. Globally, the operators inside a same laboratory sometimes apply different rules for the enumeration of the colonies, and the laboratories have individual rules for the interpretation of the lactose-sulphite test. Different practices were identified for the following steps: - For total flora, some operators enumerate all the colonies, while others consider a Petri dish uncountable (“> 300 colonies”) if the next decimal dilution contains more than 30 colonies. This decision rule induced a high variability of the enumeration and final results in the present virtual PT. - The lactic bacteria colonies can be either lenticular, or circular if they grow on the bottom of the plate. Some operators enumerated only the lenticular colonies, yielding significantly lower results than the others. Also, some operators counted only the large colonies, while others took also the tiny colonies into account. - For Cl. perfringens, when a Petri dish is completely blackened by the colonies, most laboratories do not count it but some laboratories place those dishes one night in the fridge to remove the black halo and count the colonies. - For the confirmation of Cl. perfringens, when a LS tube is doubtful (no precipitate or little gas), some laboratories consider it negative, whereas others consider it positive or re-incubate it for 24 hours.


39

All these differences in practices result in a high variability of the final analytical results, as shown with this virtual PT. 5.2 Contribution of each step to the global standar d deviation Considering that the four steps of the post-analytical process can introduce variability in the final result, the main purpose of this virtual PT was to quantify the contribution of each of these steps to the measurement uncertainty. Table 5 summarizes the average standard deviations observed when comparing (all analytical parameters together) the results of the colony counts, confirmation test and final result.

Table 5. Average variability (quadratic mean of the relative SD) observed at each post-analytical step.

Step sR intralab (%) sR interlab (%) Colony counting 9,72 11,37 Confirmation test 8,12 21,82

Final result 16,24 27,75 The variation in the colony count results seems to be mainly linked to the operator, as the intra-laboratory variability (9,72 %) is almost as high as the inter-laboratory variability (11,37 %). On the other hand, the interpretation of the confirmation test (LS tubes) is highly dependent on the laboratory instructions, as we observe 21,82 % of variability between the laboratories. Eventually, the variability of the final results can be seen as the conjunction of the variability linked to the enumeration, confirmation and calculation steps. The contribution of the latter (sR due to calculation) could not be quantified with the design of this virtual PT. 5.3 Contribution of the calculation step The calculation of the final result is a critical step where the laboratories can have different practices: the formula used and the selection of the dilutions to retain for the calculation can have a huge impact on the final result. Therefore, to better assess the contribution of the calculation step to the global error of the result, a specific virtual PT will be conducted by REQUASUD in 2020, in which the results of colony counts will be provided and the participants will only have to calculate the final results. 5.4 Conformity statements Despite the high variability of some final results, at the end, no big differences in the conformity statements provided by the laboratories were noticed. The results variability had only an impact on the conformity statement for the parameter Cl. perfringens, for which half of the laboratories concluded that the sample conformed to the microbiological criteria, whereas the other laboratories deemed the sample non-conform.


40

The very clear instructions provided by the Belgian Federal Agency for the Safety of the Food Chain (FASFC) can explain these very coherent results: the legal criteria for food microbiology are compiled in a detailed Excel table called “Action limits for the microbiological contaminants in foodstuffs” which is transmitted by FASFC to the Belgian laboratories. 6 CONCLUSIONS This virtual PT enabled to quantify the individual contribution of four post-analytical steps to the global measurement uncertainty of the food microbiology analyses. This trial demonstrated that the enumeration, interpretation and calculation errors contribute to the analytical uncertainty, whereas the statements of conformity were globally harmonized among the 16 participating laboratories from Belgium and Luxemburg. This study shows the usefulness of virtual PT schemes, which represent a privileged observation post to highlight and quantify the individual sources of errors during the analytical and post-analytical process. REFERENCES

[1] Juhos I. (2017) “Traditional vs. virtual PT items.” Eurachem 9th Workshop Proficiency testing in analytical chemistry, microbiology and laboratory medicine. www.eurachem2017.eu

[2] Abdel Massih M., Planchon V., Polet M., Dierick K., Mahillon J. (2016) “Analytical performances of food microbiology laboratories – critical analysis of 7 years of proficiency testing results.” J. Appl. Microbiol. 120, 346-354. DOI 10.1111/jam.13009.

[3] Augustin J.C., Carlier V. (2006) “Lessons from the organization of a proficiency testing program in food microbiology by interlaboratory comparison: analytical methods in use, impact of methods on bacterial counts and measurement uncertainty of bacterial counts.” Food Microbiol. 23, 1-38.

The Seventh



41

APPROACH AND OVERVIEW OF THE PT SCHEME IN ACOUSTIC

FIELD

Danica Boljevi ć1, Milenko Mari čić2, Nikola Mari čić3 1B.Sc.EE., 2B.Sc.ChE, 3Tech, Institute for materials testing (IMS Institute), Bulevar

vojvode Misica 43, 11000 Belgrade, Serbia

[email protected], [email protected], [email protected]

Abstract During 2015 IMS Institute constituted the Proficiency Testing Provider (PIMS), accredited in 2018 by Accreditation Body of Serbia. Among a large number of subjects of testing (primarily building materials), PIMS organizes inter-laboratory comparisons in the field of acoustical testing, primarily in the field of environmental noise level measurement (outdoors and/or indoors), sound power measurements of noise sources and field measurements of building elements sound insulation. These measurements are designed in a manner that the subject of testing is the same for all participants. Implemented PT cycles were conducted in Belgrade, except sequential scheme for sound power measurement of a noise source, where the same sample travels from participant to participant. In the previous years, PIMS conducted four PT cycles of environmental noise measurements (two outdoors and two indoors), two cycles of field measurement of building elements sound insulation and one cycle of sound power measurements of noise sources. An overview of the participant results, robust statistics analysis (assigned value, its uncertainty, and cycle standard deviation) as well as the conclusions for one conducted cycle will be presented in this paper. Key words PT scheme, environmental noise, robust statistics, uncertainty. 1 INTRODUCTION For several years backward, the IMS Institute performed the services of the organization of PT schemes in the field of physical, mechanical and chemical testing of building materials, as well as the acoustic testing of building structures. The acoustic tests that IMS Institute organizes are sequential type, as presented one, or a

Danica Boljević, Milenko Maričić, Nikola Maričić: Approach and overview of the PT scheme in acoustic field

42

simultaneous type. The acoustic testing refers to the environmental noise level measurements (outdoors and indoors), in accordance with the standard method defined in [1], is the basis for the PT cycle presented in this paper. 2 EXPERIMENTAL ENVIRONMENT In May 2018 the PT scheme PIO BUK/a 01/18 was organized indoors, in a room of a parallelepiped form, with a relatively small absorption, as a sequential quantitative scheme. The room has been selected with the aim that the residual noise level is low enough (<20 dB), not to have an effect on the measurement results. Also, the noise source was selected having on mind that its noise level in the receiving room (>60 dB) is sufficiently higher than the residual noise level (difference>10 dB). The noise source was simulated by the reproduction of the signal in a .wav format (sound) of uniform noise level created by the operation of the air compressor on the loudspeaker. The noise source and its level is almost identical for all participants and stable and homogeneous in time (by its origin) and was located in the source room. All participants performed measurements in the receiving room, regarding the choice of elevations of measurement points (Fig. 1).

Fig. 1. Sketch of the receiving room and measurement positions of the participants The objective of testing the laboratory's competence through the inter-laboratory comparison was to evaluate the performance of the participants. The subject of proficiency testing in this cycle were parameters that are of interest in measuring the level of noise in the environment, primarily indoors, which are the total A-weighted equivalent sound pressure level LAeq,T, as well as the frequency noise content at the one-third octave band frequencies in the range of 50 Hz to 10 kHz (LAeq,50Hz,T ... LAeq,10kHz,T), 25 parameters in total. The implementation of the test was conducted in three repeated measurements at each of the three required positions in the time interval of T=1 minute. For the duration of the measurement, the homogeneity and stability were confirmed by monitoring the effective value of the voltage at the input of the loudspeaker through which the signal is reproduced and whose differences, in the room where the noise source is located, lead to the variation of the total A-weighted equivalent sound pressure level measured at elevation of 1.2 m, at a distance of 1 m from the sound


43

source (loudspeaker) of less than 0.15 dB. The average value of this level for the duration of the measurement was 96.8 dB. Environmental conditions data (temperature, relative air humidity and atmospheric pressure) were recorded informatively before the start of measurement of each participant, in order to examine their possible impact on the obtained results. It was estimated that the variation of conditions didn’t influence the measured results (variation of temperature 1%, relative humidity 10% and atmospheric pressure 0%). 3 STATISTICAL ENVIRONMENT Statistical processing and analysis of the obtained results were performed according to the requirements stated in [2], and [3]. The assigned value is determined by the participants' consensus. The assigned value (xpt) and standard deviation for assessment (spt) are determined according to robust statistics algorithm A+S [2]. The following aspects of standard deviations with classical and robust statistics (*) were analyzed: withinlab standard deviation of repeatability attributable to the subject cycle – the extent of the repeatability in quantification of the content of a property (sr / sr*); betweenlab standard deviation – the measure of results repeatability of all the participants in quantifying the content of a property (sL / sL*); standard deviation of reproducibility – the measure of results reproducibility of all participants in quantifying the content of a property (sR / sR*), which represents the standard deviation for assessment (spt / spt*); standard deviation of assigned values (s / s*) for the standard (measurement) uncertainty estimation of the assigned values. For the methodology of evaluating participants' performance, z-score is used, where the absolute value of z-score less than or equal to 2.0 is considered to be satisfactory, the absolute value greater than 2.0 and less than or equal to 3.0 is considered to be suspicious, while the absolute value greater than 3.0 indicates the unsatisfactory performance of the participant. Those values were expressed analytically within the table as > |3.0| (+ or -). 4 RESULTS AND DISCUSSIONS All participants except one expressed results for all parameters (LAeqT, LAeq,50Hz,T ... LAeq,10kHz,T – 25 parameters in total). One participant expressed results only for the parameter LAeqT. At each of the three measurement points, the measurements were performed three times and the mean value of the results at each point is presented. The input parameters to which the methodology from the previous chapter is applied are the mean values of noise levels, which is a total average value of three results for each of the measurement points. The measured total A-weighted equivalent sound pressure level, as well as the frequency content of the noise, measured by the participants (dB) are presented in Table 1. Table also contains the assigned values (m) according to [3], marked in blue colour, assigned values by robust algorithm A (m*) according to [2], marked in green colour, as well as the corresponding standard deviations of the assigned values (s & s*), standard deviations of withinlab and betweenlab repeatability (sr & sr*, othervise sL & sL*), standard deviations of reproducibility (sR & sR*) and uncertainties of assigned values (upt & upt*).


44

Participants results and their standard deviations are presented in Fig. 2 to Fig. 26. Evaluation of participants performance was carried out via z-score using robust mean values (m*) and robust standard deviations of reproducibility (sR*) and are presented in Table 2.

Table 1. Average of measured values (dB)

Laboratory code LAeq 50Hz 63Hz 80Hz 100Hz 125Hz 160Hz 200Hz 250Hz 315Hz 400Hz 500Hz 630Hz 8 64,62 10,78 11,62 24,37 34,12 43,44 45,33 50,09 49,63 52,91 59,14 53,63 52,94 12 64,37 9,85 13,03 24,69 35,78 45,41 45,50 51,42 47,50 52,65 58,64 51,40 53,58 24 65,70 10,33 14,51 27,85 35,59 44,39 46,50 53,09 49,02 53,43 59,07 53,92 55,22 33 64,31 11,33 11,93 23,29 35,11 44,63 46,79 49,86 47,21 52,32 58,17 51,50 53,77 43 63,46 42,42 37,45 46,01 52,91 60,37 58,70 60,57 55,06 58,81 62,10 56,00 54,38 45 64,54 10,79 15,09 27,33 35,10 45,13 44,82 50,72 48,35 53,52 57,23 52,77 54,37 52 64,43 10,88 16,29 30,20 37,75 45,71 45,41 50,80 49,16 52,44 58,19 51,48 53,17 66 64,40 11,79 15,96 25,45 34,15 45,21 45,90 52,01 49,99 52,98 56,55 52,29 54,23 67 64,47 - - - - - - - - - - - - 73 63,81 11,06 14,93 25,71 34,10 45,07 45,47 49,91 48,29 53,22 55,99 50,69 53,16 81 64,65 12,38 17,82 28,88 36,56 43,60 46,16 51,83 47,76 54,67 57,12 51,78 54,30 84 64,44 10,60 16,28 26,81 35,78 44,23 46,43 50,96 47,98 54,06 56,83 53,36 54,22 86 64,74 11,72 13,99 27,12 34,92 44,14 46,10 48,96 49,29 54,20 58,61 52,31 54,47 93 64,44 10,10 14,20 28,10 36,55 45,29 45,40 50,70 49,27 53,95 57,32 52,68 54,04 99 63,36 10,07 11,82 27,04 35,54 45,02 45,58 48,78 49,01 51,77 56,08 51,68 53,12

mean (m) 64,38 13,15 16,07 28,06 36,71 45,83 46,72 51,41 49,11 53,64 57,93 52,54 53,93 robust mean (m*) 64,39 11,01 14,70 26,96 35,60 44,82 45,88 50,88 48,78 53,36 57,77 52,41 53,91

stdev (s) 0,55 8,46 6,43 5,49 4,78 4,24 3,49 2,88 1,90 1,69 1,60 1,36 0,66 robust stdev (s*) 0,39 0,94 2,41 2,39 1,29 0,87 0,71 1,52 1,08 1,04 1,37 1,21 0,70 repeatability sr 0,52 2,66 2,78 4,37 2,19 1,29 1,54 2,07 1,59 0,95 1,60 1,32 0,95 repeatability sr* 0,54 2,50 2,93 4,61 2,25 1,34 1,56 2,10 1,67 0,98 1,64 1,31 0,79 betweenlab sL 0,69 2,34 2,85 4,14 2,14 1,30 1,38 2,16 1,53 1,14 2,01 1,69 1,03 betweenlab sL* 0,23 0,00 1,34 0,00 0,00 0,11 0,00 0,45 0,17 0,74 0,99 0,94 0,54

reproducibility sR 0,86 3,55 3,98 6,02 3,06 1,84 2,07 2,99 2,20 1,49 2,57 2,14 1,40 reproducibility sR* 0,59 2,50 3,22 4,61 2,25 1,35 1,56 2,15 1,68 1,23 1,92 1,61 0,96

upt 0,14 0,21 0,54 0,53 0,30 0,20 0,16 0,34 0,24 0,23 0,43 0,36 0,18 upt 0,13 0,28 0,75 0,72 0,38 0,27 0,22 0,45 0,34 0,32 0,46 0,40 0,24

Laboratory code 800Hz 1kHz 1.25kHz 1.6kHz 2kHz 2.5kHz 3.15kHz 4kHz 5kHz 6.3kHz 8kHz 10kHz

8 54,44 56,77 53,64 47,80 48,23 49,74 26,06 15,80 11,19 12,02 9,63 8,84 12 54,23 56,56 54,62 48,14 47,40 48,38 25,18 14,98 10,20 8,44 7,86 7,33 24 56,22 58,46 55,92 49,58 48,21 48,66 26,99 17,99 13,90 13,30 12,96 12,11 33 54,98 57,13 54,15 47,67 47,63 48,14 25,81 16,73 13,08 11,85 11,74 9,69 43 55,24 56,98 53,60 46,58 45,96 47,18 24,00 13,72 9,10 7,23 6,80 6,69 45 55,21 57,37 54,93 48,66 47,35 47,71 25,83 15,67 11,53 9,80 9,18 8,40 52 55,07 57,42 54,61 48,25 47,10 46,21 25,54 19,45 16,60 16,55 16,89 12,56 66 55,29 57,33 54,26 48,02 47,65 48,21 26,01 18,51 15,62 13,67 12,73 10,38 67 - - - - - - - - - - - - 73 55,01 56,78 54,21 48,04 46,89 46,21 25,17 14,75 10,16 8,87 7,62 6,11 81 55,59 57,41 55,13 48,75 47,28 47,42 25,34 15,72 11,21 10,17 9,32 8,62 84 55,57 56,84 54,41 48,18 47,60 47,65 26,24 15,46 10,16 8,47 7,43 5,95 86 55,10 57,15 54,16 48,25 48,26 49,96 25,86 17,14 13,70 12,33 10,57 8,83 93 55,10 57,06 54,66 49,05 47,08 46,34 25,77 16,83 13,58 11,60 10,59 9,41 99 54,70 55,82 53,21 47,11 46,80 47,52 24,94 17,91 14,83 12,72 10,95 9,49

mean (m) 55,13 57,08 54,39 48,15 47,39 47,81 25,62 16,48 12,49 11,22 10,31 8,89 robust mean (m*) 55,11 57,07 54,36 48,16 47,43 47,79 25,65 16,46 12,47 11,11 10,09 8,86

stdev (s) 0,49 0,58 0,68 0,76 0,63 1,16 0,70 1,61 2,30 2,53 2,71 1,99 robust stdev (s*) 0,45 0,40 0,65 0,72 0,62 1,17 0,57 1,74 2,57 2,61 2,52 2,19 repeatability sr 0,67 0,95 0,66 0,49 0,87 1,88 0,84 1,04 1,37 1,60 1,73 1,20 repeatability sr* 0,68 0,95 0,67 0,52 0,81 1,84 0,79 0,56 0,68 0,63 0,65 0,42 betweenlab sL 0,73 0,97 0,86 0,84 0,94 1,87 0,96 1,69 2,34 2,52 2,62 2,07 betweenlab sL* 0,23 0,00 0,52 0,66 0,41 0,70 0,34 1,71 2,54 2,59 2,49 2,17

reproducibility sR 0,99 1,36 1,09 0,97 1,28 2,65 1,27 1,99 2,71 2,99 3,14 2,39 reproducibility sR* 0,71 0,95 0,85 0,83 0,90 1,97 0,87 1,80 2,63 2,67 2,57 2,21

upt 0,13 0,16 0,18 0,20 0,17 0,31 0,19 0,43 0,62 0,68 0,72 0,53 upt 0,15 0,13 0,22 0,24 0,21 0,42 0,19 0,58 0,86 0,87 0,84 0,73


45

Fig. 2. Participants original results and review of obtained results and corresponding

standard deviations for LAeq,T

Fig. 3. For LAeq,50Hz,T








Series1; 8;

64,62Series1; 12;

64,37

Series1; 24;

65,70

Series1; 33;

64,31

Series1; 43;

63,46

Series1; 45;

64,54Series1; 52;

64,43

Series1; 66;

64,40

Series1; 67;

64,47

Series1; 73;

63,81

Series1; 81;

64,65Series1; 84;

64,44

Series1; 86;

64,74

Series1; 93;

64,44

Series1; 99;

63,36

A-weighted equivalent sound pressure

level over a time interval of 1 minute LAeq

(dB)

Sta

nd

ard

de

via

tio

n (

dB

)

LAeq (dB)

Illustrative review of the obtained results

and corresponding standard deviations

Serie

s1; 8;

10,78

Serie

s1;

12;

9,85

Serie

s1;

24;

10,33

Serie

s1;

33;

11,33

Serie

s1;

43;

42,42

Serie

s1;

45;

10,79

Serie

s1;

52;

10,88

Serie

s1;

66;

11,79

Serie

s1;

73;

11,06

Serie

s1;

81;

12,38

Serie

s1;

84;

10,60

Serie

s1;

86;

11,72

Serie

s1;

93;

10,10

Serie

s1;

99;

10,07

A-weighted

equivalent sound

pressure level …

Sta

nd

ard

…

LAeq, 50 Hz (dB)

Illustrative review

of the obtained

results and …

Serie

s1; 8;

11,62

Serie

s1;

12;

13,03

Serie

s1;

24;

14,51

Serie

s1;

33;

11,93

Serie

s1;

43;

37,45

Serie

s1;

45;

15,09

Serie

s1;

52;

16,29

Serie

s1;

66;

15,96

Serie

s1;

73;

14,93

Serie

s1;

81;

17,82

Serie

s1;

84;

16,28

Serie

s1;

86;

13,99

Serie

s1;

93;

14,20

Serie

s1;

99;

11,82

A-weighted

equivalent sound

pressure level …

Sta

nd

ard

…

LAeq, 63 Hz (dB)

Illustrative review

of the obtained

results and …

Serie

s1; 8;

24,37

Serie

s1;

12;

24,69

Serie

s1;

24;

27,85

Serie

s1;

33;

23,29

Serie

s1;

43;

46,01Serie

s1;

45;

27,33

Serie

s1;

52;

30,20

Serie

s1;

66;

25,45

Serie

s1;

73;

25,71

Serie

s1;

81;

28,88

Serie

s1;

84;

26,81

Serie

s1;

86;

27,12

Serie

s1;

93;

28,10

Serie

s1;

99;

27,04

A-weighted

equivalent sound

pressure level …

Sta

nd

ard

…

LAeq, 63 Hz (dB)

Illustrative review

of the obtained

results and …Serie

s1; 8;

34,12

Serie

s1;

12;

35,78

Serie

s1;

24;

35,59

Serie

s1;

33;

35,11

Serie

s1;

43;

52,91Serie

s1;

45;

35,10

Serie

s1;

52;

37,75

Serie

s1;

66;

34,15

Serie

s1;

73;

34,10

Serie

s1;

81;

36,56

Serie

s1;

84;

35,78

Serie

s1;

86;

34,92

Serie

s1;

93;

36,55

Serie

s1;

99;

35,54

A-weighted

equivalent sound

pressure level …

Sta

nd

ard

…

LAeq, 63 Hz (dB)

Illustrative review

of the obtained

results and …

Serie

s1; 1;

43,44

Serie

s1; 2;

45,41

Serie

s1; 3;

44,39

Serie

s1; 4;

44,63

Serie

s1; 5;

60,37

Serie

s1; 6;

45,13

Serie

s1; 7;

45,71

Serie

s1; 8;

45,21

Serie

s1; 9;

45,07

Serie

s1;

10;

43,60

Serie

s1;

11;

44,23

Serie

s1;

12;

44,14

Serie

s1;

13;

45,29

Serie

s1;

14;

45,02

A-weighted

equivalent sound

pressure level …

Sta

nd

ard

…

LAeq, 125 Hz (dB)

Illustrative review

of the obtained


s1; 8;

45,33

Serie

s1;

12;

45,50

Serie

s1;

24;

46,50

Serie

s1;

33;

46,79

Serie

s1;

43;

58,70

Serie

s1;

45;

44,82

Serie

s1;

52;

45,41

Serie

s1;

66;

45,90

Serie

s1;

73;

45,47

Serie

s1;

81;

46,16

Serie

s1;

84;

46,43

Serie

s1;

86;

46,10

Serie

s1;

93;

45,40

Serie

s1;

99;

45,58

A-weighted

equivalent sound

pressure level …

Sta

nd

ard

…

LAeq, 168 Hz (dB)

Illustrative review

of the obtained

results and …

Serie

s1; 8;

50,09

Serie

s1;

12;

51,42

Serie

s1;

24;

53,09

Serie

s1;

33;

49,86

Serie

s1;

43;

60,57Serie

s1;

45;

50,72

Serie

s1;

52;

50,80

Serie

s1;

66;

52,01

Serie

s1;

73;

49,91

Serie

s1;

81;

51,83

Serie

s1;

84;

50,96

Serie

s1;

86;

48,96

Serie

s1;

93;

50,70

Serie

s1;

99;

48,78

A-weighted

equivalent sound

pressure level …

Sta

nd

ard

…

LAeq, 200 Hz (dB)

Illustrative review

of the obtained


s1; 8;

49,63

Serie

s1;

12;

47,50

Serie

s1;

24;

49,02

Serie

s1;

33;

47,21

Serie

s1;

43;

55,06Serie

s1;

45;

48,35

Serie

s1;

52;

49,16

Serie

s1;

66;

49,99

Serie

s1;

73;

48,29

Serie

s1;

81;

47,76

Serie

s1;

84;

47,98

Serie

s1;

86;

49,29

Serie

s1;

93;

49,27

Serie

s1;

99;

49,01

A-weighted

equivalent sound

pressure level …

Sta

nd

ard

…

LAeq, 200 Hz (dB)

Illustrative review

of the obtained

results and …


46






Fig. 16. For LAeq,1kHz,T

Fig. 17. For LAeq,1.25kHz,T






Serie

s1; 8;

52,91

Serie

s1;

12;

52,65

Serie

s1;

24;

53,43

Serie

s1;

33;

52,32

Serie

s1;

43;

58,81

Serie

s1;

45;

53,52

Serie

s1;

52;

52,44

Serie

s1;

66;

52,98

Serie

s1;

73;

53,22

Serie

s1;

81;

54,67

Serie

s1;

84;

54,06

Serie

s1;

86;

54,20

Serie

s1;

93;

53,95

Serie

s1;

99;

51,77

A-weighted

equivalent sound

pressure level …

Sta

nd

ard

…

LAeq, 315 Hz (dB)

Illustrative review

of the obtained

results and …

Serie

s1; 8;

59,14

Serie

s1;

12;

58,64

Serie

s1;

24;

59,07

Serie

s1;

33;

58,17

Serie

s1;

43;

62,10Serie

s1;

45;

57,23

Serie

s1;

52;

58,19

Serie

s1;

66;

56,55

Serie

s1;

73;

55,99

Serie

s1;

81;

57,12

Serie

s1;

84;

56,83

Serie

s1;

86;

58,61

Serie

s1;

93;

57,32

Serie

s1;

99;

56,08

A-weighted

equivalent sound

pressure level …

Sta

nd

ard

…

LAeq, 400 Hz (dB)

Illustrative review

of the obtained

results and …

Serie

s1; 8;

53,63Serie

s1;

12;

51,40

Serie

s1;

24;

53,92Serie

s1;

33;

51,50

Serie

s1;

43;

56,00

Serie

s1;

45;

52,77

Serie

s1;

52;

51,48

Serie

s1;

66;

52,29

Serie

s1;

73;

50,69

Serie

s1;

81;

51,78

Serie

s1;

84;

53,36

Serie

s1;

86;

52,31

Serie

s1;

93;

52,68

Serie

s1;

99;

51,68

A-weighted

equivalent sound

pressure level …

Sta

nd

ard

…

LAeq, 400 Hz (dB)

Illustrative review

of the obtained

results and …

Serie

s1;

8;

52,…

Serie

s1;

12;

53,…

Serie

s1;

24;

55,…

Serie

s1;

33;

53,…

Serie

s1;

43;

54,…

Serie

s1;

45;

54,…

Serie

s1;

52;

53,…

Serie

s1;

66;

54,…

Serie

s1;

73;

53,…

Serie

s1;

81;

54,…

Serie

s1;

84;

54,…

Serie

s1;

86;

54,…

Serie

s1;

93;

54,…

Serie

s1;

99;

53,…

A-weighted

equivalent sound

pressure level …

Sta

nd

ard

…

LAeq, 630 Hz (dB)

Illustrative review

of the obtained

results and …

Serie

s1; 8;

54,44

Serie

s1;

12;

54,23

Serie

s1;

24;

56,22

Serie

s1;

33;

54,98

Serie

s1;

43;

55,24

Serie

s1;

45;

55,21

Serie

s1;

52;

55,07

Serie

s1;

66;

55,29

Serie

s1;

73;

55,01

Serie

s1;

81;

55,59

Serie

s1;

84;

55,57

Serie

s1;

86;

55,10

Serie

s1;

93;

55,10

Serie

s1;

99;

54,70

A-weighted

equivalent sound

pressure level …

Sta

nd

ard

…

LAeq, 800 Hz (dB)

Illustrative review

of the obtained

results and …

Serie

s1; 8;

56,77

Serie

s1;

12;

56,56

Serie

s1;

24;

58,46

Serie

s1;

33;

57,13

Serie

s1;

43;

56,98

Serie

s1;

45;

57,37

Serie

s1;

66;

57,42

Serie

s1;

73;

57,33

Serie

s1;

81;

56,78

Serie

s1;

84;

57,41

Serie

s1;

86;

56,84

Serie

s1;

93;

57,15

Serie

s1;

99;

57,06Serie

s1; ;

55,82

A-weighted

equivalent sound

pressure level …

Sta

nd

ard

…

LAeq, 1 kHz (dB)

Illustrative review

of the obtained

results and …

Serie

s1; 8;

53,64

Serie

s1;

12;

54,62

Serie

s1;

24;

55,92

Serie

s1;

33;

54,15

Serie

s1;

43;

53,60

Serie

s1;

45;

54,93

Serie

s1;

52;

54,61

Serie

s1;

66;

54,26

Serie

s1;

73;

54,21

Serie

s1;

81;

55,13

Serie

s1;

84;

54,41

Serie

s1;

86;

54,16

Serie

s1;

93;

54,66Serie

s1;

99;

53,21

A-weighted

equivalent sound

pressure level …

Sta

nd

ard

…

LAeq, 1,25 kHz (dB)

Illustrative review

of the obtained

results and …

Serie

s1; 8;

47,80

Serie

s1;

12;

48,14

Serie

s1;

24;

49,58

Serie

s1;

33;

47,67Serie

s1;

43;

46,58

Serie

s1;

45;

48,66

Serie

s1;

52;

48,25

Serie

s1;

66;

48,02

Serie

s1;

73;

48,04

Serie

s1;

81;

48,75

Serie

s1;

84;

48,18

Serie

s1;

86;

48,25

Serie

s1;

93;

49,05Serie

s1;

99;

47,11

A-weighted

equivalent sound

pressure level …

Sta

nd

ard

…

LAeq, 1,6 kHz (dB)

Illustrative review

of the obtained

results and …

Serie

s1; 8;

48,23

Serie

s1;

12;

47,40

Serie

s1;

24;

48,21

Serie

s1;

33;

47,63Serie

s1;

43;

45,96

Serie

s1;

45;

47,35

Serie

s1;

52;

47,10

Serie

s1;

66;

47,65

Serie

s1;

73;

46,89

Serie

s1;

81;

47,28

Serie

s1;

84;

47,60

Serie

s1;

86;

48,26

Serie

s1;

93;

47,08

Serie

s1;

99;

46,80

A-weighted

equivalent sound

pressure level …

Sta

nd

ard

…

LAeq, 1,6 kHz (dB)

Illustrative review

of the obtained

results and …

Serie

s1; 8;

49,74

Serie

s1;

12;

48,38

Serie

s1;

24;

48,66

Serie

s1;

33;

48,14

Serie

s1;

43;

47,18

Serie

s1;

45;

47,71Serie

s1;

52;

46,21

Serie

s1;

66;

48,21Serie

s1;

73;

46,21

Serie

s1;

81;

47,42

Serie

s1;

84;

47,65

Serie

s1;

86;

49,96Serie

s1;

93;

46,34

Serie

s1;

99;

47,52

A-weighted

equivalent sound

pressure level …

Sta

nd

ard

…

LAeq, 2 ,5kHz (dB)

Illustrative review

of the obtained

results and …

Serie

s1; 8;

26,06

Serie

s1;

12;

25,18

Serie

s1;

24;

26,99

Serie

s1;

33;

25,81

Serie

s1;

43;

24,00

Serie

s1;

45;

25,83

Serie

s1;

52;

25,54

Serie

s1;

66;

26,01

Serie

s1;

73;

25,17

Serie

s1;

81;

25,34

Serie

s1;

84;

26,24

Serie

s1;

86;

25,86

Serie

s1;

93;

25,77

Serie

s1;

99;

24,94

A-weighted

equivalent sound

pressure level …

Sta

nd

ard

…

LAeq, 3,15kHz (dB)

Illustrative review

of the obtained

results and …

Serie

s1; 8;

15,80

Serie

s1;

12;

14,98

Serie

s1;

24;

17,99

Serie

s1;

33;

16,73

Serie

s1;

43;

13,72

Serie

s1;

45;

15,67

Serie

s1;

52;

19,45

Serie

s1;

66;

18,51Serie

s1;

73;

14,75

Serie

s1;

81;

15,72

Serie

s1;

84;

15,46

Serie

s1;

86;

17,14

Serie

s1;

93;

16,83

Serie

s1;

99;

17,91

A-weighted

equivalent sound

pressure level …

Sta

nd

ard

…

LAeq, 4kHz (dB)

Illustrative review

of the obtained

results and …


47





Table 2. Evaluating the participants’ performance via z-score

Lab. code

LAeq 50Hz 63Hz 80Hz 100Hz 125Hz 160Hz 200Hz 250Hz 315Hz 400Hz 500Hz 630Hz

8 0,4 0,1 -0,9 -0,5 -0,6 -0,9 -0,3 -0,3 0,6 -0,3 0,7 0,8 -1,0

12 0,0 -0,4 -0,4 -0,4 0,2 0,5 -0,2 0,4 -0,7 -0,5 0,5 -0,6 -0,3

24 2,2 -0,2 0,0 0,3 0,1 -0,2 0,4 1,1 0,2 0,2 0,7 0,9 1,4

33 -0,1 0,2 -0,8 -0,7 -0,1 -0,1 0,6 -0,4 -0,9 -0,8 0,2 -0,6 -0,1

43 -1,6 >|3,0| (+) >|3,0| (+) >|3,0| (+) >|3,0| (+) >|3,0| (+) >|3,0| (+) >|3,0| (+) >|3,0| (+) >|3,0| (+) 2,3 2,2 0,5

45 0,3 0,0 0,2 0,1 -0,1 0,3 -0,6 0,0 -0,2 0,2 -0,3 0,2 0,5

52 0,1 0,0 0,6 0,8 1,0 0,7 -0,3 0,1 0,3 -0,7 0,2 -0,6 -0,8

66 0,0 0,4 0,5 -0,3 -0,6 0,4 0,1 0,6 0,8 -0,2 -0,6 -0,1 0,3

67 0,1 - - - - - - - - - - - -

73 -1,0 0,1 0,2 -0,2 -0,6 0,3 -0,2 -0,4 -0,2 0,0 -0,9 -1,1 -0,8

81 0,4 0,6 1,1 0,5 0,5 -0,8 0,2 0,5 -0,5 1,2 -0,3 -0,4 0,4

84 0,1 -0,1 0,6 0,0 0,2 -0,4 0,4 0,1 -0,4 0,7 -0,5 0,6 0,3

86 0,6 0,3 -0,1 0,1 -0,2 -0,4 0,2 -0,8 0,4 0,8 0,4 -0,1 0,6

93 0,1 -0,3 -0,1 0,3 0,5 0,4 -0,3 0,0 0,4 0,6 -0,2 0,2 0,1

99 -1,8 -0,3 -0,8 0,1 0,1 0,2 -0,1 -0,9 0,2 -1,2 -0,9 -0,4 -0,8

Lab. code 800Hz 1kHz 1.25kHz 1.6kHz 2kHz 2.5kHz 3.15kHz 4kHz 5kHz 6.3kHz 8kHz 10kHz

8 -0,9 -0,3 -0,9 -0,4 0,9 1,0 0,5 -0,4 -0,5 0,3 -0,2 0,0

12 -1,2 -0,5 0,3 0,0 0,0 0,3 -0,5 -0,8 -0,9 -1,0 -0,9 -0,7

24 1,6 1,5 1,9 1,7 0,9 0,4 1,6 0,9 0,5 0,8 1,1 1,5

33 -0,2 0,1 -0,3 -0,6 0,2 0,2 0,2 0,2 0,2 0,3 0,6 0,4

43 0,2 -0,1 -0,9 -1,9 -1,6 -0,3 -1,9 -1,5 -1,3 -1,5 -1,3 -1,0

45 0,1 0,3 0,7 0,6 -0,1 0,0 0,2 -0,4 -0,4 -0,5 -0,4 -0,2

52 -0,1 0,4 0,3 0,1 -0,4 -0,8 -0,1 1,7 1,6 2,0 2,6 1,7

66 0,3 0,3 -0,1 -0,2 0,2 0,2 0,4 1,1 1,2 1,0 1,0 0,7

73 -0,1 -0,3 -0,2 -0,1 -0,6 -0,8 -0,5 -0,9 -0,9 -0,8 -1,0 -1,2

81 0,7 0,4 1,0 0,7 -0,2 -0,2 -0,4 -0,4 -0,5 -0,4 -0,3 -0,1

84 0,7 -0,2 0,1 0,0 0,2 -0,1 0,7 -0,6 -0,9 -1,0 -1,0 -1,3

86 0,0 0,1 -0,3 0,1 0,9 1,1 0,2 0,4 0,5 0,5 0,2 0,0

93 0,0 0,0 0,4 1,1 -0,4 -0,7 0,1 0,2 0,4 0,2 0,2 0,2

99 -0,6 -1,3 -1,4 -1,3 -0,7 -0,1 -0,8 0,8 0,9 0,6 0,3 0,3

Serie

s1; 8;

11,19

Serie

s1;

12;

10,20

Serie

s1;

24;

13,90

Serie

s1;

33;

13,08

Serie

s1;

43;

9,10

Serie

s1;

45;

11,53

Serie

s1;

52;

16,60

Serie

s1;

66;

15,62Serie

s1;

73;

10,16

Serie

s1;

81;

11,21

Serie

s1;

84;

10,16

Serie

s1;

86;

13,70

Serie

s1;

93;

13,58

Serie

s1;

99;

14,83

A-weighted

equivalent sound

pressure level …

Sta

nd

ard

…

LAeq, 5kHz (dB)

Illustrative review

of the obtained

results and …

Serie

s1; 8;

12,02Serie

s1;

12;

8,44

Serie

s1;

24;

13,30

Serie

s1;

33;

11,85

Serie

s1;

43;

7,23

Serie

s1;

45;

9,80

Serie

s1;

52;

16,55

Serie

s1;

66;

13,67Serie

s1;

73;

8,87

Serie

s1;

81;

10,17

Serie

s1;

84;

8,47

Serie

s1;

86;

12,33

Serie

s1;

93;

11,60

Serie

s1;

99;

12,72

A-weighted

equivalent sound

pressure level …

Sta

nd

ard

…

LAeq, 6,3kHz (dB)

Illustrative review

of the obtained

results and …

Serie

s1; 8;

9,63

Serie

s1;

12;

7,86

Serie

s1;

24;

12,96

Serie

s1;

33;

11,74

Serie

s1;

43;

6,80

Serie

s1;

45;

9,18

Serie

s1;

52;

16,89

Serie

s1;

66;

12,73

Serie

s1;

73;

7,62

Serie

s1;

81;

9,32

Serie

s1;

84;

7,43

Serie

s1;

86;

10,57

Serie

s1;

93;

10,59

Serie

s1;

99;

10,95

A-weighted

equivalent sound

pressure level …

Sta

nd

ard

…

LAeq, 8kHz (dB)

Illustrative review

of the obtained

results and …

Serie

s1; 8;

8,84

Serie

s1;

12;

7,33

Serie

s1;

34;

12,11

Serie

s1;

33;

9,69Serie

s1;

43;

6,69

Serie

s1;

45;

8,40

Serie

s1;

52;

12,56

Serie

s1;

66;

10,38

Serie

s1;

73;

6,11

Serie

s1;

81;

8,62Serie

s1;

84;

5,95

Serie

s1;

86;

8,83

Serie

s1;

93;

9,41

Serie

s1;

99;

9,49

A-weighted

equivalent sound

pressure level …

Sta

nd

ard

…

LAeq, 10kHz (dB)

Illustrative review

of the obtained

results and …


48

5 CONCLUSIONS - The unconformity established by the conducted inter-laboratory comparison of the A-weighted equivalent sound pressure level indoor measurement is given according to the participant's codes. A total of 6 disputed measured values (5 higher and 1 lower) and 9 unsatisfactory (higher values) were established. Non-conformities in the results of 12 participants were not established. The average score achieved in this cycle is 0.64, which, given the significant number of suspicious and unsatisfactory results of one participant, is considered to be very successful. - As a conclusion of PIO BUK/a 01/18 Provider gave a significant number of comments, but also suggestions and advice to the participants which can be the causes of the manifested nonconformities with the intention that the participants who have had an unsatisfactory result recognize the cause that led to that result. In this cycle, the reasons are from simple ones such as that the calibration of the measuring chain has not been carried out thoroughly according to the prescribed procedure of the equipment manufacturer; that the correction of the level due to the deviation of the reference value from the calibration certificate of the sound calibrator has not been taken into account; that the measurement is carried out with the perpetrator at an inadequate distance from the microphone that its manipulation can endanger the measurement result, especially at lower frequencies or insufficient attention when fulfilling in the test results From to the comlicated reasons such as that the measurement of the total A-weighted sound pressure level is carried out in a wider frequency range than the range specified by the test method when it can be assumed that the measuring instrument, depending on the manufacturer and type/model, calculates the total A-weighted equivalent sound pressure level in the frequency band of the instrument itself e.g. in the band from 10 Hz to 20 kHz, not in the frequency band of method’s interest from 50 Hz do 10 kHz when the total A-weighted equivalent sound pressure level, determined in the one-third octave band frequency range of 50 Hz to 10 kHz would not be the same) and such as that the measurement of the A-weighted sound pressure level by frequencies (frequency noise content) is carried out with an inadequately weighting frequency characteristic of the measuring instrument or to use one-third octave band filters of unknown characteristics. - Causes such as eventual changes of sound level in time in the source room or changes of residual sound level in the receiving room and changes of parameters of the environmental conditions have been analyzed and evaluated during the cycle so that they were not influenced by the measurement results. REFERENCES

[1] ISO 1996-2:2007, Аcoustics – Description, measurement and assessment of environmental noise – Part 2: Determination of environmental noise levels

[2] ISO 13528:2015, Statistical methods for use in proficiency testing by interlaboratory comparisons.

