1Engineering Research Center of Urban Transportation Beijing University of Technology No 100 Pin Le YuanChaoyang District Beijing 100124 China2College of Metropolitan Transportation Beijing University of Technology No 100 Pin Le Yuan Chaoyang DistrictBeijing 100124 China3College of Engineering and Applied Science University of Cincinnati Cincinnati OH 45221 USA
Correspondence should be addressed to Zhiqing Zhang zhangzhiqingbjuteducn
Received 14 April 2017 Revised 9 July 2017 Accepted 12 September 2017 Published 23 October 2017
Academic Editor Hainian Wang
Copyright copy 2017 Zhiqing Zhang et alThis is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited
With the same sources and regeneration techniques given RArsquos properties may display large variations The same single propertyindex of different sets maybe has a large difference of the whole property How shall we accurately evaluate the whole property ofRA 8 groups of RAs from pavement and building were used to research the method of evaluating the holistic characteristics of RAAfter testing and investigating the parameters of aggregates were analyzed The data of physical and mechanical properties show adistinct dispersion and instability thus it has been difficult to express the whole characteristics in any single property parameterThe Euclidean distance can express the similarity of samples The closer the distance the more similar the property The standardvariance of the whole property Euclidean distances for two types of RA is 119878119896 = 7341 and 119878119896 = 2208 respectively which showsthat the property of building RA has great fluctuation while pavement RA is more stable There are certain correlations amongthe apparent density water absorption and crushed value of RAs and the Mahalanobis distance method can directly evaluate thewhole property by using its parameters mean variance and covariance and it can provide a grade evaluation model for RAs
1 Introduction
Accurate evaluation of RA is beneficial to its rational uti-lization The current problems in evaluating the variabilityof RA are focused on the application of RA mixture onlyregardless of the study and conclusion [1] How shall weaccurately express the whole property of RA and how shallwe effectively evaluate the grade of RA So far there isstill no reasonable method to be seen and most of theevaluation methods are the evaluation of mixtures withRAs
The single properties of aggregates are usually describedby the apparent relative density water absorption crushedvalue wear value the adhesion index while the comprehen-sive property is the whole property of aggregate showed inapplications The single property of different RAs shows alarge variability so the comprehensive evaluation method ofeach kind of RA needs to be studied
Over the past decades the United States Europe andother developed countries have promulgated some qualitystandards and technical specifications of RA according tothe RA characteristics of their own countries Although eachof them has their own characteristic it is still insufficientfor the evaluation method The main problem is that theyfocused more on the single property index and neglected thewhole property evaluation for the RA [2ndash4] For examplethe American Society for Testing and Materials (ASTM)made clear provisions for the extremum of the evaluationindex of coarse aggregate in different engineering fieldsThe British Standard (BS) divided RA into three gradesbased on different sources [3 5] V W Y Tam and C MTam further put forward the RA should be divided into 7grades in view of the evaluation indices of apparent relativedensity water absorption flakiness content impact value thecontents of chloride and sulfate in 2007 [6] The commonproblem is that the grading standard is too absolute and does
HindawiAdvances in Materials Science and EngineeringVolume 2017 Article ID 1048172 9 pageshttpsdoiorg10115520171048172
2 Advances in Materials Science and Engineering
Table 1 Properties of RA and NA
Type of aggregate Density (gcm3) Water absorption () Crushing value () Abrasion value () Adhesion gradeRA
Commonly acceptable ge2600 le2000 le2800 le3000 ge4e analysis data of RA properties
e single property index ofRA is discrete
Express the dierence of the wholeproperty by distance method
Whether the correlation of property index is considered
Euclidean distance Mahalanobis distance
No Yes
Figure 1 RA property evaluation flow chart
not give the whole characteristics requirement in differentapplication fields In 1998 Germany promulgated ldquoRegen-eration Aggregate Concrete Application Guiderdquo where thequality requirements of RA must comply with the technicalrequirements of natural aggregate (NA) [7 8] It ignoresthe characteristics of RA and it is not conducive to theapplication of concrete aggregate Although Spain and othercountries have not given a clear classification and gradingstandard of the RA only relevant qualificationwas performedwith recycled coarse aggregate [9 10] The standard did notinclude many details The common features are that thesespecifications only gave a different range of values for a singleproperty index of the RA and did not make provisions for thewhole properties
Because of the complexity of RA sources and regenera-tion techniques an RArsquos properties show large variations inits microstructure and physical mechanical and chemicalproperties [1] The test values show a large variability thesame single property index of different sets maybe has a largedifference of the whole property so the whole property ofRA could not be only expressed by a single property indexAdopting the Euclidean distance method and Mahalanobisdistance method respectively in evaluating the single andwhole property of RA is proposed in this paperTheflow chartis demonstrated in Figure 1
2 Test and Analysis of the Characteristic ofRA Used
21 Aggregate Characteristic Test Two types of aggregateswere used in the present study (1) RA produced from the old
building foundation concrete on the twentieth century 1980sIts strength was equivalent to C20 and the coarse aggregatewas comprised of pebbles The production process stages ofRA were pretreatment crushing grinding screening andwashing (2) NA is made from granite
The RAs used in the present study were screenedinto three different sizes (475ndash950mm 475ndash132mm and950ndash190mm) with the same processing method After sam-pling the aggregates were sievedwith 475mm standard sieveand then rinsed with water to ensure that they are free of dirtand dust (see Figure 2) [11]
The properties of the RA and NA such as apparent rela-tive density water absorption crushed value wear value andthe adhesion were tested according to the Chinese standardTable 1 presents the results from the above-described tests
As indicated in Table 1 the properties of NA testedsatisfy all commonly acceptance while RAs had higher waterabsorption and lower adhesion to asphalt values than thelisted commonly accepted values
22 Aggregate Characteristic Analysis As shown in Table 1the apparent relative densities of RA with different particlesizes are similar and they are all slightly lower than those ofNA In addition the water absorption of RA is much higherthan that of NA for the same size of aggregate These may bemainly attributed to the high porosity or low density of themortar adhered on the surface of the aggregate In additionduring the production process of RA mechanical crushingparticle collision and sliding also producemicrocracks in theaggregate and these will result in an increase in the aggregateporosity and a decrease in the apparent relative density Theporous mortar and fine cracks are also responsible for thehigh water absorption of the RAs Furthermore the high siltcontent on the surface of the RA particles may also help toincrease the rate and amount of water absorption
The crush values of RAs are clearly lower than those ofNAAnd the degree of abrasion of RAs is all lower than that ofNA Amain reason for these is that the RAs had already gonethrough a crushing process during recycling After crushingwashing and sieving the RAs had high pebble content withgood integrity As a result the RAs used in this study had ahigher crushing and abrasion resistance than the NA Thissuggests that the lower the strength of the original concretethe smaller the damage degree in the process of the aggregateregeneration and the better the property of resistance tocrushing of the recycled aggregate
Advances in Materials Science and Engineering 3
(a) RA as received (b) RA after cleaning and sieving
Figure 2 RA before and after washing
(a) NA 30-time SEM picture (b) RA 30-time SEM picture
(c) NA 100-time SEM picture (d) RA 100-time SEM picture
Figure 3 RANA SEM pictures
Adhesion of aggregate to asphalt is critical for the prop-erty of asphalt concrete pavements Table 1 shows that amongthree RA samples tested two had adhesion level values lessthan 4 as required It was observed during the adhesiontesting that the general peeling area was around 10 andlocalized peeling area summed up to 30Theweak adhesionof RA may be related to the reaction of water soluble ionsleaching out from the adhered mortar on the aggregatesurface with the carboxylic acids of asphalts
A scanning electron microscope (SEM) was used toobserve the microstructure of RA and NA as shown inFigure 3
It can be seen that RA possesses a large number ofpores and microcracks internally caused by the accumulateddamage during crushing leading to an increase in the voidratio water absorption and crush value (shown in Figure 3)Compared to the surface structure ofNA the asphalt bondingmaterial can penetrate into the RA surface more easilyFurther the strength of interface may be improved
In summary for the RA from low strength concreteexcept for the water absorption rate the other propertiesof RA appear better than that of NA Perhaps this is aspecial case more extensive investigation may need to bedone
4 Advances in Materials Science and Engineering
Table 2 Property data of RA from buildings
Property index 119860 119861 119862 119863 119864 119865 119866 119867Density (gcm3) 2742 2529 25 2535 275 232 2265 2658Absorption () 194 482 215 450 249 989 570 450Flakiness () 698 370 250 340 300 250 310 260Crushing value () 1360 1380 1410 787 620 2420 3000 1300Stone species Pebbles Pebbles Pebbles Pebbles Gravel Gravel Gravel GravelStrength Low Low Low Low Low Low Low LowSources Foundation Floor Building Beam Foundation Building Building Building
31 Property Data of RA Tested and Investigated RA prop-erties can be evaluated by some experimental indices whichinclude property density water absorption needle and flaki-ness content and crushing value [12ndash14]
Sixteen groups of RAs were selected from pavements andbuildings respectively [15 16]Most of the property data wereinvestigated except for groups 119864 119865 and NAs which weretested by ourselves in the laboratory All of the RAs werecrushed by the jaw-crush method (see Tables 2 and 3)
As shown in Tables 2 and 3 the basic property data ofthe NA are generally property density at 263 gcm3 waterabsorption at 16 needle and flakiness content at 52 andcrushing value at 96
32 Variability Analysis of the NA Property Data As canbe seen from Tables 1 and 2 there is a little distinction inapparent relative density between RA and NA Most of theRArsquos apparent relative density is 88sim97 of the NAs about231ndash261 kgm3The shapes of RA fromdifferent sources havea high degree of similarity some of them are even better thanthe NA The shapes of different sources of RA are similarand for some the quality is better than the NA the apparentrelative density and needle-flakiness content are relativelystable while the water absorption and crushing values arerelatively higher The water absorption rate of RA is between2sim10 the highest value even is up to 12 which is 2 to 7times that of the NAThe crushing value of RA varies greatlyand its crush resistance ability is significantly lower than theNA as is shown in Figure 4
After being crushed the RA might produce a largenumber of microcracks on the aggregate surface and theinterior between aggregate and adhesive cement mortar
Besides the RA would be cracked more easily when beingpressed and it contains a variety of impurities on the surfacetherefore the RA strength would be lower than NA As aresult the water absorption and crushing value of RA weresignificantly larger [7 17] and the experimental data showeda large variability as well Since the content of impurities andmicrocrack is influenced by many factors such as the typesof concrete waste engineering construction and crushingmethod and the RA from building contains more impuritiesthan that from pavement its variability would be greater
33 Pavement Property with RA The conclusions of ourprevious study about RArsquos instability are as follows [11]
(1) Compared with NA the RA recycled from low gradeconcrete had lower apparent relative density higher waterabsorption and poorer adhesion to asphalt but had probablylower crushing and wearing value
(2) The rutting deformation of the HMAmixes generallyincreased with the level of RA replacement but the lowtemperature cracking resistance test results showed that thefailure load tensile stress and modulus of the HMA mixeswith RA are all significantly higher than those of themix withNA
(3) The water stability measured by immersion compres-sion tests showed a similar trend to that measured by freeze-thaw splitting strength tests The water stability of the HMAconcrete decreased with increasing RA replacement (up to50) When the RA content reached 75 the water resis-tance of the HMA concrete increased rather than decreasedfurther
Due to instability a large variability has been indicatedby experimental data of RA property index Therefore it isinaccurate to illustrate the whole property through a singleproperty index It is possible that the whole property is better
Advances in Materials Science and Engineering 5
22
23
24
25
26
27
28
aA bB cC dD eE fF gG hH
RA sample number
Building RAPavement RANA
Perfo
rman
ce d
ensit
y (g
cＧ
3)
(a) Contrast chart of density of NA and RA
0
2
4
6
8
10
aA bB cC dD eE fF gG hH
Abso
rptio
n (
)
RA sample number
Building RAPavement RANA(b) Contrast chart of absorption of NA and RA
Building RAPavement RANA
2
25
3
35
4
45
5
55
6
65
7
aA bB cC dD eE fF gG hH
Flak
ines
s con
tent
()
RA sample number
(c) Contrast chart of flakiness of NA and RA
4
8
12
16
20
24
28
32
aA bB cC dD eE fF gG hH
Crus
hing
val
ue (
)
RA sample number
Building RAPavement RANA
(d) Contrast chart of crushing value of NA and RA
Figure 4 Contrast charts of NA and RA properties
in the asphalt mixture while a single property index exhibitspoor results However the distance method is not affectedby the dimension and directly expresses the difference ofmultiple variables or sample set of indices According to thedistance method it is feasible to illustrate the whole propertyof RA
4 Euclidean Distance ComparisonMethod of RA
41 Euclidean Distance Calculation of RA Property (1) Thesingle property Euclidean distance 120588119896119895 is generally
(119909119895 minus 119910119896119894119895)2 (1)
in which 119909119895 is property test value of the NA 119910119896119894119895 is propertytest values of RA 119896 is pavement RAor building RAwhere 119896 =I (pavement RA) and 119896 = II (building RA) 119894 is the numberof test data groups when 119896 = I 119894 = 119860 119861 119862119863 119864 119865 119866119867119896 = II 119894 = 119886 119887 119888 119889 119890 119891 119892 ℎ 119895 is the different property test119895 = 1 (apparent density) 2 (water absorption) 3 (needle andflakiness content) and 4 (crushing value)
(2) The whole property Euclidean distance of RA 120588119896119894 isgenerally
120588119896119894 = radic 4sum119895=1
(119909119895 minus 119910119896119894119895)2 (2)
in which 119909119895 119910119896119894119895 119896 119894 and 119895 are the same as formulation(1)
6 Advances in Materials Science and Engineering
Table 4 Euclidean distance of single property of building and pavement RA
Property index Building RA (119860) Pavement RA (119861) 119860 minus 119861Property density 0516 0241 0275Absorption () 10669 8595 2074Flakiness content () 187 0943 0927Crushing value () 26355 1072 15635
052
1067
187
2636
024
860
094
1072
0
10
20
30
Performance Absorption Flakiness Crushing value
Eucli
dean
dist
ance
Single performances of building and pavement RA
Building RAPavement RA
density content
Figure 5 Single property of RA
(3) The whole property Euclidean distance standardvariance of RA 119878119896 is generally
119878119896 = radicsum119866119892119894=119860119886 (120588119896119894 minus 120588119896119894)27 (3)
in which type 119878119896 is the standard variance of whole propertyEuclidean distance for building RA and pavement RA 120588119896119894is the arithmetic mean value of whole property Euclideandistance for the different number of RA 120588119896119894 119896 and 119894 are thesame as formulation (1)