[3] ISO 5725-2:1994/Cor 1:2002, Accuracy (trueness and accuracy) of method and of measurement results – Part 2: Basic method for determining repeatability and reproducibility of standard measurement method

The Seventh



49

PRESERVATION OF PARABENS IN WATER SAMPLES

Boris Constantin 1, Eric Ziegler 1, Anne Tirard 1, Abdelkader Boubetra 1, Azziz Assoumani 2, Hugues Biaudet 2

1 Bipea – 189 rue d’Aubervilliers – 75018 Paris – France 2 INERIS, Parc Technologique ALATA BP 2, 60550 Verneuil-en-Halatte, France

[email protected]

Abstract Parabens are preservatives widely used in cosmetics, pharmaceutical and food products due to their antibacterial and antifungal activity. Their ubiquity in surface water and their potential adverse effects on human health and environment led to their inclusion into the French monitoring programme of aquatic environment. In order to help the laboratories improving their analytical skills regarding these chemicals, BIPEA organises since November 2016 proficiency testing schemes (PTS) according to ISO 17043 [1] for the quantification of butyl-, ethyl-, methyl- and propylparaben in freshwater. Parabens degradability in water by hydrolysis and photolysis raises the issue of their stability throughout the PT, especially for the samples with long delivery time or laboratories unable to perform their analyses at reception. In order to meet PT samples homogeneity and stability requirements, BIPEA conducted a comparative study on sample preservation by ascorbic acid and nitric acid. It included (i) an experimental PT for which parabens-spiked samples were proposed to the laboratories with each preservation agent separately, and (ii) a dedicated homogeneity and stability study performed by the French National Institute for Industrial Environment and Risks (INERIS). Overall preservation with nitric acid implied less dispersed results and closer to the spiking concentrations, especially for methylparaben, and a greater stability over time. Key words Parabens; preservation; proficiency testing schemes; water quality control.

Boris Constantin, Eric Ziegler, Anne Tirard, Abdelkader Boubetra, Azziz Assoumani, Hugues Biaudet : Preservation of parabens in water samples

50

1 INTRODUCTION Parabens are esters of parahydroxybenzoic acid. Due to their antibacterial and antifungal properties, they are mainly used as preservatives in the cosmetics, and in pharmaceutical and food products. Recent studies showed that they may have potential adverse effects on human health and on environment. In fact these substances may exert weak oestrogenic activity [2, 3, 4] and may present potential hazard on reproduction [5]. Their wide range of applications favors their indirect transfer into the environment, especially through the urban wastewater containing residues of cosmetics [6] and the industrial wastewater [7]. Their ubiquity in surface water and their potential adverse effects led to their inclusion into the French monitoring programme of aquatic environment as emerging contaminants. In order to help the laboratories improving their analytical skills regarding these emergent contaminants, BIPEA organises since November 2016 proficiency testing schemes (PTSs) according to ISO 17043 standard [1] for the quantification of butyl-, ethyl-, methyl- and propylparaben in freshwater. Parabens degradability in water by hydrolysis and photolysis raises the issue of their stability throughout the PT, especially for the samples with long delivery time or laboratories unable to perform their analyses at reception. In fact an experimental stability study was conducted on PT samples with parabens (butyl-, ethyl-, methyl- and propylparaben) and preserved with ascorbic acid only, before the present work. This study, performed according to the ISO 13528 standard [8], demonstrated the stability for butylparaben and the instability for the others parabens over a period of 22 days. In order to meet PT samples homogeneity and stability requirements, BIPEA conducted a comparative study on parabens preservation by ascorbic acid and nitric acid. This study presents an experimental PT for which parabens-spiked samples were proposed to the laboratories with each preservation agent separately, and a dedicated homogeneity and stability study performed by INERIS. It aims at determining whether preservation with nitric acid may be more efficient than with ascorbic acid. 2 EXPERIMENTAL The design of this experimental PT included three main steps: manufacturing of the samples, analyses by the laboratories and statistical treatment of the data, with the estimation of an assigned value. 2.1 Manufacturing of the samples For this PT, samples were prepared from surface water that was characterized by an accredited laboratory (see Table 1). The surface water was homogenized using a homogenization tank, and divided into samples by a quasi-simultaneous filling of bottles. Then each bottle was individually spiked with butylparaben (BuP), ethylparaben (EtP), methylparaben (MeP) and propylparaben (PrP) at different concentrations. Finally half of the samples were preserved with 1 g/l of ascorbic acid, and the other half with 1 ml of nitric acid concentrate (at 65%). Theoretical concentrations for each paraben and each preservative are given in Table 2. Sample


51

total volume was 1 L. Amber glass bottles were used in order to minimize any sorption on bottle wall or any photolysis that may occur until analysis by the participants. 2.2 Analysis of the samples Samples were shipped at (5±3) °C to the laboratories participating to the test. A reply form was made available to allow the laboratories to return their analysis results. Moreover, participants were invited to enter in the reply form some complementary information such as sample temperature upon receipt, date of the beginning of the analysis, and method used. Given the stability of the product, the participants were invited to analyse the samples as soon as possible after the reception.

Table 1. Characteristics of the raw surface water

Parameter Unit Method Results

pH pH unit EN ISO 10523 8.2

Conductivity at 25 °C µS/cm EN 27888 760

Suspended solids mg/L EN 872 30

Nitrates mg NO3/L EN ISO 10304-1 23

Chlorides mg Cl/L EN ISO 10304-1 27

Sulphates mg SO4/L EN ISO 10304-1 79

Calcium mg Ca/L EN ISO 11885 140

Magnesium mg Mg/L EN ISO 11885 21

Sodium mg/L EN ISO 11885 12

Haze FNU EN ISO 7027 22

Total Organic Carbon mg/L EN 1484 4.0

Hydrogen carbonates mg/L EN ISO 9963-1 347

Table 2. Theoretical concentrations in the samples for each paraben and each preservative

Spiked paraben Unit Concentration in

samples preserved with ascorbic acid

Concentration in samples preserved

with nitric acid Butylparaben µg/L 0.307 0.192

Ethylparaben µg/L 0.333 0.208

Methylparaben µg/L 0.384 0.240

Propylparaben µg/L 0.358 0.224


52

2.3 Statistical treatment The statistical treatment of the PT data was performed according to the ISO 13528 standard [8]. The assigned values, xpt, were estimated using the robust means of the results. The proficiencies of each laboratory were evaluated with tolerance values, TV, being 60 % of the assigned value for each parameter. The results, x, could be evaluated and classified through z-scores, z, (see Eq. 1): result with z ≤ |2| is considered as satisfactory, result with |2| ≤ z ≤ |3| is considered as questionable, and result with z ≥ |3| is considered as unsatisfactory, where:

(1)

2.4 Homogeneity and stability study At the same time as the test, INERIS performed an experimental study to verify the homogeneity and stability of the four parabens in the samples. Eight samples were analysed at day 0 for the homogeneity; and three samples were analysed after 1, 3 and 7 days for stability assessment. 3 RESULTS This experimental PT was set up in March 2018 gathering 9 laboratories. One laboratory gave semi-quantitative results for all chemical compounds, one laboratory gave results for EtP, MeP and PrP only, while the rest of the participants gave results for all substances. Except the laboratory with semi-quantitative results and using mass spectrometry, the laboratories used tandem mass spectrometry, MS/MS. In addition to the detection method, the participants could provide their extraction method. Four of them carried out a direct injection, two performed a solid/liquid extraction, and one used an on-line preconcentration. The low number of participants did not allow to compare relevantly the extraction and detection techniques.

Table 3. Main statistical parameters for each paraben and each preservative

Statistical parameter Concentration in samples

preserved with ascorbic acid Concentration in samples preserved with nitric acid

BuP EtP MeP PrP BuP EtP MeP PrP

Number of returned results 7 8 8 8 7 8 8 8

Assigned value xpt (in µg/L) 0.250 0.235 0.162 0.278 0.165 0.170 0.182 0.178

Number of results taken into account

for xpt estimation 6 7 6 7 7 8 8 8

Robust standard deviation of the

results s(xpt) 0.099 0.077 0.062 0.077 0.060 0.045 0.051 0.047

Coefficient of variation CV(xpt) (in %) 40 33 38 28 36 26 28 26


53

The main statistical parameters of this PT are given for each paraben and each preservation method - ascorbic acid and nitric acid - in Table 3. It should be noted that the number of results taken into account for the estimation of the assigned value is not always the same, and relatively low. The results for each paraben were relatively dispersed with coefficients of variation ranging from 26% to 40%. Considering these latter, less discrepancy among the participants was observed for EtP and MeP in the samples preserved with nitric acid compared to those preserved with ascorbic acid. For BuP and PrP, similar results dispersion were observed for both preservation method. Assigned values were compared with theoretical concentrations, Cth (Fig. 1.). All assigned values underestimated the spiking values regardless of the preservation method. With nitric acid, relative differences between xpt and Cth ranged from -14% to -24% while with ascorbic acid they ranged from -19% to -58%. xpt values were closer to Cth when nitric acid was added to the samples, especially for EtP and MeP.

Fig.1. Relative differences, in %, between the assigned values xpt and the theoretical concentrations Cth for each paraben and each preservative

At the same time as the PT, some samples were analysed by INERIS for a homogeneity and stability study. Each result was compared to the spiking value by calculating the relative difference, in %, between the individual measurement and the theroretical concentration. Results are shown in Fig. 2. and Fig. 3. For the homogeneity study, eight samples were analysed at day 0. For each paraben and each preservation method, these results were less dispersed than the PT results. Nevertheless the same trend as in the PT results was observed: results for samples preserved with nitric acid were closer to the theoretical concentrations, and this for all parabens, especially for MeP. For the stability study, three samples were analysed after 1, 3 and 7 days. For the samples with ascorbic acid, and for all parabens, relative difference with the spiking concentration started increasing (in absolute value) from day 1. Remarkably relative difference reached about -70% for MeP after 7 days. In the contrary, such “loss” was


54

not observed for the samples with nitric acid: results were relatively steady over time, even for MeP.

Fig. 2. Results of the homogeneity study performed for each paraben and each preservative, ascorbic acid and nitric acid


55

Fig. 3. Results of the stability study performed for each paraben and each preservative, ascorbic acid and nitric acid

4 CONCLUSION Parabens are emerging contaminant considering their ubiquity in the environment and their potential effects. Need for PTS with these substances is also emerging. An experimental PT was implemented in order to compare in real conditions two preservatives, ascorbic acid and nitric acid, for ensuring parabens stability in PT water samples over time. Less dispersion was observed among the participants for EtP and MeP in the samples preserved with nitric acid. Overall assigned values were closer to the spiking concentrations when nitric acid was used, especially for EtP and MeP. A homogeneity and stability study performed on these PT samples confirmed that preservation with nitric acid allowed a higher recovery rate of the parabens in water


56

samples, especially of MeP. It also indicated that nitric acid ensured more stable samples over seven days. REFERENCES

[1] ISO 17043:2015, Conformity assessment - General requirements for proficiency testing.

[2] Bazin I., Gadal A., Touraud E., and Roig B., “Hydroxy benzoate preservatives (parabens) in the environment: data for environmental toxicity assessment”. In: Fatta-Kassinos D., Bester K., Kümmerer K., editors; Xenobiotics in the urban water cycle, Vol. 16, Springer, 2010: pp. 245-257.

[3] Oishi S. “Effects of propyl paraben on the male reproductive system”. Food and Chemical Toxicology. 40 (2002):1807-1813.

[4] Soni M.G., Taylor S.L., Greenberg N.A., and Burdock G.A. “Evaluation of the health aspects of methyl paraben: a review of the published literature”. Food and Chemical Toxicology. 40 (2002):1335-1373.

[5] Soni M. G., Carabin I. G. and Burdock G. A. "Safety assessment of esters of p-hydroxybenzoic acid (parabens)". Food and Chemical Toxicology. 43 (2005):985-1015.

[6] Canosa P., Rodríguez I., Rubí E., Bollaín M. H. and Cela R. "Optimisation of a solid-phase microextraction method for the determination of parabens in water samples at the low ng per litre level”. Journal of Chromatography A. 1124 (2006): 3-10.

[7] Ramirez N., Borrull F., and Marcé R.M. “Simultaneous determination of parabens and synthetic musks in water by stir-bar sorptive extraction and thermal desorption-gas chromatography-mass spectrometry”. Journal of Separation Science. 35 (2012): 580-588

[8] ISO 13528:2015 - Statistical methods for use in proficiency testing by interlaboratory comparisons.

The Seventh



57

MICROBIOLOGY PROFICIENCY-TESTING SCHEME IN COSMETICS

Marie DANGERVILLE , Abdelkader BOUBETRA, Romain LE NEVE, Sandrine NGUYEN, Anne TIRARD

Bipea – 189 rue d’Aubervilliers – 75018 Paris, France

[email protected]; [email protected]; [email protected]; [email protected]; [email protected]

Abstract The current context of reducing preservatives in cosmetics, especially antimicrobials compounds, is compelling the cosmetics industry to find new solutions to ensure the health security of raw materials and finished products while respecting the expectations of consumers of more natural products. Laboratories performing microbiological analyses on cosmetics therefore have a key role to play in this process and must ensure their performance through regular proficiency-testing schemes (PTSs). The complexity and diversity of cosmetic matrices is a factor to be taken into account in developing a PTS, especially during the preparation of stable and homogeneous samples. Since March 2018, Bipea set up tests on yeast, mould and mesophilic aerobic bacteria in cosmetics. For each test, the statistical treatment of the data is performed according to ISO 13528 standard [1]. The assigned values are calculated from the participants’ results and the performances of the laboratories can be evaluated individually and collectively according to ISO 17043 standard [2]. Key words Proficiency-testing schemes, microbiology, cosmetics quality control, laboratory performance 1 INTRODUCTION This work describes the design and the implementation of a proficiency-testing scheme for the analyses of cosmetic samples spiked with three different strains. The goal of this PTS is to allow laboratories to demonstrate the reliability of their results and to compare each other analytical data and protocols used for the enumeration of yeast, mould and mesophilic aerobic bacteria in cosmetics.

Marie DANGERVILLE, Abdelkader BOUBETRA, Romain LE NEVE, Sandrine NGUYEN, Anne TIRARD – Microbiology Proficiency - testing scheme in cosmetics

58

2 EXPERIMENTAL The designing of a PTS can be summarized into three main steps: manufacturing of the samples, analyses by the laboratories and statistical treatment of the data, with the estimation of an assigned value. 2.1 Manufacturing of the sample One of the fundamental aspects for the implementation of a PT is the preparation of homogenous and stable samples. For this PT, samples were prepared by spiking individually each sample of dermatological cream with a suspension of Aspergillus niger, Candida albicans and Bacillus cereus in well controlled proportions. According to the requirements of the ISO 13528 standard [1], homogeneity of the samples was verified by experimental studies on 10 samples in duplo taken randomly across a batch of samples for mesophilic aerobic bacteria parameter. Stability of the product was proved by analysing a sample after 7 and 14 days. 2.2 Analysis of the samples Samples were shipped at (5±3) °C to the laboratories participating to the test together with a standard sample for monitoring the temperature. A reply form was made available to allow the laboratories to return their analysis results. Moreover, participants were invited to enter in the reply form some complementary information such as the date of the beginning of the analysis, method used, growth medium used, incubation temperature and time and the type of plating. Given the stability of the product, the participants were invited to analyse the samples as soon as possible after the reception. 2.3 Statistical treatment of the data The statistical treatment of the data was performed according to the ISO 13528 standard [1]. The assigned values (xpt) were estimated using the robust means of the results transformed in log (CFU/g). The proficiencies of each laboratory were evaluated thanks to tolerance values (TV) of twice the standard deviations. The results (x) could be evaluated and classified through z-scores: z ≤ |2|: satisfactory, |2| ≤ z ≤ |3|: questionable, z ≥ |3|: unsatisfactory,

Where:

3 RESULTS AND DISCUSSIONS The results of the homogeneity check are summarized graphically in Figure 1. These data show that the samples are homogenous enough to meet the requirements of the


59

test, with a gap between the minimal and maximal values lower than 0,5 CFU/g in log for mesophilic aerobic bacteria parameter.

Fig. 1. Results of the homogeneity check of the samples as a function of the production order for mesophilic aerobic bacteria parameter

The analyses results of the stability checks showed a satisfactory recovery rate considering the expected concentration after storing the samples at (5±3) °C for 14 days (see Table 1). The variability of the performed method can explain the difference between the results collected from D0 to D14. Table 1. Average results of the enumeration of mould, yeast and mesophilic aerobic

bacteria in dermatological cream after 14 days at (5±3) °C

Day of analysis

log(CFU/g) D

0 D

7 D

14

Mould 3.00 3.06 3.07

Yeast 3.02 2.95 2.80

Mesophilic aerobic bacteria 2.83 2.88 2.67 The first PT on microbiological analyses of cosmetics was set up in March 2018, gathering fourteen laboratories around the world. Ten to thirteen laboratories out of fourteen gave their results together with useful information for the interpretation of the data. The main statistical parameters of this PT are summarized in Table 2.


60

Table 2. Summary of the statistical treatment of the data, log (CFU/g)

Analytical parameter

Statistical parameter

Mesophilic aerobic bacteria

Yeast Mould Yeast + Mould

Number of returned results (pca

) 13 10 11 13

Assigned values for proficiency testing (x

pt) 3.575 3.156 3.047 3.458

Standard uncertainty of the assigned value (u

xpt) 0.064 0.099 0.199 0.064

Robust standard deviation of the results (x

pt), from all the results which

participated to the estimation of the assigned values

0.178 0.239 0.503 0.176

Number of results taken into account for the estimation of the assigned value (p

xpt) 12 9 10 12

Coefficient of variation (%) (CV xpt

) 5 8 17 5

Standard deviation for proficiency assessment σ

pt 0.178 0.239 0.503 0.176

Tolerance value (TV) 0.356 0.478 1.006 0.352 Number of results out of the tolerance interval p

D 2 0 1 1

An assigned value (xpt) of 3.575 log (CFU/g) was calculated for mesophilic aerobic bacteria parameter and 3.458 log (CFU/g) was calculated for “yeast + mould” parameter. These values were estimated from the robust mean of the all returned results except those considered as incoherent. The laboratories’ results are shown on Figures 2 and 3. On the histograms, assigned value and tolerance interval are indicated in the x-axis and the results of the laboratories are shown in different colours as a function of the performed method.

6

5

4

3

2

1

Class number < 1 2 3 4 5 6 >

Class

Interval

^ ^ ^

xpt - TV xpt xpt + TV

3.219 3.575 3.931

3.331 3.835 4.339

3.079 3.583 4.087 4.591

Ret

urne

d re

sult

s

Fig. 2. Distribution of the results for “mesophilic aerobic bacteria” parameter


61

6

5

4

3

2

1

Class number < 1 2 3 4 5 >

Class

Interval

^ ^ ^

xpt - TV xpt xpt + TV

3.106 3.458 3.810

Ret

urne

d re

sult

s

2.940 3.230 3.520

3.085 3.375 3.665

Fig. 3. Distribution of the results for “yeast + mould” parameter Most of the laboratories followed the standard methods. For mesophilic aerobic bacteria parameter (figure 2), 9 laboratories used ISO 21149 [3] method (in dark grey) and 4 laboratories used another method (in light grey). For “yeast + mould” parameter (figure 3), 10 laboratories used ISO 16212 [4] method (in dark grey) and 2 laboratories used another method (in light grey). However, no tendency was highlighted as a function of the method used. 4 CONCLUSIONS A PT for microbiology analysis in cosmetics, gathering fourteen laboratories around the world, was successfully implemented and the results were published in an interlaboratory comparison report distributed to the participants. This first PT has been transformed into a regular PTS, including 2 tests per year. Laboratories can now monitor punctually and/or continuously through time the reliability of their results for microbiological analyses of cosmetics. REFERENCES

[1] ISO 13528:2015 - Statistical methods for use in proficiency testing by interlaboratory comparisons

[2] ISO 17043:2010 - Conformity assessment - General requirements for proficiency testing

[3] ISO 21149: Cosmetics - Microbiology - Enumeration and detection of aerobic mesophilic bacteria

[4] ISO 16212: Cosmetics -- Microbiology -- Enumeration of yeast and mould

The Seventh



62

PROFICIENCY-TESTING SCHEME FOR

HALOANISOLES AND HALOPHENOLS IN OAK WOOD

Caterina MAZZONI, Anne TIRARD, Abdelkader BOUBETRA, François ROUSSEY, Eric ZIEGLER

BIPEA, 189 rue d’Aubervilliers, 75018 PARIS - FRANCE

[email protected]

Abstract Haloanisoles are responsible for musty or mouldy off-flavours in wine [1]. These molecules are extremely fragrant and they alter wines in an irreversible way. The origin of haloanisoles can be attributed to the biodegradation of halophenols, which can be found in winery environments. Various materials, including oak products (wood tanks, barrels, chips, staves) may be contaminated by haloanisoles and halophenols. Once polluted, these materials may release these molecules into wine. The analyses’ request of haloanisoles and halophenols in oak wood has gradually increased in recent years, above all from the coopers, who want to prove the quality of their products. However, the lack of an official testing method is an obstacle for the performance monitoring of the laboratories. In response to these challenges, Bipea organizes, since October 2013, regular proficiency-testing schemes (PTS) for the detection and quantification of these molecules in oak wood. For each test, the statistical treatment of the laboratories’ results is performed according to ISO 13528 [2]. These PTS enable the participating laboratories to compare each other, draw up a general inventory of their analytical skills and improve their performances for detection and quantification analyses of haloanisoles and halophenols in oak wood. These tests are particularly important for the laboratories to obtain recognition of their analytical procedures by costumers and, above all, the accreditation bodies according to the ISO 17025 [3]. Key words Proficiency-testing schemes, haloanisoles and halophenols, wine quality control, laboratory performance

Caterina MAZZONI, Anne TIRARD, Abdelkader BOUBETRA, François ROUSSEY, Eric ZIEGLER: Proficiency-testing scheme for haloanisoles and halophenols in oak wood

63

1 INTRODUCTION Haloanisole contamination is a serious problem for wine quality: even trace amounts of 2,4,6-tricloroanisole (TCA), 2,3,4,6-tetrachloroanisole (TeCA), pentachloroanisole (PCA) and 2,4,6-tribromoanisole (TBA) can cause musty or moldy off-flavors in wine [4]. Each of these haloanisoles has a similar odour but possesses different sensory thresholds. These compounds are not naturally occurring wine constituents. The origin of haloanisoles can be attributed to the biodegradation of 2,4,6-tricholorophenol (TCP), 2,3,4,6-tetrachlorophenol (TeCP), pentachlorphenol (PCP) and 2,4,6-tribromophenol (TBP) respectively, which can be found in winery environments. Several materials, including barrels’ oak wood, may be contaminated and release these molecules into wine. The coopers’ need to prove the quality of manufactured barrels increases the request of analyses of haloanisoles and halophenols (HAHP) in oak wood. However, coopers and laboratories face to difficulties in results interpenetration due to the lack of an official testing method. Different analytical methods, more or less comparable, have been implemented by laboratories. The main goal of setting up this PTS is to develop an evaluation process for the laboratories’ performances, considering that laboratories can perform composition or migration analyses, which results are not equivalent. Both methods have advantages and disadvantages: analyses on corks are mostly migration ones, because the product impact (natural corks or technological corks) makes the composition analysis less relevant for coopers’ needs; on the other hand, composition methods developed by each laboratory should give the same result which corresponds to the exact value of HAHP. Arguments for and against composition and migration analyses are summarized in table 1.

Table 1. Migration and Composition analyses

Analysis type Advantages Disadvantages

Migration analyses: Maceration in a model wine solution for a minimum of 24h.

It reflects the thermo-dynamic equilibrium of HAHP composition between the wood chips and the liquid: this analysis is more similar to the real migration conditions in barrels.

The migration timings are unknown and long (24h). The management of the analysis period could be a problem above all for the returning of the results to the costumers.

Composition analyses: Solvent extraction on ground wood under defined temperature conditions.

It allows the complete analysis of all the wood contaminants. Thanks to the grinding the results are more accurate, but the laboratories have to manage the losses of compound during this step. The results could be quicker.

The interpretation of the results is difficult, because the migration and equilibrium conditions of the organo-halogenated in wood and in wine/alcoholic solution are unknown.

Another challenge in developing these PTS is manufacturing stable and homogeneous samples of oak wood dipped in haloanisoles and halophenols. This work describes the design and the implementation of a PTS for the analyses of HAHP oak wood samples, with a focus on the results of the proficiency test of March 2018.


64

2 EXPERIMENTAL Production of homogenous and stable samples is a crucial point for the PT implementation. The homogeneity and stability of the samples must be demonstrated to avoid misjudging laboratory performance owing to sample inadequacy. The main steps of the manufacturing of spiked oak wood samples, setting up to reach a stable equilibrium between free and absorbed molecules, are described in Figure 1.

Dispersion of the wood chipping in a glass container

Addition of ethanol and spiking solutions

Maceration of the product in a closed container to allow the

migration of halophenols and haloanisoles in wood

Slow evaporation of the solvent

Stabilization

Homogenization

Packing of the wood powder in an Aluminium foil and then in a bag under vacuum

Fig. 1. Manufacturing schema of spiked oak wood samples

The homogeneity of the samples is verified by experimental studies on 10 samples in duplicate taken randomly across the batch of samples, according to requirements of the ANNEX B of ISO 13528 standard [2]. The results are analysed through several statistical tests: -Fisher test (variance analysis): observed F value < critical F value; -Test of significant inhomogeneity: between sample variance < critical c value; -Study of the ratio of between samples standard deviation/standard deviation for

proficiency assessment: ss/SDPA < 30%. The stability of the samples during the test period is checked on the conditioned samples stored at (5±3) °C. According to ISO 13528 standard [2], the samples can be considered stable if the absolute difference between the means at t0 and t1 (t0+3 weeks, test period accorded to participants) is inferior or equal to the 0,3*standard deviation for proficiency assessment (|`y0 - `y1 | ≤ 0.3*SDPA). Once the homogeneity and the stability demonstrated, the samples are shipped at (5±3) °C to all the participants, who, given the nature of the product, are invited to analyze the samples as soon as possible after the reception. Laboratories’ results are collected via a reply form available online over a period of 3 weeks. In this test, migration and composition analyses are divided and evaluated separately. To harmonized migration analyses between all the laboratories, participants are asked to perform the analysis according to the following conditions (inspired on ISO 20752 standard [5] for cork stoppers adapted to the coopers’ needs):

- Solvent: model wine solution (20% ethanol, pH 3.4 with tartaric acid) - Temperature: room temperature (20°C +/- 2°C), without stirring - Soaking time: 24 hours - Extreme rate: 50 g/L.


65

No instruction is given for composition analyses, because methods developed by each the laboratory should give the same result which corresponds to the exact value of HAHP in wood. The statistical treatments of the returned results are conducted according to ISO 13528 [2]. The assigned values (xpt) are estimated using the robust means of the results from application of robust algorithm A. Performances of each laboratory are evaluated using a tolerance value (TV) of twice the standard deviations. This value is used to identify an interval around the assigned value. Results out of this range are considered as untrue. Moreover, the laboratories’ results (x) are also evaluated through z-scores (z):

The results of the laboratories who have a z-score ≤2are considered satisfactory and those with a z-score >2are classified as “questionable” or “unsatisfactory”: if the laboratories’ z-score is 2< z ≤3the result is considered questionable and >3unsatisfactory. The results are published in a specific interlaboratory comparison report distributed to all the participants who can then classify their results and implement some corrective and/or preventive actions if necessary. 3 RESULTS AND DISCUSSIONS Since October 2013, 4 tests per year are proposed to the laboratories. On average, 15 laboratories have been participating to PTS, with 10 results returned for composition analyses and 9 for migration ones. The laboratories’ results are good, with, medially, less than 1 unsatisfactory result by analytical parameter. However, the range of dispersion may reach 100% according to the organo-halogenated analysed, the concentration levels and the test; this can be mainly linked to the nature of product, the extraction and analysis methods. Figure 2 shows the graphs describing the dispersion of the results (twice the robust standard deviation) as a function of the assigned value for TeCA and TeCP from the test March 2013 (for composition and migration analyses). Values of the PT from March 2018 to March 2019 are represented in a different colour, green. A reduction of the dispersion of all the returned results can be noted on the last tests. This can be due to the fact that the laboratories participating to regular PT can have a critical point of view on the performed method of analysis and improve their analytical procedures to obtain values closer to the assigned ones. As the same product is analysed for migration and composition analyses, results of each PT the difference between assigned values were estimated for each compound. Averagely, assigned values of migration analyses are between 90% and 60% lower than those of composition analyses (see Table 2), that corresponds to the % of compound that can migrate from wood in the wine standard solution at 20°C for 24 hours. It has to be noted that this difference is not linked to the contamination level.


66

Table 2. Average relative difference between composition and migration assigned values (%) of the PT performed from October 2013

Compound Average relative d iffere nce between composition and migration values (%)

TBA -80 TCA -71 TeCA -82 PCA -90 TBP -78 TCP -60 TeCP -78 PCP -87

Fig. 2. Dispersion of the results (twice the robust standard deviation) as a function of the assigned value for TeCA and TeCP, for composition and migration analyses

Results of the PT of March 2018 are examined in details. Table 3 summarizes statistical data of this test for each analysed compound. Assigned values (xpt) were estimated for all HAHP except for the migration analysis of pentachlorophenol due to the high dispersion of the results. Uncertainties, u(xpt), that quantify the confidence to the assigned values, are good and vary according, among other factors, to the concentration levels and number of results taken into account to estimate the value. Coefficients of variation CV(xpt), reflecting data dispersion according to assigned values, range from 17% to 58% for migration analyses and from 28% to 48% for composition ones. Laboratories’ results are satisfactory except for migration analyses of TCA and TeCP, with, respectively, 3 and 2 unsatisfactory results out of 8 returned.


67

Table 3. Main statistical parameters of PT of March 2018

Compound Analysis xpt1

(ng/g) u(x pt)

2

(ng/g) s(x pt)

3

(ng/g) p(x pt)

4 CV(xpt)5

(%) σpt

6

(ng/g) pD

7 pCA8

TBA Composition 2.07 0.28 0.68 9 33 0.68 0 9

Migration 0.24 0.05 0.09 6 38 0.09 0 8

TCA Composition 4.08 0.82 1.96 9 48 1.96 0 9

Migration 1.15 0.09 0.18 6 16 0.18 3 8

TeCA Composition 2.33 0.26 0.66 10 28 0.66 0 10

Migration 0.30 0.09 0.15 5 50 0.15 0 8

PCA Composition 18.37 3.08 7.8 10 42 7.8 0 10

Migration 1.07 0.29 0.51 5 48 0.51 0 8

TBP Composition 4.03 0.65 1.63 10 40 1.63 0 10

Migration 0.68 0.18 0.32 5 47 0.32 0 8

TCP Composition 5.40 0.72 1.82 10 34 1.82 1 10

Migration 1.27 0.38 0.74 6 58 0.74 0 8

TeCP Composition 8.03 1.07 2.70 10 34 2.70 0 10

Migration 1.63 0.17 0.28 4 17 0.28 2 8

PCP Composition 71.99 8.80 22.26 10 31 22.26 1 10

Migration Due to the high dispersion of the results, no assigned value

was estimated 8

1.xpt: Assigned value or conventionally true value, calculated by the robust algorithm A from ISO 13528 standard. 2.u(x pt): Standard uncertainty of the assigned value; this value permits to quantify the confidence that can be given to the

assigned value. It depends on the mathematical model applied (algorithm A) and is a function of the standard deviation and the number of results used for the estimation of the assigned value. It is calculated as indicated in § 5.6.2 of ISO 13528 standard.

3.s(x pt): Robust standard deviation of the results, calculated by the robust algorithm A from ISO 13528 from all the results which participated to the estimation of the assigned value.

4.p(x pt): Number of results taken into account for the estimation of the assigned value. 5.CV(xpt): Coefficient of variation, this value permits to measure the dispersion of the results. 6.σpt: Standard deviation for proficiency assessment. 7.pD: Number of unsatisfactory results. 8.pCA:Total number of returned results (including incoherent and qualitative ones).

Absolute and relative differences between composition and migration values of each compound are described in Table 4. Results are coherent with those obtained in other tests, with a relative difference between -69% for TeCA and -94% for PCA analysis. Values of laboratories that returned quantitative results for both methods were also compared through z-scores. Fig. 3 shows Youden plot confidence ellipse based on Jackson method comparing migration and composition laboratories’ z-scores for TCA and TCP. Results are satisfactory for TCP, with only one laboratory out of the trueness region and the confidence ellipse at 5%. Concerning TCA, 3 laboratories are out of the trueness region with one out of the confidence ellipse at 1%.