42 The Comparison and Analysis of Euclidean Distance ofRA Property According to the Euclidean distance formula(1)ndash(3) for the two types of RA the Euclidean distances ofsingle property andwhole property are shown in Tables 4 and5 and Figures 5 and 6 respectively
The greater the value of the properties the greater thedifference between the RA and theNA otherwise the smallerthe value the closer the difference The standard varianceof the whole property Euclidean distance of building RAand pavement RA was calculated to be 7341 and 2208respectively
Table 4 and Figure 5 show slightly differences of theproperty density and flakiness content of RA compared withthe NA while the crushing value and water absorption areobvious differences In addition pavement RA was signifi-cantly better than building RA in terms of single propertyindexAs shown inTable 5 andFigure 6 thewhole property of
2051
1590
450 440 438 388
091 020
789
506 400 372
247 199 150 130
02468
10121416182022
G F h e C A B f H g b c a d D EEu
clide
an d
istan
ce
Whole performance of building and pavement RA
Building RAPavement RA
Figure 6 Whole property of RA
RA with better single property may be poor while the wholeproperty of RA with poor property may be better
From the size of Euclidean distance it can be seen that thecomprehensive properties of group 119864 group 119863 and group 119889are the best whereas group 119866 shows the worst results TheEuclidean distance of crushing value and water absorptionis the largest which shows a primary need to improve theproperty index of the RA
The standard variances of the whole property Euclideandistances are 7341 and 2208 respectively which shows thatthe property of building RA has great fluctuation whilepavement RA is more stable
5 Mahalanobis Distance Evaluation andGrading Methods of RA
The analysis above indicates that the property evaluation ofRA contains multiple indicators Due to the instability of theproperties of RA its experimental data has a large variabilityFor different evaluation indices different evaluation resultsare achieved Since there exists certain dependence amongproperties of RA andEuclidean distance could not express theinfluence factors of the correlation among various propertiesof the samples another index is proposed to solve thisproblem named Mahalanobis distance
51 Mahalanobis Distance Evaluation Model Mahalanobisdistance can discriminate the spatial distribution of a pointrelative to the total by calculating the distance to a certainpoint Mahalanobis distance considers three parameters
Advances in Materials Science and Engineering 7
Table 5 Euclidean distance of whole property of building and pavement RA
mean value variance and covariance which shows thetranslational invariance and invariance under nonsingulartransformation Mahalanobis distance can express the influ-encing factors of the correlation among different variablesand it is independent of the dimension of each characteristicparameter and the judgement basis is according to thecharacteristic distribution condition of variables in the wholespace [18 19]
Suppose that there are two variables 1198661 and 1198662 whichobey the state distribution 119909 isin 119877119901 is a new sample point andthe distance from 119909 to 1198661 and 1198662 is defined as follows
119889 (119909 1198661) = radic(119909 minus 119906(1)Σminus11 ) (119909 minus 119906(1)) 119889 (119909 1198662) = radic(119909 minus 119906(2)Σminus12 ) (119909 minus 119906(2))
(4)
in which 120583(1) 120583(2) are the mean vector of 1198661 and 1198662 Σ1 Σ2are covariance matrix of set 1198661 and 1198662
Therefore it can be determined according to the followingrules
The indices which can express the property of RA aredefined as a vector 119883 (11988311198832 119883119899) the mean vector of119883 is 120583 (1205831 1205832 120583119899) and the standard deviation vector is 120590(1205901 1205902 120590119899) The property standard is defined as vector 119884(1198841 1198842 119884119899) so the vector 119884 is degenerated into a pointof the multidimensional space
The global 1198661 is defined by the vector 119883 and the global1198662 as 119884 then the distance from a sample point of global 1198661to global 1198662 is equal to the distance from a point of vector119883to vector 119884119863 (119883 119884)= radic (1198841 minus 1198831)212059021 + (1198842 minus 1198832)212059022 + sdot sdot sdot + (119884119899 minus 119883119899)21205902119899 (6)
119863 is a normalized distance of nonuniform distributionand it can reflect the comprehensive index of RA propertywithout the need to consider the dimension of each index and
Table 6 Grading methods of RA
Grading of RA Meet the conditionsGrading 1 119863(11990110158401119901) lt 119863(11990110158402119901) cup 119863(11990110158401119901) lt 119863(11990110158403119901)Grading 2 119863(11990110158402119901) lt 119863(11990110158401119901) cup 119863(11990110158402119901) lt 119863(11990110158403119901)Grading 3 119863(11990110158403119901) lt 119863(11990110158401119901) cup 119863(11990110158403119901) lt 119863(11990110158402119901)the choice of unit Based on the characteristic distributioncondition in the whole space 119863 can express the effect ofcorrelation among indices of RA Therefore 119863 can betterdescribe the similarity between RA and the set so as toprovide a basis to determine the level of RA The smaller the119863 value the better the property and vice versa
52 Mahalanobis Distance Calculation and Grading Method
521 Mahalanobis Distance Calculation Water absorptioncrushing value and adhesion are represented by 119882 119862 and119873 respectively The evaluation index of RA comprehensiveproperty can be considered as a three-dimensional vector119875(119882119862119873) and assuming that the value of the three indi-cators is 1198751015840(1198821015840119899 1198621015840119899 1198731015840119899) we can calculate the distance 119863between any RA and NA and then distinguish the wholeproperty of RA
119863(1199011015840119899119901)= radic (119882 minus1198821015840119899)21205902119882 + (119876 minus 1198761015840119899)21205902119862 + sdot sdot sdot + (119873 minus 1198731015840119899)21205902119873 (7)
in which 1198821015840119899 1198761015840119899 1198731015840119899 are the water absorption crushingvalue and adhesion of different grades of NA 119882 119876 119873are the average value of water absorption crushing valueand adhesion of a certain RA 1205902119882 1205902119862 1205902119873 are the standarddeviation of water absorption crushing value and adhesionof a certain RA
522 Grading Method of Mahalanobis Distance When the119863(11990110158401119901) 119863(11990110158402119901) 119863(11990110158403119901) represent respectively Maha-lanobis distance grade 1 grade 2 and grade 3 the RA canbe graded by the methods of Table 6 therefore Mahalanobisdistance comprehensive evaluation method can be applied toevaluate some cases which are difficult to compare each other
8 Advances in Materials Science and Engineering
Table 7 Mahalanobis distance between RA and natural aggregate
523 The Comprehensive Property Evaluation of RA Basedon the RA property test results we can calculate the Maha-lanobis Distance of RA as is shown in Table 7
As can be seen from Table 6 119863(11990110158402119901) lt 119863(11990110158403119901) lt119863(11990110158401119901) and therefore this RA is nearest to the entirety ofsecond class and its level can be represented as two
Compared with the other grading standard Mahalanobisdistance grading method is more intuitive and it is con-venient for operations without requiring a large number ofproperty tests
6 Conclusions
(1) RA is influenced by many factors such as its source andcrushingmethodThe numerical value of the property exper-iments shows obvious variability and instability Comparedto the property of pavement RA building RA has a widervariability and lower stability
(2) The distance method of comparison is not affectedby the dimensions and it can be used to express the wholedifference of various sample sets simply The greater thedistance value the lower the similarity of the two samples
(3)The calculation results of Euclidean distance show thatthe comprehensive property of the pavement RA (119878119896 = 2208)is superior to that of building RA (119878119896 = 7341) And forthese two kinds of RA the Euclidean distance of the property(120588119896119895 = 052) and apparent relative density (120588119896119895 = 187) areslightly different whereas the change of water absorption(120588119896119895 = 1067) and crushing value (120588119896119895 = 2636) is quiteremarkable Thus the water absorption rate and crushingvalue can approximately reflect the property differences ofRA
(4) Euclidean distance calculations show that there is acertain correlation among apparent density water absorptionrate and crushing value The lower the apparent density thelarger the water absorption rate and crushing value Howeverthe correlation for flakiness content is low so the mostimportant property indices for the RA are crushing value andwater absorption and at same time the major improvementfor these two property indices is needed as well
(5) The Mahalanobis distance method can evaluate thecomprehensive property of RA directly and provide gradeevaluation guidance for RAs by considering the influencefactors of the correlation among variables such as parametersof mean variance and covariance
Additional Points
Highlights (i) It is clear that there has been a large variabilityabout the individual property of recycled aggregate (RA) andit has been difficult to express thewhole property in any singleproperty parameter (ii) A method is proposed to expressthe differences of the whole property of RAs using Euclidean
distance method (iii) A calculation model and evaluationmethod are provided for the whole property of RAs withMahalanobis distance method
Conflicts of Interest
The authors declare that they have no conflicts of interest
Acknowledgments
This study was supported by the National Natural ScienceFoundation of China (Project 51578027)
References
[1] F Xiao B Putman and S Amirkhanian ldquoRheological char-acteristics investigation of high percentage RAP binders withWMA technology at various aging statesrdquo Construction andBuilding Materials vol 98 pp 315ndash324 2015
[2] Y C Guo and K W Sun ldquoConsideration on new specificationsof road materials based on properties of recycled ag-gregaterdquoJournal of Changrsquoan University (Natural Science Edition) vol 3pp 1671ndash8879 2010
[3] S Nagataki A Gokce T Saeki and M Hisada ldquoAssessmentof recycling process induced damage sensitivity of recycledconcrete aggregatesrdquoCement and Concrete Research vol 34 no6 pp 965ndash971 2004
[4] Z Q Shi Jian guang ldquoEffects of different materials in theconstruction waste on the properties of concreterdquo Concrete vol6 pp 11ndash13 2008
[5] A R Chini S-S Kuo J M Armaghani and J P Duxbury ldquoTestof recycled concrete aggregate in accelerated test trackrdquo Journalof Transportation Engineering vol 127 no 6 pp 486ndash492 2001
[6] V W Y Tam and C M Tam ldquoA new approach in assessingcement mortar remains on recycled aggregaterdquo Magazine ofConcrete Research vol 59 no 6 pp 413ndash422 2007
[7] M Etxeberria A R Marı and E Vazquez ldquoRecycled aggregateconcrete as structural materialrdquo Materials and Structures vol40 no 5 pp 529ndash541 2007
[8] S YWang PH Zhu XWang J Zhou andH E Xia ldquoResearchon some key issues for applicat ion of recycled concreterdquoConcrete vol 4 pp 39ndash41 2008
[9] G R Robinson Jr W D Menzie and H Hyun ldquoRecycling ofconstruction debris as aggregate in the Mid-Atlantic RegionUSArdquo Resources Conservation amp Recycling vol 42 no 3 pp275ndash294 2004
[10] X Yuezhou The Numerical Simulation and Influence of Aggre-gate Gradation on the Compressive Strength of Recycled Aggre-gate Concrete Xiamen University 2007
[11] Z Zhang K Wang H Liu and Z Deng ldquoKey performanceproperties of asphalt mixtures with recycled concrete aggregatefrom low strength concreterdquo Construction and Building Materi-als vol 126 pp 711ndash719 2016
[12] H Sung Do P Hee Mun and R Suk keun ldquoA study onengineering characteristics of asphalt concrete using filler withrecycled waste limerdquoWaste Management vol 28 no 1 pp 191ndash199 2008
[13] Y D Wong D D Sun and D Lai ldquoValue-added utilisationof recycled concrete in hot-mix asphaltrdquo Journal of WasteManagement vol 27 no 2 pp 294ndash301 2007
Advances in Materials Science and Engineering 9
[14] S Wu J Zhong J Zhu and D Wang ldquoInfluence of demolitionwaste used as recycled aggregate on performance of asphaltmixturerdquo Road Materials and Pavement Design vol 14 no 3pp 679ndash688 2013
[15] L Huan The Research for Recycled Concrete Aggregate and ITSApplied in Asphalt Pavement Beijing University of TechnologyBeijing China 2011
[16] R HuiThe Evaluation for Property of Recycled Coarse ConcreteAggregate Basing on Distance Method Beijing University ofTechnology Beijing China 2012
[17] A Stimilli G Ferrotti A Graziani and F Canestrari ldquoPerfor-mance evaluation of a cold-recycled mixture containing highpercentage of reclaimed asphaltrdquo Road Materials and PavementDesign vol 14 no 1 pp 149ndash161 2013
[18] K L Vasconcelos A Bhasin and D N Little ldquoInfluence ofreduced production temperatures on the adhesive propertiesof aggregates and laboratory performance of fine aggregate-asphalt mixturesrdquo Road Materials and Pavement Design vol 11no 1 pp 47ndash64 2011
[19] Z-G Zhao and F Liu ldquoApplication research of statistical mon-itoring index based on mahalanobis distancerdquo Acta AutomaticaSinica vol 34 no 4 pp 493ndash495 2008
Commonly acceptable ge2600 le2000 le2800 le3000 ge4e analysis data of RA properties
e single property index ofRA is discrete
Express the dierence of the wholeproperty by distance method
Whether the correlation of property index is considered
Euclidean distance Mahalanobis distance
No Yes
Figure 1 RA property evaluation flow chart
not give the whole characteristics requirement in differentapplication fields In 1998 Germany promulgated ldquoRegen-eration Aggregate Concrete Application Guiderdquo where thequality requirements of RA must comply with the technicalrequirements of natural aggregate (NA) [7 8] It ignoresthe characteristics of RA and it is not conducive to theapplication of concrete aggregate Although Spain and othercountries have not given a clear classification and gradingstandard of the RA only relevant qualificationwas performedwith recycled coarse aggregate [9 10] The standard did notinclude many details The common features are that thesespecifications only gave a different range of values for a singleproperty index of the RA and did not make provisions for thewhole properties
Because of the complexity of RA sources and regenera-tion techniques an RArsquos properties show large variations inits microstructure and physical mechanical and chemicalproperties [1] The test values show a large variability thesame single property index of different sets maybe has a largedifference of the whole property so the whole property ofRA could not be only expressed by a single property indexAdopting the Euclidean distance method and Mahalanobisdistance method respectively in evaluating the single andwhole property of RA is proposed in this paperTheflow chartis demonstrated in Figure 1
2 Test and Analysis of the Characteristic ofRA Used
21 Aggregate Characteristic Test Two types of aggregateswere used in the present study (1) RA produced from the old
building foundation concrete on the twentieth century 1980sIts strength was equivalent to C20 and the coarse aggregatewas comprised of pebbles The production process stages ofRA were pretreatment crushing grinding screening andwashing (2) NA is made from granite
The RAs used in the present study were screenedinto three different sizes (475ndash950mm 475ndash132mm and950ndash190mm) with the same processing method After sam-pling the aggregates were sievedwith 475mm standard sieveand then rinsed with water to ensure that they are free of dirtand dust (see Figure 2) [11]
The properties of the RA and NA such as apparent rela-tive density water absorption crushed value wear value andthe adhesion were tested according to the Chinese standardTable 1 presents the results from the above-described tests
As indicated in Table 1 the properties of NA testedsatisfy all commonly acceptance while RAs had higher waterabsorption and lower adhesion to asphalt values than thelisted commonly accepted values
22 Aggregate Characteristic Analysis As shown in Table 1the apparent relative densities of RA with different particlesizes are similar and they are all slightly lower than those ofNA In addition the water absorption of RA is much higherthan that of NA for the same size of aggregate These may bemainly attributed to the high porosity or low density of themortar adhered on the surface of the aggregate In additionduring the production process of RA mechanical crushingparticle collision and sliding also producemicrocracks in theaggregate and these will result in an increase in the aggregateporosity and a decrease in the apparent relative density Theporous mortar and fine cracks are also responsible for thehigh water absorption of the RAs Furthermore the high siltcontent on the surface of the RA particles may also help toincrease the rate and amount of water absorption