68

Table 4. Absolute and relative difference between composition and migration values of the test of March 2018

Compound

Difference between composition and migration values

Minimum Median Maximum Assigned value of the test

Number of results taken into account

TBA Absolute (ng/g) -1.08 -1.57 -2.60 -1.73

6 Relative (%) -86 -88 -88 -88

TCA Absolute (ng/g) -1.50 -2.00 -2.77 -1.91

6 Relative (%) -88 -87 -83 -86

TeCA Absolute (ng/g) -1.06 -2.50 -4.66 -2.52

7 Relative (%) -84 -70 -73 -69

PCA Absolute (ng/g) -10.26 -15.28 -31.20 -17.04

7 Relative (%) -99 -94 -94 -94

TBP Absolute (ng/g) -1.88 -3.43 -4.91 -3.54

6 Relative (%) -95 -86 -82 -84

TCP Absolute (ng/g) -3.50 -4.06 -7.68 -3.92

8 Relative (%) -95 -77 -76 -76

TeCP Absolute (ng/g) -4.18 -7.59 -9.56 -6.48

7 Relative (%) -96 -85 -83 -80

PCP Absolute (ng/g) -47.16 -77.08 -108.60 -

7 Relative (%) -100 -94 -85 -

Fig. 3. Youden plot confidence ellipse based on Jackson method comparing migration and composition laboratories’ data for TCA and TCP, test of March 2018

Moreover, a complementary study was performed on the composition samples of this PT. Two samples of the same product were shipped to laboratories at different dates, t0 and at t1 (t0 + 20 days), and participants were invited to return their results in two independent reply forms. Values given by the laboratories at t0 and t1 correspond to


69

results obtained under intermediate precision conditions. Youden graphs have been elaborated for each parameter showing the laboratories’ results at t0 as a function of the t1 results to allow to each participant to place its results simultaneously for the two samples according to the bisector, which indicates the two same results. Fig. 4 presents these Youden graphs for TeCA and TeCP.

Fig. 4. Youden graphs comparing laboratories’ results obtained under intermediate precision conditions for TeCA and TeCP, composition analyses, test of March 2018

4 CONCLUSIONS A PTS for migration and composition analyses of haloanisoles and halophenols in oak wood has been implemented successfully, both from homogeneous and stable samples production and statistical point of view. This PT program has been approved and accredited by COFRAC (Comité Français d’Accréditation—French Accreditation Body). These proficiency tests are an important tool for laboratories, answering to a lack of official methods and enabling participants to draw up a general inventory of their analytical skills in term of composition and migration analyses. Laboratories can now monitor the reliability of their results and obtain recognition of their analytical procedures by coopers and accreditation bodies. REFERENCES

[1] Buser H. R., Zanier C. and Tanner H. Identification of 2,4,6-trichloroanisole as a potent compound causing cork taint in wine. Agric. Food Chem. 30 (1982): 359-362.

[2] ISO 13528:2015 Statistical methods for use in proficiency testing by interlaboratory comparisons.

[3] ISO/CEI 17025:2005 General requirements for the competence of testing and calibration laboratories.


70

[4] Callejón R.M., Ubeda C., Ríos-Reina R., Morales M.L. and Troncoso A.M. Recent developments in the analysis of musty odour compounds in water and wine: A review. Journal of Chromatography A. 1428 (2016): 72-75.

[5] ISO 20752:2014 Cork stoppers -- Determination of releasable 2, 4, 6-trichloroanisol (TCA).

The Seventh



71

30 YEARS OF INTER-LABORATORY TESTS FOR CEMENT IN

ROMANIA

Virginia -Graziela GUSLICOV

CEPROCIM S.A., Blvd. Preciziei, no. 6, Sector 6, Bucharest, Romania

[email protected]

Abstract Beginning with 1988, 30 years have passed since CEPROCIM had been organizing annual interlaboratory schemes for cement. At the time of initiating the interlaboratory attempts for cement in Romania, there were no standards in the world. The first International Guide, with two parts (ISO / IEC Guide 43 / 1.2), concerning intercomparison schemes, appeared about 10 years after the onset of the cement scheme in 1997. The Scheme of Interlaboratory Testing for Cement has been carried out for 30 years annually, involving in this period of time over 138 laboratories that had participated at least once in these inter-comparisons. According to their specificity, the participating laboratories belong to the following fields of activity: research, cement plants, precast units, hydro plant construction companies, building companies generally. Nowadays the cement interlaboratory tests scheme has an international character of one or more laboratories from 16 countries other than Romania, namely Austria, Bulgaria, Cyprus, Croatia, Lebanon, Lithuania, Macedonia, Republic of Moldova, Poland, Serbia, Sparska, Sri Lanka, Sudan, Turkey, Ukraine, Hungary, participated in the inter-comparison scheme. 38 standardized analysis and determinations are subject to cement inter-comparison tests. The statistical analysis of the data provided by the participants in the scheme according to ISO 15328:2015 [1] and the evaluation of the results is carried out according to the specific standards ISO/IEC 17043:2010 [2] that allow the results to be classified as satisfactory, questionable or unsatisfactory, and finally the appreciation of the competence of the laboratories. Global evaluation of the results of the interlaboratory test scheme in 30 years shows that the reproducibility of the international competence tests of compressive strength at 2 and 7 days expressed by the coefficient of variation shows that only in the last 2-3 years the condition of 5.5 and 4.5 imposed by the standard EN 196:1-2016 [3] was fulfilled. The coefficient of variation at the last edition is about 3 times lower than at the beginning years, however, the condition imposed by the same standard for the

Virginia-Graziela Guslicov : 30 years of Interlaboratory tests for cement in Romania

72

Reproducibility of the results of compressive strength at 28 days set in the beginning of 6% and currently of 4%, was rarely met. The results obtained during the 30 years of activity prove that the Interlaboratory Testing is one of the most effective tools for verifying and improving the working procedures and establishing a common language for all the participating laboratories which perform cement analyses and tests. Key words Interlaboratory tests for cement, reproducibility of the international competence tests of compressive strength for cement 1 HISTORY Beginning with 1988, 30 years have passed since CEPROCIM had been organizing annual interlaboratory schemes for cement. This activity was welcomed, especially at the beginning, and it contributed to the establishment of a common language between cement producers and users, among specialists in different fields (education, research, production). The cement quality complaints from the owners have gradually gone unnoticed by their understanding that poor or inadequate endowment of cement testing in their own laboratories is one of the major causes for non-appropriate results. For 15 years, the interlaboratory tests were carried out without a participation fee. In 2002-2003 this activity was carried out within a project (INFRAS) funded by the Ministry of Research and awarded with a diploma of excellence. At the time of initiating the interlaboratory tests for cement in Romania, there were no standards in the world. The first International Guide, with two parts (ISO / IEC Guide 43 / 1.2), concerning intercomparison schemes, appeared about 10 years after the onset of the cement scheme in 1997. 2 PARTICIPANTS The Scheme of Interlaboratory Testing for Cement has been carried out for 30 years annually, involving in this period of time over 138 laboratories that had participated at least once in these inter-comparisons (Fig. 1).

44

61

50

40

46 4644

5051

57

5356

26

43

37 3743

44

39 39

4745

3839

3232

3030

2327

0

10

20

30

40

50

60

70

Number of participants

Fig. 1. The number of participants in the cement inter-comparison scheme for 30 years


73

According to their specificity, the participating laboratories belong to the following fields of activity: research, cement plants, precast units, hydro plant construction companies, building companies in general (Fig. 2).

15 laboratories from 11 countriesAustria (2 laboratories)Cyprus (1 laboratory)Croatia (2 laboratories)Lithuania (2 laboratories)Poland (1 laboratory)Republic of Moldova (1 laboratory)Turkey (2 laboratories)Sparska (1 laboratory)Sri Lanka (1 laboratory)Sudan (1 laboratory)Hungary (1 laboratory)

11 laboratories from 8 countriesBulgaria (1 laboratory)Croatia (2 laboratories)Lebanon (1 laboratory)Macedonia (1 laboratory)Republic of Moldova (2 laboratories)Poland (1 laboratory)Turkey (1 laboratory)Serbia (1 laboratory)Ukraine (1 laboratory)

25

22

1713

61

Types of laboratories (138 participants)

Research Laboratories

Cements Plants

Precast Units

Hydro Plant Construction

Companies

Building Companies

Fig. 2 Number of participating laboratories according to their specific activity

Nowadays the cement interlaboratory tests scheme has an international character of one or more laboratories from 16 countries other than Romania, namely Austria, Bulgaria, Cyprus, Croatia, Lebanon, Lithuania, Macedonia, Republic of Moldova, Poland, Serbia, Republic of Srpska, Sri Lanka, Sudan, Turkey, Ukraine, Hungary, participated in the intercomparison scheme. 3 NATURE OF THE SAMPLES AND THEIR PREPARATION a) Prepare the cement lot At each edition of the interlaboratory tests a batch of industrial manufactured cement, always another, of about 800-1000 kg, is homogenized, tested from uniformity point of view and packaged in about 100 parcels.

b) Establishment of cement homogeneity For the determination of the cement homogeneity, it was chosen for repeatability tests based on the Blaine specific surface test according to EN 196:6 [4], carried out on different samples, taken from the homogenization cement lot. Checking the homogeneity of cement is done with the same equipment, by the same operator, in a minimum time.


74

For the determination of homogeneity, it was chosen for the repeatability tests, regarding the Blaine-specific surface determination performed in duplicate on 10 samples taken from 10 different packets. Checking the homogeneity of the cement is done with the same equipment, by the same operator, in a minimum amount of time. The homogeneity test is performed according to Annex B (B 2.2) of ISO 13528: 2015 [1], when: ss ≤ 0.3σpt

in which:

ss = the standard deviation calculated for the homogeneity tests

σpt = the standard deviation of repeatability = 50 cm2/g, required by the standard

EN 196-6 [4].

See for example the results of the homogeneity test of the 30th edition of Interlabora- tories tests (Fig.3).

4640

4645

4650

4655

4660

4665

4670

4675

4680

Sam

ple

1

Sam

ple

2

Sam

ple

3

Sam

ple

4

Sam

ple

5

Sam

ple

6

Sam

ple

7

Sam

ple

8

Sam

ple

9

Sam

ple 1

0Blai

ne S

pecif

ic Su

rface

are

a (cm

2/g)

Upper limit (Average + 0.3 · 50 cm2/g) Values 1-1 … 10-1

Values 1-2 … 10-2 Average value of specific surface

Lower limit (Average - 0.3 ·50 cm2/g)

Fig. 3 Results of the homogeneity test c) Sample packaging After checking homogeneity, cement is packaged in about 100 parcels. At the same time, from a standard SR EN 196-1 sand batch, about 100 parcels comprising 4 sandbags each of 1350 g each are prepared. Prior to packing, the sand is verified for its chemical, physical and granulometry characteristics and compliance with the requirements of the standard EN 196:1[3]. Packages are sent to the participants accompanied by the Work Instructions and the forms on which the results obtained by each laboratory are completed. According to the previously announced program, the participants submit their own results to CEPROCIM for statistical interpretation and evaluation. 4 DETERMINATIONS A number of 38 determinations are subject to interlaboratory tests, as follows: A number of 13 Chemical analysis: •Loss on ignition, according to EN 196-2 •SiO2, according to EN 196-2


75

•Al2O3, according to EN 196-2 •Fe2O3, according to EN 196-2 •CaO, according to EN 196-2 •MgO, according to EN 196-2 •SO3, according to EN 196-2 •Free CaO, according to BS EN 196-2 •Insoluble residue in HCl and Na2CO3, according to EN 196-2 •Na2O, according to EN 196-2 •K2O, according to EN 196-2 •Cl- , according to EN 196-2 •Water-soluble chromium (VI) content, according to EN 196-10 A number of 7 Physical Determinations: •Residue on the 90 µm sieve, according to EN 196-6 •Density, according to EN 196-6 •Blaine specific surface area, according to EN 196-6 •Initial setting time, according to EN 196-3 •Final setting time, according to EN 196-3 •Soundness (Le Chatelier needles), according to EN 196-3 A number of 9 Mechanical determinations performed according to EN 196-1 a) with sand from CEPROCIM •weight at demoulding •flexural strength at 1, 2, 7 and 28 days •compressive strength at 1, 2, 7 and 28 days b) with sand from the laboratory of every participant •weight at demoulding •flexural strength at 1, 2, 7 and 28 days •compressive strength at 1, 2, 7 and 28 days 5 STATISTICAL CALCULATION Establish of the assigned value for each determination (the consensus value), was made based on statistical calculation (algorithm A) indicated in the annex C from the standard ISO 13528:2015. The general principle of the above mentioned robust statistics means to apply an iterative calculation until data convergence. The value obtained for the robust mean after the last iteration represents the assigned value (of consensus) (xpt). z-score is calculated with the formula:

z =pt

pti xx

σ−

where: x i = the result reported by a participating laboratory (xlab); xpt = the assigned value; σpt = the standard deviation for competency evaluation (s*). The evaluation of the results is made according to EN ISO/CEI 17043:2010, as follows:


76

- SATISFACTORY, when 2≤z

- QUESTIONABLE, when 32 << z

- UNSATISFACTORY, when 3≥z

6 GLOBAL EVALUATION OF THE RESULTS OF THE INTERLABO RATORY TEST SCHEME IN 30 YEARS The reproducibility of the international competence tests of compressive strength at 2 and 7 days expressed by the coefficient of variation shows that only in the last 2-3 years the condition of 5.5 and 4.5 respectively provided in EN 196:1-2016 [3,5] was fulfilled (Fig.4 and Fig. 5).

12,52

15,18

13,4

14,25

17,41

20,51

14,24

11,99

12,91

20,99

13,613,41

9,498,18

7,216,68

8,31

11,6913,28

7,697,32

8,157,63

6,7

3,96

7,248,71

6,64

3,874,62

0

5

10

15

20

25

Ed

itio

n 1

(1

98

9)

Ed

itio

n 2

(1

99

0)

Ed

itio

n 3

(1

99

1)

Ed

itio

n 4

(1

99

2)

Ed

itio

n 5

(1

99

3)

Ed

itio

n 6

(1

99

4)

Ed

itio

n 7

(1

99

5)

Ed

itio

n 8

(1

99

6)

Ed

itio

n 9

(1

99

7)

Ed

itio

n 1

0 (

19

98

)

Ed

itio

n 1

1 (

19

99

)

Ed

itio

n 1

2 (

20

00

)

Ed

itio

n 1

3 (

20

01

)

Ed

itio

n 1

4 (

20

02

)

Ed

itio

n 1

5 (

20

03

)

Ed

itio

n 1

6 (

20

04

)

Ed

itio

n 1

7 (

20

05

)

Ed

itio

n 1

8 (

20

06

)

Ed

itio

n 1

9 (

20

07

)

Ed

itio

n 2

0 (

20

08

)

Ed

itio

n 2

1 (

20

09

)

Ed

itio

n 2

2 (

20

10

)

Ed

itio

n 2

3 (

20

11

)

Ed

itio

n 2

4 (

20

12

)

Ed

itio

n 2

5 (

20

13

)

Ed

itio

n 2

6 (

20

14

)

Ed

itio

n 2

7 (

20

15

)

Ed

itio

n 2

8 (

20

16

)

Ed

itio

n 2

9 (

20

17

)

Ed

itio

n 3

0 (

20

18

)

Coefficient of

variation obtained,

V (%)

Coefficient of

variation provided

in the standard, V

(%)

Fig. 4 The coefficient of variation for the reproducibility of international competence tests at 2 days according to EN 196-1: 2016

8,63

10,58

8,05

9,9410,98

13,98

10,07

9,84

9,81

11,87

9,04

9,188,41

7,88

5,35,92

6,32

9,01

5,914,79

8,21

6,01

7,64

5,01

3,14

6,535,57

4,36

4,29

3,74

0

2

4

6

8

10

12

14

16

Ed

itio

n 1

(1

98

9)

Ed

itio

n 2

(1

99

0)

Ed

itio

n 3

(1

99

1)

Ed

itio

n 4

(1

99

2)

Ed

itio

n 5

(1

99

3)

Ed

itio

n 6

(1

99

4)

Ed

itio

n 7

(1

99

5)

Ed

itio

n 8

(1

99

6)

Ed

itio

n 9

(1

99

7)

Ed

itio

n 1

0 (

19

98

)

Ed

itio

n 1

1 (

19

99

)

Ed

itio

n 1

2 (

20

00

)

Ed

itio

n 1

3 (

20

01

)

Ed

itio

n 1

4 (

20

02

)

Ed

itio

n 1

5 (

20

03

)

Ed

itio

n 1

6 (

20

04

)

Ed

itio

n 1

7 (

20

05

)

Ed

itio

n 1

8 (

20

06

)

Ed

itio

n 1

9 (

20

07

)

Ed

itio

n 2

0 (

20

08

)

Ed

itio

n 2

1 (

20

09

)

Ed

itio

n 2

2 (

20

10

)

Ed

itio

n 2

3 (

20

11

)

Ed

itio

n 2

4 (

20

12

)

Ed

itio

n 2

5 (

20

13

)

Ed

itio

n 2

6 (

20

14

)

Ed

itio

n 2

7 (

20

15

)

Ed

itio

n 2

8 (

20

16

)

Ed

itio

n 2

9 (

20

17

)

Ed

itio

n 3

0 (

20

18

)

Coefficient of

variation

obtained, V (%)

Coefficient of

variation

provided in the

standard, V (%)

Fig. 5 The variation coefficient for the reproducibility of the 7-day international competence tests according to EN 196-1: 2016


77

Although the overall trend is very low, the coefficient of variation at the last edition is about 3 times lower than in the beginning years, however, the condition imposed by the standard EN 196:1-2016 [3,5] for the Reproducibility of the results of compressive strength at 28 days set in the beginning of 6% and currently of 4%, was rarely met (Fig. 6). The fact is mainly due to the less experienced participants who participated in the inter-comparison tests, each year always others.

9,07

9,88

8,42

8,93

6,9

10,45

6

8,9

9,87

7,45

7,868,49

6,33

6,556,58

4,655,28

6,066,33

4,04

6,86

5,845,5

4,53

3,45

5,785,94

4,43

2,86

3,61

0

2

4

6

8

10

12

Ed

itio

n 1

(1

98

9)

Ed

itio

n 2

(1

99

0)

Ed

itio

n 3

(1

99

1)

Ed

itio

n 4

(1

99

2)

Ed

itio

n 5

(1

99

3)

Ed

itio

n 6

(1

99

4)

Ed

itio

n 7

(1

99

5)

Ed

itio

n 8

(1

99

6)

Ed

itio

n 9

(1

99

7)

Ed

itio

n 1

0 (

19

98

)

Ed

itio

n 1

1 (

19

99

)

Ed

itio

n 1

2 (

20

00

)

Ed

itio

n 1

3 (

20

01

)

Ed

itio

n 1

4 (

20

02

)

Ed

itio

n 1

5 (

20

03

)

Ed

itio

n 1

6 (

20

04

)

Ed

itio

n 1

7 (

20

05

)

Ed

itio

n 1

8 (

20

06

)

Ed

itio

n 1

9 (

20

07

)

Ed

itio

n 2

0 (

20

08

)

Ed

itio

n 2

1 (

20

09

)

Ed

itio

n 2

2 (

20

10

)

Ed

itio

n 2

3 (

20

11

)

Ed

itio

n 2

4 (

20

12

)

Ed

itio

n 2

5 (

20

13

)

Ed

itio

n 2

6 (

20

14

)

Ed

itio

n 2

7 (

20

15

)

Ed

itio

n 2

8 (

20

16

)

Ed

itio

n 2

9 (

20

17

)

Ed

itio

n 3

0 (

20

18

)

Coefficient of

variation

obtained, V (%)

Coefficient of

variation

provided in the

standard, V (%)

Fig. 6 Coefficient of variation for the reproducibility of 28 days competence tests according to EN 196-6: 2016

7 CONCLUSIONS •The Scheme of Interlaboratory Testing for Cement initiated since 1988 by the CEPROCIM laboratory is the first and the only scheme of inter-laboratory cement testing in Romania. •The scheme has been carried out annually, to date, involving in this period of time over 138 laboratories that had participated at least once in these inter-comparisons. •Nowadays the cement interlaboratory tests scheme has an international character in one or more laboratories from countries such as Austria, Bulgaria, Cyprus, Croatia, Lebanon, Lithuania, Macedonia, Republic of Moldova, Poland, Republic of Sparska (inlocuit cu Srpska), Serbia, Sri Lanka, Sudan, Ukraine, Hungary, participating in the inter-comparison scheme. •38 standardized analyses and determinations are subject to cement inter-comparison tests. •The statistical analysis of the data provided by the participants in the scheme and the evaluation of the results are carried out according to the specific standards that allow the results to be classified as satisfactory, questionable or unsatisfactory, and finally the appreciation of the competence of the laboratories. •Global evaluation of the results of the interlaboratory test scheme in 30 years showed that the reproducibility of the international competence tests of compressive


78

strength at 2 and 7 days expressed by the coefficient of variation shows that only in the last 2-3 years the condition of 5.5 and 4.5 imposed by the standard EN 196:1was fulfilled. The coefficient of variation at the last edition is about 3 times lower than in the beginning years, however, the condition imposed by the same standard for the Reproducibility of the results of compressive strength at 28 days set in the beginning of 6% and currently of 4%, was rarely met. •The results obtained during the 30 years of activity prove that the Interlaboratory Testing is one of the most effective tools for verifying and improving the working procedures and establishing a common language for all the participating laboratories which perform cement analyses and tests. REFERENCES

[1] ISO 13528:2015. Statistical methods for use in proficiency testing by interlaboratory comparison

[2] ISO/IEC 17043:2010. Conformity assessment — General requirements for proficiency testing

[3] EN 196:1. Methods of testing cement - Part 1: Determination of strength [4] EN 196:6. Methods of testing cement - Part 6: Determination of fineness

[5] All the 30 General Reports in period 1988 -2018 regarding Interlaboratory Tests for Cement, organized by CEPROCIM (ICPILA) Bucharest, Romania

Acknowledgments are due to all team which have worked at Interlaboratory tests for cement in the last 30 years, especially to eng Cristina Vlad and geol. Nicoleta Vlad.

The Seventh



79

RESEARCH ON THE PROFICIENCY TESTING PROGRAM OF

DETERMINATION OF SOLUBLE LEAD IN SYNTHETIC MATERIAL S TRACK SURFACES

Ke-xiao Y u, Sha-sha Wu, H ui -chao Liang, Ying Xu, Xiao -ling Zhu, Zhong-bao Guo, Jian-bo Zhou, Ruo-lin Li

China Building Material Test & Certification Group Co., Ltd. No.1 Guanzhuang Dongli Chaoyang District, Beijing, China

[email protected] Abstract Proficiency testing is an important technical means to understand or monitor laboratory capacity, as well as an important external quality assurance for laboratories. All kinds of toxic heavy metals in plastic track will not only affect human health seriously, but also poison the water and soil, which will do great harm to the environment. In order to learn about the laboratory technical ability of lead detection in the surface layer of synthetic material runway, the capability verification plan is carried out. In this PT plan, the sample selection, preparation, crushing methods, uniformity and stability, the data statistics and evaluation are studied in detail. The result shows that the uniformity and stability of samples are critical to the PT plan, and the crushing methods, particle size and shape have more effects to soluble lead detection. Key words Plastic track; soluble lead; proficiency testing 1 INTRODUCTION In recent years, the problem of "poisonous running track" has become a hot issue of social concern, affecting Beijing, Jiangsu, Hebei, Guangdong and other provinces and cities. The plastic track contains lead salt, a heavy metal drier, which can promote the solidification of the track, but lead can cause permanent pollution. When human skin is exposed to lead on a plastic track, lead penetrates inside the body, causing serious damage to the body. Lead salt is not only harmful to human body, but also poisons water and soil, and seriously pollutes the environment. There is no such proficiency testing program in China. The purpose of this proficiency testing is to strengthen the supervision and management of the laboratory response

Ke-xiao Yu, Sha-sha Wu, Hui-chao Liang, Ying Xu, Xiao-ling Zhu,Zhong-bao Guo, Jian-bo Zhou, Ruo-lin Li: Research on the Proficiency Testing Program of Determination of Soluble Lead in Synthetic Materials Track Surfaces

80

to test the determination of soluble lead in plastic track, and to improve mastering ability of the testing methods and testing level. Certification and Accreditation of the People’s Republic of China (CNCA) organized the PT program, and implemented by China Building Material Test & Certification Group Co., Ltd. This PT program number is CNCA-17-A16. This paper gives a detail relevant research of the PT Program. 2 SAMPLE SELECTION AND PREPARATION 2.1 Sample selection The plastic track can be classified into four types: mixed type plastic track, breathable plastic track, compound type plastic track, all plastic racetrack, etc. according to the construction method of the material and construction of the plastic racetrack. Compared with other types of runway, the most prominent feature of the full-plastic runway is the artificial synthesis of pure PU (polyurethane). It doesn't contain any black particles. As long as in the construction process to ensure the uniformity of the thinner, you can ensure the uniformity of the sample. Finally, the full-plastic self-textured runway was selected as the sample for this PT. 2.2 Sample preparation 2.2.1 Sample design Two samples with different soluble lead contents were designed and prepared: sample A and sample B. Two samples were distributed to each participating laboratory at the same time, in which sample B was used as interference sample, which could effectively prevent data collusion between laboratories. 2.2.2 Sample preparation, packaging, signage and storage Sample preparation: After fully mixing the slurry, the samples were prepared according to the normal construction procedures of the full-plastic runway. After 14 days of solidification, the packaging is divided; Sample size: 150mm*150mm* actual thickness; Specifications: 1 piece/bag; Packaging requirements: The sample shall be encapsulated with a single layer of plastic bag, and the sample identification shall be affixed outside the plastic bag; Storage conditions: The samples were stored in a cool, dry, ventilated and clean room temperature environment.


81

3 DETECTION METHOD GB/T 14833-2011 Synthetic materials track surfaces. GB 9758.1-88 Paints and varnishes—Determination of “soluble” metal content—Part 1: Determination of lead content—Flame atomic absorption spectrophotometric method and dithizone spectrophotometric method. 4 SELECTION OF SAMPLE TREATMENT 4.1 Sample test location In the implementation of this PT plan, the detection positions are also specified in the operation instruction. This can not only ensure the sample test consistency, but also to a certain extent to prevent the impact of sample edge pollution on the test results. The detection position of samples is shown in Fig. 1.

Detection position

Fig.1. Schematic diagram of sample detection position 4.2 Sample fragmentation According to the requirements of GB/T 14833-2011 standard, the samples should be processed into small particles of about 1mm3. No heat can be generated during the sample processing of the plastic runway. There are two kinds of treatment: one is frozen grinding, one is hand-cut. The crushed samples should be screened to select about 1mm3 of small particles for testing. Six samples were randomly selected from sample A, and the samples were processed by means of freeze grinding and manual cutting. After screening, three sample particles with different particle sizes were obtained, and the soluble lead content in the sample particles with different particle sizes was tested. The detection data is shown in Table 1.


82

Table 1. Sample test data of different treatment methods and different particle sizes

Processing method Hand cut Refrigeration-rubbing

Particle sizes 0.85mm≤φ≤1.4mm Φ≥1.4mm 0.85mm≤φ≤1.4mm

1 203.8 109.3 312.5

2 194.8 112.5 330.1

3 202.6 118.6 315.7

4 198.6 108.9 322.2

5 196.6 120.5 316.7

6 198.5 116.2 323.4

Average 200.0 114.3 320.1

The following conclusions can be obtained from the data in table 1: (1) In the process of plastic track detection, different sample treatment methods and particle sizes have great influence on the detection results. (2) In the same treatment mode, the smaller the sample particle diameter, the larger the specific surface area of the sample, the larger the contact area of the leaching liquid in the process of soluble lead leaching, so that the soluble lead content in the sample with small particle size is relatively large; (3) Under the condition of different treatment methods and the same particle size, the detection results of frozen ground samples were larger. This is because the shape of samples after freezing grinding is highly irregular, and the specific surface area is much larger than that of samples with the same particle size under the condition of manual cutting. Therefore, in order to make the laboratory test results consistent, the following provisions are made for the crushing treatment and particle size selection of samples in the work instruction of PT plan: a) Using hand-cut processes to crush samples; b) Fine particles with particle size ranging from 0.85mm to 1.4mm (20 mesh to 14 mesh) were selected as test samples. 5 HOMOGENEITY TEST AND STABILITY TEST 5.1 Homogeneity test

According to CNAS-GL003:2018 Guidance on Evaluating the Homogeneity and Stability of samples Used for Proficiency Testing, F-test method was used to investigate the homogeneity of samples. When F ≤ F0.05 (9,10) = 3.02, the sample is judged to be uniform. The homogeneity test results and determination are shown in Table 2.


83

Table 2. Test results and evaluation of sample homogeneity

Number Soluble Lead（mg/kg）

First Second

001 203.8 196.6 007 194.8 201.9 028 202.6 204.1 043 198.6 200.5 052 196.6 203.6 139 198.5 202.5 157 201.1 194.9 177 197.6 195.2 185 192.3 195.6 202 198.2 196.4

Average 198.8

F 1.15

Conclusion F≤F0.05（9,10）=3.02， The sample is judged to be uniform.

Homogeneity verification

Ss 0.94

0.3σPT 3.71

Conclusion Ss≤0.3σPT， The homogeneity test of samples is qualified.

The results of the homogeneity test show that the characteristic values of the sample are uniform. After the data statistics are completed, the homogeneity test results are verified by Ss≤0.3σPT, and the results show that the homogeneity test of samples is qualified. 5.2 Stability test

According to CNAS-GL003:2018, t-test method was used to evaluate the stability of samples. On August 14, 2017 and September 20, 2017, six samples were randomly taken to detect the soluble lead content. The stability test results and judgment are shown in table 3. The results of the stability test showed that the soluble lead content in the sample was stable in the process of the proficiency testing.


84

Table 3 Test results and evaluation of sample stability

Soluble Lead（mg/kg）

Number 2017.7.10 Number 2017.8.14 Number 2017.9.20

001 203.8 018 200.3 051 198.8

007 194.8 046 198.5 081 197.6

028 202.6 077 196.4 119 199.9

043 198.6 133 199.8 134 203.2

052 196.6 156 201.6 139 200.7

139 198.5 188 203.1 211 201.1

157 201.1 / / / /

177 197.6 / / / /

185 192.3 / / / /

202 198.2 / / / /

t / 0.958 1.161

Conclusion /

t＜t 0.05（14）=2.145 As of August 14, 2017,

the soluble lead content in the samples was

stable.

t＜t 0.05（14）=2.145 As of September 20,

2017, the soluble lead content in the

samples was stable.

6 DATA STATISTICAL DESIGN AND EVALUATION METHOD 6.1 Data Statistical Design According to CNAS-GL002:2018 Guidance on Statistic Treatment of Proficiency Testing Results and Performance Evaluation, the testing results were evaluated by robust statistical method A. Statistical analysis was performed on the results obtained from the labs in the PT

program, including the total number, robust average value，robust standard deviation

，robust Cv, robust mean designated uncertainty, minimum, maximum, range. 6.2 Evaluation Method The robust average value is as the assigned value and the robust standard deviation is as the standard deviation for proficiency assessment, and then calculate the Z-score for each laboratory results.

( ) ˆ/Z x X σ= −

In formula: x--laboratory test results; X-- Robust average value; σ --standard deviation for proficiency assessment. The proficiency of each lab is evaluated by Z-score.


85

∣Z∣≤2 Satisfactory results

2＜∣Z∣＜3 Questionable results

∣Z∣≥3 Unsatisfactory results

7 RESULT EVALUATION 41 labs participated in the PT program. The z-score of laboratory soluble lead test results was calculated. In order to clearly show the results of the verification of the participating abilities of the laboratories, the z-scores are arranged as a histogram in the order of size, each of which is marked with the number of the laboratory.

Fig.2. Z-score histogram

From z-score histogram, see Fig.2. We can see that the result satisfaction rate is over 90%. This suggests that most of the laboratory in the aspects of soluble lead has higher detection ability. 8 CONCLUSIONS (1) The technical key of PT plan is to ensure the homogeneity and stability of PT samples. It can be seen from the verification that the homogeneity of the samples can be effectively guaranteed by using the full-plastic self-textured runway. (2) Particle size and particle shape after sample crushing can directly affect the detection results of soluble lead. It is suggested to select fine particles with particle size between 0.85mm~1.4mm (20 mesh~14 mesh) as test samples after crushing treatment. (3) The crushing treatment of plastic runway samples has a great influence on the determination of soluble lead. It is recommended to specify the crushing treatment method in relevant standards.


86

REFERENCES

[1] GB/T 27043-2012 Conformity assessment -- General requirements for proficiency testing.

[2] GB/T 28043-2011 Statistical methods for use in proficiency testing by interlaboratory comparisons.

[3] CNAS-GL002:2018 Guidance on Statistic Treatment of Proficiency Testing Results and Performance Evaluation.

[4] CNAS-GL003:2018 Guidance on Evaluating the Homogeneity and Stability of Samples Used for Proficiency Testing.

[5] GB/T 14833-2011 Synthetic materials track surfaces. [6] GB/T 9758.1-1988 Paints and varnishes – Determination of “soluble” metal

content— Part 1: Determination of lead content—Flame atomic absorption spectrometric method and dithizone spectrophotometric method.

The Seventh



87

OVERVIEW OF PROFICIENCY TESTING IMPLEMENTATION OF

STYRENE BUTADIENE STYRENE MODIFIED BITUMIOUS SHEET MATERIALS

Sha-sha Wu , Ke-xiao Yu , Hui-chao Liang , Ying Xu , Jian-bo Zhou, Ruo-lin Li

China Building Material Test & Certification Group Co., Ltd., No.1, Guanzhuang Dongli, Chaoyang District, Beijing, China

[email protected]

Abstract This paper summarizes the proficiency testing programme of styrene butadiene

styrene （ SBS ） modified bitumious sheet materials, and focuses on the critical control points in the process of proficiency testing,including sample design, on-site inspection design, statistical design, etc. These critical control points effectively guarantees the authenticity of the data. This proficiency testing adopts a combination of routine proficiency testing and on-site inspection,which effectively prevent data collusion between laboratories. This proficiency testing truly reflects the laboratory's detection level, and help the laboratory to identify differences between other laboratories, which can promote the improvement of laboratory's detection level. Key words Styrene butadiene styrene modified bitumious sheet materials, proficiency testing, critical control point, sample design, on-site inspection design, statistical design 1 INTRODUCTION

The proficiency testing programme of styrene butadiene styrene（SBS）modified bitumious sheet materials was organized by Beijing Municipal Administration of Quality and Technology Supervision in 2018.This proficiency testing adds an on-site inspection mode. On-site inspections include on-site inspection process inspections and related data verification. Compared with the conventional proficiency testing mode, this mode can prevent data collusion, and it can help the laboratory to find problems in the detection process. It is easier to find some problems in the detection process and daily management by the on-site inspections. This paper analyzed and summarized some critical control points in The proficiency testing programme.

Sha-sha Wu, Ke-xiao Yu, Hui-chao Liang, Ying Xu, Jian-bo Zhou, Ruo-lin Li: Overview of proficiency testing implementation of styrene butadiene styrene modified bitumious sheet materials

88

2 ON-SITE INSPECTION DESIGN AND IMPLEMENTATION The use of proficiency testing combined with on-site inspection mode is actually to increase on-site inspection activities in the conventional proficiency testing mode. The on-site inspections include on-site inspection process inspections and related data verification. This mode can effectively prevent data collusion than the conventional proficiency testing mode. It is easier to find some problems in the detection process and daily management through the verification of on-site data. The on-site inspection implementation plan is as follows: 2.1 On-site inspection team Four on-site inspection teams were established to meet the requirements of on-site inspections. In order to ensure impartiality, each on-site inspection team consists of two people. At the specified time, the on-site inspection team goes to the laboratory for on-site inspection. 2.2 On-site inspection of the inspection process According to the detection characteristics of the samples, different schemes are planned. If the samples without prior state adjustment are needed, the on-site inspection team will take the samples directly to the laboratory; if there is a state adjustment request, the laboratory is required to receive samples at the specified time. And the state adjustment, the laboratory can not start testing without the consent of the on-site inspectors. On-site inspectors need to take photos of the key aspects of the inspection process. 2.3 On-site inspection of relevant witness material s The inspection contents include: inspection process, environmental facilities, equipment, original records, etc. On-site inspection personnel are required to fill in the “on-site inspection form” during the on-site inspection to verify the relevant witness materials provided by the laboratory and record them. Finally, the person in charge of the laboratory confirms the signature and seal. The on-site inspection form and all witness materials shall be sealed and sealed on the spot, and the laboratory seal shall be stamped at the seal and brought back by the on-site inspection personnel. 3 SAMPLE DESIGN AND IMPLEMENTATION 3.1Sample overview

This proficiency test uses two homogeneous styrene butadiene styrene（SBS）modified bitumious sheet materials as samples. The sample size is 400mm*1000mm. To ensure that the sample is not exposed to moisture, the sample is packaged in a plastic bag with a unique identifier. The initial test will randomly distribute the A sample or the B sample to each participating laboratory; the supplementary test will issue a different sample to the laboratory than the initial test.


89

3.2 Sample design and implementation

The items of the proficiency testing are the lateral maximum tensile force and the lateral elongation at maximum tensile force. GB/T 328.8-2007 "Test method for building waterproofing - Part 8: Bitumen sheets for owaterproofing-tensile properties"[1]. 3.3 Homogeneity and stability of samples

3.3.1 Homogeneity test

In the prepared samples A and B, 10 bags of samples were randomly selected for uniformity test. According to GB/T 328.8-2007, the lateral maximum tensile force and the lateral elongation at maximum tensile force are measured under repeated conditions. According to CNAS-GL003: 2018 " Guidance on Evaluating the Homogeneity and Stability of Samples Used for Proficiency Testing "[2], the results of uniformity test of

samples were evaluated and verified by F test. When F ≤ F0.05 (9, 10) = 3.02, it was judged that the sample was uniform, as shown in Table 1.

Table 1. The homogeneity test data

Test parameters

sample A sample B

the lateral maximum tensile force

the lateral elongation at maximum tensile force

the lateral maximum tensile force

the lateral elongation at maximum tensile force

First Second First Second First Second First Second

01 964 956 44.2 45.2 1072 1078 56.9 55.1

02 954 962 44.5 44.7 1068 1062 55.6 55.5

03 961 947 44.3 43.5 1058 1065 53.6 54.1

04 977 956 45.1 43.4 1095 1072 53.7 53.5

05 959 972 46.7 43.9 1077 1075 54.3 55.6

06 970 969 46.6 45.6 1066 1068 56.2 56.8

07 962 957 45.3 44.8 1051 1076 56.8 56.4

08 972 965 45.9 47.5 1068 1083 54.0 56.6

09 955 958 45.3 47.4 1072 1050 56.4 55.1

10 965 970 45.3 45.5 1069 1053 54.1 56.1

Everage 963 45.2 1069 55.3

F 1.19 1.64 1.13 2.21

Conclusion F＜F0.05（9，10）=3.02,samples are homogeneous.

homogeneity test

Ss 7.0 0.83 11.0 1.28

σPT 38.2 5.04 41.4 5.65

0.3σPT 11.5 1.51 12.4 1.70

Conclusion Ss≤0.3σPT ,samples are homogeneous.


90

3.3.2 Stability test According to CNAS-GL003 Guidance on Statistic Treatment of Proficiency Testing

Results and Performance Evaluation[2]，six samples were randomly selected from

the prepared samples，test parameters are consistent with homogeneity test. The

results were analyzed by t-test in CNAS-GL003, When t≤t0.05（14）=2.145，then the samples were judged to be stable. The Stability test data are shown in table 2.

Table 2. The Stability test data(sample A)

Table 3. The Stability test data(sample A)

the lateral maximum tensile force the lateral elongation at maximum tensile force

NO. 2018.5.9 NO. 2018.9.3 NO. 2018.5.9 NO. 2018.9.3

A01 964 A11 955 A01 44.2 A11 43.9

A02 954 A12 967 A02 44.5 A12 44.6

A03 961 A13 976 A03 44.3 A13 45.5

A04 977 A14 958 A04 45.1 A14 43.8

A05 959 A15 982 A05 46.7 A15 44.8

A06 970 A16 969 A06 46.6 A16 46.3

A07 962 / / A07 45.3 / /

A08 972 / / A08 45.9 / /

A09 955 / / A09 45.3 / /

A10 965 / / A10 45.3 / /

t / 0.892 t / 1.076

Conclusion t＜t 0.05（14）=2.145， sample A is stable at the 0.05 significance level

the lateral maximum tensile force the lateral elongation at maximum tensile force

NO. 2018.5.9 NO. 2018.9.3 NO. 2018.5.9 NO. 2018.9.3

B01 1072 B11 1058 B01 56.9 B11 55.8

B02 1068 B12 1073 B02 55.6 B12 54.3

B03 1058 B13 1048 B03 53.6 B13 57.1

B04 1095 B14 1053 B04 53.7 B14 56.8

B05 1077 B15 1065 B05 54.3 B15 55.1

B06 1066 B16 1078 B06 56.2 B16 56.8

B07 1051 / / B07 56.8 / /

B08 1068 / / B08 54.0 / /

B09 1072 / / B09 56.4 / /

B10 1069 / / B10 54.1 / /

t / 1.184 t / 1.257

Conclusion t＜t 0.05（14）=2.145， sample A is stable at the 0.05 significance level


91

4 STATISTICAL DESIGN AND EVALUATION METHODS The proficiency testing received results from 74 laboratories. The z-score was calculated and the Satisfactory, the suspect and the outlier were judged according to “3. statistical design and evaluation methods”.The z-scores were shown in Fig.1 in the form of a histogram.

Fig. 1. Sample A：The histogram of z-scores (the lateral maximum tensile force)

Fig. 2. Sample B： The histogram of z-scores (the lateral maximum tensile force)

key

x number of laboratories

y z-score for the lateral maximum tensile force


92

Fig. 3. Sample A： The histogram of z-scores (lateral elongation at maximum tensile force)

Fig. 4. Sample B：The histogram of z-scores (lateral elongation at maximum tensile force)

key x number of laboratories y z-score for the lateral elongation at maximum tensile force


93

5 SUPPLEMENTARY TEST DESIGN AND IMPLEMENTATION According to the requirements of the article “JING ZHI JIAN FA [2018] No. 60”, “If the test result is outlier or suspicious, the inspection and testing institution shall submit an application for retesting to the project undertaker within 10 days after receiving the notification of the first test result of the capability verification. Each project undertaker shall carefully review the application for retesting and rectification of the relevant institution. If the rectification or rectification is invalid, it is not allowed to participate in the supplementary test. If the applicant fails to apply for a supplementary test or fails to participate in the supplementary test, the result of this capability verification is regarded as “Outlier”. The laboratory with suspect or Outlier test can perform a supplementary test to supplement the test result as the final result. The results of the supplementary test are statistically processed using the first result statistic, and the results of the supplementary test are determined according to the result evaluation method. The first test results for the 10 laboratories with problems or dissatisfaction submitted the retest application and rectification materials on time. For the laboratories No. 15, 18, 30, and 56 that participated in the supplementary test, the test sample B was issued, and the test sample A was issued to the laboratories 40, 43, 46, 63, 67, and 69 that participated in the supplementary test. After rectifications, the results of the retests were satisfactory. After the retest, the number of laboratories with satisfactory results rose to 74, and the satisfaction rate rose to 100%.

6 SUMMARY During the implementation of this proficiency test, the key control points were fully considered, including sample design, on-site inspection design, statistical design, and re-test design. This proficiency test uses a combination of routine proficiency testing and on-site inspection to effectively prevent data collusion between laboratories and truly reflect the laboratory's detection level, which helps the laboratory to identify differences between peers. The successful implementation of this capability verification provided real and accurate data support for the supervision and management of the Beijing Municipal Bureau of Quality Supervision and Inspection. REFERENCES

[1] Test method for building waterproofing - Part 8: Bitumen sheets for owaterproofing-tensile properties GB/T 328.8

[2] Guidance on Evaluating the Homogeneity and Stability of Samples Used for Proficiency Testing CNAS-GL003

[3] Guidance on Statistic Treatment of Proficiency Testing Results and Performance Evaluation CNAS-GL002

The Seventh



94

ESTABLISH A NATIONWIDE CEMENT TESTING COMPARISON

PLATFORM TO REALIZE INFORMATION-BASED MANAGEMENT

Ying Xu, Kexiao Yu, Shasha Wu, Xiaoling Zhu, Huicha o Liang

China National Center Quality Supervision and Test of Cement No.1, GuanZhuang Dongli, Chaoyang District, Beijing, China

[email protected]

Abstract Nationwide Cement Testing Comparison (NCTC) dates back to 1983. China National Center Quality Supervision and Test of Cement (CCQTC), as the agency of highest-level metrological traceability of cement testing in China, organizes nationwide cement quality items testing comparison and nationwide cement chemical analysis comparison once every two years. In the past three decades, CCQTC has successfully held 16 nationwide cement quality items testing comparisons and 16 nationwide cement chemical analysis comparisons. Comparison has played an important role in improving test abilities of laboratories of chinese cement enterprises and has made significant contributions in cement product quality improvement. This paper introduces the background, basis and technical solutions of the NCTC platform, and presents detailed functions of the NCTC platform including publishing comparison schemes, uploading files, online registration, data query, test result announcement, etc. Clients can register, log in, enroll, confirm receipt of sample, submit data, inquire, download files and print files all in one system. Moreover, this paper also introduces the current status of the platform promotion and application, and the actual results of clients services. Key words NCTC, information-based management, platform, functions 1 INTRODUCTION China National Center Quality Supervision and Test of Cement (“CCQTC”), with the objective of continuously improving operational efficiency, has been optimizing its processes and management system by adopting the fast developing information technologies. As one major initiative, it was decided to build up an official website, in which a critical module is the Nationwide Cement Testing Comparison ( “NCTC”) platform.

Ying Xu, Kexiao Yu, Shasha Wu, Xiaoling Zhu, Huichao Liang, Establish a Nationwide Cement Testing Comparison Platform to Realize Information-based Management

95

Information Technology has been rapidly developing and implementing, in the mean time, the scale of cement production has also increased dramatically, and Big Data application systems have gained rich experiences. It not only adopts increasingly mature information technology, enhances central data resource integration and implement further applications such as data sharing and statistical analysis, but also strengthens its relationship with clients and rapidly respond to client’s needs by providing an on-line platform. Mature information technology and client’s needs are the fundamental drivers for the innovation of this informational operation system in the Center. 2 TECHNICAL SOLUTION Based on Microsoft SQL Server 2008 R2, a comprehensive database platform, the Comparison conducts enterprise-level data management. Microsoft SQL Server 2008 R2 as a data engine, is the core of the data management solution. Split architecture is used to ensure effecitve table split and data distribution. In the Comparison, same user’s data are stored in the same partition whevenr possible; After distribution, it provides multiple options to save the data of user relationship. It can set up and deploy cost-effective smart data solutions, which introduces data based applications into various business fields. 3 FUNCTIONS FOR SERVICE PROVIDERS Functions designed for service providers include: establish official website and major modules, system management, publish comparison and proficiency testing projects, issue notices and announcements, promote policies, regulations and relevant standards, operate and manage NCTC platform. 3.1 Set up official website with principal modules include business center, news information, resource center and member center.

3.2 System management includes user permission management, data security design, system operation log management, content management, etc. 3.3 Publish NCTC projects, issue notice and announcement, share center dynamics and industry news, promote relevant information including policies, regulations and standards. 3.4 Establish NCTC management system, publish comparison schemes, manage enterprise’ registration, data query, data comparison statistics and generate comparison reports.


96

4 FUNCTIONS FOR CLIENTS Functions designed for clients: register and log in, online regeistration, tracking courier number, confirm receipt of sample, submit data online, query progress, find comparison awards, download reports and other important client services. 4.1 Register and log in Clients register and fill in with relevant information on the official website. Clients can log in by inputting user name and password after successful registration. 4.2 Sign up online

Fig.2. Sign up Page 4.3 Query progress

Clients can query the progress of the implementation of comparison project in the member center. Clients go into the next step after completing and confirming the previous step.

Fig.3. The Implementation of Progress

4.4 Confirm receipt of sample

4.4.1 The CCQTC sends out samples after preparation. Clients can query the date of shipment, express company and tracking number.


97

4.4.2 After receive the sample, click “confirm receipt”. If the sample is intact, click “intact” and submit; if the sample is damaged and damp which may affect inspection results, click “not intact”, upload the picture of sample and submit. After confirmation, the center will send out a new sample. 4.5 Fill in data online

4.5.1 Participate in sample testing and then fill in the test data on online. The data cannot be modified after submission. 4.5.2 Submit testing result, the scanning copy of the paper-based report of test result of comparison should also be uploaded. 4.6 Query result 4.6.1 Query awards Query total score and award status 4.6.2 Query reports of comparison Reports of comparison can be queried, downloaded and printed out. 5 FEATURES OF NCTC PLATFORM 5.1 The platform can send email and text message automatically, support form and data list printing with external printers, do statistical analysis of imported data and export data in word and excel documents. 5.2 The platform is equipped with a self-service module, by which users can query comparison projects released and their progress, track samples shipped. In the member center, users can query, download and print out test data, comparison reports and awards over the years, which guarantees the continuity and integrity of client data. 5.3 The platform provides Important Client services. It provides large-scale cement enterprises with the data statistics and analysis reports of their subordinate units participating in nationwide cement comparison so as to further provide Important Clients with tailored services. 6 IMPLEMENTATION AND ACHIEVEMENTS NCTC platform was officially launched in Mar 2017. By May 2019, it has nearly two thousand registered enterprise members. The platform has successfully processed the data of nationwide cement chemical analysis comparison in 2016, and has supported the entire operation of nationwide cement quality items testing comparison in 2017 and 2018, with 1212 and 1219 enterprises attended respectively. The services include online registration, data upload, statistical summary, awards query, report download, etc. All functions of NCTC platform have been well accepted and highly commented by clients. NCTC platform has provided a highly efficient mechanism for clients to communicate with CCQTC, and greatly improved the productivity of the Comparison. With further development and improvement of modules of the official website, the platform will provide clients with more comprehensive, efficient and high-quality services, to achieve mutual benefit and win together.


98

REFERENCES

[1] Lin Ziyu. Principles and Applications of Big Data Technology: Concept, Storage, Processing, Analysis and Application. Posts and Telecocm Press, 2017.

[2] Zhong Qiuyan. Principles and Applications of Database. Tsinghua University Press, 2016.

[3] Zhang Ruoyu Python Scientific Calculation. Tsinghua University Press, 2012.

The Seventh



99

THE 10TH EDITION OF INTERLABORATORY TESTS FOR

ADHESIVES FOR CERAMIC TILES - AN ANNIVERSARY EDITIO N

Cristina Stancu

CEPROCIM S.A., Blvd. Preciziei, no. 6, Sector 6, Bucharest, Romania

[email protected]

Abstract The aim of this paper is to present the proficiency testing scheme called Interlaboratory Test for adhesive for ceramic tiles at the moment of celebrating 10 editions. The scheme begun in 2007 with only one test: initial adhesion strength according EN 1348 and with 9 participants, majority from Romania, working the sample twice. In 2014 the scheme was extended at two tests: initial adhesion strength and tensile adhesion strength after water immersion according to EN 1348 and most of the participants were from Europe. In 2018, at the anniversary edition, the scheme was extended again, at the request of the participants, at three tests: initial adhesion strength, tensile adhesion strength after water immersion and open time according to EN 12004-2. The organizers of the schemes thought-out from the beginning the scheme to be dynamic and to meet the needs of the participants. The paper also presents the technical problems the organizers faced during the running of the programs and how they solved them taking into account the facts that the compositions and the requirements of the adhesives for ceramic tiles were very different. The paper aims to present, by using several case studies, the influence of continuous participation at the proficiency testing programs on the performance improvement of the participants, affecting also the coefficient of variation, as well as the influence brought by the joining of new participants. The case studies represent both accredited laboratories according to EN ISO/IEC 17025 and non-accredited laboratories whose performance were assessed during their participation of the all 10 editions of the Interlaboratory Test for adhesive for ceramic tiles. Key words Proficiency testing scheme, adhesive for ceramic tiles, initial adhesion strength, tensile adhesion strength after water immersion, z-score.

Cristina Stancu: The 10th edition of interlaboratory tests for adhesives for ceramic tiles - an anniversary edition

100

1 INTRODUCTION The Interlaboratory tests on adhesives for ceramic tiles begun in 2007 as the response from the need to meet the requirements of the accreditation standard SR EN ISO/ IEC 17025 [1] to participate in PT/ ILC schemes. At that time in Europe there weren't any institution that organized such a scheme. In this context, CEPROCIM SA through a small team started organizing a new ILC scheme whose object was a single test, initial adhesion strength according to EN 1348/ EN 12004-2 on adhesive for ceramic tiles. At the first round, only 9 laboratories participated, majority from Romania, and they worked the samples twice. In 2014, for meeting the needs of the participants, the ILC scheme has been extended at two tests: initial adhesion strength and tensile adhesion strength after water immersion according to EN 1348/ EN 12004-2 [2]. In 2018, at the anniversary of 10 editions of ILC scheme, the scheme was extended again, at three tests: initial adhesion strength, tensile adhesion strength after water immersion and open-time according EN 12004-2. 2 ORGANIZING THE INTERLABORATORY TEST PROGRAM The team that organized this scheme for adhesive for ceramic tiles has more than 30 years’ experience in organizing ILC schemes. This vast experience, as well as the participation of the team as a participant in other international schemes, allowed the team to organize an ILC scheme for a material for which there weren’t enough data at international level to establish a reproducibility of the test results obtained. At the ten editions (rounds) of Interlaboratory Tests on Adhesives for ceramic tiles, it’s have been participated at least once 67 laboratories from 25 countries from Europe and Asia [3]. Each participant was randomly given a number which is used as laboratory code to enable confidentiality of results. Reference to each laboratory in all the general reports is made by its code number. More than that, from one round to the other the lab code was changed and there wasn’t any connection between lab name and its code allocation system over time. Through the ten editions (rounds) of the Interlaboratory Tests on Adhesives for ceramic tiles different types of adhesives according to EN 12004 produced by different manufacturers from Romania were use (i.e. C2FTE, C2TE, etc.). The test used in all the rounds of the Interlaboratory tests on Adhesives for ceramic tiles for establishing the homogeneity of the adhesive for ceramic tiles was determination of the residue on the 250 µm sieve [3]. For establishing the homogeneity of the adhesives used, after the homogeneity was performed, we took ten samples which were tested in the same day, by the same operator, using the same equipment, every sample have been worked in double. The samples were considered as being homogeneous when all the results have been placed in the range: mean value of the residue on the 250 µm sieve ± 2s (%). The value of “s” represents the standard deviation of repeatability. 3 RESULTS AND DISCUSSIONS For the statistical calculation, algorithm A in Annex C from the standard ISO 13528:2015 was applied. This implies an iterative calculation of the robust values for mean and standard deviation based on all the participants’ results [4].


101

Step 1: The initial values for x* and s* were calculated: , i = (1, 2, …, p)(1)

, i = (1, 2, …, p)(2) Step 2: The values of x* and s* was update:

(3) Step 3: For each xi (i = 1, 2, ..., p), was calculate:

(4)

Step 4: The value obtained after the last iteration represents the assigned value (xpt) chosen to be the consensus value. The values of xpt and σpt were calculated as follows:

(5) (6)

(7)

Step 5: The z-score is calculated as follows: (8)

Step 6: The evaluation of the results was made according to EN ISO/IEC 17043: -satisfactory, when |z| ≤ 2 (9) -questionable, when 2 < |z| <3(10) -unsatisfactory, when |z| ≥ 3 (11) Where: p the number of laboratories that took part at interlaboratory tests; xi the result reported by one participant laboratory i x* robust average of the results reported by all participant laboratories, calculated according to algorithm A method s* robust standard deviation of the results reported by all participant laboratories, calculated according to algorithm A method xpt assigned value (consensus value); σpt standard deviation for proficiency assessment; CV coefficient of variation. The robust estimates xpt and σpt is derived by an iterative calculation, by updating the values of x* and s* several times using the modified data, until the process converges. Convergence was assumed when there wasn’t change from one iteration to the next in the third significant figure of the robust standard deviation and of the equivalent figure in the robust average [4]. For an assessment of how the scheme helped the participating laboratories to improve the way they work and the results obtained, the coefficient of variation was selected and its evolution was followed from one round to another, depending on the number of participants, the history of participation to the scheme and sensitivity of the test. The most common use of the coefficient of variation is to assess the precision of a technique (test, in our case). The coefficient of variation (CV) is a statistical measure of the dispersion of data points in a data series around the mean. It is a useful statistic for comparing the degree of variation from one data series to another, even if the means are drastically different from one another. The higher the coefficient of variation, the greater the level of dispersion around the mean.


102

The Fig. 1 shows the evolution of the coefficient of variation for the initial adhesion strength and variation for the tensile adhesion strength after water immersion.

a) b)

Fig 1. The evolution of the coefficient of variation for: a) the initial adhesion strength; b) the tensile adhesion strength after water

immersion

It can be seen from Fig. 1 that the values of the coefficient of variation don’t exceed 50%, except for the first round, which means that the values of the tests obtained by the participants are homogeneous. Analysing the evolution of coefficient of variation is can be seen an improvement from one round to another, as follows: a) for the initial adhesion strength using the materials send by the organizers the biggest drop of the coefficient of variation was about 80 % compared to the first organized round; b) in the case of using the tensile adhesion strength after water immersion the evolution of the coefficient of variation was between 14 – 31 %. The values of the coefficient of variation, respectively the homogeneity of the obtained results is influenced by several factors, such as: the homogeneity of the materials that are the object of the scheme, the participants' experience, the number of participations in the scheme, etc. Fig. 2 shows the number of laboratories participating in the ILC scheme during the 10 rounds organized so far.


103

Fig 2. The participation of the laboratories in the ILC schemes

In Fig. 2 it can be seen that many laboratories have participated only once or twice to the ILC scheme and only 4 laboratories participated in all 10 organized rounds. The reasons for which the laboratories hadn’t participated constantly in the ILC scheme were various: financial impediments, overcrowding of staff, obtaining a satisfy z-score after one round or on the contrary, non-attaining satisfy results, etc. But their management should take all measures to overcome them and use the ILC scheme in real assessment of the capability of their own lab. Fig. 3 shows the distribution of the laboratories that had participated in the ILC scheme.

Fig 3. The distribution of the laboratories that had participated in the ILC scheme

It can easily be noticed form Fig. 3 that more than 50% of all laboratories participated only in 1 or 2 rounds of the ILC scheme and less than a quarter participated in more than 6 rounds. Synthetic assessment of the results obtained for each test by the participants at the all rounds of ILC Scheme, on the z-score basis is presented in Fig. 4.


104

Fig. 4. Overall assessment of the results of ILC scheme based on z score

Evaluating the results reported for each test by the participating laboratories based on the z-score obtained, it can be seen that 94% of them are classified as satisfactory, 3% questionable and 3% unsatisfactory. The management of a laboratory, whether it is accredited according to EN ISO/IEC 17025 or not, must constantly evaluate its capability, and the participation in an ILC scheme is one of the routes indicated to achieve this. Figures 5 and 6 shows the evolution of z-score of the 4 laboratories presents at the all ILC schemes organized by CEPROCIM SA: 10 rounds for initial adhesion strength and 4 rounds for tensile adhesion strength after water immersion. Following the evolution of the z-score of the 4 laboratories participating in all the round of the ILC scheme, 2 accredited laboratories according to EN ISO/IEC 17025 and 2 laboratories which weren’t accredited, showed in Fig. 5 and 6, it can be seen that the results are for the most part satisfactory, although there were situations where the value of the z-score was raised, but for these 4 laboratories it wasn’t recorded any unsatisfactory results (|z|≥3), although the participants varied. But following the trends of the 4 laboratories, it can be noticed that in the case of the laboratory no 17 (an accredited laboratory) the reported results had quite high scores, there were also recorded results as questionable. In contrast, in the cases of laboratories no 13 (accredited laboratory) and no 27 (unaccredited laboratory), the reported results were quite close to the average.


105

Fig. 5. The evolution of z-score for the 4 laboratories present at the all 10 rounds of the ILC scheme

Fig. 6. The evolution of z-score for the 4 laboratories present at the all rounds of the ILC scheme

Monitoring the evolution of the z-score by the laboratory is very important because it can find the trends that persist over several rounds and it can take the necessary actions if is the case [5]. However, at the laboratory level it has to be investigated right from the first abnormal result obtained and to take the necessary measurements to remedy the situation. In conclusion, following the evolution of z-scores recorded for the 4 laboratories used as case studies, it can be concluded that only 3 of them, 1 accredited laboratory and 2 unaccredited laboratories, have reached a certain maturity to allow them to provide reliable results for their customers.


106

4 CONCLUSIONS Since 2007 CEPROCIM had yearly organized the ILC scheme called Interlaboratory test program on adhesives for ceramic tiles. At all the ten editions (rounds) of Interlaboratory Tests on Adhesives for ceramic tiles, it’s have been participated at least once 67 laboratories from 25 countries from Europe and Asia. The ILC scheme started with one test, the initial adhesion strength, reaching after 10 rounds organized to include three tests: the initial adhesion strength, tensile adhesion strength after water immersion and open-time after 20 minutes. Using the experience accumulated over time in the organization of ILC schemes, the CEPROCIM SA team succeeded in organizing an ILC scheme to help participants to evaluate their own capabilities and to provide reliable results to their clients. Over 94% of the test results provided by the participant laboratories could be framed as “satisfactory” (having IzI ≤ 2) according to EN ISO/CEI 17043, which proves that overall the participating laboratories performed well and tend to improve their working procedure. REFERENCES

[1] EN ISO/IEC 17025:2005 + EN ISO/ IEC 17025:2005/AC:2006 General requirements for the competence of testing and calibration laboratories

[2] Adhesives for ceramic tiles - Part 2: Test methods [3] General Report of “Interlaboratory Test on Adhesives for ceramic tiles” organised

yearly by CEPROCIM: Ist edition (2008-2009) until the 10th edition (2018-2019) [4] ISO 13528:2015 Statistical methods for use in proficiency testing by interlaboratory

comparisons [5] EN ISO/CEI 17043:2010 Conformity assessment - General requirements for

proficiency testing

The Seventh



107

SCOPE OF PROFICIENCY TESTING PROGRAMMES PROVIDED BY

UKZUZ FOR AGRICULTURAL MATRICES

Martin Vana, Eva Cizmarova

Central Institute for Supervising and Testing in Agriculture (UKZUZ), Hroznova 2, CZ-65606 BRNO, Czech Republic

e-mail: [email protected], [email protected]

Abstract The proficiency testing (PT) is one of the best ways for an analytical laboratory to carry out its external quality control and to monitor own performance. PT helps to highlight not only repeatability and reproducibility performance among laboratories, but also systematic errors. The laboratories can demonstrate and verify their good performance. Since 1996 UKZUZ has been organizing proficiency testing programmes for analytical laboratories (MPZ UKZUZ). At the very beginning only 2 programmes were offered to participants (40 laboratories took part in). Nowadays UKZUZ provides 8 accredited and 6 non-accredited programmes for more than 130 laboratories including participants from abroad. Most laboratories have been registered in more programmes. Different matrices (soils, sludges and sediments, plants, feedstuffs) are used in wide scale of PT programmes. All samples are of native origin. Each laboratory can determine any number of parameters being evaluated. A complete anonymity of participants within the MPZ UKZUZ is ensured by the system of numeric coding. The participants use web interface to get registered, to submit their results and to access to the result reports, which are available electronically and can be downloaded or printed. Tailor-made software for all the actions connected with PT (statistical evaluation, sample database, stability evaluation, long-term evaluation of the results for a particular parameter(s), list of participant(s), method indicating codes, etc.) is used. The poster shows different matrices and overview of parameters used in PT programmes provided by UKZUZ. Keywords Proficiency testing, matrix, soil, sludge and sediment, plant, feedstuff.

Martin Vana, Eva Cizmarova: Scope of proficiency testing programmes provided by UKZUZ for agricultural matrices

108

1 INTRODUCTION The proficiency testing (PT) is one of the best ways for an analytical laboratory to carry out its external quality control and to monitor own performance. PT helps to highlight not only repeatability and reproducibility performance among laboratories, but also systematic errors. The laboratories can demonstrate and verify their good performance. Another advantage is to use the benefit of PT participation for validation purposes and for in-house quality control procedures in laboratory management system. Since 1996 UKZUZ has been organizing proficiency testing programmes for analytical laboratories (MPZ UKZUZ). The content and organization of MPZ UKZUZ are based on the requirements for the accredited proficiency tests providers given in EN ISO/IEC 17043:2010 Conformity assessment - General requirements for proficiency testing and on the requirements of participants. Since the year 2008 the Department of Proficiency Testing Programmes has been accredited by the Czech Accreditation Institute as the proficiency testing provider No. 7005. The Certificate of Accreditation was issued on the basis of assessment of fulfilment of the accreditation criteria in accordance with CSN EN ISO/IEC 17043:2010. 2 SCOPE OF MPZ UKZUZ At the very beginning only 2 programmes were offered to participants (40 laboratories took part in). Nowadays UKZUZ provides 8 accredited and 6 non-accredited programmes for more than 130 laboratories including participants from abroad. Most laboratories have been registered in more programmes. The Central Institute for Supervising and Testing in Agriculture organizes regular proficiency testing programmes:

- MPZ UKZUZ - Analysis of Soils - MPZ UKZUZ - Analysis of Sludge and Sediments - MPZ UKZUZ - Analysis of Plants - MPZ UKZUZ - Analysis of Feedstuffs - MPZ UKZUZ - Determination of Mycotoxins in Feedstuffs and Food - MPZ UKZUZ - Analysis of Oil Plant Seeds - MPZ UKZUZ - Determination of Additives in Feedstuffs - Coccidiostats - MPZ UKZUZ - Determination of Additives in Feedstuffs – Vitamins - MPZ UKZUZ - Detection of Animal Proteins in Feedstuffs* - MPZ UKZUZ - Elisa - Detection of the Viral lnfection in Plants* - MPZ UKZUZ - Detection of Phytoplasmas by PCR method* - MPZ UKZUZ - Identification of Phytoplasmas by PCR method* - MPZ UKZUZ - Detection of Globodera spp. nematodes in soil samples* - MPZ UKZUZ - Identification of selected species of Tilletia on cereals*

*labeled programs are not accredited by CAI according to EN ISO/IEC 17043:2010. Preparing of the samples for PTs, including the homogeneity and stability tests, is covered by well trained staff with experience in sample preparation. The distributed


109

quantity of the sample enables all determinations included in the proficiency tests. Each laboratory can determine any number of parameters being evaluated. A complete anonymity of participants within the MPZ UKZUZ is ensured by the system of numeric coding. 3 STATISTICAL EVALUATION OF PARTICIPANTS RESULTS Tailor-made software for all actions connected with PT (statistical evaluation, results transfer, sample database, stability evaluation, long term evaluation of the results for a particular parameter(s), list of participant(s) and method indicating codes, etc.) is used. The software is well-established for PT provider purposes as well as is user-friendly for participants. All procedures of the statistical evaluation of the PT MPZ UKZUZ results are robust so that the influence of outliers would be excluded. The assigned value and the standard deviation of individual comparisons are determined from reported results obtained from the MPZ UKZUZ participants in a particular round of the proficiency testing programme. The principle of the calculation of z-score values are robust values of the average x* and robust standard deviation s* calculated according to algorithm A for the robust analysis. The robust analysis according to algorithm A with iteration - estimation of mean value and standard deviation of data is primarily used for the statistical processing and evaluation of participant´s results. The detailed description can be found in standard ISO 13528 [1]. Z-score is calculated and expressed both in the numerical and the graphical forms. The determination of the reference quantity and the confidence interval for small number of laboratories (4-7) are also provided by Horn procedure [2]. In case of problematic results, it is possible to use the reference values determined by the evaluation of homogeneity or stability tests as the assigned values (ATV - assigned target value and sATV - assigned target value of reference standard deviation). 4 REPORT The statistical evaluation is carried out and results reports are issued after finishing of each MPZ UKZUZ period. The results of participants are presented under the anonymous numeric code. The part of the report forms the overview of z-score for each participating laboratory. These reports also contain brief information on MPZ UKZUZ, description of the samples distributed in the period, the list of the participants, the list of the method indicating codes and other necessary information. The results reports are available in the web interface (*.pdf file). The participants have a possibility to download or to print them. The schedule of one period MPZ UKZUZ is displayed in Fig. 1.


110

Fig. 1. Schedule of one period

5 REFERENCES

[1] ISO 13528:2015 Statistical methods for use in proficiency testing by interlaboratory comparisons.

[2] Horn, P. S.: "Some Easy t Statistics". Journal of the American Statistical Association, December 1983, Volume 78, Number 384, Theory and Methods Section

[3] Meloun, M., Militký, J.: "Statistické zpracování experimentálních dat". Praha; Plus Praha, 1994

The Seventh



111

CONCLUSIONS OF A BLIND PROFICIENCY TESTING SCHEME I N

MANURE ANALYSIS

Thibaut Cugnon 1, Florence Ferber ², Jacques Mahillon 3, Richard Lambert 1 1 UCLouvain - Centre de Michamps, Horritine 1, 6600 Bastogne, Belgium

2 ASBL REQUASUD, Rue de Liroux 9, 5030 Gembloux, Belgium 3 UCLouvain - Laboratory of Food and Environmental Microbiology, Croix du Sud

L7.05.12, 1348 Louvain-la-Neuve, Belgium

[email protected] Abstract

In order to test the complete analytical process in laboratories, a “blind” PT was organized by the REQUASUD PTprovider. The Pt items distributed to the laboratories were indistinguishable from normal customer samples. Two different samples of liquid manures were prepared and distributed to six different laboratories. Each laboratory received three samples of six liters volume: one sample of a liquid cow manure and two samples of the same and homogenous liquid pig manure, in order to test the repeatability of the laboratories. No problem was identified for the cow manure that had a quite high dry matter content (11%). The dry matter of the pig manure was relatively low, as were his content in phosphorus and magnesium. Several unsatisfactory results were observed for the pig manure on following parameters: dry matter, nitrogen, phosphorus and magnesium. The post-analysis of the results, in collaboration with the unsatisfactory laboratories, highlighted that insufficient homogenization of the sample before subsampling was the cause of theses unsatisfactory results. This observation shows the benefit of “blind PTS” in order to assess the whole analytical process in laboratories, including subsampling and homogenization. Key words Blind PT, Manure, Homogeneity

Thibaut Cugnon, Florence Ferber, Jacques Mahillon, Richard Lambert: Conclusions of a blind proficiency testing scheme in manure analysis

112

1 INTRODUCTION Proficiency testing are commonly used for the objective evaluation of a laboratory, and consists in comparing laboratories’ results with each other (laboratory’s results to those of other laboratories). This is achieved by the PT scheme provider that distributes homogeneous and stable samples to participants for analysis, and then reports the results [1]. The main objective of a PT scheme is to help the participants to assess the accuracy of their measurements. However, it does not only give information on the performance of the analytical system, but also on other aspects of the management system, such as reception/handling of the sample, process of the data, results reporting etc. Nevertheless, as test samples are already homogenous, their handling by the lab is often reduced, and the usual entire pretreatment of the sample is not tested (homogenization, drying, subsampling…). Moreover, when PT items are received, the laboratories are aware that it is a PT. Samples might then be treated with more attention compared to classical customer items, as already identified in the late 70’s [2]. 2 Material and methods In order to test the entire analytical process, a special application of PT, a “blind” PT [1], was organized in the REQUASUD network. REQUASUD is a Belgian PT provider. The Pt items distributed to the laboratories were indistinguishable from normal customer samples. Two different samples of liquid manures were prepared and distributed to six different laboratories. Each laboratory received three samples of six liters volume: one sample of a liquid cow manure and two samples of the same and homogenous liquid pig manure, in order to test the repeatability of the laboratories. In the classical way of PT, liquid manures submitted by REQUASUD are one-liter volume samples. Eight analytical parameters were asked for analyses: dry matter (DM), organic matter (OM), total nitrogen (Nt), ammonium (N-NH4

+) and minerals (P, K, Mg, and Ca). The target values (TV) were determined by the average results of all peer participants after discarding the outliers. The PT provider confirmed this target value by analysis of the submitted material. The standard deviation (SD) used for proficiency assessment matched the SD commonly used in the classical organic matter’s REQUASUD PT, as shown in table 1. These SD are established to provide a constant in low and high values in order to avoid excessive evaluation (too restrictive in low content and too permissive in high content) of the analytical performance of laboratories.

Table 1. Standard deviation used for proficiency assessment of the blind PT Measurands DM OM Nt N-

NH4 P K Mg Na Ca

SD (% of TV) 5 6 6 20 10 10 10 10 10

SD Min (absolute value)

0.25% 0.25% 0.12 kg/T

0.25 kg/T

0.25 kg/T

0.25 kg/T

0.15 kg/T

0.25 kg/T

0.25 kg/T

SD Max (absolute value)

5% 4% 3kg/T 3kg/T 4kg/T 4kg/T 2kg/T 10kg/T

15kg/T


113

3 RESULTS AND DISCUSSIONS As shown in table 2, no bad z-score was identified for the cow manure,that had a quite high dry matter content (11%). Only four results were received for OM and five for N-NH4, Mg and Ca. This is not resulting from analytical problems, but from misunderstanding of the analytical request. Indeed, almost all laboratories have different analytical menus, and as samples were distributed by potential customer, some of them didn’t specify the missing measurand in the analytical menu of the laboratory. This is not a problem for the treatment of the PT, but it shows the importance of a correct formulation of the request from the customer.

Table 2. Performance of the PT for the cow manure

Measurand Nb of results

TV SD number of warning or

action z-scores DM 6 11,98 0,54 0 OM 4 9,24 0,78 0 Nt 6 6,01 0,28 0 N-NH4 5 2,18 0,36 0 P 6 2,82 0,11 0 K 6 5,29 0,34 0 Mg 5 1,35 0,12 0 Ca 5 3,22 0,27 0

The repeatability was not evaluated for the cow manure, as a single sample was distributed. Nevertheless, all the single results were compliant as shown in Fig. 1.

Fig. 1. Results of the blind PT for Nt content of the cow manure

The dry matter of the pig manure was relatively low (3%), as were his content in organic matter (1.85%) phosphorus (1.08 ppm) and magnesium (0.53 ppm). As shown in table 3, several unsatisfactory results were observed on the pig manure for all measurands except N-NH4. As two similar pig manure samples were provided to the six laboratories, 12 results could be received for each measurand. Only 8 results

0

1

2

3

4

5

6

7

Lab1

Lab2

Lab3

Lab4

Lab5

Lab6

target value (TV) TV +/- 2SD


114

for the OM were received, as only four laboratories performed this analyses, as already explain for the cow manure. The unsatisfactory results for OM might be due its very low content in the pig manure, which could be below the quantification limits of the laboratories for such products.