The crush values of RAs are clearly lower than those ofNAAnd the degree of abrasion of RAs is all lower than that ofNA Amain reason for these is that the RAs had already gonethrough a crushing process during recycling After crushingwashing and sieving the RAs had high pebble content withgood integrity As a result the RAs used in this study had ahigher crushing and abrasion resistance than the NA Thissuggests that the lower the strength of the original concretethe smaller the damage degree in the process of the aggregateregeneration and the better the property of resistance tocrushing of the recycled aggregate
Advances in Materials Science and Engineering 3
(a) RA as received (b) RA after cleaning and sieving
Figure 2 RA before and after washing
(a) NA 30-time SEM picture (b) RA 30-time SEM picture
(c) NA 100-time SEM picture (d) RA 100-time SEM picture
Figure 3 RANA SEM pictures
Adhesion of aggregate to asphalt is critical for the prop-erty of asphalt concrete pavements Table 1 shows that amongthree RA samples tested two had adhesion level values lessthan 4 as required It was observed during the adhesiontesting that the general peeling area was around 10 andlocalized peeling area summed up to 30Theweak adhesionof RA may be related to the reaction of water soluble ionsleaching out from the adhered mortar on the aggregatesurface with the carboxylic acids of asphalts
A scanning electron microscope (SEM) was used toobserve the microstructure of RA and NA as shown inFigure 3
It can be seen that RA possesses a large number ofpores and microcracks internally caused by the accumulateddamage during crushing leading to an increase in the voidratio water absorption and crush value (shown in Figure 3)Compared to the surface structure ofNA the asphalt bondingmaterial can penetrate into the RA surface more easilyFurther the strength of interface may be improved
In summary for the RA from low strength concreteexcept for the water absorption rate the other propertiesof RA appear better than that of NA Perhaps this is aspecial case more extensive investigation may need to bedone
4 Advances in Materials Science and Engineering
Table 2 Property data of RA from buildings
Property index 119860 119861 119862 119863 119864 119865 119866 119867Density (gcm3) 2742 2529 25 2535 275 232 2265 2658Absorption () 194 482 215 450 249 989 570 450Flakiness () 698 370 250 340 300 250 310 260Crushing value () 1360 1380 1410 787 620 2420 3000 1300Stone species Pebbles Pebbles Pebbles Pebbles Gravel Gravel Gravel GravelStrength Low Low Low Low Low Low Low LowSources Foundation Floor Building Beam Foundation Building Building Building
31 Property Data of RA Tested and Investigated RA prop-erties can be evaluated by some experimental indices whichinclude property density water absorption needle and flaki-ness content and crushing value [12ndash14]
Sixteen groups of RAs were selected from pavements andbuildings respectively [15 16]Most of the property data wereinvestigated except for groups 119864 119865 and NAs which weretested by ourselves in the laboratory All of the RAs werecrushed by the jaw-crush method (see Tables 2 and 3)
As shown in Tables 2 and 3 the basic property data ofthe NA are generally property density at 263 gcm3 waterabsorption at 16 needle and flakiness content at 52 andcrushing value at 96
32 Variability Analysis of the NA Property Data As canbe seen from Tables 1 and 2 there is a little distinction inapparent relative density between RA and NA Most of theRArsquos apparent relative density is 88sim97 of the NAs about231ndash261 kgm3The shapes of RA fromdifferent sources havea high degree of similarity some of them are even better thanthe NA The shapes of different sources of RA are similarand for some the quality is better than the NA the apparentrelative density and needle-flakiness content are relativelystable while the water absorption and crushing values arerelatively higher The water absorption rate of RA is between2sim10 the highest value even is up to 12 which is 2 to 7times that of the NAThe crushing value of RA varies greatlyand its crush resistance ability is significantly lower than theNA as is shown in Figure 4
After being crushed the RA might produce a largenumber of microcracks on the aggregate surface and theinterior between aggregate and adhesive cement mortar
Besides the RA would be cracked more easily when beingpressed and it contains a variety of impurities on the surfacetherefore the RA strength would be lower than NA As aresult the water absorption and crushing value of RA weresignificantly larger [7 17] and the experimental data showeda large variability as well Since the content of impurities andmicrocrack is influenced by many factors such as the typesof concrete waste engineering construction and crushingmethod and the RA from building contains more impuritiesthan that from pavement its variability would be greater
33 Pavement Property with RA The conclusions of ourprevious study about RArsquos instability are as follows [11]
(1) Compared with NA the RA recycled from low gradeconcrete had lower apparent relative density higher waterabsorption and poorer adhesion to asphalt but had probablylower crushing and wearing value
(2) The rutting deformation of the HMAmixes generallyincreased with the level of RA replacement but the lowtemperature cracking resistance test results showed that thefailure load tensile stress and modulus of the HMA mixeswith RA are all significantly higher than those of themix withNA
(3) The water stability measured by immersion compres-sion tests showed a similar trend to that measured by freeze-thaw splitting strength tests The water stability of the HMAconcrete decreased with increasing RA replacement (up to50) When the RA content reached 75 the water resis-tance of the HMA concrete increased rather than decreasedfurther
Due to instability a large variability has been indicatedby experimental data of RA property index Therefore it isinaccurate to illustrate the whole property through a singleproperty index It is possible that the whole property is better
Advances in Materials Science and Engineering 5
22
23
24
25
26
27
28
aA bB cC dD eE fF gG hH
RA sample number
Building RAPavement RANA
Perfo
rman
ce d
ensit
y (g
cＧ
3)
(a) Contrast chart of density of NA and RA
0
2
4
6
8
10
aA bB cC dD eE fF gG hH
Abso
rptio
n (
)
RA sample number
Building RAPavement RANA(b) Contrast chart of absorption of NA and RA
Building RAPavement RANA
2
25
3
35
4
45
5
55
6
65
7
aA bB cC dD eE fF gG hH
Flak
ines
s con
tent
()
RA sample number
(c) Contrast chart of flakiness of NA and RA
4
8
12
16
20
24
28
32
aA bB cC dD eE fF gG hH
Crus
hing
val
ue (
)
RA sample number
Building RAPavement RANA
(d) Contrast chart of crushing value of NA and RA
Figure 4 Contrast charts of NA and RA properties
in the asphalt mixture while a single property index exhibitspoor results However the distance method is not affectedby the dimension and directly expresses the difference ofmultiple variables or sample set of indices According to thedistance method it is feasible to illustrate the whole propertyof RA
4 Euclidean Distance ComparisonMethod of RA
41 Euclidean Distance Calculation of RA Property (1) Thesingle property Euclidean distance 120588119896119895 is generally
(119909119895 minus 119910119896119894119895)2 (1)
in which 119909119895 is property test value of the NA 119910119896119894119895 is propertytest values of RA 119896 is pavement RAor building RAwhere 119896 =I (pavement RA) and 119896 = II (building RA) 119894 is the numberof test data groups when 119896 = I 119894 = 119860 119861 119862119863 119864 119865 119866119867119896 = II 119894 = 119886 119887 119888 119889 119890 119891 119892 ℎ 119895 is the different property test119895 = 1 (apparent density) 2 (water absorption) 3 (needle andflakiness content) and 4 (crushing value)
(2) The whole property Euclidean distance of RA 120588119896119894 isgenerally
120588119896119894 = radic 4sum119895=1
(119909119895 minus 119910119896119894119895)2 (2)
in which 119909119895 119910119896119894119895 119896 119894 and 119895 are the same as formulation(1)
6 Advances in Materials Science and Engineering
Table 4 Euclidean distance of single property of building and pavement RA
Property index Building RA (119860) Pavement RA (119861) 119860 minus 119861Property density 0516 0241 0275Absorption () 10669 8595 2074Flakiness content () 187 0943 0927Crushing value () 26355 1072 15635