Table 3. Performance of the PT for the pig manure

Measurand Nb of

results TV SD of

the results

number of warning signal

number of action signal

DM 12 3,00 0,5 1 2 OM 8 1,85 0,57 2 2 Nt 12 3,94 0,51 5 1 N-NH4 10 2,62 0,25 0 0 P 12 1,08 0,13 2 0 K 12 3,29 0,38 1 0 Mg 10 0,53 0,1 3 0 Ca 10 0,98 0,35 1 0

Figs 2 and 3 represent the average results for two critical measurands. Indeed nitrogen is the most important component of manure in fertilization concerns, and dry mater directly influences the results of mineral analysis. Those two measurands show the most important unsatisfactory results of the PT, especially for lab2 and lab4. They both have warning or action signal for the two measurands. Lab3 also has a warning signal for Nt, but most of all, it is the only one that did not meet requirements of the repeatability test, as represented by the standard errors of its results for dry matter.

Fig. 2. Results of the blind PT for MS content of the pig manure

Fig. 3. Results of the blind PT for Nt content of the pig manure

0

0,5

1

1,5

2

2,5

3

3,5

4

Lab1

Lab2

Lab3

Lab4

Lab5

Lab6

%

Target Value TV +/- 2SD

0

0,5

1

1,5

2

2,5

3

3,5

4

4,5

5

Lab

1

Lab

2

Lab

3

Lab

4

Lab

5

Lab

6

kg

/T

Target Value TV +/-2SD


115

The post-analysis of the results, in collaboration with the unsatisfactory laboratories, pointed out that insufficient homogenization of the sample before subsampling was the cause of theses unsatisfactory results. 4 CONCLUSIONS These observations prove the benefit of “blind PTS” in order to assess the whole analytical process in laboratories, including subsampling and homogenization. Indeed, as the volume of distributed samples were quite huge, and as the consistence of the product could lead to sedimentation, the homogenization before subsampling is critical. Classical PT with small amount of very homogenous samples are necessary to improve the analytical procedures of the laboratories, but they are not really relevant to improve the homogenization and subsampling routine procedure of these laboratories. REFERENCES

[1] Mann I. and Brookman B. “Selection, Use and Interpretation of Proficiency Testing (PT) Schemes.” Eurachem, Second Edition 2011. 52

[2] Louis C. LaMotte, Jr., Gordon O. Guerrant, D. Sue Lewis, and Charles T. Hall. "Comparison of Laboratory Performance with Blind and Mail-Distributed Proficiency Testing Samples." Public Health Reports (1974-) 92, no. 6 (1977): 554-60.

The Seventh



116

BRAZILIAN’S NMI PROFICIENCY TESTING PROGRAMS

AN IMPORTANT TOOL FOR COMPETITIVENESS, INNOVATION A ND INDUSTRY DEVELOPMENT

José Ricardo Bardellini da Silva, Adelcio Rena Lemo s, Carla Thereza Coelho, Paulo Roberto da Fonseca Santos

Instituto Nacional de Metrologia, Qualidade e Tecnologia – Inmetro, Duque de Caxias, Brazil

[email protected], [email protected], [email protected], [email protected]

Abstract The insertion of Brazil into the globalized market requires a strong metrological basis

to promote exports and avoid imports of substandard products. Brazil is a signatory of

Mutual Recognition Agreements (MRA) that establish practices for acceptance of

calibration and tests results produced by accredited bodies from other countries. In

order to sustain these agreements, it is fundamental that Brazilian laboratories have

tools that make it possible to demonstrate their competence to carry out

measurements and tests.

Inmetro's GT-PEP Proficiency Testing Working Group has a characteristic that

distinguishes it from other national PT providers: being part of a National Metrology

Institute (NMI). The proficiency testing/interlaboratory comparison - PT/IC - is an

activity to be developed in partnership with any technical organizational unit of

Inmetro. Due to these characteristics GT-PEP can meet the national requirements of

proficiency tests, where other providers would have great difficulty in acting.

As a NMI, Scientific Metrology and Technology Directorate - Dimci's - laboratories

meet the requirements of structure, resources, process and management system of

ISO/IEC 17025:2017, and are evaluated by specialists from other NMI in the peer

evaluation modality, thus obtaining international recognition by the International

Committee of Weights and Measures (CIPM). As a provider of proficiency testing,

José Ricardo Bardellini da Silva, Adelcio Rena Lemos, Carla Thereza Coelho, Paulo Roberto da Fonseca Santos: Brazilian’s NMI Proficiency Testing Programs - an Important Tool for Competitiveness, Innovation and Industry Development

117

Dimci meets the requirements of ABNT NBR ISO/IEC 17043:2011, both GT-PEP and

laboratories being submitted to an internal auditing process composed of auditors

mostly by experts from the General Coordination of Accreditation of Inmetro, the

Brazilian accreditation body, who act in the accreditation of providers of proficiency

testings.

The realization of PT in the country is fundamental to increase the credibility of the

measurement results and, consequently, contributes to facilitate international trade

and to prevent technical barriers. Until July 2019, there are 67 institutions that

organize proficiency testing programs in the country registered on Eptis database,

totalizing approximately 565 programs in different scopes.

This work presents the several PT programs led by Inmetro actually, with highlights

for the areas of activity and socioeconomic impacts in Brazil and abroad.

Key words Proficiency testing, National Metrology Institute, Eptis, PT Providers 1 HISTORY OF INMETRO’S PT/IC ACTIVITIES By means of GT-PEP, since the beginning of its operation until June 30th, 2019,

Inmetro organized 81 PT rounds of 30 PTs, involving 1610 participating laboratories

and 19 IC, involving 208 participating laboratories. Table 1 shows these PT/IC

activities, correlated to the respective impact areas. These PTs are grouped into 16

proficiency testing schemes (PTS), according to the PT area and are listed in Table

2. These programs were created from the demands for proficiency tests of the

technical areas of Inmetro and became permanent from the beginning of their first

rounds.


118

Table 1. PT/IC activities and respective impact areas. IC PT name

There are 30 Proficiency testing (PT) and 22 Inter laboratory comparison (IC) registred

PT in CachaçaPT for Determination of Agrochemicals in FoodsPT for Determination of Mycotoxins in FoodsPT in Food MicrobiologyPT in JuicesIC for Certification of Reference Material (MR) - ICMICROAL Project - Salmonella in Freeze-dried milkIC to Characterize a Candidate for Certified Reference Material in the Infant FormulaIC to Characterize a Candidate for Certified Reference Material in the Freeze-dried Chicken Brest Matrix Nitrofuran Metabolites ParameterIC in Dynamometric Test Otto Cycle MotorIC in Emission Tests on Diesel Cycle MotorsIC of Torque Measurement, Power and Specific Consumption in Otto Cycle MotorsPT of Gas Water HeatersPT of Gas Stoves and OvensPT in Electroacoustic Equipment CalibrationPT in SpectrophotometryPT in Determination of Glass CapacityPT in Photometry Luminous Flux MeasurementPT in Temperature and HumidityPT in UltrasoundPT of High Voltage in Continuous and Alternating CurrentPT in Thermal ConductivityPT of Nanovoltmeter CalibrationPT of Active and Reactive Power 60 HzPT in High Accuracy Multimeter CalibrationPT in Nanoparticle MeasurementIC with Type K Thermocouple from -40 to 300 °CIC of Glass Liquid Thermometer IC of HygrometryPT for Analysis of Gas Mixture CompositionPT in WaterPT of Vehicles EmissionsPT in Electrolytic ConductivityPT in pHPT of Motorcycles EmissionsIC of Vehicles Emissions - Motorcycles Emissions - PROMOT M4 CycleIC of Vehicles Emissions - Motorcycles Emissions - M3 CycleIC of Turbidity Measurement PT for Steel Composition AnalysisIC for Residual Voltage AnalysisPT in Ethyl Alcohol Anhydrous FuelPT of ARLA 32PT in ViscosityIC of Water Content in Alcohol FuelIC of Biodiesel - Obtaining the precision data of the ABNT NBR 15343 Standard - Determination of the Concentration of Methanol and/or ethanol by Gas ChromatographyIC for Electrode Performance Evaluation in pH Measurement in Fuel Ethanol - Exclusive to the Ethanol Fuel Study Commission (CEEC)IC in Bioethanol Anhydrous Bioethanol - Project PEAALIC of Electrolytic Conductivity in Ethanol Fuel "Exclusive to the Ethanol Fuel Study Commission (CEEC)"IC in Biodiesel - Determination of Property Value and Uncertainty of Characterization of Parameters: Acidity Index, Specific Mass and Water Content in Candidate for Certified Reference Material (MRC)

Health PT for Analysis of Polymorphic Phases by X-Ray Diffraction

Metallurgy

Oil and Derivatives, Natural Gas, Alcohol and Fuels

Impact area

Beverage and Food

Automotive and transport equipment

Household appliance

Machinery and equipment for measuring and control

Environment


119

Table 2. Inmetro’s Proficiency Testing Schemes

Inmetro’s Proficiency Testing Schemes – updated Jun e 30th, 2019

1 PTS in Household Appliance

2 PTS in Vehicles and Motors

3 PTS in Water

4 PTS in Beverage and Food

5 PTS Fuels, Additives and Lubricants

6 PTS in Electrolytic Conductivity

7 PTS in pH Measurement

8 PTS in Optics

9 PTS in Temperature, Humidity and Thermophysical Properties

10 PTS in Volume and Density

11 PTS in Electroacoustic Equipment Calibration

12 PTS for Material Analysis

13 PTS in Mechanical Properties

14 PTS in Ultrasound

15 PTS in Electricity

16 PTS for Health Area

2 EXTRACT OF INMETRO’S PT/IC One of the characteristics of Inmetro’s PT/IC organization is to collect demands of laboratories in the country and to submit them to the technical areas of the institution, indicating which fields of interest can be considered as a new exercise of PT or IC. Inmetro is also involved in partnerships with some entities that are interested in the development and improvement of laboratory activities by means of PT/IC. As examples, following are presented extracts of 4 PT/IC recently organized by Inmetro and the relation existing with national or international commitments. 2.1 PT of Vehicles Emissions Due to regulatory and accreditation bodies requirements, there is an increasing need to perform better measurement of pollutant gases. Besides that, due to constant emissions limits reductions, measurement methods should adequate to new needs. Pollutants analysis is one of the most delicate items of a vehicle or engine emission test. In this sense, the execution of PT in vehicles emissions aims to evaluate the performance of laboratories in determining the amount of compounds present in vehicle emissions, providing subsidies for the identification and solution of analytical problems and contributing to the harmonization of measurement results in the country, besides being a tool for data generation that can support the preparation of new regulations. [1] This regular PT program is organized in partnership with the Brazilian Automotive Engineering Association (AEA), which is a non-profit organization that aims to be a neutral forum for discussion on strategic issues related


120

to national automotive engineering with direct involvement of the automotive industry, government agencies, education and research institutions, international entities and society in general. This partnership began in early 2005 and is actually in its 11th round. 2.2 PT for Analysis of Gas Mixture Composition Within the framework of the PTB-SIM-COPANT-IAAC project "Promoting innovation in the green economy through the inclusion of quality infrastructure in Latin America and the Caribbean", a pilot project was launched to increase the reliability of atmospheric pollutant measurements in urban centers. Under the leadership of the NMI in Brazil and Mexico, Inmetro and CENAM, respectively, cooperation was initiated between NMI from different countries in Latin America and the Caribbean and the competent authorities for air quality monitoring. In order to strengthen the cooperation activity between the organisms, a Proficiency Testing was requested to Inmetro to measure the atmospheric component of carbon monoxide. With the implementation of this PT, coordinated by Inmetro, participants had the opportunity to perform their measurements with their routine methods and compare their results with the designated value of the certified gaseous reference material (CRM) of a national metrology institute (Inmetro). [2] This was de 7th round on Analysis of Gas Mixture Composition PT. 2.3 PT in Water Brazil is a country rich in groundwater, including in regions affected by drought. It is the seventh world mineral water producer with about 10 billion liters in 2007. Data from 1995 showed 319 mining concessions for the exploration of mineral water sources in the entire territory obtained from the National Mineral Production Department of the former Ministry of Mines and Energy, the body responsible for regulation and which, in 2006, increased to 814. According to the National Health Surveillance Agency of the Ministry of Health (ANVISA/MS) under Resolution No. 54/2000, mineral water is that obtained directly from natural sources or artificially collected, of underground origin, characterized by the defined content and constituted by mineral salts (ionic composition) and the presence of trace elements and other constituents. Following the Consultative Committee for Amount of Substance (CCQM) recommendations, Inmetro’s Inorganic Analysis Laboratory (Labin) assigned reference values to analytes of interest by performing measurements with traceable results to the SI through an uninterrupted chain of comparisons with certified reference materials (MRC). [3-4] Inmetro has organized 8 rounds of PT in water, since May, 2004. 2.4 IC to Characterize a Candidate for Certified Re ference Material in the Freeze-dried Chicken Brest Matrix Nitrofuran Metabo lites Parameter Issues related to food safety, as well as the search for improvement through the enhancement of international trade relations, have stimulated nations in the search for quality of measurements in the food area. Certified Reference Materials (CRMs) play a key role in this improvement because they give these measures traceability to


121

the International System (SI). The important role of the CRM in face of the problems of trade restrictions in some countries, such as the measures adopted by the European Union with Directive 96/23/EC, to control residues in the field of veterinary drugs in products of animal origin. In April 2002, a new testing methodology detected the presence of nitrofuran residues in batches of poultry meat exported by Brazil, banned in the EU. Some estimates suggest a loss of around US$ 40 million per year with additional controls. According to data from the United States Department of Agriculture (USDA), in 2017 Brazil became the second largest producer of chicken meat in the world and remained the largest exporter. Given this scenery, it is necessary to maintain and improve measures to guarantee the quality of this product, since its embargo by other countries would compromise the Brazilian trade balance. In order to carry out the control, it is necessary to use analytical methods that are applicable and that meet the national and international provisions. The use of CRM is recommended in guidelines on performance of analytical methods for unambiguous detection of chemical residues in foods of animal origin, such as the European Union’s Commission Decision 2002/657/EC and the Analytical Quality Assurance Manual of the Ministry of Agriculture, Livestock and Food Supply. This IC aimed to support the characterization value of a candidate CRM of nitrofurans metabolites in freeze-dried chicken muscle. The mass fraction values of the metabolites and their respective uncertainties were determined, and the measurements of residual moisture content of the lyophilized material were optional. [5] 3 INSERTION OF BRAZIL IN EPTIS DATABASE Brazil figures on Eptis Database since 2005 covering several impact areas, as shown in table 3. It can be noticed the commitment to accreditation, which is growing in a moderate rate, but is remarkable.

Table 3. Brazilian Inserts on Eptis Database (updated July 10th, 2019)

Accredited TotalN° of Brazilian PTS on EPTIS database 67 565

Impact Area Accredited N° of PTSFood and drink 8 57Automotive and transport equipment 41 212Household appliance 1 8Machinery and equipment for measuring and control 1 4Environment 0 3Metallurgy 0 1Oil and Derivatives, Natural Gas, Alcohol and Fuels 2 46Health 26 219Others 15TOTAL 79 565

NOTE: the different numbers of accredited PTS above is due to the overlapping areas


122

4 CONCLUSIONS Since 2005 a lot of effort have been done in Brazil concerning PT, particularly by Inmetro. Today, Brazil occupies the second position in PT EPTIS database, where 49 countries are listed. However, there are too much to be done. Regarding the number of Accredited PTS in EPTIS database, Inmetro figures in the 21st position with 12 % of accredited PTS, while 5 countries occupy the 1st position with 100 % of accredited PTS. Although this big challenge, Inmetro is working very hard to attend its clients. Periodically, satisfaction survey is conducted in order to hear the external clients’ opinion and suggestions upon how to improve PT services. All information on IC and PT rounds are disposed on Inmetro’s site and the link http://www.inmetro.gov.br/metcientifica/ensaio-proficiencia/ensaioProficiencia.asp REFERENCES [1] Final Report of the Proficiency Testing in Vehicles Emissions – 10th round –

Diesel Vehicles – April, 2019, available at http://www.inmetro.gov.br/metcientifica/pdf/Final_Report_of_the_Proficiency_ Testing_in_Vehicles_Emissions_-_10%C2%AA_round.pdf.

[2] Preliminary Report of Proficiency Testing of Gas Mixture Composition Analysis - 7th Round - Carbon Monoxide in Nitrogen – July, 2019.

[3] Final Report of Proficiency Test in Water - 3rd round - Measurement of Metals in Mineral Water - April, 2010, available at http://www.inmetro.gov.br/metcientifica/ensaio-proficiencia/profiAgua.asp.

[4] Final Report of Proficiency Test in Water - 4th round - Measurement of Anions in Mineral Water - February, 2011, available at http://www.inmetro.gov.br/metcientifica/ensaio-proficiencia/profiAgua.asp.

[5] Preliminary Report of the Interlaboratorial Comparison to Characterize a Certified Reference Material Candidate for Nitrofuran Metabolites in Chicken Meat, July, 2019.

The Seventh



123

IMPLEMENTATION MODE FOR GOVERNMENT ORDERED

PROFICIENCY TESTING PROGRAM

Huichao Liang, Kexiao Yu , Shasha Wu, Jianbo Zhou, Ying Xu ，Xiaoling Zhu,

Ruolin Li ，Ying Xu

China Building Material Test & Certification Group Co., Ltd. No.1 Guanzhuang Dongli Chaoyang District, Beijing, China

[email protected] Abstract In China, government always entrust proficiency testing provider to hold a proficiency

testing program，in order to know the technological capability and management level of the laboratories within their jurisdiction. Using the normal implementation mode

can only know the testing accuracy of the laboratories，but cannot get the other information. This paper presents a new implementation mode, which include proficiency and site inspection. This mode can not only comprehensively master the

capability and level of the laboratories，but also effectively avoid data collusion. Key words

Proficiency Testing (PT)，implementation mode 1 INTRODUCTION Proficiency testing is one of the important means to evaluate the ability of conformity assessment body, and is becoming more of a concern to authorized institutions and

the government. China Building Material Test &Certificate Co., Ltd.（CTC for short）

is the proficiency test provider （ PTP for short ） which is certificated by China National Accreditation Service for Conformity Assessment. CTC also is one of the earlist and main PTPs. Except the normal proficiency test program, CTC also organize custom PT that is entrusted by the government or industry association. The practical range of this kind of PT always in the province. The participate laboratories

Huichao Liang, Kexiao Yu , Shasha Wu, Jianbo Zhou, Ying Xu，Xiaoling Zhu, Ruolin Li，Ying Xu : Implementation mode for goverment ordered proficiency testing program

124

know each other, so how to avoid data collusion is very important. In addition, as an organizer, government or industry association often needs to understand not only the accuracy of laboratory test results, but also whether the laboratory personnel, equipment, and test environment meet the standard requirements and whether the report is received from the sample to meet the management requirements. To this end, we have developed an implementation model for government ordered proficiency testing program. 2 IMPLEMENTATION MODE for GOVERMENT ORDERED PROFICI ENCY TESTING PROGRAM For government organized and CTC implemented proficiency testing program, we propose three modes:. 2.1 Mode one: Request participated laboratories to provide video evidence for sample testing process For government organized large-scale proficiency testing program(the number of participating laboratories is more than 100), through requesting participated laboratories to provide video evidence for sample testing process to verify the authenticity and validity of the proficiency test results. The video usually includes but is not limited to the following:

①Sample preparation process;

② Maintenance or state adjustment process, containing sample status and environmental conditions;

③Testing process, containing Sample number, environmental conditions, equipment calibration mark, digital display test result and so on. This mode would simultaneously control the test time to reduce the chance of data collusion. It implements supervision of the testing process without increasing personnel costs. The video includes the entire process, making it easy to identify weak points in unsatisfactory and problematic laboratories. In turn, technical analysis provides a basis for improvement for these laboratories. CTC implemented the proficiency testing program of the thermal conductivity test of thermal insulation materials and tensile properties test of elastomer modified asphalt waterproofing coiled material commissioned by the Construction Engineering Quality and Safety Association of Hubei Province separately in 2017 and 2018. The number of participating laboratories was 123 and 278. This two program all adopt this mode. The test time was less than three days. 2.2 Mode two: Carry out the mode of “on-site inspec tion (mainly for test process)” The on-site inspection described in this model is mainly for the test process, which is applicable to the medium scale (the number of participating laboratories is 50-100) and the small scale(the number of participating laboratories is less than 50). Mainly check the following and form a witness material storage:

①Laboratory personnel list;


125

②Tester profile;

③Tester ID card and social security record;

④Tester signature handwriting(Check whether it is consistent with the previous test

record of the person’s test);

⑤File of testing equipment;

⑥ Calibration certificate and calibration result confirmation record of the test equipment;

⑦Whether the tester is the equipment operation authorized person. This model confirms the personnel, equipment, and environmental conditions of the proficiency testing laboratory. The test process was witnessed on site, and the test data was obtained at the test site, avoiding collusion data between laboratories. The on-site witness can also identify problems in the laboratory testing process and include the technical analysis section of the proficiency testing report to provide a basis for technical improvement in the laboratory. Compared with mode one, this mode increases the personnel cost and transportation expenses of on-site inspection. If sample preparation and state adjustment are involved, this mode is usually no longer on-site witness this two part, so the on-site information is often not comprehensive enough. In 2018, CTC undertook the implementation of the proficiency testing program for the tensile properties of elastomer-modified asphalt waterproofing coiled material and physical properties of cement orgnized by the Beijing Municipal Bureau of Quality and Technical Supervision and the Zhejiang Provincial Bureau of Quality and Technical Supervision. Due to the scattered laboratory, it took a lot of time on the road. Although the final project was successfully completed, the whole process was long and costly. 2.3 Mode three: Carry out the model of "on-site tes t (the whole process from the commissioned test)" This mode is the most detailed check content in the three modes, including the scope of qualification, contract of attorney, the test process, the test report and the original record, etc. Usually, the on-site inspector carries the sample to the laboratory, so it is not suitable for testing witch containing the process for sample preparation and state adjustment. A capability verification project that adjusts or detects long periods. Taking into account the entire operating costs, this model is suitable for small scale proficiency testing program with no more than 50 laboratories. The specific inspection contents are as follows:

①Environment and facilities, such as the environment, layout, and placement of reagents;

②On-site commissioned test, such as the time from sample collection to sample preparation, whether to paste the unique identifier, whether there is a flow order, etc.;

③Fill in the order, such as the sample name and specification model filling, testing items, testing standards, etc.;


126

④The consistency between the test report and the order, such as the conformity of the qualification scope, the timeliness of the report, and the consistency with the order;

⑤Whether the content of the test report is normative and complete;

⑥Testing personnel 's on-the-job management, such as whether there is a certificate of employment, whether the on-the-job personnel and the list of documents will be consistent, etc.

⑦Verification management of testing equipment, if there is no equipment calibration certificate, operation documents, etc.

⑧ Management of waste liquid (if any), such as management documents and processing records for accidental waste disposal. In 2018, CTC implemented the competency comparison of indoor air test organized by Tianjin Binhai New Area Market and Quality Supervision Administration. Since the tester's entire inspection process was witnessed and scored on the spot, and the technical ability of the tester was fully evaluated, the project was combined with the vocational skill competition. 3 SUMMARIZE Integrating the implementation modes of the above three types of government-customized proficiency testing projects, the first mode, which requires the participating units to provide video data, not only supervises the testing process, but also avoids the laboratory to change the data, and the implementation efficiency and cost are also the highest. It is the main mode of our implementation. As laboratory supervision becomes more stringent, proficiency testing technology is becoming more widely used in government regulation of laboratories. CTC will also continue to research and develop new proficiency testing models to meet the needs of market and government management.

The Seventh



127

TROUBLESHOOTING OF CHLORITE OVERLAPPING OVER

BROMATE PEAK IN DRINKING WATER

Mohammed Z. Alrabeh , Mohammed A. Almutairi , Mostafa A. Alsamti .

Saudi Food & Drug Authority, Riyadh, Saudi Arabia


Abstract Proficiency test is part of quality requirements objectives for laboratories to meet their specific needs.sample of disinfection by-products is likely to contain both anions chlorite and bromate. This paper studies determination of Bromate in water by Ion chromatography. Moreover describes the chlorite overlapping over bromate peak PT sample in details and what the possible solutions to overcome the issue. After analysis, the chromatogram of the PT sample shows that part of bromate peak was overlapped by high concentrated chlorite peak This overlapping was attributed to poor peaks resolution of the separation column as well as the high response of chlorite .Furthermore, different techniques such as conditioning of separation column, filtration by silver cartridge and reducing flow rate was applied, however, neither of these solutions was resolve the overlapping. Although, there are other solutions like increasing eluent concentration strength or using another detector type such as UV detector, a statistical tool of calculation using peak height has proven to achieve accurate result where a z-score of 0.3 was obtained. Key words Ion chromatography, Bromate, Chlorite, drinking water 1 INTRODUCTION Chlorite and Bromate are disinfection by-products (DBPs) that are generated during chemical disinfection process of drinking water. Bromate is mainly formed after the ozonation of the bottled water [1]. Bromate is commonly separated and quantified using Dionex ICS-3000 Ion exchange Chromatography (IC). Occasionally, routine bottled water samples received in our laboratories contain fluoride, sulphate, bromide, nitrate and nitrite whereas chloride peak is identified in

Mohammed Z. Alrabeh, Mohammed A. Almutairi, Mostafa A. Alsamti: Troubleshooting of chlorite overlapping over bromate peak in drinking water

128

almost all samples. Proficiency testing schemes of potable water (e.g. FAPAS program) contain the most common disinfectant by-products of chlorine dioxide; namely chlorite and chlorate. The challenge is that chlorite peak is frequently elutes closely to the bromate peak [2,3]. In this paper, we highlight our strategy in resolving such overlapping. 2 EXPERIMENTAL Bottled water samples are injected directly to IC that is equipped with conductivity detector. In few cases, sample pre-treatment is carried out to remove any contamination using SPE cartridges or filtrate [2]. Liquid chromatographic separation of anions was carried out by separation using (Ion Pac AS19 column). 2.1 EQUIPMENT AND MATERIALS In this method, samples were analysed by Dionex Model ICS-3000 Dual system (Fig. 1).

Fig. 1. Dionex Model ICS-3000 2.2 Reagents and standards

Deionized water Type 1 (18 µΩ.cm), potassium hydroxide (10 mM) provided by (Dionex) and bromate stock standard (1000 mg/L BrO3) were obtained from Ultra Scientific. Intermediate standard solution (1 mg/L) was prepared by adding 0.1 mL of 1000 mg/L bromate stock solution up to 100 mL De-ionized water in volumetric flask. For the spiked samples, aliquots of the intermediate stock solution were accurately weighed and then diluted to obtain the final bromate concentrations. These concentrations were set as reference values to calculate the analytical recovery. Water samples were provided by FAPAS. 2.3 Procedures Prior the analysis, the instrument was stabilized under the specified conditions for at least 1 hour. Ideally, the detector signal should be less than 1 µS. The sample was injected directly to IC where a conductivity detector and chemical suppressor were used. The measurement of bromate was made in the range of 5 to 100 µg/L. Bromate was separated by liquid chromatographic separation column. Di-


129

ionised water is used as a mobile phase and anion exchange resin as the stationary phase [2]. Operating conditions used during the current method are described in Table 1.

Table1. Instrumental conditions

The details for calibration standards are shown in Table 2. Anions peaks were identified by comparing sample’s retention time with the corresponding reference standards.

Table 2. Calibration Standards

STD

Volume of BrO3

-

intermediate std*, mL

Final Volume, mL

Final Concentration, µg/L

STD 0 0 50 0 STD 1 0.25 50 5 STD 2 0.5 50 10 STD 3 1 50 20 STD 4 2.5 50 50 STD 5 5 50 100

Each batch should meet quality control requirements described in Table 3.

Table 3. Quality Control Requirements

Parameters Acceptance criteria Reagents blank Confirmed no interferences with

bromate retention time and free of contamination

Calibration check standards Rec 70-110 % RSD 20%

Duplicate Sample deviation between the duplicate sample not more than 20%

Calibration standards R = 0.999

Mobile phase Mili-Qwater Flow rate 1.0 mL/min

Eluent generator 10mM potassium hydroxide Column temperature 30.0 ˚C. Suppressor current 25mA Maximum pressure 3000 psi Minimum pressure 200 psi

Column Dionex IonPac AS 19 Analytical 4 x 250 mm

Detector Conductivity Detector


130

3 RESULTS AND DISCUSSIONS Calibration curve at 5 levels (5 – 100 µg/L) demonstrated linear response where r2 was 0.999 using bracket calibration (Fig.2). Analysis of quality control sample (9.80 ± 0.39 µg/L) has achieved excellent recovery (99.2%) as shown in Table 4 and Fig. 3. Additionally, blank sample chromatogram was free from any interferences or carryover analyte (Fig.4). Potable water sample was analyzed 7 times to achieve accuracy parameter (calculated by peak area and peak height) (Table 5).

Fig. 2. Linearity plot for bromate standards at µg/L levels

Table 4. Bromate internal standard 9.80 ± 0.39 µg/L

y = 0.003x - 0.004

R² = 0.9992

-0,05

0

0,05

0,1

0,15

0,2

0,25

0,3

0,35

0 20 40 60 80 100 120

Pe

ak

hig

h u

S

Concentration ug/l

Calibration standards of Bro3

Bromate (Bro3) QC = 9.80 ± 0.39 µg/L

Sample (Potable water)

Peak high (µS) Results ( µg/L)

1 9.3

2 9.3

3 10

4 10.32

Average( µg/L) 9.73

STD 0.51

RSD (%) 5.27

Recovery (%) 99.2


131

Table 5. Results of sample analysis

Fig. 3. Bromate QC chart (mean of 9.80 µg/L).

Bromate (Bro3) FAPAS Assigned value × 9.65 ±0.19 µg/L

Sample (Potable water)

Peak area (µS*min) Results ( µg/L)

Peak high (µS) Results ( µg/L)

1 6.94 9

2 6.89 9

3 7.16 9.3

4 6.92 9

5 7.8 10.3

6 8.28 11.48

7 8.28 11.48

Average ( µg/L) 7.46 9.9*

STD 0.637 1.17

RSD (%) 8.5 11.8


132

The retention time of target analyte was separated with high resolution at 7.40 min in bromate standards (Fig. 5).Figure 6 shows that chlorite peak was partially overlapping with bromate peak. This overlapping created difficulties in the calculation of exact amount for bromate in the PT sample. In order to resolve this issue, techniques such as conditioning of separation column (Fig.7), filtration by silver cartridge and reducing flow rate [4,5] have all been used but none of them was able to correct the problem. Moreover, standard addition technique was used, however this attempt was unsuccessful resulting in underestimation of the correct value. After re-evaluation of the tried methods, it was found that the calculation using peak height was the most successful technique [4,5]. Statically, there were significant differences (p ≤ 0.01) in Bro3 concentrations between the peaks (area and height) of the examined sample.

In the current investigation, difference was observed in the calculation of bromate content in PT sample using peak area and peak height. This may be attributed to concentration dependent co-elution co-incidenced when the chlorite response was higher than typical concentrations of the neighboring bromate peak, which overlapped into the retention window of the target analyte.

Fig.4. Blank sample chromatogram

Fig.5. Standard bromate peak at 7.40 min.


133

Fig.6. Chlorite peak (7.1 min) overlapping over bromate peak (7.4 min).

Fig.7. Conditioning of separation column shows better separation and resolution of

bromate.


134

4 CONCLUSIONS Annually, SFDA food laboratories participating in proficiency testing of DBPs in drinking water to evaluate the competency of the staff, validity of reference material used and precision of the laboratory equipments. Analysis of PT sample by Ion exchange chromatography equipped with conductivity detector shows overlapping of chlorite peak over part of bromate peak. This overlapping resulted in poor resolution of separation peak. Different techniques were utilized to resolve the issue but none helped. Nevertheless, calculation by peak height was the most successful technique even though that overlapping exist. REFERENCES [1] Determination of Chlorite, Bromate, Bromide, and Chlorate in Drinking Water by

Ion Chromatography with an On-Line-Generated Postcolumn Reagent for Sub-µg/L Bromate Analysis” DIONEX Application Note 149.

[2] “Water quality – Determination of dissolved bromate” ISO 15061:2001. [3] Determination of Bromate in Drinking Water Using Two-Dimensional Ion

Chromatography With Suppressed Conductivity Detection, EPA, 2009. [4] Dabeka, R.W.; Conacher, H.B.S.; Lawrence, J.F.; Newsome,W.H.; McKenzie,

A.;Wagner, H.P.; Chadha, R.K.H.; Pepper, K. Survey of bottled drinking waters sold in Canada for chlorate, bromide, bromate, lead, cadmium and other trace elements. Food Addit. Contam. 2002, 19, 721–732.

[5]

Liu Y.J., Mou S.F. Determination of bromate and chlorinated haloacetic acids in bottled drinking water with chromatographic methods. Chemosphere. 55, 1253, 2004.

The Seventh



135

IMPLEMENTATION AND VALIDATION OF FORMALDEHYDE

ANALYSIS AT SFDA LABORATORIES

Majda Alenezi 1, Adnan S. Almussallam 1, Abdullah T. Bawazir 1, Ahmed I. Al -Ghusn 1, Nawaf F. Alsowidan 1 and Ahmad Hanafi 2.

1 National Drug and Cosmetic Control Laboratories (NDCCL), Saudi Food and Drug Authority, Riyadh, Saudi Arabia 2 Ain Shams University, Egypt

[email protected], [email protected], [email protected], [email protected], [email protected], [email protected]

Abstract Formaldehyde is a pungent toxic chemical and potential carcinogen. It is frequently added into cosmetics products as a preservative. Due to its health hazards, many regulatory authorities set a limit for its concentration. In Saudi Arabia, and according to Annex VI of the Cosmetic Directive 76/768/EC, it is obligatory for cosmetic manufacturers to abide the limit for formaldehyde concentration (not more than 0.2%). The current paper describes how the Saudi Food & Drug Authority (SFDA) developed its formaldehyde testing in hair straightener products based on previous published methods and validated the newly implemented method. Additionally, SFDA tested the newly adapted method against real samples and the results demonstrated that some samples exceeded the regulatory limits. The method is based on Hydrophilic Interaction Liquid Chromatography (HILIC) followed by post column derivatization using acetyl-acetone. The commercial formaldehyde reagent (38-40%) was standardized using sodium thiosulfate and then calibration curves was generated. Quality Control (QC) charts were produced using Certified Reference Material (CRM) to test the accuracy and stability for this method. Statistical parameters linearity, selectivity, precision, uncertainty, limit of detection, and limit of quantitation were calculated and validated. The developed method was used to assess formaldehyde level in 25 hair straightener products. Results revealed that 36% (9 products) violated the limit for formaldehyde. Key words Formaldehyde, cosmetics, FDA, toxicity

Majda H. Alenezi , Fahad S. Al Dawsari , Ahmed I. Al-Ghusn : Implementation and validation of formaldehyde analysis at SFDA laboratories

136

1 INTRODUCTION Formaldehyde (HCHO) is a colourless gas that is commonly used as preservative [1]. It is used in many preservation applications such as organ, clothes, woods, fabrics, papers and cosmetics. It can be used as a gas dissolved in water or it can be released from certain chemicals (known as formaldehyde releasers or donors) [2]. However, its potential carcinogenicity and allergy reactions prompted regulatory authorities worldwide to control its use [1,2]. Consequently, limits for this chemical have been set in various goods. Multiple analytical methods have been suggested to measure formaldehyde concentration in different matrices. Pre-column derivatization, post-column derivatization, semi-quantitative evaluation and UV detection are some of these reported methods [2,3]. In Europe, maximum allowable limit of formaldehyde in cosmetic products is 0.2% [2]. 2 EXPERIMENTAL The aim of this study was to establish a method for formaldehyde quantification to apply it in testing potential cosmetics at Saudi Food & Drug Authority (SFDA) laboratories. A published method which is based on Hydrophilic Interaction Liquid Chromatography (HILIC) followed by post column derivatization using acetyl-acetone was chosen [2]. 2.1 Materials Anhydrous ammonium acetate, pentanedione (acetylaceton) kept at 4 oC, disodium hydrogen phosphate dihydrate (Na2HPO4. 2H2O), Phosphoric acid 85%, Glacial acetic acid, Certificate reference material (CRM) 651 from Netherlands Food and Cosmetic Product Safety Authority (contains 0.302% formaldehyde), Certificate reference material (CRM) Formaldehyde in Water, Product ID =QC1380-20ML containing 10200 mg/L (Sigma Aldrich).

2.2 Instrumentation and chromatographic conditions The development and validation of the method was performed on HPLC (Shimadzu) with degasser unit (DGU-20ASR), Pump A &B LC-20AD, diode array detector (SPD-M20A), auto sampler (SIL-20AC), column oven (CTO-20AC), chemical reaction unit (CRB-6A) and LC solution software was applied for data collecting and processing. The stationary phase used was Waters NovaPak® C18 (3.9mmx300mm,4µm) column. The mobile phase was 0.006 M phosphate buffer (pH 2.1), diluent was mobile phase Phosphate buffer (pH 2.1). The detection wavelength was 420 nm with a flow rate of 0.6 and 0.3 mL/min for pump A and B respectively.