052
1067
187
2636
024
860
094
1072
0
10
20
30
Performance Absorption Flakiness Crushing value
Eucli
dean
dist
ance
Single performances of building and pavement RA
Building RAPavement RA
density content
Figure 5 Single property of RA
(3) The whole property Euclidean distance standardvariance of RA 119878119896 is generally
119878119896 = radicsum119866119892119894=119860119886 (120588119896119894 minus 120588119896119894)27 (3)
in which type 119878119896 is the standard variance of whole propertyEuclidean distance for building RA and pavement RA 120588119896119894is the arithmetic mean value of whole property Euclideandistance for the different number of RA 120588119896119894 119896 and 119894 are thesame as formulation (1)
42 The Comparison and Analysis of Euclidean Distance ofRA Property According to the Euclidean distance formula(1)ndash(3) for the two types of RA the Euclidean distances ofsingle property andwhole property are shown in Tables 4 and5 and Figures 5 and 6 respectively
The greater the value of the properties the greater thedifference between the RA and theNA otherwise the smallerthe value the closer the difference The standard varianceof the whole property Euclidean distance of building RAand pavement RA was calculated to be 7341 and 2208respectively
Table 4 and Figure 5 show slightly differences of theproperty density and flakiness content of RA compared withthe NA while the crushing value and water absorption areobvious differences In addition pavement RA was signifi-cantly better than building RA in terms of single propertyindexAs shown inTable 5 andFigure 6 thewhole property of
2051
1590
450 440 438 388
091 020
789
506 400 372
247 199 150 130
02468
10121416182022
G F h e C A B f H g b c a d D EEu
clide
an d
istan
ce
Whole performance of building and pavement RA
Building RAPavement RA
Figure 6 Whole property of RA
RA with better single property may be poor while the wholeproperty of RA with poor property may be better
From the size of Euclidean distance it can be seen that thecomprehensive properties of group 119864 group 119863 and group 119889are the best whereas group 119866 shows the worst results TheEuclidean distance of crushing value and water absorptionis the largest which shows a primary need to improve theproperty index of the RA
The standard variances of the whole property Euclideandistances are 7341 and 2208 respectively which shows thatthe property of building RA has great fluctuation whilepavement RA is more stable
5 Mahalanobis Distance Evaluation andGrading Methods of RA
The analysis above indicates that the property evaluation ofRA contains multiple indicators Due to the instability of theproperties of RA its experimental data has a large variabilityFor different evaluation indices different evaluation resultsare achieved Since there exists certain dependence amongproperties of RA andEuclidean distance could not express theinfluence factors of the correlation among various propertiesof the samples another index is proposed to solve thisproblem named Mahalanobis distance
51 Mahalanobis Distance Evaluation Model Mahalanobisdistance can discriminate the spatial distribution of a pointrelative to the total by calculating the distance to a certainpoint Mahalanobis distance considers three parameters
Advances in Materials Science and Engineering 7
Table 5 Euclidean distance of whole property of building and pavement RA
mean value variance and covariance which shows thetranslational invariance and invariance under nonsingulartransformation Mahalanobis distance can express the influ-encing factors of the correlation among different variablesand it is independent of the dimension of each characteristicparameter and the judgement basis is according to thecharacteristic distribution condition of variables in the wholespace [18 19]
Suppose that there are two variables 1198661 and 1198662 whichobey the state distribution 119909 isin 119877119901 is a new sample point andthe distance from 119909 to 1198661 and 1198662 is defined as follows
119889 (119909 1198661) = radic(119909 minus 119906(1)Σminus11 ) (119909 minus 119906(1)) 119889 (119909 1198662) = radic(119909 minus 119906(2)Σminus12 ) (119909 minus 119906(2))
(4)
in which 120583(1) 120583(2) are the mean vector of 1198661 and 1198662 Σ1 Σ2are covariance matrix of set 1198661 and 1198662
Therefore it can be determined according to the followingrules
The indices which can express the property of RA aredefined as a vector 119883 (11988311198832 119883119899) the mean vector of119883 is 120583 (1205831 1205832 120583119899) and the standard deviation vector is 120590(1205901 1205902 120590119899) The property standard is defined as vector 119884(1198841 1198842 119884119899) so the vector 119884 is degenerated into a pointof the multidimensional space
The global 1198661 is defined by the vector 119883 and the global1198662 as 119884 then the distance from a sample point of global 1198661to global 1198662 is equal to the distance from a point of vector119883to vector 119884119863 (119883 119884)= radic (1198841 minus 1198831)212059021 + (1198842 minus 1198832)212059022 + sdot sdot sdot + (119884119899 minus 119883119899)21205902119899 (6)
119863 is a normalized distance of nonuniform distributionand it can reflect the comprehensive index of RA propertywithout the need to consider the dimension of each index and
Table 6 Grading methods of RA
Grading of RA Meet the conditionsGrading 1 119863(11990110158401119901) lt 119863(11990110158402119901) cup 119863(11990110158401119901) lt 119863(11990110158403119901)Grading 2 119863(11990110158402119901) lt 119863(11990110158401119901) cup 119863(11990110158402119901) lt 119863(11990110158403119901)Grading 3 119863(11990110158403119901) lt 119863(11990110158401119901) cup 119863(11990110158403119901) lt 119863(11990110158402119901)the choice of unit Based on the characteristic distributioncondition in the whole space 119863 can express the effect ofcorrelation among indices of RA Therefore 119863 can betterdescribe the similarity between RA and the set so as toprovide a basis to determine the level of RA The smaller the119863 value the better the property and vice versa
52 Mahalanobis Distance Calculation and Grading Method
521 Mahalanobis Distance Calculation Water absorptioncrushing value and adhesion are represented by 119882 119862 and119873 respectively The evaluation index of RA comprehensiveproperty can be considered as a three-dimensional vector119875(119882119862119873) and assuming that the value of the three indi-cators is 1198751015840(1198821015840119899 1198621015840119899 1198731015840119899) we can calculate the distance 119863between any RA and NA and then distinguish the wholeproperty of RA
119863(1199011015840119899119901)= radic (119882 minus1198821015840119899)21205902119882 + (119876 minus 1198761015840119899)21205902119862 + sdot sdot sdot + (119873 minus 1198731015840119899)21205902119873 (7)
in which 1198821015840119899 1198761015840119899 1198731015840119899 are the water absorption crushingvalue and adhesion of different grades of NA 119882 119876 119873are the average value of water absorption crushing valueand adhesion of a certain RA 1205902119882 1205902119862 1205902119873 are the standarddeviation of water absorption crushing value and adhesionof a certain RA
522 Grading Method of Mahalanobis Distance When the119863(11990110158401119901) 119863(11990110158402119901) 119863(11990110158403119901) represent respectively Maha-lanobis distance grade 1 grade 2 and grade 3 the RA canbe graded by the methods of Table 6 therefore Mahalanobisdistance comprehensive evaluation method can be applied toevaluate some cases which are difficult to compare each other
8 Advances in Materials Science and Engineering
Table 7 Mahalanobis distance between RA and natural aggregate
523 The Comprehensive Property Evaluation of RA Basedon the RA property test results we can calculate the Maha-lanobis Distance of RA as is shown in Table 7
As can be seen from Table 6 119863(11990110158402119901) lt 119863(11990110158403119901) lt119863(11990110158401119901) and therefore this RA is nearest to the entirety ofsecond class and its level can be represented as two
Compared with the other grading standard Mahalanobisdistance grading method is more intuitive and it is con-venient for operations without requiring a large number ofproperty tests
6 Conclusions
(1) RA is influenced by many factors such as its source andcrushingmethodThe numerical value of the property exper-iments shows obvious variability and instability Comparedto the property of pavement RA building RA has a widervariability and lower stability
(2) The distance method of comparison is not affectedby the dimensions and it can be used to express the wholedifference of various sample sets simply The greater thedistance value the lower the similarity of the two samples
(3)The calculation results of Euclidean distance show thatthe comprehensive property of the pavement RA (119878119896 = 2208)is superior to that of building RA (119878119896 = 7341) And forthese two kinds of RA the Euclidean distance of the property(120588119896119895 = 052) and apparent relative density (120588119896119895 = 187) areslightly different whereas the change of water absorption(120588119896119895 = 1067) and crushing value (120588119896119895 = 2636) is quiteremarkable Thus the water absorption rate and crushingvalue can approximately reflect the property differences ofRA
(4) Euclidean distance calculations show that there is acertain correlation among apparent density water absorptionrate and crushing value The lower the apparent density thelarger the water absorption rate and crushing value Howeverthe correlation for flakiness content is low so the mostimportant property indices for the RA are crushing value andwater absorption and at same time the major improvementfor these two property indices is needed as well
(5) The Mahalanobis distance method can evaluate thecomprehensive property of RA directly and provide gradeevaluation guidance for RAs by considering the influencefactors of the correlation among variables such as parametersof mean variance and covariance
Additional Points
Highlights (i) It is clear that there has been a large variabilityabout the individual property of recycled aggregate (RA) andit has been difficult to express thewhole property in any singleproperty parameter (ii) A method is proposed to expressthe differences of the whole property of RAs using Euclidean
distance method (iii) A calculation model and evaluationmethod are provided for the whole property of RAs withMahalanobis distance method
Conflicts of Interest
The authors declare that they have no conflicts of interest
Acknowledgments
This study was supported by the National Natural ScienceFoundation of China (Project 51578027)
References
[1] F Xiao B Putman and S Amirkhanian ldquoRheological char-acteristics investigation of high percentage RAP binders withWMA technology at various aging statesrdquo Construction andBuilding Materials vol 98 pp 315ndash324 2015
[2] Y C Guo and K W Sun ldquoConsideration on new specificationsof road materials based on properties of recycled ag-gregaterdquoJournal of Changrsquoan University (Natural Science Edition) vol 3pp 1671ndash8879 2010
[3] S Nagataki A Gokce T Saeki and M Hisada ldquoAssessmentof recycling process induced damage sensitivity of recycledconcrete aggregatesrdquoCement and Concrete Research vol 34 no6 pp 965ndash971 2004
[4] Z Q Shi Jian guang ldquoEffects of different materials in theconstruction waste on the properties of concreterdquo Concrete vol6 pp 11ndash13 2008
[5] A R Chini S-S Kuo J M Armaghani and J P Duxbury ldquoTestof recycled concrete aggregate in accelerated test trackrdquo Journalof Transportation Engineering vol 127 no 6 pp 486ndash492 2001
[6] V W Y Tam and C M Tam ldquoA new approach in assessingcement mortar remains on recycled aggregaterdquo Magazine ofConcrete Research vol 59 no 6 pp 413ndash422 2007
[7] M Etxeberria A R Marı and E Vazquez ldquoRecycled aggregateconcrete as structural materialrdquo Materials and Structures vol40 no 5 pp 529ndash541 2007
[8] S YWang PH Zhu XWang J Zhou andH E Xia ldquoResearchon some key issues for applicat ion of recycled concreterdquoConcrete vol 4 pp 39ndash41 2008
[9] G R Robinson Jr W D Menzie and H Hyun ldquoRecycling ofconstruction debris as aggregate in the Mid-Atlantic RegionUSArdquo Resources Conservation amp Recycling vol 42 no 3 pp275ndash294 2004
[10] X Yuezhou The Numerical Simulation and Influence of Aggre-gate Gradation on the Compressive Strength of Recycled Aggre-gate Concrete Xiamen University 2007
[11] Z Zhang K Wang H Liu and Z Deng ldquoKey performanceproperties of asphalt mixtures with recycled concrete aggregatefrom low strength concreterdquo Construction and Building Materi-als vol 126 pp 711ndash719 2016
[12] H Sung Do P Hee Mun and R Suk keun ldquoA study onengineering characteristics of asphalt concrete using filler withrecycled waste limerdquoWaste Management vol 28 no 1 pp 191ndash199 2008
[13] Y D Wong D D Sun and D Lai ldquoValue-added utilisationof recycled concrete in hot-mix asphaltrdquo Journal of WasteManagement vol 27 no 2 pp 294ndash301 2007
Advances in Materials Science and Engineering 9
[14] S Wu J Zhong J Zhu and D Wang ldquoInfluence of demolitionwaste used as recycled aggregate on performance of asphaltmixturerdquo Road Materials and Pavement Design vol 14 no 3pp 679ndash688 2013
[15] L Huan The Research for Recycled Concrete Aggregate and ITSApplied in Asphalt Pavement Beijing University of TechnologyBeijing China 2011
[16] R HuiThe Evaluation for Property of Recycled Coarse ConcreteAggregate Basing on Distance Method Beijing University ofTechnology Beijing China 2012
[17] A Stimilli G Ferrotti A Graziani and F Canestrari ldquoPerfor-mance evaluation of a cold-recycled mixture containing highpercentage of reclaimed asphaltrdquo Road Materials and PavementDesign vol 14 no 1 pp 149ndash161 2013
[18] K L Vasconcelos A Bhasin and D N Little ldquoInfluence ofreduced production temperatures on the adhesive propertiesof aggregates and laboratory performance of fine aggregate-asphalt mixturesrdquo Road Materials and Pavement Design vol 11no 1 pp 47ndash64 2011
[19] Z-G Zhao and F Liu ldquoApplication research of statistical mon-itoring index based on mahalanobis distancerdquo Acta AutomaticaSinica vol 34 no 4 pp 493ndash495 2008
(a) RA as received (b) RA after cleaning and sieving
Figure 2 RA before and after washing
(a) NA 30-time SEM picture (b) RA 30-time SEM picture
(c) NA 100-time SEM picture (d) RA 100-time SEM picture
Figure 3 RANA SEM pictures
Adhesion of aggregate to asphalt is critical for the prop-erty of asphalt concrete pavements Table 1 shows that amongthree RA samples tested two had adhesion level values lessthan 4 as required It was observed during the adhesiontesting that the general peeling area was around 10 andlocalized peeling area summed up to 30Theweak adhesionof RA may be related to the reaction of water soluble ionsleaching out from the adhered mortar on the aggregatesurface with the carboxylic acids of asphalts
A scanning electron microscope (SEM) was used toobserve the microstructure of RA and NA as shown inFigure 3
It can be seen that RA possesses a large number ofpores and microcracks internally caused by the accumulateddamage during crushing leading to an increase in the voidratio water absorption and crush value (shown in Figure 3)Compared to the surface structure ofNA the asphalt bondingmaterial can penetrate into the RA surface more easilyFurther the strength of interface may be improved
In summary for the RA from low strength concreteexcept for the water absorption rate the other propertiesof RA appear better than that of NA Perhaps this is aspecial case more extensive investigation may need to bedone
4 Advances in Materials Science and Engineering
Table 2 Property data of RA from buildings
Property index 119860 119861 119862 119863 119864 119865 119866 119867Density (gcm3) 2742 2529 25 2535 275 232 2265 2658Absorption () 194 482 215 450 249 989 570 450Flakiness () 698 370 250 340 300 250 310 260Crushing value () 1360 1380 1410 787 620 2420 3000 1300Stone species Pebbles Pebbles Pebbles Pebbles Gravel Gravel Gravel GravelStrength Low Low Low Low Low Low Low LowSources Foundation Floor Building Beam Foundation Building Building Building
31 Property Data of RA Tested and Investigated RA prop-erties can be evaluated by some experimental indices whichinclude property density water absorption needle and flaki-ness content and crushing value [12ndash14]
Sixteen groups of RAs were selected from pavements andbuildings respectively [15 16]Most of the property data wereinvestigated except for groups 119864 119865 and NAs which weretested by ourselves in the laboratory All of the RAs werecrushed by the jaw-crush method (see Tables 2 and 3)
As shown in Tables 2 and 3 the basic property data ofthe NA are generally property density at 263 gcm3 waterabsorption at 16 needle and flakiness content at 52 andcrushing value at 96