137

2.3 Procedures 2.3.1 Preparation of derivatizing reagent: Derivatizing reagent was prepared by dissolving 62.5 g of anhydrous ammonium acetate in 800 mL of deionized water. Then, 7 mL of glacial acetic acid and 5 ml acetylacetone were added. Finally, mixture was completed to 1 L by deionized water. 2.3.2 Preparation of calibration solutions Serial dilutions in water were made from CRM formaldehyde in water in the concentration ranges of 1 to 100 ppm. 2.3.3 Preparation sample preparation Into 50 mL volumetric flask, approximately 1 g of the sample was weighted followed by mixing (with sonication) with 40 mL of diluent for 10 min. Later, mixture was transferred into a 50 mL volumetric flask and filled to the mark with diluent. 2.3.4 Preparation of quality control samples Quality control (QC) samples were prepared from the (CRM) containing 0.302% formaldehyde by suitable dilutions using the mobile phase. 3 RESULTS AND DISCUSSIONS Several chromatographic methods have been discussed to identify free and total formaldehyde concentration at different matrices [2,3]. Due to the volatile properties of formaldehyde, derivatizations methods were suggested [2]. There are available published methods for both pre- and post-column derivatization strategies. In the current study, post-column derivatization was used as this method is gaining popularity among researchers in the recent years. In order to achieve optimum recovery, the influence of the extraction solvent was studied. First the QC sample was extracted with dichloromethane: hydrochloric acid, 0,002 M (50:50, v/v). The recovery was unexpectedly as high as 185.61%. When the mobile phase was used for extraction, the recovery was 104.75%. High recovery was explained by presence of interference between formaldehyde and dichloromethane at the same retention time.

The linearity of the implemented method was statistically assessed by evaluating the generated results (Table 1), making it possible to test the assumption of non-validity of the linear dynamic range (using the Fischer-Snedecor test) [4]. These errors are given as a standard deviation, resulting from the square root of the ratio between a sum of squares and a degree of freedom [4,5].The linear regression analysis data are summarized in Table 2.


138

The selectivity of a method is expressed by its ability to measure only the compound under testing [5]. From linearity study data, the parameters of the regression line were used to estimate selectivity. These parameters are Standard deviation on the slope (Sb) and standard deviation on the intercept point (Sa) [5]. The purpose was to conclude the absence of any interference and to achieve an acceptable specificity. This is true if the overlap line y = a + bx is equivalent to the line y = x.

Table 11 Calibration data in term of formaldehyde concentrations (µg/mL)

Table 2. linear regression parameters of formaldehyde in standard solution

a number of degrees of freedom is np-2. b number of degrees of freedom is n-2.

c The ratio 2exp

2

S

SF def

obs= obeys the Fischer-Snedecor law with the degrees of freedom n-2, np-n, compared

to the limit value: F1-α (n- 2,np-n), extracted from the Snedecor law table. The value for α used in practice is generally 5%.


139

In order to satisfy these requirements, two tests were carried out namely; Test of the assumption that slope b of the overlap line is equal to 1, and the Test of the assumption that intercept point a is equal to 0 [4,5]. These assumptions are tested using the Student test, generally associated with a risk of error of 1%. Let Tcritical, bilateral [dof; 1%] was used as a Student bilateral variable associated with a risk of error of 1% for a number of degrees of freedom (dof) [5]. The selectivity data analysis is summarized in Table 3.

Table 3. The selectivity data analysis.

That means both conditions (If Tobs is lower than Tcritical, then the slope of the regression line is equivalent to 1 or If T’obs is lower than Tcritical, then the intercept point of the regression line is equivalent to 0) are true, and so the method is deemed to be specific [5]. Detection and quantification limit (LOD and LOQ) were determined by the linearity study. The data to be recovered from the linearity study are: 1) Slope of the regression line and 2) Residual standard deviation (Sres) to calculate the standard deviation at the intercept point (Sa). The Detection and quantification limit data analysis are summarized in Table 4.

Table 4: The Detection and quantification limit data


140

Precision parameters were estimated using four concentration levels (n) analyzed using five replicates (p) [5]. For each replica, the measurement was made with three repetitions (K) as shown in (Table 5) by Fischer-Snedecor test; the Fobs is higher than F1- α, which mean the repeatability value of the alternative method is significantly higher than that of the reference value [5].

Table 5. The precision analysis results

The accuracy of the procedure was demonstrated by the recovery studies, which were carried out by spiking aliquots of drug substance stock solution with formaldehyde at four levels corresponding to 10.2, 40.8, 81.6 and 102 µg/ml as shown in the Table 6.


141

Table 6. Accuracy (Recovery) data

For the uncertainty estimation, there are several approaches for its calculation based on single laboratory validation approach [4-6]. Intermediate precision and bias were investigated as two independent uncertainty components. These uncertainty components were combined as relative standard uncertainties, with other components that also contributed to uncertainty measurement [6-7]. The results demonstrated that for formaldehyde at concentration level of (40.8 µg/mL); the uncertainty level was 2.73 µg/mL. After successful application of the implemented validation parameters, the method was used to quantify formaldehyde in real hair straightener samples. The samples were all imported and were sent to the SFDA laboratory as post-market surveillance study. Out of 25 samples tested, 9 samples exceeded the regulatory limit of formaldehyde (0.2%). Some of these out of range samples contained more than 10-fold formaldehyde compared to the maximum limit. The SFDA laboratory had successfully participated into a PT scheme organized by Deutsche Referenzbüro für Lebensmittel-Ringversuche und Referenzmaterialien GmbH (DRRR, Germany) and the laboratory achieved a satisfactory result regards formaldehyde analysis. 4 CONCLUSIONS A method for formaldehyde quantification was successfully transferred and validated at SFDA laboratory. Avoiding interference with formaldehyde during solvent extraction is very crucial to counteract unexpected recovery calculations. The choice of proper method (pre- vs post-) column is critical and factors such as matrix and extraction solvent should be taken into scientist considerations.


142

REFERENCES

[1] Cosmetic legislation Method Analysis of cosmetic product –Volume2 –page 50, 1999 Edition.

[2] Mona M. Alshehri, Maha A. AlMeshal. Precolumn derivatization HPLC method for rapid and sensitive determination of free and total formaldehyde in hair straightening products. Arab. J. Chem. 2018. In press, https://doi.org/10.1016/j.arabjc.2018.03.008

[3] Soman, Y. Qiu and Q Chan. HPLC-UV Method Development and Validation for the Determination of Low Level Formaldehyde in a Drug Substance. J. of Chromatogr. Sci., 46, 2008.

[4] USP 40-NF 35 V2017: <1225> Validation of Compendial Procedures and <1210> Statistical Tools for Procedure Validation.

[5] OIV-MA-AS1-12 : R2005 : Practical guide for the validation, quality control, and uncertainty assessment of an alternative oenological analysis method.

[6] JCGM 100:2008: Evaluation of measurement data - Guide to the expression of uncertainty in measurement.

[7] ISO 21748:2010: Guidance for the use of repeatability, reproducibility and trueness estimates in measurement uncertainty estimation.

The Seventh



143

VALIDATION OF A METHODOLOGY FOR ASSESSING THE LEVEL OF RESPIRABLE CRYSTALLINE SILICA (RCS) IN BULK POWD ERS

RELEVANT TO THE CEMENT INDUSTRY

Dr. Peter Kruspan 1, Dr. Markus Arndt 2, Dr. Klaus Lipus 3 1 Holcim (Schweiz) AG, TEC Lab, Zementweg 1, 5303 Würenlingen, Switzerland,

2 HeidelbergCement AG, Oberklamweg 2-4, 69181 Leimen, Germany, 3 Verein Deutscher Zementwerke e.V., Tannenstrasse 2, 40476 Düsseldorf,

Germany


Abstract In Europe, standard prEN 17289 was elaborated by CEN TC 137 for the characterization of bulk materials. One element of this standard is the determination of the so-called Size-Weighted Fine Fraction (SWFF) and more specifically its crystalline silica content (SWFFCS) via a proposed calculation method. SWFFCS can be used as a measure for assessing the health-relevant level of respirable crystalline silica (RCS) in bulk powder products. As input to calculate SWFFCS the three key parameters specific density, bulk powder fineness as well as crystalline silica content are experimentally determined. With a specific focus on the analytical variation of those three parameters, the European cement industry carried out a detailed and comprehensive validation process over a total of three years. In a first round, 22 plant laboratories – covering the great majority of cement producers in Europe – assessed two less complex limestone powders (fully crystalline) and two more complex coal fly ash samples (partially X-ray amorphous). This assessment revealed that variation in SWFFCS is mainly linked to variation in crystalline silica results, but not so much to specific density or powder fineness. Attempts were therefore made in a second comparative testing program among 13 industrial labs to improve crystalline silica analytics on the two very same fly ash samples of the first round. Finally, three central laboratories of the cement industry validated the outcome on two cement samples (i.e. not only on cement constituents) and also compared the statistical findings with the SWFFCS reference sample used for the elaboration of European standard prEN 17289. Keywords cement, bulk powder, fineness, X-Ray Diffraction (XRD), respirable crystalline silica (RCS), Size-Weighted Fine Fraction (SWFF), validation of testing methods

Dr. Peter Kruspan, Dr. Markus Arndt, Dr. Klaus Lipus: Validation of a methodology for assessing the level of RCS in bulk powders relevant to the cement industry

144

1 INTRODUCTION The decision of the United Nations (UN) made in Rio 1992 on the ‘Globally Harmonized System of Classification and Labelling of Chemical (GHS)’ was adapted in Europe as Regulation No. 1272/2008 of the European Commission (EC), the so-called directive on the ‘Classification, Labelling and Packaging of Substances and Mixtures (CLP)’. Based on this CLP directive, substances and mixtures that contain Respirable Crystalline Silica (RCS) above 1%, must be labelled accordingly [10]. In order to have a measure for the RCS level in a (not yet airborne) bulk material [10], the Industrial Minerals Association (IMA) of Europe developed the so-called SWFF (‘Size-Weighted Fine Fraction’) methodology [13]. This methodology is now available as European pre-standard prEN 17289 [7] and has been successfully applied to various industrial minerals such as quartz, clay, kaolin, talc, feldspar, mica, cristobalite, vermiculite, diatomaceous earth, barite and andalusite. No validation so far exists for cement and its constituents. Therefore, the members of the European Cement Association CEMBUREAU decided to launch a thorough investigation program. In this paper, the key elements of this validation process from cement perspective are presented. 2 FIRST COMPARATIVE TESTING PROGRAM In order to promptly set a common ground for being able to identify both, current lab abilities of the cement industry as well as methodological issues where to focus further efforts, not cement, but four material samples having less complex characteristics were selected (see Table 1). All four material types are industrially used as cement constituents / commercial clinker replacement materials and are either fully crystalline (both limestone samples) or partially X-ray amorphous (both fly ash samples).

Table 1. Overview of the four samples investigated in the first testing program

Sample Number Material Type Source Pre-Treatment

P.1271 Crushed Siliceous Limestone 0/2 mm

Switzerland, Commercial

Ground in lab mill to powder fineness

P.1272 Limestone Powder Germany,

Commercial None, used as-is

P.1273 Coal Fly Ash Germany, Commercial None, used as-is

P.1274 Coal Fly Ash Germany,

Commercial None, used as-is

Closely following the principles described in [1, 8, 9, 11, 12], powder aliquots of 500 g were individually generated from each material type, mainly via (i) homogenization of a 25 kg master sample for at least two hours in a Turbula T50A mixer (W.A.


145

Bachofen AG, Muttenz, Switzerland) and (ii) sample splitting over three runs with a spin riffler PT100 (Retsch GmbH, Haan, Germany). A total of 22 participating labs of the European cement industry received four sample aliquots each, representing the companies and associations listed in Table 2.

Table 2. Companies and associations participating in the first testing program

Buzzi Unicem Cementa AB Cemex CEMKUT

CIMPOR CRIC Dyckerhoff HeidelbergCement

Holcim (3 labs) Italcementi Jura / CRH Lafarge (4 labs)

SECIL (2 labs) Titan VDZ VOEZ At this stage, no strict specifications were imposed on the labs about how exactly to carry out the measurements, i.e. own Standard Operating Procedures (SOP) were to be applied. Apart from the individual testing results on (i) crystalline silica (quartz) content [%], (ii) specific density [kg/m3], (iii) particle size distribution of the powder / d50 value [µm], as well as (iv) the final SWFFCS value [%] calculated according to the official template available on the web [15], the labs were additionally asked to provide any details about specific methodologies used locally. In Table 3 all individual results of the first comparative testing program are listed. The great majority of the participating labs provided single measurements. However, the average of two measurements were provided by labs 3, 11 and 15; the average of three measurements by lab 1; and the average of five measurements by labs 10, 17 and 18. For the determination of the crystalline silica / quartz content, 21 labs applied XRD Rietveld methodology, 1 lab Raman spectroscopy. The specific density was analysed in 14 labs with Helium pycnometry and in 6 labs via the Le Chatelier flask method according to EN 196-6; one lab did not provide any information, one lab did not measure. Powder fineness (whole granulometric curve / d50 value) was measured in one lab via Flow Particle Image Analyser FPIA, in 21 labs via laser diffractometry based on either Fraunhofer or Mie model; from these 21 labs, 8 labs used wet dispersion (usually in isopropanol), 6 labs dry dispersion at various pressures applied and from 7 labs no detailed information was available. From the 4 datasets of each parameter (see Table 3), a statistical evaluation was performed, starting with identifying outliers according to the principles described in [3-6, 14], by using a confidence level of 95% (α=0.05); those outliers are indicated red in Table 3. By excluding those outliers from the dataset, one of three pre-requisites for calculating meaningful statistical indicators was fulfilled [2], the remaining ones being (ii) normal distribution of the measurement series and (iii) random variation of the dataset (i.e. showing no trends). Based on the statistical indicators obtained, the related z-score of one particular single value X could be calculated via the relation ‘(X-Median)/Standard Deviation’, as a means for assessing data precision. In Table 3, results having z-scores above |2.0|, corresponding to 2σ standard deviations, are highlighted orange and are considered insufficient in terms of precision.


146

Table 3. Compilation of all individual results of the first testing program: crystalline silica (quartz) content at top left, specific density of the sample at top right, fineness (d50-value) at bottom left and calculated SWFFCS-value at bottom right. Outliers in red cells, values with z-score >|2.0| in orange cells, no measurement in grey empty cells.

Lab

P.1271 P.1272 P.1273 P.1274

Limestone Limestone Fly Ash Fly Ash

01 35.0 3.0 5.4 7.6

02 36.8 2.3 10.5 14.7

03 34.8 2.8 12.5 16.5

04 39.6 3.2 12.3 16.5

05 43.2 2.9 14.3 13.0

06 39.0 2.8 12.6 18.1

07 32.9 1.9 10.8 16.2

08 24.3 1.7 10.4 16.0

09 37.1 2.5 11.1 14.9

10 36.1 3.2 9.4 14.0

11 36.5 3.0 10.4 14.5

12 21.0 1.9 4.8 8.3

13 35.8 3.5 20.3 24.0

14 3.6 1.0 7.8 7.4

15 2.5 18.6

16 39.0 2.8 12.7 16.9

17 37.1 2.4 13.6 18.5

18 40.6 3.2 9.4 12.2

19 35.8 3.4

20 34.0 3.0 8.5 11.0

21 40.4 3.8 15.5 19.1

22 38.0 4.0 8.0 13.0

Crystalline Silica / Quartz[%]

P.1271 P.1272 P.1273 P.1274Limestone Limestone Fly Ash Fly Ash

2721 2762 2360 2260

2705 2755 2350 2275

2698 2752 2339 2267

2663 2736 2306 2215

2570 2710 2330 2230

2690 2740 2330 2250

2700 2750 2340 2270

2710 2760 2350 2230

2730 2760 2360 2290

2690 2718 2298 2224

2711 2758 2359 2293

2690 2740 2340 2260

2710 2730 2310 2170

2700 2740 2330 2260

2750 2250

(2700) (2700) (2700) (2700)

2710 2760 2340 2280

2706 2758 2356 2278

(3000) (3000)

2708 2757 2344 2263

2690 2740 2310 2220

2693 2746 2355 2291

[kg/m3]

Density (Sample)

Lab


01 7.7 13.9 16.1 16.8

02 8.1 15.5 20.5 20.5

03 7.8 13.8 18.2 17.9

04 7.7 14.3 18.4 18.1

05 9.0 21.8 23.1 21.9

06 8.0 14.1 16.0 15.7

07 6.6 13.8 16.6 16.6

08 14.3 26.7 25.8 24.8

09 9.5 16.8 15.6 18.7

10 9.5 13.6 19.7 20.4

11 7.8 16.5 18.7 18.9

12 11.5 26.8 20.6 22.7

13 10.6 22.0 19.8 18.4

14 6.0 5.2 4.2 4.5

15 15.0 19.5

16 4.5 10.9 15.3 15.5

17 9.8 19.1 23.7 23.8

18 7.3 14.4 22.0 22.6

19 8.9 14.6

20 11.1 16.6 16.2 17.0

21 6.6 12.0 16.2 15.6

22 7.9 16.8 17.8 17.4

Fineness (d50)

[um]


7.4 0.3 0.4 0.6

7.7 0.2 0.9 1.0

7.4 0.4 1.3 1.8

9.4 0.4 1.4 1.7

8.8 0.3 1.3 1.5

7.8 0.3 1.6 2.4

8.8 0.2 1.1 1.9

2.2 0.1 0.4 0.8

6.2 0.2 1.1 1.4

5.2 0.2 0.7 1.0

8.8 0.3 1.0 1.5

3.2 0.1 0.4 0.5

5.1 0.2 1.0 1.0

0.6 0.3 1.9 1.6

0.3 1.2

11.7 0.3 1.1 1.7

6.4 0.1 0.7 1.0

9.0 0.4 0.7 0.8

7.2 0.3

0.5 0.0 0.1 0.2

9.4 0.5 1.5 2.0

8.5 0.5 1.0 1.7

SWFFCS ('RCS Bulk')

[%]


147

From the results listed in Table 3, the following key findings can be derived:

• Crystalline silica (quartz): On pure crystalline material with low quartz content (sample P.1272), very robust results could be achieved. On the other hand, much more fluctuating results were obtained on the same material type having elevated quartz levels (sample P.1271): a typical range between MIN and MAX of up to 10% points (%pt.) must be taken into account. The same width between MIN and MAX applies for both non-crystalline materials (samples P.1273 and P.1274). This means that a discrimination between for instance 5% and 15% quartz content is for the time being not feasible when such a particular sample is analysed by different labs (i.e. in terms of reproducibility).

• Specific density: very stable results were achieved in all cases. However, striking is the fact that density results identified as outliers or with insufficient z-scores could directly be attributed to those labs applying the Le Chatelier flask method. This points towards using only Helium pycnometry as harmonized method for SWFFCS purposes.

• Fineness / d50 value: data are rather scattering, most probably due to the multitude of (insufficiently disclosed) influence parameters such as type of dispersion (wet or dry), air pressure used (in case of dry dispersion) and model applied (Fraunhofer or Mie). Still, when excluding outliers and unprecise results (having z-scores >|2.0|), an indeed satisfying range between MIN and MAX of ~5 µm d50 can be taken into account.

• SWFFCS: an increasing influence on this parameter is along following order, specific density (least improving influence as precision is already high) – fineness – quartz content (strongest improving influence as precision is low). Consequently, improvements in quartz analytics would also result in better accuracy of SWFFCS values. However, at this stage, final SWFFCS values of, for instance, 0.5% and 1.5% (= the range of both fly ash samples, serving as analogue to cement) cannot be discriminated in between labs (i.e. in terms of reproducibility).

3 SECOND COMPARATIVE TESTING PROGRAM Based on the outcome of the first comparative testing program (see chapter 2), it was decided to run a second testing program on the very same fly ash samples of the first testing program (i.e. P.1273 and P.1274), however, by strictly following a pre-defined standard operating procedure (SOP). This SOP included the following items:

• For each parameter, five repeatability measurements shall be realized, i.e. taking out five times sample aliquots and having them measured in sequence by the same lab technician.

• All analyses shall be performed at constant lab room conditions, i.e. at (20 ± 2) °C as well as ≥50% relative humidity.

• Specific density shall be determined with Helium pycnometer. To equilibrate the sample prior to testing for at least 24h at (20 ± 2) °C. The final result shall be the average of at least three and a maximum of ten measurements (until acceptable deviation is achieved).


148

• Fineness/granulometry shall be determined with laser diffraction (dry), as this type of method is mostly used in cement plant labs. Air pressure of 2 bar to be applied. Start of measurement if optical concentration ≥1%, stop of measurement if optical concentration ≤1% for more than three seconds. Fraunhofer model to be applied (Mie model for the time being not sufficiently established or widespread in cement plant labs).

• Crystalline silica (quartz) content to be determined with X-ray diffraction (XRD). A sufficient mass of the sample shall be thoroughly mixed with 20 wt.-percent of an appropriate (i.e. fully crystalline, sufficiently fine) internal standard, such as corundum (α-Al2O3). Nine crystalline phases (quartz, mullite, hematite, magnetite, lime, periclase, rutile, anhydrite, and calcite) plus the ‘X-ray amorphous content’ shall be quantified by Rietveld refining the data. Sample rotation activated, Kβ radiation eliminated. Preferred step size of 0.02° and counting time of 60 s/step.

In this second testing program, again 13 labs from those listed in Table 2 participated. All results were statistically evaluated in the same manner as described in chapter 2. In Table 4, the main statistical indicators for each of the four parameters (quartz content, specific density, fineness, SWFFCS) obtained on the two fly ash samples are shown for both testing programs.

Table 4. Statistical indicators of the first and second comparative testing programs, run on both fly ash samples. Outliers, identified according to Dean-Dixon [3-6, 14], are excluded from the datasets. n=number of labs; MED=median; AVG=average;

STDEV=standard deviation; COV [%]=coefficient of variation; MAX=maximum value; MIN=minimum value; all calculations according to Excel (Microsoft©).

P.1273 P.1274 P.1273 P.1274

Fly Ash Fly Ash Fly Ash Fly Ash

n 19 21 13 13

MED 10.5 14.9 9.9 14.2

AVG 10.5 14.8 10.3 13.9

STDEV 2.8 4.1 3.0 3.1

COV (in %) 26.9 27.6 28.6 22.5

MAX 15.5 24.0 15.8 19.1

MIN 4.8 7.4 4.2 6.2Cry

sta

llin

e S

ilic

a /

Qu

art

z [%

]

First Testing Second Testing

P.1273 P.1274 P.1273 P.1274


19 20 11 11

2340 2260 2352 2272

2337 2254 2347 2269

19 31 18 18

0.8 1.4 0.8 0.8

2360 2293 2370 2292

2298 2170 2310 2236De

nsi

ty (

Sa

mp

le)

[kg

/m3]


P.1273 P.1274 P.1273 P.1274


n 19 20 9 9

MED 18.4 18.5 17.3 17.2

AVG 19.0 19.1 16.8 17.1

STDEV 3.0 2.8 1.1 1.4

COV (in %) 16.0 14.7 6.6 8.4

MAX 25.8 24.8 17.8 19.5

MIN 15.3 15.5 14.2 15.0

Fin

en

ess

(d

50)

[um

]


P.1273 P.1274 P.1273 P.1274


20 21 10 11

1.0 1.4 0.9 1.0

1.0 1.3 0.8 1.2

0.5 0.6 0.3 0.5

46.5 42.5 31.3 38.3

1.9 2.4 1.3 2.1

0.1 0.2 0.4 0.6

SW

FF

CS

('R

CS

Bu

lk')

[%

]


It can be seen that overall the precision of all four parameters could be improved, what is reflected in generally decreasing standard deviations from first to second testing (see cells highlighted yellow). However, for the calculated parameter SWFFCS the improvement of precision (or reduction of standard deviation, or reduction of


149

coefficient of variation) was only marginal: still a spread of some ±0.5%pt. must be taken into account. This is due to the fact that – although (i) precision on fineness was substantially improved and (ii) precision on specific density was again kept at high level – the spread on quartz results could not be significantly narrowed down. Hence, this confirms the conclusion of the first comparative testing program (see chapter 2): with current lab abilities the error on quartz results does have the strongest influence on the final error of calculated SWFFCS values. 4 COMPLEMENTARY TESTS In order to further establish a tight link between material types used in the cement industry and European pre-standard prEN 17289 [7], three expert labs from those listed in Table 2 carried out two additional testing series:

1. Two synthetic blended cement samples were generated by mixing and homogenizing a commercial CEM I (Portland cement) with 15% and 30% of siliceous limestone powder P.1271 (see Table 1). Apart from applying exactly the same procedure as described in chapter 3 for determining SWFFCS, the sedimentation method in isopropanol as defined in [7] was performed as well. While ease of measurement, robustness and accuracy of calculated SWFFCS of both samples was as previously encountered (1.4 ± 0.1) % and (3.2 ± 0.3) %, respectively), the sedimentation method occurred being much more challenging for those two synthetic cement samples, mainly due to the only very small residue available after all steps of the procedure have been carried out. Furthermore, a very strong dependence between sample weighed-in (e.g. 2 g or 5 g) and final result was observed. However, still, the requirement of [7] for validating new material types – ‘values for SWFFCS obtained by calculation must be equal or higher than the results by sedimentation’ – was entirely met.

2. The 100% quartz powder reference sample of [7] used in the fundamental Round Robin of [13] was also tested via both methodologies, i.e. the calculated SWFFCS and sedimentation method, carrying out three measurements per lab. It was again observed that SWFFCS values calculated with Fraunhofer model ((33.3 ± 1.6) %) lead to higher values and better precision than sedimentation ((31.9 ± 5.7) %), showing lower values and worse precision.

5 CONCLUSIONS Unlike in other mineral resource sectors, the cement industry is confronted with the fact that cement constituents used or cement products sold generally have SWFFCS values well below 10%, usually even below 1%. At this low absolute SWFFCS-levels, narrow ranges for standard deviation will yet translate into wide ranges for coefficient of variation, which in turn may question lab precision or abilities as such. Various in-depth comparative studies carried out among up to 22 labs have shown that for the time being an error of up to ±0.5%pt. (in terms of reproducibility) must be accepted for final SWFFCS-values. Improvement of this error can only be realized if the precision of XRD Rietveld analytics is improved, namely in every lab calculating SWFFCS values. In the light of the low-CO2-driven trend towards cement products


150

with an increasing number of material types of complex composition, the key challenge is to be able to determine still robust SWFFCS-values on cement products, even without knowing in advance the full list of crystalline and X-ray amorphous phases present. REFERENCES

[1] Allen, Terence. Powder sampling and particle size determination. Amsterdam: Elsevier, 2003.

[2] Danzer, Klaus. Analytical chemistry, theoretical and metrological fundamentals. Berlin: Springer, 2007.

[3] Dean, R.B., Dixon, W.J. “Simplified statistics for small numbers of observations.” Analytical Chemistry 23/4 (1951), 636-638.

[4] Dixon, W.J. “Analysis of extreme values.” The Annals of Mathematical Statistics 21/4 (1950), 488-506.

[5] Dixon, W.J. “Ratios Involving Extreme Values.” The Annals of Mathematical Statistics 22/1 (1951), 68-78.

[6] Dixon, W.J. “Processing data for outliers.” Biometrics 9/1 (1953), 74-89. [7] prEN 17289. Characterization of bulk materials – Determination of a size-

weighted fine fraction and crystalline silica content. Part 1: General information and choice of test methods. Part 2: Calculation method. Part 3: Sedimentation method. Bruxelles: European Committee for Standardization, CEN/TC 137. Planned final voting date: 29 Nov 2019.

[8] Ferraris, Chiara F., Hackley, Vincent A., Ivelisse Avilés, Ana, Buchanan Jr., Charles E. Analysis of the ASTM Round-Robin Test on Particle Size Distribution of Portland Cement: Phase I (NISTIR 6883) and Phase II (NISTIR 6931). Washington: NIST, 2002.

[9] Ferraris, Chiara F., Hackley, Vincent A., Ivelisse Avilés, Ana. “Measurement of particle size distribution in Portland cement powder: Analysis of ASTM Round Robin studies.” Cement, Concrete, and Aggregates 26/2 (2004).

[10] Friede, Bernd, Wyart-Remy, Michelle. “Regulatory assessment of Respirable Crystalline Silica in Europe: REACh, GHS, CLP, and Carcinogen Directive.” Silica and Associated Respirable Mineral Particles. Harper, Martin and Lee, Taekhee (Eds.), ASTM STP 1565 (2013): 1-11

[11] Hackley, Vincent A., Lum, Lin-Sien, Gintautas, Vadas, Ferraris, Chiara F. Particle size analysis by Laser diffraction spectrometry: Application to cementitious powders (NISTIR 7097). Gaithersburg: NIST, 2004.

[12] Jillavenkatesa, Ajit, Dapkunas, Stanley J., Lum, Lin-Sien. Particle size characterization. NIST Recommended Practice Guide, Special Publication 960-1, Washington: National Institute of Standards and Technology, 2001.

[13] Pensis, Ingeborg, Luetzenkirchen, Frank, Friede, Bernd. “SWeRF – A method for estimating the relevant fine particle fraction in bulk materials for classification and labelling purposes.” Ann. Occup. Hygiene (2013): 1-11

[14] Rorabacher, David B. “Statistical treatment for rejection of deviant values: critical values of Dixon's "Q" parameter and related subrange ratios at the 95% confidence level.” Analytical Chemistry 63/2 (2009), 139-146.

[15] SafeSilica, web platform for information about crystalline silica, operated by IMA Europe: https://safesilica.eu/safety-and-measurement/

The Seventh



151

REVISITING A ROUND ROBIN TEST TOWARDS THE EVALUATIO N

OF IMAGING SPECTROSCOPY SYSTEMS FOR CULTURAL HERITAGE APPLICATIONS

Irina -Mihaela Ciortan

Norwegian University for Science and Technology, Gjovik, Norway

[email protected]

Abstract The application of imaging spectroscopy (IS) in cultural heritage has escalated with ascending interest for the last decade. The main advantage of IS with respect to traditional spectrometry is that it provides non-contact measurements of reflectance distributed over an entire surface as opposed to limited sampling for only a limited collection of points. For this reason, many IS configurations have been proposed in the scientific community for analytical studies of artworks. However, up to date, there is a gap of standardized procedures and reference materials in the spectral imaging protocols, especially as regards the acquisition, documentation and analysis of cultural heritage objects. In order to select a suitable device for various kinds of applications in the cultural heritage sector, it is necessary to assess the available spectral imaging systems. This paper revisits the preliminary results of a Round Robin Test (RRT) conducted within the COST project Colour and Space in Cultural Heritage (active between 2012-2016) and designed to interchange 5 items (out of which 4 reference materials and 1 Russian icon) towards cross-measurement between 19 spectral imaging laboratories. The previously published evaluation tests of the RRT included quantitative methods (pixel-wise and areal colorimetric and spectral differences) and qualitative methods (visual check of the alignment of the spectral distributions of the same material, as well as the capability of resolving details spatially). This present study aims to revisit the comparison of the IS systems in the RRT, by introducing novel metrics triggered mainly from the field of image quality evaluation. the approach is a data-driven one, meaning to evaluate the quality of the setups based on the quality of output images, by taking into account a set of device-specific parameters (sensor characteristics, spectral sensitivity, spectral range, noise, dynamic range, optics, data formats) as well as the image capture conditions (illuminant intensity and spectral power distribution, imaging geometry, illuminant non-uniformities etc.). The goal of this work is to provide proficient testing that enables a ranking of state-of-the-art IS setups according to their appropriateness for cultural heritage applications.

Irina-Mihaela Ciortan: Revisiting a Round Robin Test towards the Evaluation of Imaging Spectroscopy Systems for Cultural Heritage

152

Key words image spectroscopy, round robin test, cultural heritage 1 INTRODUCTION Recent advances have made spectral imaging systems accessible and portable solutions for the study of cultural heritage objects. A number of works in the literature have studied the advantages of spectral imaging systems over color imaging systems and punctual spectroscopic instruments on various applications and test objects. In [1], several tasks concerning the readability and legibility of hand-written historic documents were resolved with segmentation based on spectral-domain distances. Pigment-mapping was performed by using divergence computations in a hyperspectral cube of Edward Munch’s The Scream painting, with a precision validated by traditional conservation methods (X-Ray Fluorescence) [2]. By means of hyperspectral imaging that acquire hundreds of channels in the electromagnetic spectrum, reflectance can be recovered from a frequent sampling of discrete responses, such that it allows for accurate simulations of the acquired object under different lighting conditions and can improve the quality of printed reproductions [3]. Imaging spectroscopy systems come in many set-ups and configurations. For this reason, the Color and Space in Cultural Heritage (COSCH) project [4] funded by the EU COST action, carried out a Round Robin Test (RRT) to test the quality of various multispectral and hyperspectral systems [5]. A RRT is a widely accepted tool to study the influence of various parameters, which may vary between individual laboratories and instruments. The RRT experiment during the COSCH project implied the circulation of five test targets to 19 laboratories, where they were measured with various imaging spectroscopy set-ups. 2 ROUND ROBIN TEST DESIGN A round-robin test needs to be designed by considering the requirements of the measuring systems, the application field and the relevance of the underlying studied materials for the given application. Thus, the test targets need to be selected accordingly. Moreover, given that they have to be transported for the inter-laboratory acquisition, it is recommended to opt for an operable size and weight. In the COSCH project, special care has been given to the selection of the test targets that will be described below. Test target 1: Spectralon reference. The Spectralon® multi-component wavelength calibration standard is one of the most commonly used reference standards to keep track of the reflectance/absorbance of the measured object. Spectralon is impregnated with a combination of three rare earth oxides (holmium oxide, erbium oxide, dysprosium oxide), is a stable and chemically inert reflectance standard, commonly used for establishing the accuracy of the wavelength scale of reflectance spectrophotometers. Spectralon has a size of 90mm of diameter, and opaque, homogenous, smooth surface, and exhibits sharp absorptions at specific wavelengths span over UV-VIS-NIR region of the electromagnetic spectrum. This object was


153

selected as one of the reference objects in this RRT with the purpose to assess the wavelength accuracy and the spectral resolution of each spectral system.

Fig. 1. RRT test objects (a) Spectralon reference (b) ColorChecker Classic (c) White

balance (d) Painted panel (e) Icon Test target 2: X-rite ColorChecker Classic. The X-Rite® ColorChecker Classic target is a twenty-four patch color chart with standard colors which are the representation of true colors of natural matter (such as skin, foliage and sky), additive and subtractive primary colors, various steps of grey, and black and white. Each square patch has a size of 40mm and total size of 279.4mm x 215.9mm. The ColorChecker is a very commonly used item for conventional photography and also as a standard color reference to evaluate color reproduction processes, to guarantee that the information obtained is valuable and repre-sents the true colors of the object that has to be studied and documented (Berns 2001). Test target 3: X-rite White Reference. The ColorChecker® white balance target is a spectrally neutral reference standard to deliver accurate white reproduction. White balance reference is used to measure the light distribution in the objects location and it guarantees precise, uniform, neutral white under any lighting condition. Test target 4: Painted panel. In addition to the above listed reference targets, a painted panel is also included in the test objects. The panel was reconstructed by Elena Prandi and Marina Ginanni (Restoration Laboratories of the Soprintendenza SPSAE e per il Polo Museale della città di Firenze, Italy), so as to reproduce the medieval Tuscan technique as described by Cennino (1954). This panel was made on a wooden support and includes several layers of different pigments and underdrawings. The painted panel was included in the RRT items to evaluate the


154

relevant spectral region for pigment analysis and to unveil underdrawings especially in the NIR region. Test target 5: Icon. Test targets 1-4 were mainly proposed as reference objects for spectral imaging in cultural heritage. However, it is also desirable to study the current challenges in this field, which arises when dealing with objects of different properties like metallic or glossy surfaces. This will help to form best practices for imaging such items [6]. A Russian icon, “Virgin of Kazan” dated 1899 owned by Lindsay MacDonald (University College London, UK) was chosen as a test target. The icon is in the form of a colored image printed on copper plate, with wooden support as the substrate. Due to the metallic plating, the illuminant generates specular reflections, which often reduces the quality of imaging. The aim of including this icon in this RRT was to understand the imaging instruments response to such challenging objects and to investigate ways to address such issues. As part of the RRT procedure, the test objects were sent to each of the participating laboratories. Each participant was provided with a set of general instructions of the test and was asked to perform the image acquisitions in the same manner as they usually operate with their instrument. Details of the spectral instrument and the image acquisition conditions have been collected by means of a report form. Every laboratory had to compute the spectral reflectance data, which was the main data derivative necessary to be collected as a common starting point for comparison. Ground truth data for the test objects was measured using point spectrometers and used as reference for comparison for the data sets. 3 REVISITING THE ROUND ROBIN TEST RESULTS This paper aims to revisit previous results of the RRT as a starting point for discussing possible future analysis. For clarity purposes, this study will focus on the comparison of three spectral setups that were involved in the RRT. Table 1 presents an overview of these systems based on a set of quality attributes. Given that there is no established standard to evaluate the spectral imaging systems, these quality attributes are proposed with the view of building such standard. The quantitative analysis was split into several experiments. The first experiment aimed to evaluate the spectroscopic intra-variation provoked by the spatial inhomogeneity of the samples. This was performed by measuring the colorimetric differences, with the CIELAB distance, between areas with different locations inside a patch and the center of the patch, based on the images of the ColorChecker target provided by System 2. Selected patches shown in Fig. 2 were considered in this experiment. As the results in Table 2 indicate, the spatial factor plays an important factor in the processing and comparison of the data and so, in the scope of a fair quantitative evaluation of the different systems, a consistency regarding the studied spatial locations should be preserved.


155

Table 1. Quality specifications of the 3 spectral systems.

System 1 System 2 System 3

System Attributes

Spatial sampling 275 ppi 150 ppi 300 ppi

Spectrum coverage

Visible +

near infrared

(400-1000nm)

Visible +

near infrared

(400-1000nm)

Visible +

near infrared

(400-960nm)

Number of bands 61 160 450

Band separation ~10 nm ~3.7nm ~1.25nm

Number of bits 18 12 16

Set-up Information

Position of the target Horizontal Horizontal Vertical

Complexity Medium Medium Medium

Portability of the system Low Low Low

Cost High High High

Post-processing Compatibility with ENVI No Yes Yes

The second experiment assessed the spectral and colorimetric inter-variation of the ColorChecker and the Painted Panel test targets between the three systems. The metric chosen for computing the colorimetric difference was the CIELAB distance, while the difference in reflectance was computed with the Root Mean Square Error (RMSE) indicator. The values of the test objects extracted from the spectral images acquired with the three systems in Table 1 were compared with the ground-truth reflectance values and color coordinates measured by the point-based spectroscopic device based on the above-mentioned error metric. Smaller values of these metrics imply more accurate results, and this means that the reflectance from the captured images and the estimated reflectance spectrum are similar to the ground-truth. The results are shown in Table 3 and Table 4.

Fig. 2.: The patches selected for spectroscopic analysis from the colorChecker and Painted Panel targets.


156

Table 2: CIELAB color differences using as reference the LAB coordinates of the central 1mm square for selected Color Checker patches measured with

System 2.

Patch number

Spatial position 3 5 7 10 13 14 15 16 17 18

LAB central 35 mm square 148px 0.2 0.4 0.6 1.0 0.3 0.4 0.3 0.2 0.4 0.2 LAB central 8 mm square 34px 0.2 0.2 0.2 0.2 0.2 0.1 0.2 0.3 0.2 0.1

LAB central 1px 0.1 0.2 0.1 0.3 0.2 0.2 0.2 0.2 0.1 0.2 LAB top right 1 mm square 6px 0.7 1.1 2.5 2.3 2.0 1.4 1.1 0.8 1.8 0.9

LAB bottom right 1 mm square 6px 0.9 1.8 3.1 3.0 3.1 3.0 1.1 0.7 1.7 0.9 LAB bottom left 1 mm square 6px 0.5 1.5 2.1 1.3 0.9 1.5 1.7 1.0 1.8 1.0

LAB top left 1 mm square 6px 1.0 1.4 1.8 1.5 0.9 1.7 0.7 1.3 2.5 1.5 On average, the quantitative results suggest that the lowest spectral and color differences belong to System 3, for both studied test targets. Overall, the colorimetric and spectral differences are proportional. In the case of the panel stripe, the highest CIELAB value coincides with the highest RMSE value and these are associated the first stripe, which according to Fig. 2, is the gypsum ground, that has a coarse texture and, hence, a surface characterized by increased non-uniformity. The ivory black stripe and the burnt umber have the next biggest color and spectral differences. This might be due to the camera’s decreased sensitivity for dark materials, which results in a noisy, non-homogeneous signal. Table 3: RMSE and CIELAB color difference computed for the each of the 24 patches of the Color Checker, between the values measured with the three imaging systems

and the ground truth data.

Color Checker

Patch 1 2 3 4 5 6 7 8 9 10 11 12 13

S 1

RMSE 0.11 0.12 0.02 0.03 0.09 0.10 0.12 0.09 0.09 0.05 0.11 0.11 0.04

CIELAB diff 8.24 5.44 1.43 9.54 4.11 6.28 17.76 7.77 11.55 3.90 13.61 17.34 9.93

S 2

RMSE 0.05 0.03 0.05 0.02 0.01 0.05 0.07 0.03 0.04 0.04 0.02 0.09 0.03

CIELAB diff 1.33 1.54 1.76 2.22 1.43 6.38 6.89 4.52 2.18 4.14 3.59 7.42 6.41

S 3

RMSE 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01

CIELAB diff 2.31 1.84 1.03 1.46 1.54 1.94 2.19 2.25 1.20 1.42 2.68 2.00 0.69

Color Checker Patch

14 15 16 17 18 19 20 21 22 23 24 Mean

S 1

RMSE 0.11 0.11 0.14 0.10 0.04 0.14 0.05 0.02 0.02 0.01 0.01 0.08

CIELAB diff 15.12 15.55 17.42 12.03 6.84 6.88 5.55 4.25 3.21 4.03 4.81 8.86

S 2

RMSE 0.02 0.02 0.03 0.08 0.03 0.14 0.09 0.04 0.02 0.04 0.02 0.04

CIELAB diff

5.37 3.52 2.18 3.07 5.93 7.72 2.86 0.93 2.14 2.67 3.65 3.74

S 3

RMSE 0.01 0.02 0.02 0.02 0.01 0.02 0.01 0.01 0.01 0.01 0.01 0.01

CIELAB diff 2.11 2.18 1.51 2.84 3.28 0.43 0.51 0.40 0.69 0.34 1.20 1.58


157

Table 4. RMSE and CIELAB color difference computed for each stripe of the Painted Panel, between the values measured with the three imaging systems and the

ground truth data.

Painted Panel 1 2 3 4 5 6 7 8 9 10 11 12 Mean

S 1 RMSE 0.04 0.03 0.12 0.14 0.12 0.11 0.03 0.02 0.06 0.05 0.08 0.02 0.07

CIELAB diff 6.52 4.20 11.18 12.96 15.15 15.94 7.08 5.92 8.70 7.19 6.48 7.57 9.08

S 2 RMSE 0.04 0.01 0.09 0.04 0.07 0.12 0.06 0.04 0.02 0.02 0.21 0.04 0.06

CIELAB diff 1.87 3.15 3.61 7.16 7.19 7.23 8.96 2.57 4.37 5.35 6.39 4.29 5.18

S 3 RMSE 0.01 0.01 0.01 0.01 0.02 0.01 0.00 0.01 0.01 0.01 0.02 0.01 0.01

CIELAB diff 3.73 4.97 5.56 3.54 2.31 0.83 0.78 1.00 1.05 1.99 1.01 4.39 2.60

This paper revisited partial results of a Round Robin Test pursued for characterizing the performance of spectral systems on acquiring cultural heritage objects. The results discussed in the current article give an idea of the complexity of such a project and, at the same time, they insinuate the still unexplored resources of the RRT. The RRT is an effective tool for comparing the quality of imaging spectroscopy systems. Nonetheless, the huge amount of heterogeneous data gathered needs presents several challenges towards standardizing this comparison. A way to standardize the comparison is to select metrics that are robust to measurement and material error. As intended future work, the collected dataset is going to be explored with novel metrics from the field of image quality that go beyond color and spectral differences and take into consideration the semantic content of the reference objects. ACKNOWLEDGEMENTS This work was partially supported by the Color and Space in Cultural Heritage project. The work presented here is a result of a collaborative work of the RRT group within COSCH consortium, with a substantial contribution of Sony George and Jon Yngve Hardeberg from the Norwegian Color and Visual Computing Laboratory. REFERENCES

[1] Ciortan, Irina, Hilda Deborah, Sony George, and Jon Y. Hardeberg. "Color and hyperspectral image segmentation for historical documents." In 2015 Digital Heritage, vol. 1, pp. 199-206. IEEE, 2015.

[2] Deborah, Hilda, Sony George, and Jon Yngve Hardeberg. "Spectral-divergence based pigment discrimination and mapping: A case study on The Scream (1893) by Edvard Munch." Journal of the American Institute for Conservation (2019): 1-18.

[3] Berns, Roy S., Lawrence A. Taplin, Philipp Urban, and Yonghui Zhao. "Spectral color reproduction of paintings." In Conference on Colour in Graphics, Imaging, and Vision, vol. 2008, no. 1, pp. 484-488. Society for Imaging Science and Technology, 2008.


158

[4] Color and Space in Cultural Heritage project, Avaiable online: http://cosch.info/

[5] George, S. O. N. Y., J. O. N. Y. Hardeberg, J. O. Ã. O. Linhares, L. I. N. D. S. A. Y. MacDonald, Cristina Montagner, S. É. R. G. I. O. Nascimento, Marcello Picollo, R. Pillay, T. Vitorino, and E. K. Webb. "A Study of Spectral Imaging Acquisition and Processing for Cultural Heritage." Digital Techniques for Documenting and Preserving Cultural Heritage (2017): 141-158.

[6] MacDonald, Lindsay W., Tatiana Vitorino, Marcello Picollo, Ruven Pillay, Michał Obarzanowski, Joanna Sobczyk, Sérgio Nascimento, and João Linhares. "Assessment of multispectral and hyperspectral imaging systems for digitisation of a Russian icon." Heritage Science 5, no. 1 (2017): 41.

The Seventh



159

PHTHALATES DETERMINATION IN COSMETICS AT SFDA

LABORATORIES

Adnan S. Al -Mussallam , Fahad S. Aldawsari and Nawaf F. Alsowidan

Saudi Food and Drug Authority, Riyadh, Saudi Arabia


Abstract Phthalates are widely used in cosmetics industries as plasticizers and /or fixative agents in perfumes. Recent scientific reports had alarmed the public about the phthalates’ endocrine-disrupting potential. In this study, a GC-MS method was implemented at SFDA laboratories to achieve accurate and precise results regards phthalates analysis in nail polish and perfumes products. The most important challenges in the current study were 1) the presence of phthalate contaminants in pure organic solvents and 2) overlapping of some phthalate peaks (e.g. DEHP and DCHP). Results of this study showed recoveries of phthalates by more than 99 % in hexane. Additionally, more than 150 commercial samples were analysed. It was found that DEHP and DBP were the most frequently encountered phthalates in perfumes and nail polish respectively. Key words Phthalates, DEP, cosmetics, perfumes, fragrance, nail polish. 1 INTRODUCTION Phthalates are plasticizers widely used in consumer products. They can be found in personal care products, toys, food packages, pharmaceutical or other household products [1]. In cosmetic products, they are mainly used for the purpose of solvent and fixative properties [2,3]. Moreover, they are use in nail polish as to reduce cracking by making them less brittle. Recent scientific reports had alarmed the public about the phthalates’ endocrine-disrupting potential [4]. Consequently, many medicine and environmental regulatory authorities have initiated plans to reduce or prohibit few (if not all) of these chemicals [4,5]. Phthalates regulatory status in cosmetics can be classified into three categories [1,3]. First, is the ones are common in cosmetics and had a relatively broad range of safety profile (e.g. diethyl phthalates

Adnan S. Al-Mussallam, Fahad S. Aldawsari and Nawaf F. Alsowidan: Phthalates determination in cosmetics at SFDA laboratories

160

(DEP)). Second, are phthalates that are banned and should be absent in any cosmetic products (e.g. di-n-butyl phthalate (DBP), benzyl butyl phthalate (BBP) and di-(2-ethylhexyl) phthalates (DEHP)) [2]. Third category are phthalates that constituted recent safety issues and are under regulatory assessment (e.g. di-n-octyl phthalate (DOP)). According to Gulf Standards Organization (GSO) 1943/2016 Safety Requirements of Cosmetics and Personal Care Product, under List of Substances Prohibited in Cosmetic Products Annex 2; Benzyl butyl phthalate (BBP), Diethyl hexyl phthalate (DEHP), Dicyclohexyl phthalate (DCHP), and Dibutyl phthalate (DBP) are prohibited to use in the cosmetics products due to their documented toxicity [6]. In contrast, Dimethyl phthalate (DMP), Diethyl phthalate (DEP) and Di-isobutyl phthalate (DiBP) in cosmetics products and fragrance pose low risk for human health. The aim of the current study was to transfer and implement phthalate testing in cosmetic products to ensure the safety of these products available in Saudi Arabia market. 2 EXPERIMENTAL The current method was adapted from Netherlands Food and Consumer Product Safety Authority (NVWA) [1]. The scope of original method was to analyse plasticizers in PVC using gas GC-MS. The method described the identification and quantification of plasticizers; diisobutyl phthalate (DIBP), dibutyl phthalate (DBP), benzylbutyl phthalate (BBP), diethylhexyl phthalate (DEHP), di-n-octyl phthalate (DNOP), diisodecyl phthalate (DIDP), diisononyl phthalate (DINP), di(2-methoxyethyl) phthalate (DMEP), diisoheptyl phthalate- C6,7,8 (DIHP), di-n-pentyl phthalate (DPP), n-pentylisopentyl phthalate (PIPP) and di-n-hexyl phthalate (DHxP).

For the purpose of testing phthalates in cosmetics at SFDA laboratories, four phthalates in the original method (DBP, DiBP, BBP, DEHP) were selected in addition to three others namely; Dimethyl phthalate (DMP), Diethyl phthalate (DEP) and Dicyclohexyl phthalate (DCHP). Determination of phthalates were analysed by GC-MS (Agilent GC7890 B, 5977A MSD) with select ion monitoring (SIM) mode by using deuterium-labelled phthalate as internal standard.

Table 1. GC conditions used in the current study * mass range that used in identifications of phthalates

Column DB-5MS (30 m x .25 mm x 0.25 µm) Agilent Injector temperature 285°C Injection type Splitless Pressure 9 psi constant pressure Injection volume 1.0 µl Rinse solvent Tetrahydrofuran and hexane Carrier gas Helium Oven program

Temperature °C

hold time min

Rate °C / min

50 1 20 220 1 5

270 5 - Transfer line temperature 280 °C Mass range 45-450 m/z*


161

DMP, DEP, DBP, DiBP, BBP, DEHP and DCHP were purchased from (Sigma Aldrich, USA). Tetrahydrofuran inhibitor free, tetra butyl methyl ether (TMBE), puriss and hexane puriss were purchased from (Sigma Aldrich, USA). Di-n-Pentyl phthalate-3,4,5,6-D4 used as an internal standard and it was purchased from (Canadian isotopes, Canada). Certified Reference Material (CRM) was purchased from Netherlands food and consumer product safety authority (NVWA). 3 RESULTS AND DISCUSSIONS An important consideration in phthalates extraction was to minimize contaminations from the used laboratory materials (e.g. glassware, solvents, pipettes, micro pipettes, GC vials and GC caps). Therefore, it is advisable to wash glassware with solvents at high temperature and to avoid using plastic materials as they may contains phthalates and could possibly interfere with the results. During linearity study, the range used was 1.0-10 µg/ml using five concentrations along with 4.0 µg/ml of DnPP-d4 internal standard. When acetone was used as a solvent, the linearity gave unsatisfactory regression according to method validation requirements (Fig.1(a)). It was suspected that acetone was contaminated with some phthalates that may altered the linearity results. Indeed, using acetone as a blank, signal indications of phthalates were noticed which confirmed contamination hypothesis. After that, acetone was replaced by hexane puriss as a solvent which gave a satisfactory results (Fig.1(b)). Phthalate samples were extracted by using tetrahydrofuran in order to dissolve the nail polish matrix which would enhance the extraction in this matrix under sonication. The main solvent hexane was used to precipitate polymers from nail polish.

Fig.1(a). Calibration curve of Dimethyl phthalate (DMP) in acetone. For the mass detection, two modes were utilized; scanning and selected ion monitoring (SIM). The purpose for using the scan mode was to identify and quantify phthalates which often difficult to detect due to the complexity of nail polish matrix. Phthalates have similar ion fragmentation (such as ions 149 and 167 m/z) hence ion ratio comparison become very useful. For instance, DEHP and DCHP shares ions 149 and 167 m/z however they give different ion ratio between these two ions. Thus, selected ion monitoring (SIM) has been applied along with selected characteristic ion for both DEHP and DCHP (249 and 279 m/z). Table 2 illustrates the SIM mode ions:

Fig.1(b). Calibration curve of Dimethyl phthalate (DMP) in hexane.


162

Table 2. Selected ion monitoring (SIM) for phthalates

Phthalate Ions DMP 77 , 163 , 194 DEP 149, 105 , 167 DBP 205 , 149 , 167, DiBP 223.1, 149 , 167, IS (DNPP-d4) 153 BBP 149 , 91, 206 DEHP 149 ,249 , 167 , DCHP 279 ,149 , 167

During the method development, the challenge was to separate DEHP and DCHP due to two factors; similarity in retention time and the common mass fragmentation ions (149, 167 m/z) (Fig.2). Therefore, the target ions were selected according to their selectivity rather than their sensitivity [5]. Fragmentation ion at 249 m/z for DCHP had lower intensity than the base peak for phthalate ester at 149 m/z (Fig.3).

Fig.2. Primary structure of phthalates and major fragmentation of phthalate ester on

EI ionization

Fig.3: SIM mode chromatogram and mass to charge ions for DCHP and DEHP


163

The oven temperature in the GC instrument was thermally programed at fixed rate (20 °C / min) throughout the chromatographic run. Unfortunately, both of DEHP and DCHP were co-eluted along with low resolution. To overcome this problem, oven rate was modified to (5 °C / min) to achieve optimum resolution in separation. To satisfy method validation parameters, certified reference material (CRM) was analysed using the implemented method and the same matrix. Phthalates present in the CRM were DEHP, DBP and BBP (according the certificate from the supplier NVWA). Method validation parameters (linearity, selectivity, repeatability and measurement of uncertainty) were calculated. In addition, recoveries for phthalates, relative standard deviation, Limits of detection (LOD) for phthalates and limits of quantitation LOQ were calculated and the results are presented in Table 3. Overall, the method demonstrated no major differences between CRM and solvent in linearity study, repeatability and measurement of uncertainty.

Table 3. Validation data for different phthalates according to ICH guidelines:

Practical application of the implemented method on commercial products was performed. More than 150 samples were analysed to identify and quantify phthalates in perfumes and nail polish, particularly prohibited compounds, as per standards regulations. DEHP was the most frequently found phthalate in perfumes while the DBP was the mostly found phthalate in nail polish. For 90 nail polish samples and 60 perfumes samples the percentage of phthalates is illustrated in Fig.4.

Fig.4. the results of phthalates found in commercial products (nail polish and

perfumes).

Phthalates validation data Compound r2 Recovery

% LOD µg/ml

LOQ µg/ml

SD U %

DMP 0.9998 99.6 0.13 0.43 1.63 11 DEP 0.9995 99.5 0.11 0.33 0.99 15 DBP 0.9997 99.88 0.12 0.35 0.66 15 DiBP 0.9994 99.7 0.11 0.34 0.97 15 BBP 0.9998 99.8 0.13 0.43 1.5 12

DCHP 0.9999 98.78 0.08 0.258 1.73 11 DEHP 0.9999 100.9 0.11 0.33 0.89 9


164

4 CONCLUSIONS Appropriate sample extraction and recovery studies are critical to achieve higher recoveries of phthalates especially from nail polish samples. In method development challenge was the resolution between Diethyl hexyl phthalate (DEHP) and Dicyclohexyl phthalate (DCHP) peaks. Also, the selection of characteristic ions for individual phthalate is important to minimize the matrix interference and other phthalates interference. To that end, minimizing contamination that may come from laboratory materials is critical aspect to ensure a representative result.

REFERENCES [1] Determination and quantification of plasticizers in PVC and nail polish using GC-

MS, CHE01-WV487, V5, 2015. [2] Li D and Suh S. Health risks of chemicals in consumer products: A review.

Environ. Int. 2019, Feb.123:580-587. [3] Koo HJ and Lee BM. Estimated exposure to phthalates in cosmetics and risk

assessment. J. Toxicol. Environ. Health A. 2004, Dec,67(23-24);1901-14. [4] Mario Russo, Pasquale Avino, Luisa Peruginia and Ivan Notardonatoa.

Extraction and GC-MS analysis of phthalate esters in food matrices: a review. RSC Advances, 2015,5, 46, p. 37023.

[5]

Hao-Yu S, Hai-Liang J, Hong-Lei M, Gang Pan, Lu Zhou, Yun-Feng C. Simultaneous determination of seven phthalates and four parabens in cosmetic products using HPLC-DAD and GC-MS methods. J. Sep. Sci. 2007, 30, 48–54

[6] https://www.gso.org.sa/store/gso/standards/GSO:724703/GSO%201943:2016?lang=en

The Seventh



165

MODELLING THE VARIATION OF MEASUREMENT UNCERTAINTY

WITH THE MEASURED VALUE FROM PROFICIENCY TEST RESUL TS

Carla Palma 1, Vanessa Morgado 1,2, Ricardo J. N. Bettencourt da Silva 2* 1Instituto Hidrográfico, Rua das Trinas, 49, 1249-093 Lisboa, Portugal

2Centro de Química Estrutural – Faculdade de Ciências da Universidade de Lisboa – Campo Grande, 1749-016 Lisboa, Portugal

[email protected]

Abstract It is only possible to decide if a measurement result is fit for the intended use or to objectively interpret a result, such as the compliance of the characterised item with a limit value, if the measurement uncertainty was evaluated. The most popular approach for the evaluation of the measurement uncertainty is the top-down approach based on global performance parameters collected in precision and trueness tests performed on the procedure. Measurement precision component is quantified by the intermediate precision standard deviation and measurement trueness component is quantified from the analysis of reference materials such as proficiency test samples. Models of measurement precision variation with the measured value have been developed based on well-known precision variation. Precision standard deviation or relative standard deviation tends to be approximately constant below or above two times de Limit of Quantification, respectively. Many measurements in chemistry are affected by systematic effects (e.g. matrix effects). By identifying cases where these effects are relevant, for instance due to matrix effects variability, it is possible to decide if improving measurement robustness to these effects reduces the measurement uncertainty significantly. This work presents models of trueness uncertainty variation with the measured level based on proficiency test results. The developed methodology was applied to the analysis of metals in sediments by atomic spectroscopy. Key words Uncertainty models; Proficiency test; Heavy Metals; Sediments; Atomic Spectrometry

Carla Palma, Vanessa Morgado, Ricardo J. N. Bettencourt da Silva: Modelling the variation of measurement uncertainty with the measured value from proficiency test results

166

1 INTRODUCTION A measurement result is only fit for the intended purposed if traceable to an adequate reference and associated with an adequately low uncertainty. The selection of the measurement reference, such as an internal or certified reference material, sets the measurement traceability and the comparison of the calculated uncertainty with a target uncertainty allows concluding about its adequacy [1]. Different approaches for the evaluation of the measurement uncertainty have been developed that differ from the type of data used. The bottom-up approach involves the separate quantification of the uncertainty components distinguishable from the measurement procedure and effects that impact on the measurement result. The bottom-up evaluations are more informative about the relevance of the various components when more uncertainty components are distinguished (i.e. the percentage contribution of the identified uncertainty components). However, this approach involves complex measurement modelling and can easily drive to uncertainty underestimation. The top-down assessments of the measurement uncertainty are more popular since involve the identification of few and almost the same uncertainty components regardless of the studied procedure. The uncertainty components are the precision, trueness and any additional components not expressed in the first two such as the sampling uncertainty. This approach relies more on experimental data about the measurement performance than on uncertainty components modelling. This makes this evaluation strategy demanding regarding the collection of experimental data, in particular if the procedure is applicable to a wide range of studied property values (e.g. analyte concentration). Pragmatic approaches have been proposed for developing continuous models of measurement uncertainty variation with the studied parameter values. Since it is well known how measurement precision varies below and above the Limit of Quantification, LOQ, step models of intermediate precision variation have been developed [1-3]. This work discusses how to define models of trueness uncertainty variation with the measured quantity when the trueness is assessed from the analysis of proficiency test samples. The developed approach is also applicable when other reference materials are used such as spiked samples and certified reference materials. 2 UNCERTAINTY EVALUATION 2.1 Identification and combination of components fo rward to uncertainty expansion The top-down evaluation of the measurement uncertainty involves combining precision, trueness and additional uncertainty components using the law of propagation of uncertainty [2,3]. If the absolute uncertainty components, i.e. expressed in measured value units (e.g. analyte concentration units) or relative uncertainty components, i.e. unitless, are constant, Eq.(1) or (2) are used to calculate the absolute, u(q), or relative, u’(q), combined standard uncertainty, respectively.


167

(1)

(2)

Where uP and u’P are the absolute and relative precision standard uncertainty, uT and u’T the absolute and relative trueness standard uncertainty and, uA and u’A the absolute and relative uncertainty from additional components, respectively. When u’(q) is calculated, it must be multiplied by the measured value, q, to obtain the u(q) (i.e. u(q) = qu’(q)). Whenever the uncertainty components are associated or expected to be associated with a large number of degrees of freedom, the standard uncertainty can be expanded to approximately 95 % or 99 % confidence level by using a coverage factor of 2 or 3, respectively. 2.2 Intermediate precision component This component is quantified by the absolute or relative intermediate precision standard deviation. When the LOQ is quantified under intermediate precision condition, it can be assumed that precision at two times the LOQ, 2qLOQ, slightly overestimates the absolute and relative precision below and above this value, respectively [2,3] (Fig. 1).

Fig. 1. The absolute, sI, (a) or relative, s’I, (b) intermediate precision standard deviation observed at two times the LOQ, 2qLOQ, slightly overestimates the precision

below and above 2qLOQ, respectively. The overestimation of the uncertainty is reduced if instrumental precision is estimated at more quantity levels. For instance, if the relative intermediate precision standard deviation, s’I, is estimated at two additional quantity values above 2qLOQ (e.g. qA and qB), the respective intermediate precision relative standard deviations, s’I(qA) and s’I(qB), can be used to improve precision model (Fig. 2).

Fig. 2. Step model of the variation of the relative intermediate precision standard deviation, s’I, above 2qLOQ based on precision estimated at three quantity values

(2qLOQ, qA and qB).


168

2.3 Trueness component The trueness uncertainty is assessed from results of the analysis of reference materials, i.e. materials with a reference value. Proficiency test items with adequate reference value are fit for this assessment if several results from the analysis of reference materials are available. If each reference material is analysed once, the mean analyte recovery, , and the respective standard uncertainty, , also designated trueness uncertainty, uT, is estimated by Eq.(3) and (4), respectively.

(3)

(4)

where qi and Qi are the estimated and reference value of the i-th proficiency test sample (i = 1 to N), respectively, s(qi) is the intermediate precision standard deviation estimated for qi from models of section 2.1 and u(Qi) is the standard uncertainty associated with Qi. If matrix effects vary significantly between analysed items, the term (s(qi)/ qi) should be substitute by [3], where s(Ri) is the standard deviation of measurement recovery (Ri = qi/Qi).

The is subsequently compared with the ideal 100 % recovery value to decide if the measured values should be corrected for the observed recovery. The is metrologically equivalent to 100 % when the following condition is valid:

(5)

where t(0.05, N – 1) is the t value of the Student's t-distribution for 95 % confidence level and N-1 degrees of freedom. If the condition is valid, the original measured value is reported and the mean recovery relative and absolute standard uncertainties are equivalent since (i.e. ). When the condition is not valid, the reported measured value, , is the original value, q, corrected for recovery

( ) and the mean recovery relative standard uncertainty: . In some cases, the way measurements are defined and the type of reference material used does not allow the correction of the measured value with observed recovery. This work does not discuss those cases. The relative and absolute trueness standard uncertainties used in Eq.(1) and (2) are estimated from recovery uncertainty: and , respectively, where q is the measured value.


169

2.4 Additional component The additional uncertainty components, such as the sampling uncertainty, are expressed as absolute or relative standard uncertainties in measured value units or unitless, and combined with the other uncertainty components as describe in Eq.(1) and (2), respectively [2,3]. 3 EXPERIMENTAL In this work, it was studied the determination of the total mass fraction of five elements in a marine sediment using a procedure developed by OSPAR [4] or the operationally defined extractable mass fraction of two elements using EPA 3051A standard [5]. It was studied the analytes Cr, Cu, Li, Mn, Ni, Pb and Zn.

Fig. 3. Variation of analyte recovery, Ri, with the mass concentration, w. Determinations of total Cr (a), Cu (b), Li (c), Mn (d) and Zn (g), and acid extractable

Ni (e) and Pb (f) according to the EPA 3051A standard, in marine sediments.


170

3.1 Measurement procedures The samples from marine sediments are freeze-dried, homogenised and reduced to a fine powder in a mortar and pestle. For the OSPAR procedure, an aliquot of 0.45 g to 0.55 g of sediment is placed in a Teflon bomb with a mixture of 2 mL hydrofluoric acid and 6 mL aqua regia. For the EPA procedure an equivalent mass of sediment is placed in a Teflon bomb with 10 mL of ultrapure nitric acid. The bombs are heated in a microwave oven using a specific temperature and power programme. For the OSPAR procedure, after the digestion, the solution is neutralised with boric acid and diluted to 50 mL with ultra-pure water. In the EPA procedure, the digested sample is simply diluted to 50 mL with ultra-pure water. The elements are quantified in the sample solution by flame atomic absorption spectrometry using a calibration curve obtained from external calibrators. The mass fraction of the elements in the sediments are calculated on a dry weight basis by taking a correction factor estimated from the mass loss of an independent analytical portion subject to 103 °C to 105 °C. 3.2 Procedure validation The validation of the measurement procedure involved the following steps: 1) Determination of the LOQ under intermediate precision conditions; 2) Assessment of the linearity of atomic spectrometric determinations; 3) Determination of the intermediate precision standard deviation for the analysis of homogeneous or heterogeneous samples; 4) Assessment of measurement trueness from the analysis of proficiency test samples; 5) Evaluation of the measurement uncertainty including modelling uncertainty variation with the mass fraction value; 6) Comparison of the measurement uncertainty with a target (i.e. maximum admissible) uncertainty value.


171

Table 1. Developed uncertainty model and summary of their validation through the analysis of proficiency tests samples (Compatibility test) and comparison with a target

standard uncertainty (uctg and u’c

tg ).

Element (procedure):

Cr (OSPAR)

Cu (OSPAR)

Li (OSPAR)

Mn (OSPAR)

Ni (EPA3051)

Pb (EPA3051)

Zn (OSPAR)

Matrix: Marine sediments

wLOQ (mg kg-1): 5 5 0.5 5 7.5 10 2

uP (mg kg-1): 0.633 0.347 0.0675 0.680 1.18 1.08 0.159

u'P (%): 6.33 3.47 6.75 6.80 7.97 6.37 3.97

uT(absolute): 0.0189 0.0111 0.0395 0.0150 0.0131 0.0141 0.0145

(%): 110 100 104 99.3 88.9 93.1 103

Is No Yes Yes Yes No No Yes

uc (mg kg-1): 1.05 0.612 0.127 0.920 1.25 1.90 0.365

u'c : 10.5 6.13 12.7 9.20 8.44 10.1 9.13

U (mg kg-1): 2.10 1.22 0.254 1.84 2.51 3.79 0.731

U' (%): 21.0 12.3 25.4 18.4 16.9 20.2 18.3

Comp. test: 20 in 20 20 in 20 15 in 15 20 in 20 20 in 20 20 in 20 20 in 20

uctg (mg kg-1): 1.63 1.125 0.1125 0.675 1.44 2.25 1.500

u'ctg (%): 22.5 17.5 17.5 13.0 15.8 17.5 43.8

wLOQ – Limit of Quantification in mass fraction units; uP and u'P – precision absolute and relative standard uncertainties below and above 2wLOQ, respectively; uc and u'c – combined absolute and relative standard uncertainties below and above 2wLOQ, respectively; U and U' – expanded absolute and relative standard uncertainties below and above 2wLOQ, respectively; Comp. test – number of estimated proficiency test results compatible with the reference value; uc

tg and u'ctg – absolute and relative target standard uncertainties

below and above 2wLOQ, respectively. 4 RESULTS Fig. 3 presents the variation of analyte recovery with the mass fraction observed in proficiency test for the analysis of various parameters in marine sediments. It can be observed that no strong variation of recovery value with measured value is observed. Therefore, Eq.(4) was used to estimate uT where s(qi)/qi was substituted by . Table 1 presents the measurement performance data and summarises the validation of the uncertainty models through the analysis of other proficiency test samples and comparison with a target uncertainty. The uncertainty models were 100 % successful when applied to the analysis of 15 to 20 samples from proficiency tests independent from those considered in the evaluation of the measurement uncertainty. The calculated measurement uncertainty is not significantly larger than the target value defined by taking the reference standard deviation used by QUASIMEME [6] to calculate the z-score of proficiency test results. A tolerance of 30 % was considered in the comparison between the calculated and target standard uncertainties [1].


172

5 CONCLUSION No relevant variations of analyte recovery with the studied mass fraction of elements in marine sediments was observed allowing using the same trueness uncertainty in the measurement uncertainty model. The developed uncertainty models were successfully applied to 135 analysis of proficiency test samples independent from the samples used to develop the uncertainty models. The quantified uncertainties are not significantly larger than the target (i.e. maximum admissible) uncertainty inferred from how z-score are calculated in QUASIMEME proficiency tests. REFERENCES

[1] R. Bettencourt da Silva, A. Williams (Eds.), Eurachem/CITAC Guide: Setting and Using Target Uncertainty in Chemical Measurement, 2015.

[2] R. Cordeiro, C. Rosa, R. J. N. Bettencourt da Silva. “Measurements recovery evaluation from the analysis of independent reference materials: Analysis of different samples with native quantity spiked at different levels.” Accred. Qual. Assur. 23 (2018): 57-71.

[3] C. Palma, V. Morgado, R. J. N. Bettencourt da Silva. “Top-down evaluation of matrix effects uncertainty.” Talanta 192 (2018): 278-287.

[4] OSPAR COMMISSION, JAMP Guidelines for Monitoring Contaminants in Sediments, Agreement Ref. no. 2002-16, (2015) (111 pp).

[5] EPA, Method 3051A – Microwave Assisted Acid Digestion of Sediments, Sludges, Soils and Oils, EPA, USA, 2007.

[6] QUASIMEME, Quasimeme Laboratory Performance Studies – Programme 2017, Wageningen University, Wageningen, 2017.

The Seventh



173

STATISTIC CONTROL OF THE RESULTS FOR THE DETERMINATIONS OF MAJOR OXIDES FROM A PORTLAND

CEMENT, BY USING OF SHEWHART DIAGRAMS

Gherghina Ciortan , Marina Martin

CEPROCIM S.A., # 6, Preciziei Blvd., Postal Code 062203, Sect. 6, Bucharest


Abstract In SR EN ISO / CEI 17025: 2018, the requirement to ensure the quality of the results requires more laboratory practice, including the use of control diagrams. Internal quality control at the chemical laboratory implies a continuous, critical assessment of the laboratory's own analytical methods and work routines. Control means that the analytical process begins with the entry of the sample into the laboratory and ends with the analytical report. The most important tool in this quality control is the use of Shewhart control charts. The paper will present the quality control of the results for the major cement oxides (SiO2, Fe2O3, Al2O3, CaO, MgO), over the three years 2016-2018. Five chemical analyzes SiO2, Fe2O3, Al2O3, CaO, MgO were considered. The control sample was "In House Cement". Six determinations (6 samples equivalent to 6 rounds) were performed for each major element. The control charts built were "Shewhart" type . Control limits were "Statistics". The central line was the "average value". 14 statistical parameters were calculated as follows: N = number of values; x med. = average of values; s = calculated standard deviation; CV = calculated coefficient of variation; x min. = lowest value; x max. = highest value; x med. - 2s = media - 2s = warning limit (WL) x med. + 2s = average + 2s = upper limit of warning (WL) x med. - 3s = average - 3s = lower limit of action (action limit, AL) x med. + 3s = average + 3s = upper limit of action (action limit, AL) N '= the number of values recalculated after the elimination of values in the operating areas; x 'med. = average of values recalculated after removing values in the operating areas;

Gherghina Ciortan, Marina Martin: Statistic control of the results for the determination of major oxides from a Portland cement , by using of Shewhart diagrams.

174

s' = the standard deviation recalculated after removing the values in the operating areas; CV '= the coefficient of variation recalculated after the elimination of the values in the operating areas. Key words Intern control of quality, control diagrams. 1 INTRODUCTION Statistic control of the quality is an important tool for continuous improvement of the results performances in laboratory and consists mainly in the elaboration of control diagrams. The statistic control has the following components: a) actual statistic control – passive statistic method , that follows the process, and when the process is under control, passive remark does not produce utile information, but signal problem appearance or the trend b) experimental design – active statistic method – performing of the tests from which result the modifications of input variables which determine modifications of output variables and utile information in order to improve the process or for solving of the problem appeared in process. 2 SHEWHART CONTROL DIAGRAMS They are traced by drawing up more lines with different significance [1] : Median or central line (CL) represents average value of the appropriate characteristics for control zone, in which just contingent causes are present; 2 horizontal lines: upper control limit LCS and lower control limit LCI or lines of warning CL ±2s (root-mean-square) 2 horizontal lines: at distance of ±3s against central line, named lines of action The limits of control are chosen so that the process to be under control, which means that almost all points of the specimens to be within these limits [2] . If all the points are between these limits, the process is under control and no action is necessary. If the points are outside limits, the process is not under control and correction actions are necessary in order to found and eliminate the causes [6] . The relation between Gauss normal curve of distribution of the tests results an control graphic is presented in the figure 1.


175

Fig. 1. Gauss Normal distribution curve of the tests results In the above figure, three areas are distinguished. Area 1 – stable regime – current fluctuations are found; Area 2 – warning – results quality signal the appearance of some causes which had determined these variations; Area 3 – action – results quality imposes investigations and actions for the remedying of the causes which had determined quality depreciation [3] . 3 THE EXPERIMENTAL DESIGN In the following tables and figures,are presented the control diagrams for major oxides in a cement sample which was analyzed and qualitative controlled on a period of three years: 2016-2018


176

Table 1. Values for making the silicium dioxide diagram and values obtained between 2016-2018

Fig. 2. Repeatability of silicium dioxide in cement sample

Sample 1 Sample 2 Sample 3 Sample 4 Sample 5 Sample 6 2016 2016 2017 2017 2018 2018Lower

action limit, AL

x med. - 3s (19,88%) 19,88 19,88 19,88 19,88 19,88 19,88 19,88 19,88 19,88 19,88 19,88 19,88

Lower warning limit, WL

x med. - 2s (20,08%) 20,08 20,08 20,08 20,08 20,08 20,08 20,08 20,08 20,08 20,08 20,08 20,08

x med. (20,48%) 20,48 20,48 20,48 20,48 20,48 20,48 20,48 20,48 20,48 20,48 20,48 20,48

determination values 20,42 20,35 20,20 20,65 20,75 20,52 20,13 20,45 20,25 20,53 20,60 20,74Higher warning limit, WL

x med. + 2s (20,88%) 20,88 20,88 20,88 20,88 20,88 20,88 20,88 20,88 20,88 20,88 20,88 20,88

Higher action limit,

ALx med. + 3s (21,09%) 21,09 21,09 21,09 21,09 21,09 21,09 21,09 21,09 21,09 21,09 21,09 21,09

20,4220,35

20,20

20,6520,75

20,52

19,80

20,00

20,20

20,40

20,60

20,80

21,00

21,20

Sample 1 Sample 2 Sample 3 Sample 4 Sample 5 Sample 6

Va

lue

s, %

Samples

Shewhart charts for SiO2

x med. - 3s (19,88%)

x med. - 2s (20,08%)

x med. (20,48%)

determination values

x med. + 2s (20,88%)

x med. + 3s (21,09%)


177

Fig. 3. Variation of silicium dioxide according to the framing limits

Fig. 4. Variation of silicium dioxide between 2016-2018

20,13

20,45

20,25

20,5320,60

20,74

19,20

19,40

19,60

19,80

20,00

20,20

20,40

20,60

20,80

21,00

21,20

2016 2016 2017 2017 2018 2018

Va

lue

s, %

Years

Shewhart charts for SiO2

x med. - 3s (19,88%)

x med. - 2s (20,08%)

x med. (20,48%)


x med. + 2s (20,88%)

x med. + 3s (21,09%)


178

Table 2. Values for making the iron oxide diagram and values obtained between 2016-2018

Fig. 5. Repeatability of iron oxide in cement sample


action limit, AL

x med. - 3s (3,62&%) 3,62 3,62 3,62 3,62 3,62 3,62 3,62 3,62 3,62 3,62 3,62 3,62


x med. - 2s (3,64%) 3,64 3,64 3,64 3,64 3,64 3,64 3,64 3,64 3,64 3,64 3,64 3,64

x med. (3,68%) 3,68 3,68 3,68 3,68 3,68 3,68 3,68 3,68 3,68 3,68 3,68 3,68

determination values 3,68 3,66 3,70 3,66 3,70 3,67 3,65 3,70 3,66 3,65 3,68 3,67Higher warning limit, WL

x med. + 2s (3,72%) 3,72 3,72 3,72 3,72 3,72 3,72 3,72 3,72 3,72 3,72 3,72 3,72


ALx med. + 3s (3,73%) 3,73 3,73 3,73 3,73 3,73 3,73 3,73 3,73 3,73 3,73 3,73 3,73

3,68

3,66

3,70

3,66

3,70

3,67

3,60

3,62

3,64

3,66

3,68

3,70

3,72

3,74


Va

lue

s, %

Samples

Shewhart charts for Fe2O3

x med. - 3s

(3,62&%)

x med. - 2s

(3,64%)

x med. (3,68%)

determination

values

x med. + 2s

(3,72%)

x med. + 3s

(3,73%)


179

Fig. 6. Variation of iron oxide according to the framing limits

Fig. 7. Variation of iron oxide between 2016-2018

3,65

3,70

3,663,65

3,683,67

3,56

3,58

3,60

3,62

3,64

3,66

3,68

3,70

3,72

3,74

3,76

2016 2016 2017 2017 2018 2018

Va

lue

s, %

Years

Shewhart charts for Fe2O3

x med. - 3s (3,62&%)

x med. - 2s (3,64%)

x med. (3,68%)


x med. + 2s (3,72%)

x med. + 3s (3,73%)

Fe2O3

3,6

3,65

3,7

2016 2016 2017 2017 20182018

Va

lue

s,

%

Years

Fe2O3

Fe2O3


180

Table 3. Values for making the aluminium oxide diagram and values obtained between 2016-2018

Fig. 8. Repeatability of aluminium oxide in cement sample


action limit, AL

x med. - 3s (5,04%) 5,04 5,04 5,04 5,04 5,04 5,04 5,04 5,04 5,04 5,04 5,04 5,04


x med. - 2s (5,09%) 5,09 5,09 5,09 5,09 5,09 5,09 5,09 5,09 5,09 5,09 5,09 5,09

x med. (5,19%) 5,19 5,19 5,19 5,19 5,19 5,19 5,19 5,19 5,19 5,19 5,19 5,19determination values 5,18 5,11 5,25 5,21 5,17 5,22 5,12 5,13 5,19 5,20 5,24 5,12

Higher warning limit, WL

x med. + 2s (5,29%) 5,29 5,29 5,29 5,29 5,29 5,29 5,29 5,29 5,29 5,29 5,29 5,29


ALx med. + 3s (5,34%) 5,34 5,34 5,34 5,34 5,34 5,34 5,34 5,34 5,34 5,34 5,34 5,34

5,18

5,11

5,25

5,21

5,17

5,22

5,00

5,05

5,10

5,15

5,20

5,25

5,30

5,35


Va

lue

s, %

Samples

Shewhart charts for Al2O3

x med. - 3s (5,04%)

x med. - 2s (5,09%)

x med. (5,19%)

determination

valuesx med. + 2s (5,29%)


181

Fig. 9. Variation of aluminium oxide according to the framing limits

Fig. 10. Variation of aluminium oxide between 2016-2018

5,12 5,13

5,19 5,205,24

5,12

4,85

4,90

4,95

5,00

5,05

5,10

5,15

5,20

5,25

5,30

5,35

5,40

2016 2016 2017 2017 2018 2018

Va

lue

s, %

Years

Shewhart charts for Al2O3

x med. - 3s (5,04%)

x med. - 2s (5,09%)

x med. (5,19%)


x med. + 2s (5,29%)

x med. + 3s (5,34%)


182

Table 4. Values for making the calcium oxide diagram and values obtained between 2016-2018

Fig. 11. Repeatability of calcium oxide in cement sample


action limit, AL

x med. - 3s (63,36%) 63,36 63,36 63,36 63,36 63,36 63,36 63,36 63,36 63,36 63,36 63,36 63,36


x med. - 2s (63,63%) 63,63 63,63 63,63 63,63 63,63 63,63 63,63 63,63 63,63 63,63 63,63 63,63

x med. (64,16%) 64,16 64,16 64,16 64,16 64,16 64,16 64,16 64,16 64,16 64,16 64,16 64,16determination values 64,16 63,80 63,90 64,50 64,25 64,34 64,25 63,94 64,33 63,76 64,53 63,99


x med. + 2s (64,69%) 64,69 64,69 64,69 64,69 64,69 64,69 64,69 64,69 64,69 64,69 64,69 64,69


ALx med. + 3s (64,96%) 64,96 64,96 64,96 64,96 64,96 64,96 64,96 64,96 64,96 64,96 64,96 64,96

64,16

63,8063,90

64,50

64,2564,34

63,20

63,40

63,60

63,80

64,00

64,20

64,40

64,60

64,80

65,00


Va

lue

s, %

Samples

Shewhard charts for CaO

x med. - 3s

(63,36%)

x med. - 2s

(63,63%)

x med.

(64,16%)

determination

values

x med. + 2s

(64,69%)


183

Fig. 12. Variation of calcium oxide according to the framing limits

Fig. 13. Variation of calcium oxide between 2016-2018

64,25

63,94

64,33

63,76

64,53

63,99

62,50

63,00

63,50

64,00

64,50

65,00

65,50

2016 2016 2017 2017 2018 2018

Va

lue

s, %

Years

Shewhard charts for CaO

x med. - 3s (63,36%)

x med. - 2s (63,63%)

x med. (64,16%)


x med. + 2s (64,69%)

x med. + 3s (64,96%)


184

Table 5. Values for making the magnesium oxide diagram and values obtained

between 2016-2018

Fig. 14. Repeatability of magnesium oxide in cement sample


action limit, AL

x med. - 3s (1,13%) 1,13 1,13 1,13 1,13 1,13 1,13 1,13 1,13 1,13 1,13 1,13 1,13


x med. - 2s (1,14%) 1,14 1,14 1,14 1,14 1,14 1,14 1,14 1,14 1,14 1,14 1,14 1,14

x med. 1,18 1,18 1,18 1,18 1,18 1,18 1,18 1,18 1,18 1,18 1,18 1,18determination values 1,18 1,16 1,20 1,18 1,20 1,16 1,18 1,17 1,19 1,21 1,17 1,2


x med. + 2s (1,22%) 1,22 1,22 1,22 1,22 1,22 1,22 1,22 1,22 1,22 1,22 1,22 1,22


ALx med. + 3s (1,23%) 1,23 1,23 1,23 1,23 1,23 1,23 1,23 1,23 1,23 1,23 1,23 1,23

1,18

1,16

1,20

1,18

1,20

1,16

1,12

1,14

1,16

1,18

1,20

1,22

1,24

Sample 1Sample 2Sample 3Sample 4Sample 5Sample 6

Va

lue

s, %

Samples

Shewhart charts for MgO

x med. - 3s

(1,13%)x med. - 2s

(1,14%)x med.

determination

valuesx med. + 2s

(1,22%)


185

Fig. 15. Variation of magnesium oxide according to the framing limits

Fig. 16. Variation of magnesium oxide between 2016-2018

1,181,17

1,19

1,21

1,17

1,2

1,06

1,08

1,10

1,12

1,14

1,16

1,18

1,20

1,22

1,24

1,26

2016 2016 2017 2017 2018 2018

Va

lue

s, %

Years

Shewhart charts for MgO

x med. - 3s (1,13%)

x med. - 2s (1,14%)

x med.


x med. + 2s (1,22%)

x med. + 3s (1,23%)


186

4 CONCLUSIONS For results interpretation the most used modality is the diagram Levey- Jennings [2], where control values which are not within in acceptability limits are identified and adequate corrective measures are taken. Another way for results interpretation is the using of Westgard rules [2]. This approach is directed more to avoiding than to correction. Through these rules may be identified trends in laboratory results which not involve an immediately exclusion of the results, but if these are not remedied may lead to an exclusion. From presented graphics it may concluded that the results are under control, as follow [4] : • There are no values higher than upper limit and no values smaller than lower limit ( ±3α) • 2 from 3 consecutive points do not exceed the limits ±2α • 4/5 from consecutive values do not exceed ±1α • The results situated between the zone 2 and 3 are not trends • There are no more than 2/3 from consecutive values of same part of central line

By elaboration of control diagrams we perform [5] : • Fast detection of appearance random causes of non-conformities into the laboratory, so that the laboratory can take corrective action in due time before the results are elaborated to clients. • Provide useful information for improvements into the laboratory and eliminate the variability of results as much as possible. • Comparison data with acceptance limits from the standards of product or limits imposed by mathematic statistics. • The functioning of the equipment used.

REFERENCES

[1] Havard Hovind, Bertil Magnusson, Mikael Krysell, Ulla Lund, Irma Makinen-Nordtest report Tr 569- Internal quality control.

[2] Bertil Magnusson, Teemu Naykki, Havard Hovind , Mikael Krysell- Nord test Report TR 537.

[3] Caulcutt, R. (2004) Control charts in practice. Significance, 1(2): 81–84. [4] Laitinen, H.A. Statistics in Quantitative Analysis. In Chemical Analysis: An

Advanced Text and Reference; McGraw-Hill: New York, N.Y., 1960; Chapter 26.

[5]

[6]

Wernimont, G. Use of control charts in the analytical laboratory. Ind. Eng. Chem. Anal. Ed. 1946, 18(10), 587–92 [doi: 10.1021/i560158a001]. SR ISO 8258:1991+C1: 1998 “Fise de control Shewhart”.

The Seventh



187

ABSOLUTE STANDARDIZATION OF THE RADIONUCLIDE 54Mn AND

PARTICIPATION AT INTERNATIONAL COMPARISONS

Maria Sahagia , Aurelian Luca

Horia Hulubei National Institute of R&D for Physics and Nuclear Engineering, IFIN-HH, P.O. Box MG-6, Bucharest, RO-77125, Romania


Abstract IFIN-HH, as a Designated Institute (DI) in the field of Ionizing Radiations and owner of the National Standard of Activity (of a radionuclide) unit, Becquerel, is registered in the Key Comparison Data Base (KCDB) of the CIPM-MRA with a CMC referring at 54Mn, an important radionuclide for radionuclide metrology. The radionuclide is decaying by electron capture followed by gamma-ray emission; it was standardized by the 4π(PC) - γ coincidence counting. In this work a special treatment of the prepared sources was used, allowing to improve the efficiency and to obtain a very good result in the standardization. Two comparisons regarding 54Mn are reported in the paper. First comparison is a key comparison, organized within the International Committee of Weights and Measures, CIPM, via the Bureau International of Weights and Measures (BIPM), codified as CCRI(II).K2.Mn-54, referring at the absolute standardization. The second comparison was with the PTB-Germany and referred at measurements performed with the secondary standard, ionization chamber. Key words Primary Standard, Key Comparison CCRI(II)-K2.Mn-54 , 4πPC-γ coincidence method, Ionisation chamber measurement 1 INTRODUCTION 54Mn decays by electron capture, emitting Auger electrons and x-rays with energies close to that emitted by 55Fe and 65Zn, followed by the emission of mono energetic 834.5 keV gamma rays. Due to these special emitted radiations it is an important radionuclide for radionuclide metrology, in two applications: (i) Absolute standardization of radionuclides 55Fe and 65Zn by the liquid scintillation method in the variant CIEMAT/NIST, being a better alternative to 3H usually used as a tracer [1], or absolute standardization of pure decaying electron capture radionuclides, like 55Fe, by the 4π(PPC)β-γ - coincidence counting efficiency tracing. (ii) Calibration of the gamma-ray spectrometers, HPGe or NaI(Tl) detector based, in this energy region, as

Maria Sahagia, Aurelian Luca: Absolute standardization of the radionuclide 54Mn and participation at international comparisons

188

a single gamma energy emitter; a 54Mn standard source is a basic component of any gamma-ray spectrometry calibration set. Due to these uses, there is a necessity to perform very precise measurements of its activity; this was the reason to be organized international comparisons regarding its standardization. The participations in two such international comparisons are reported in this paper. Regarding the key comparison, organized within the International Committee of Weights and Measures (CIPM) codified as CCRI(II).K2. Mn-54, the results of the comparison were first distributed to the participants, primary standard laboratories, as Draft A report versions and finally as a Draft B, in 2018.The second comparison, with the Physikalisch-Technische Bundesanstalt (PTB) - Germany, regarded the measurements carried out at PTB and at IFIN-HH, as presented in the papers [2, 3]. This paper describes the absolute standardization by using the 4π(PC) - γ coincidence counting in the variant of efficiency extrapolation and its validation by the excellent result obtained in these comparisons, which will be presented in detail. 54Mn was standardized in several laboratories by various methods, some of them described in the papers mentioned below. The proportional counter – NaI(Tl) scintillation crystal, 4π(PC) - γ coincidence counting in the variant of efficiency extrapolation was applied in [4]. The liquid scintillation (LS) CIEMAT/NIST method was applied in [5], while the coincidence counting based on liquid scintillator and coincidence, without extrapolation, was applied in [6]. 2 ABSOLUTE STANDARDIZATION 2.1 Decay scheme of 54Mn. The decay scheme from Figure 1 is in agreement with the data from Chechev and Kuzmenco, [7]. 54Mn decays 100% by electron capture, having a half life of 312.2(4) d, followed by the emission of the γ- rays with energy 834.850 keV, intensity 99.975%.

Fig.1. Decay scheme of 54Mn


189

The electron capture is followed by the emission of low energy, 5.3 keV K Auger electrons (63.3%) and 5.41 to 5.95 keV K x-rays of Cr (25.7%). 2.2 Experimental conditions The 54Mn solution was standardized within the frame of the CCRI(II)-K2.Mn-54, by the 4πPC-γ coincidence method. The classical coincidence installation of the laboratory was used. On the beta channel, a pure methane flow proportional counter (PC), operated at a 0.1 MPa pressure, was used. One may use very easily the efficiency extrapolation variant of the coincidence method, with an appropriate choice of the gamma energy interval, such as to get a low slope of extrapolation curve. We used a PC counting threshold of 1 keV. On the γ –channel, provided with a NaI(Tl) crystal, the energy interval of γ-rays, 250 keV- 1000 keV, was counted. The coincidence equations, connecting the counting rates in the PC, gamma and coincidence channels, NPC, Nγ, Nc, and activity N0, in the case of this simple decay scheme, are: , 1 ,

(1) , 1 ,

εX,A is the PC counting efficiency for K- Auger electrons and x-rays (X,A) ,

are the efficiencies of PC and NaI(Tl) detectors for γ-rays, is the efficiency for registration of Compton rays and γ- γ coincidences When an efficiency extrapolation procedure is applied, the equations (1) reduce at: 1 1

1 (2)

Where

,

Expression (2) becomes ( 1 when the PC efficiency

1 .

The radioactive solution for the key comparison was prepared and distributed by the pilot laboratory, PTB - Germany, in the form of MnCl2 in a solution of 0.1 N HCl, in a PTB type ampoule, code: 2003-1355, having a mass of M = (2.0289±0.0001) g of solution. On its receipt in the laboratory, the ampoule was introduced in our secondary activity standard, CENTRONIC IG12/20A ionisation chamber, and the ionisation current produced by the 834.850 keV γ- rays was registered. After opening the ampoule, a number of 9 solid sources were prepared on VYNS foils and left to dry in free air. Due to the low energy of K- Auger electrons and x-rays (X,A)


190

the flow gas PC counting efficiency, εX,A , is expected to be low; in our case, the maximum value, εX,A = 0.11. Our idea was to measure the sources normally prepared, to obtain a number of extrapolation points, and then to re-dissolve them with LUDOX 10-4, re-crystallize, and re-measure them, obtaining other efficiency values, for experimental points. The maximum efficiency raised at εX,A = 0.15 (increment 36%). The radionuclide impurity content was verified using our high resolution gamma-ray spectrometric system, based on a HPGe detector. No γ- ray emitting impurity, above the detection limit, 0.01% relative activity, was detected. 3 RESULTS AND DISCUSSIONS 3.1 Extrapolation curves The principle of extrapolation was applied for all measured sources, obtaining linear extrapolations. An example of extrapolation expressions, Eq.(2), is presented in Table 1 for source code D121 in conditions (a) of initially prepared source and (b) for re dissolved source; extrapolation intervals and total number of measurement points, used to obtain linear extrapolation curves, are also presented.

Table 1. Source D121 extrapolation curve

Mode of measurement

1 X,A/ X,A variation interval /

number of points

Extrapolation curve: ""γ/"c N0 1 1 ;

(a) Initially

prepared source

(8.3-73.4)/5 13456 (30) [1+0.00199(10)E]

(b) Re dissolved source

(5.8-73.4)/8 13439(23) [1+0.00225(8)E]

From Table 1 one can deduce that by re dissolving the sources (b) the efficiency was substantially improved, the extrapolation was done on a shorter interval and the statistical uncertainty of parameters was lowered. The slope of the linear extrapolation curve is very low, demonstrating that the γ- ray energy interval was optimally chosen. 3.2 Standardization results

The radioactive concentration of solution, a, was calculated as the arithmetic mean of the individual results obtained after the extrapolation of the nine sources, as

$ "$ %$& , where %$ is the mass of each source, mg. The uncertainty budget

(relative units), and its calculation manner, is presented in Table 2.


191

Table 2. Uncertainty budget of 54Mn radioactive concentration

Uncertainty component Value,% Remarks Counting statistics 0.11 From the mean of the 9

measured sources, Type A evaluation;

Weighing 0.05 Previous experience Dead time 0.035 Uncertainty propagation Background 0.01-beta;

0.025-gamma Uncertainty propagation

Resolving time 0.030 Uncertainty propagation Adsorption 0.01 From previous experience Impurities 0.01 Detection limit Decay scheme parameters 0.13 Uncertainty propagation Extrapolation of efficiency

curve 0.19 Uncertainty of slope, 8.7%,

multiplied by mean correction 0.023

Combined standard uncertainty

0.27 Calculated as quadratic sum of components

3.3 Results reported to the key comparison CCRI(II )-K2.Mn-54 organizers and

agreement with the Key Comparison Reference Value (KCRV)

The reported activity concentration was: a = (312.69 ± 0.84) kBq g-1 on the comparison reference time, for a coverage factor k=1. If one compares this value with the calculated Key Comparison Reference Value (KCRV), reported in [8], using the Power Moderate Mean (PMM) calculation method, as reported in [9], our result has an equivalence degree at international scale, expressed as: Difference from CRV, Di = - 0.40 kBq g-1 (- 0.13%); Expanded uncertainty of difference (k = 2), Ui = ± 0.98 kBq g-1 (±0.31%). These values demonstrate a very high degree of equivalence of our result with the CRV. 3.4 Results of the comparison with ionisation cham ber measurements.

The calibration factor of the CENTRONIC IG12/20A ionisation chamber, expressed in terms of ratio between the registered ionisation current I, pA, and the calculated activity of solution in ampoule, A, in MBq, was determined according to the formula: ' ( )& (3) Activity A was calculated as: ) *+ (4) Where a is the activity concentration determined by the absolute method, Section 3.3, and M is the mass of solution. The results were compared with those determined at PTB, using their standard solution, and also with the CRV value of concentration, as reported in the paper [8]. The agreement degree was expressed in terms of En numbers, calculated according to the formula:


192

,=-$

./ 012 3022 (5)

Table 3. Presents the results obtained in the two ionisation chamber comparisons Determination mode of F value

F value, pA MBq-1 En numbers

Calibration using our reported concentration, a

25.58 ± 0.09 (0.35%) -

Calibration at PTB , [9] 25.80 ± 0.13 (0.50%) PTB – IFIN-HH: 0.69

Calibration using CRV of concentration, a

25.52 ± 0.04 (0.16%) CRV - IFIN-HH: 0.34

In both situations, although very small uncertainties were reported, the Deviations Normalized with Respect to the Stated Uncertainty (EAL-P7 Z Eal Interlaboratory Comparisons, February 2002), En numbers, En <1.00, reflecting the very good agreement of the three results. 4 CONCLUSIONS - A 54Mn solution was standardized absolutely by the, 4πPC-γ coincidence method, within a key comparison, codified as CCRI(II).K2.Mn-54. - In spite of its simple decay scheme, due to low energy of its emitted Auger electrons and x-rays, the counting efficiency is expected to be small. By using a treatment of the solid sources, consisting in their re dissolving, the efficiency was improved and a very small combined standard uncertainty of activity concentration, only 0.27%, was obtained. - The results obtained in two international comparisons were very satisfactory: a difference of only (-0.13%) from the CRV of the key comparison in absolute standardization and small values of En numbers, 0.69 and 0.34 respectively in interlaboratory comparisons with PTB and BIPM, in ionisation chamber measurements were obtained. ACKNOWLEDGEMENT The authors express many thanks to their colleagues, Mrs. Cristina Watjen and Dr. E.L. Grigorescu, for their contribution at the measurements.

The paper was written and presented in the Conference with the support of the European Metrology Programme for Innovation and Research (EMPIR), JRP-Contract 16ENV10 MetroRADON (www.euramet.org).The EMPIR initiative is co-funded by the European Union’s Horizon 2020 research and innovation progamme and the EMPIR Participating States.

The support of Program Nucleu – PN-IFIN-HH, project code 19 09 06 02 04 is also acknowledged.


193

REFERENCES

[1]

[2] [3]

[4]

[5]

[6]

[7] [8]

[9]

Günther, E. “Standardization of the EC Nuclides 55Fe and 65Zn with the CIEMAT/NIST LSC Tracer Method”. Appl. Radiat. Isot. 49, 9-11 (1998)1055 – 1060 Schrader, H., Luca, A., Druker, Al. “Internal PTB and IFIN-HH Reports”, 2001 Sahagia, M., Watjen, A.C., Luca, A., Ivan, C. “IFIN-HH ionization chamber calibration and its validation; electrometric system improvement “Appl. Radiat. Isot. 68 (2010)1266-1269 Havelka, M., Auerbach, P., Sochorova´, J. ”Standardisation of 54Mn and 65Zn using a software coincidence counting system”. Appl. Rdiat. Isot. 64 (2006) 1215-1219 Rodriguez, L., Los Arcos, J.M., Grau, A. “LSC standardization of 54Mn in inorganic and organic samples by the CIEMAT/NIST efficiency tracing method”. Nuclear. Instr. Methods in Phys. Res. A312 (1992) 124-131 Simpson, B.R.S., Morris, W.M. “Direct activity determination of 54Mn and 65Zn by a non-extrapolation liquid scintillation method”. Appl. Radiat. Isot. 60 (2004) 475-479 Chechev, V.P., Kuzmenco, N.C. “LNE-LNHB, Table de Radionucléides, http://www.nucleide.org/DDEP_WG/Nuclides/Mn-54_tables.pdf (2014) Ratel, G., Michotte, C. Bureau International des Poids et Mesures.“Final report of the CCRI(II)-K2.Mn-54 comparison”, Key Comparison Database of the CIPM MRA, https://kcdb.bipm.org/AppendixB/appbresults/CCRI(II)-K2.Mn-54/CCRI(II)-K2.Mn-54.pdf Pommé, S. (2015) Determination of a reference value and its uncertainty through a power-moderated mean. Metrologia 52(2015) S200-S212

The Seventh



194

COMPARISON OF PT PERFORMANCE IN RADIOACTIVITY MEASUREMENT LABORATORIES USING SCINTILLATION

AND SEMICONDUCTOR DETECTORS

Jarosl aw Rachubik

National Veterinary Research Institute, Department of Radiobiology, Partyzantow 57, 24-100 Pulawy, Poland

[email protected] Abstract The successful performance in PTs is one of the most important elements in obtaining accreditation. Radioactivity measurement laboratories usually use scintillation and semiconductor detectors for detection of gamma-emitting radioisotopes, mainly 137Cs. The aim of the present study was to compare the PT performance of laboratories using these two types of measurement devices. IAEA food and feed reference materials (RMs) were chosen as test objects. They are homogenous and stable for a long time, thus a sequential PT model with circulating items could be adopted. An evaluation method, elaborated by IAEA, was used in the assessment of laboratory results [1]. This approach takes into account the uncertainty of measurements estimated by the participants. There were no differences in the performance of laboratories using scintillation and semiconductor detectors. Based on the PT results, it can be concluded that both types of detectors are suitable for determining 137Cs in routinely tested samples. Key words PT, radiocaesium, food, feed REFERENCES

[1] Shakhashiro A., Fajgelj A., Sansone U. Comparison of different approaches to evaluate proficiency test data. In: Combining and Reporting Analytical Results. RSC Publishing, Cambridge 2006, pp. 220–228.

The Seventh



195

DEPARTMENT OF CHEMISTRY MALAYSIA PROFICIENCY TESTIN G

PROVIDER (JKM PTP): CURRENT PRACTICES AND FUTURE CHALLENGES

Hui Ling Li , Khairul Anuar Abdul Aziz

Ms, Jalan Sultan, Petaling Jaya, Malaysia, Code 46661

[email protected]

Abstract Department of chemistry Malaysia aka Jabatan Kimia Malaysia (JKM) is the first organization in Malaysia to receive MS ISO/IEC 17043 Malaysia Proficiency Testing Provider (MyPTP) accreditation scheme, from Department of Standards Malaysia. A total of thirty (35) Proficiency Testing (PT) schemes from twelve (12) areas are accredited under the field of Chemical, Biology, Forensic Science and Halal Testing. In Malaysia, the main purpose to join the PT program is to ensure that the laboratory quality system meets the requirements of MS ISO/IEC 17025 SAMM accreditation scheme. Thus, JKM PTP is supporting most of the government and private chemical analysis laboratory. JKM PTP program is divided into two schemes, i.e. Forensic scheme and Kimia scheme (Chemical and Biology). Forensic scheme, includes analysis in the area of narcotic, toxicology, criminalistics, DNA and document examination, are mostly qualitative scheme whereas Kimia scheme, includes analysis in the area of food, water, environment, industry, GMO, microbiology and Halal testing, are mostly quantitative scheme. For the qualitative scheme, 100% matching with the standard profile is required for satisfactory results. For the quantitative scheme, to achieve the minimum requirement of the MS ISO/IEC 17043, the assigned value is taken as the consensus results of the participants in a PT scheme. Using consensus result as the assigned value is the most economical practice now. However, the consensus value was not necessarily identical with the true value. That is the most challenging task for JKM PTP to determine the assigned value. Technical aspects such as a lack of certified reference materials and justification of expert laboratory in Malaysia are the limitation for JKM PTP establish of assigned value. Practical considerations such as the various matrices and the logistics of the distribution for the cold condition samples to the international country. Key words Proficiency Testing, Assigned value

The Seventh



196

INTER-LABORATORY COMPARISON TEST FOR

ELECTROCHEMICAL ANALYTICAL METHODS IN REAL WASTEWATER SAMPLE: Z-SCORE MODEL AND PARTICIPANTS

EFFECT ABSTRACT

Hulusi Demircio ğlu1, İlgi Karapınar 2, 1Dokuz Eylul University, Institute of Marine Sciences and Technology,

Narlidere- Izmir 2Dokuz Eylul University, Department of Environmental Engineering,

Haydar Aliyev Bul. No:32, Izmir/Balçova-Inciralti, Code 35340, Turkey

[email protected]

Abstract The study was conducted to evaluate the z-scores of accredited and non accredited laboratories for the electrochemical analysis namely pH, conductivity and salinity through an inter-laboratory comparison test carried out on real wastewater samples. The participants were small scale 16 non-accredited and 9 accredited wastewater analysis laboratories in Turkey. The samples from a wastewater treatment plant were collected from influent, effluent point and these samples were mixed to obtain three different levels of these measurements namely high, low and medium, respectively. pH, conductivity or salinity are the measurements based on electrochemical analytical methods in which the influence of the other factors such as technical staff ability on the measurement is limited. However, the method used in z-score calculation could have a substantial effect on the z-score performance of the laboratory. Therefore, four different Z-score calculation approaches using SDPA and different assigned value were applied to the inter-laboratory comparison test results. The assigned value was determined through average, and median of accredited laboratories results and by robust analysis. The performances of the non-accredited ones compared with respect to accredited laboratories. The results indicated that the method used in z-score calculation does not substantially affect the scores of the laboratory. No population quality effect on the z-score exists. The laboratory with low performance will not pass the comparison test whatever the z-score calculation method. Key words: z-score, inter-laboratory comparison test, pH, conductivity, salinity, wastewater. Key words z-score, inter-laboratory comparison, wastewater

The Seventh



197

DEVELOPMENT OF PROFICIENCY TESTING PROGRAM IN

SOFTWARE SAFETY VERIFICATION THROUGH NUMERICAL EQUIVALENCE TEST

Javier Luiso , Mariano Palacios,

ILT - Interlaboratory Test S.A., Spiro 918 2ºA, Adrogue, Argentina, Code1846


Abstract The core objective of equivalence testing is to validate the model-to-code translation process by demonstrating numerical equivalence between the model and the generated code. A valid model-to-code translation requires that the execution of the object code exhibit the same observable effects as the simulation of the model for any given set of test vectors. In this proficiency test program, equivalence testing can be carried out by stimulating both the model used for production code generation and the executable derived from the generated code with identical test vectors. Testing for numerical equivalence is unique in that the expected outputs for the test vectors do not have to be provided. The software under testing is a module software which computes the Optical Flow Field from a sequence of RGB images acquires from a digital camera. This module is a component of a Adaptive Cruise Control System (ACC). ILT-Interlaboratory Test S.A has develop a PT Program (ILT-U-1749) on software safety validation of a software module called Optical Flow Module (OFM) according to IEC 61508-3 and ISO 26262. ILT has selected one theoretical model of OFM which has been translated from model to code for a specific hardware Compute Module (CM). The participants receive the Optical Flow Module theoretical model, the Compute Module programed with the OFM implementation, and a set of Test Vectors. A Numerical Equivalence Test has to be performed in order to demonstrate numerical equivalence between the model provided and the generated code which has been programed in the CM provided. The results obtained by the participating laboratory is evaluated as being consistent or inconsistent with respect to the reference value. Key words Proficiency Testing, Software Verification, Automotive Industry, Numerical Equivalence Test

The Seventh



198

DEVELOPMENT OF PROFICIENCY TESTING PROGRAM IN

RANDOM NUMBER GENERATOR VERIFICATION ACCORDING TO REMOTE GAMBLING AND SOFTWARE TECHNICAL STANDARD

Javier Luiso , Mariano Palacios,



Abstract To ensure that games and other virtual events operate fairly. Random number generation and game results must be “acceptably random”. Acceptably random here means that it is possible to demonstrate to a high degree of confidence that the output of the RNG, game, lottery and virtual event outcomes are random through, for example, statistical analysis using generally accepted tests and methods of analysis. Adaptive behaviour (ie a compensated game) is not permitted. ILT-Interlaboratory Test S.A. has develop a PT Program (ILT-U-1746) on the Random Number Generator (RNG) verification, according to Remote Gambling and Software Technical Standards (June 2017), RTS 7 – Generation of random outcomes. The participants receive a set of six software implemented RNGs is provided to each participant to be verified by statistical annalists of its number sequences output. The objective of the program is verify that RNG compliance a number of qualities enumerated in RTS 7. The results obtained by the participating laboratories is evaluated as being consistent or inconsistent with respect to the reference values. Key words Proficiency Testing, Software Verification, Gambling, RNG

The Seventh



199

NEW APPROACH PT-PROGRAMS FOR SOFTWARE TESTING

LABORATORIES IN INNOVATIVE TECHNOLOGIES

Mariano Palacios, Javier Luiso



Abstract Artificial intelligence is currently attracting considerable interest and attention from industry, researchers, governments as well as many other areas. Increasingly sophisticated algorithms are being employed to support human activity. The increasing use of machines to help humans make adequate decisions is also generating a number of risks and threats that businesses, governments and policy makers need to understand and tackle carefully. Safety, Security, Privacy, Trust, are all aspects of increasing interest and we are going to see the emerging of many standards in the next years. ILT offer a new way to learn and develop new skills that will allow any laboratory to know and train many types of test that related standards requires for software products in different areas such as medicine devices, automotive industry. In Artificial Intelligence area, ILT has develop the ILT-U-1745 Proficiency Test Program to characterize the AI-ML system performance measuring the F1Score. This kind of program give to Laboratories the possibility to get into AI through Test Programs which will focused on AI aspects that it is well understood today that will have a crucial role in safety and the behavior of AI systems. Key words Proficiency Testing, Artificial Intelligence, Machine Learning Software Verification

INDEX OF AUTHORS

A

Marleen Abdel Massih 32 Fahad S.Aldawsari 159 Adnan S. Al-Mussallam 135; 159 Majda Alenezi 135 Ahmed I. Al-Ghusn 135 Mohammed A. Almutairi 127 Mohammed Z. Alrabeh 127 Markus Arndt 143 Mostafa A. Alsamti 127 Nawaf F. Alsowidan 135; 159 Azziz Assoumani 49 Jean-Christophe Augustin 24 Khairul Anuar Abdul Aziz 195 B

José Ricardo Bardellini da Silva

116

Abdullah T. Bawazir 135 Ricardo J.N. Bettencourt da Silva 165 Hugues Biaudet 49 Danica Boljević 41 Abdelkader Boubetra 10; 49; 57; 62 Mirella Maria Buzoianu 1 C

Vincent Carlier 24 Gherghina Ciortan 173 Irina-Mihaela Ciortan 151 Eva Cizmarova 107 Carla Thereza Coelho 116 Bertrand Colson 16 Boris Constantin 49 Thibaut Cugnon 111

D Marie Dangerville 57 Hulusi Demircioğlu 196 F Florence Ferber

32; 111

Paulo Roberto da Fonseca Santos 116 G

Zhong-bao Guo 79 Virginia-Graziela Guslicov 71 H

Ahmad Hanafi 135 K

Ilgi Karapinar 196 Peter Kruspan 143 L

Richard Lambert 111 Adelcio Rena Lemos 116 Romain Le Neve 57 Hui Ling Li 195 Ruo-lin Li 79; 87; 123 Hui-chao Liang 79; 87; 94; 123 Klaus Lipus 143 Aurelian Luca 187 Javier Luiso 197; 198; 199 M

Jacques Mahillon 32; 111 Milenko Maričić 41 Nikola Maričić 41 Marina Martin 173 Caterina Mazzoni 62 Vanessa Morgado 165

INDEX OF AUTHORS

N Sandrine Nguyen 57 P

Mariano Palacios 197; 198; 199 Carla Palma 165 Elena Pitchugina 32 Viviane Planchon R

32

Jaroslaw Rachubik 194 Eloise Riouall 24 Francois Roussey S

62

Maria Sahagia 187 Cristina Stancu 99 T

Anne Tirard 10; 49; 57; 62 U

Steffen Uhlig 16 V Martin Vana 107 W

Shasha Wu 79; 87; 94; 123 X

Ying Xu 79; 87; 94; 123 Y

Ke-xiao Yu 79; 87; 94; 123

Z Jian-bo Zhou 79; 87; 123 Xiaoling Zhu 79; 94; 123 Eric Ziegler 10; 49; 62

Date post:	15-Jul-2020
Category:	Documents
Upload:	others
View:	1 times
Download:	0 times

PROCEEDINGS - PT-CONF · Romania (INM) is mainly responsible for developing, maintaining and...

Documents