32 Variability Analysis of the NA Property Data As canbe seen from Tables 1 and 2 there is a little distinction inapparent relative density between RA and NA Most of theRArsquos apparent relative density is 88sim97 of the NAs about231ndash261 kgm3The shapes of RA fromdifferent sources havea high degree of similarity some of them are even better thanthe NA The shapes of different sources of RA are similarand for some the quality is better than the NA the apparentrelative density and needle-flakiness content are relativelystable while the water absorption and crushing values arerelatively higher The water absorption rate of RA is between2sim10 the highest value even is up to 12 which is 2 to 7times that of the NAThe crushing value of RA varies greatlyand its crush resistance ability is significantly lower than theNA as is shown in Figure 4
After being crushed the RA might produce a largenumber of microcracks on the aggregate surface and theinterior between aggregate and adhesive cement mortar
Besides the RA would be cracked more easily when beingpressed and it contains a variety of impurities on the surfacetherefore the RA strength would be lower than NA As aresult the water absorption and crushing value of RA weresignificantly larger [7 17] and the experimental data showeda large variability as well Since the content of impurities andmicrocrack is influenced by many factors such as the typesof concrete waste engineering construction and crushingmethod and the RA from building contains more impuritiesthan that from pavement its variability would be greater
33 Pavement Property with RA The conclusions of ourprevious study about RArsquos instability are as follows [11]
(1) Compared with NA the RA recycled from low gradeconcrete had lower apparent relative density higher waterabsorption and poorer adhesion to asphalt but had probablylower crushing and wearing value
(2) The rutting deformation of the HMAmixes generallyincreased with the level of RA replacement but the lowtemperature cracking resistance test results showed that thefailure load tensile stress and modulus of the HMA mixeswith RA are all significantly higher than those of themix withNA
(3) The water stability measured by immersion compres-sion tests showed a similar trend to that measured by freeze-thaw splitting strength tests The water stability of the HMAconcrete decreased with increasing RA replacement (up to50) When the RA content reached 75 the water resis-tance of the HMA concrete increased rather than decreasedfurther
Due to instability a large variability has been indicatedby experimental data of RA property index Therefore it isinaccurate to illustrate the whole property through a singleproperty index It is possible that the whole property is better
Advances in Materials Science and Engineering 5
22
23
24
25
26
27
28
aA bB cC dD eE fF gG hH
RA sample number
Building RAPavement RANA
Perfo
rman
ce d
ensit
y (g
cＧ
3)
(a) Contrast chart of density of NA and RA
0
2
4
6
8
10
aA bB cC dD eE fF gG hH
Abso
rptio
n (
)
RA sample number
Building RAPavement RANA(b) Contrast chart of absorption of NA and RA
Building RAPavement RANA
2
25
3
35
4
45
5
55
6
65
7
aA bB cC dD eE fF gG hH
Flak
ines
s con
tent
()
RA sample number
(c) Contrast chart of flakiness of NA and RA
4
8
12
16
20
24
28
32
aA bB cC dD eE fF gG hH
Crus
hing
val
ue (
)
RA sample number
Building RAPavement RANA
(d) Contrast chart of crushing value of NA and RA
Figure 4 Contrast charts of NA and RA properties
in the asphalt mixture while a single property index exhibitspoor results However the distance method is not affectedby the dimension and directly expresses the difference ofmultiple variables or sample set of indices According to thedistance method it is feasible to illustrate the whole propertyof RA
4 Euclidean Distance ComparisonMethod of RA
41 Euclidean Distance Calculation of RA Property (1) Thesingle property Euclidean distance 120588119896119895 is generally
(119909119895 minus 119910119896119894119895)2 (1)
in which 119909119895 is property test value of the NA 119910119896119894119895 is propertytest values of RA 119896 is pavement RAor building RAwhere 119896 =I (pavement RA) and 119896 = II (building RA) 119894 is the numberof test data groups when 119896 = I 119894 = 119860 119861 119862119863 119864 119865 119866119867119896 = II 119894 = 119886 119887 119888 119889 119890 119891 119892 ℎ 119895 is the different property test119895 = 1 (apparent density) 2 (water absorption) 3 (needle andflakiness content) and 4 (crushing value)
(2) The whole property Euclidean distance of RA 120588119896119894 isgenerally
120588119896119894 = radic 4sum119895=1
(119909119895 minus 119910119896119894119895)2 (2)
in which 119909119895 119910119896119894119895 119896 119894 and 119895 are the same as formulation(1)
6 Advances in Materials Science and Engineering
Table 4 Euclidean distance of single property of building and pavement RA
Property index Building RA (119860) Pavement RA (119861) 119860 minus 119861Property density 0516 0241 0275Absorption () 10669 8595 2074Flakiness content () 187 0943 0927Crushing value () 26355 1072 15635
052
1067
187
2636
024
860
094
1072
0
10
20
30
Performance Absorption Flakiness Crushing value
Eucli
dean
dist
ance
Single performances of building and pavement RA
Building RAPavement RA
density content
Figure 5 Single property of RA
(3) The whole property Euclidean distance standardvariance of RA 119878119896 is generally
119878119896 = radicsum119866119892119894=119860119886 (120588119896119894 minus 120588119896119894)27 (3)
in which type 119878119896 is the standard variance of whole propertyEuclidean distance for building RA and pavement RA 120588119896119894is the arithmetic mean value of whole property Euclideandistance for the different number of RA 120588119896119894 119896 and 119894 are thesame as formulation (1)
42 The Comparison and Analysis of Euclidean Distance ofRA Property According to the Euclidean distance formula(1)ndash(3) for the two types of RA the Euclidean distances ofsingle property andwhole property are shown in Tables 4 and5 and Figures 5 and 6 respectively
The greater the value of the properties the greater thedifference between the RA and theNA otherwise the smallerthe value the closer the difference The standard varianceof the whole property Euclidean distance of building RAand pavement RA was calculated to be 7341 and 2208respectively
Table 4 and Figure 5 show slightly differences of theproperty density and flakiness content of RA compared withthe NA while the crushing value and water absorption areobvious differences In addition pavement RA was signifi-cantly better than building RA in terms of single propertyindexAs shown inTable 5 andFigure 6 thewhole property of
2051
1590
450 440 438 388
091 020
789
506 400 372
247 199 150 130
02468
10121416182022
G F h e C A B f H g b c a d D EEu
clide
an d
istan
ce
Whole performance of building and pavement RA
Building RAPavement RA
Figure 6 Whole property of RA
RA with better single property may be poor while the wholeproperty of RA with poor property may be better
From the size of Euclidean distance it can be seen that thecomprehensive properties of group 119864 group 119863 and group 119889are the best whereas group 119866 shows the worst results TheEuclidean distance of crushing value and water absorptionis the largest which shows a primary need to improve theproperty index of the RA
The standard variances of the whole property Euclideandistances are 7341 and 2208 respectively which shows thatthe property of building RA has great fluctuation whilepavement RA is more stable
5 Mahalanobis Distance Evaluation andGrading Methods of RA
The analysis above indicates that the property evaluation ofRA contains multiple indicators Due to the instability of theproperties of RA its experimental data has a large variabilityFor different evaluation indices different evaluation resultsare achieved Since there exists certain dependence amongproperties of RA andEuclidean distance could not express theinfluence factors of the correlation among various propertiesof the samples another index is proposed to solve thisproblem named Mahalanobis distance
51 Mahalanobis Distance Evaluation Model Mahalanobisdistance can discriminate the spatial distribution of a pointrelative to the total by calculating the distance to a certainpoint Mahalanobis distance considers three parameters
Advances in Materials Science and Engineering 7
Table 5 Euclidean distance of whole property of building and pavement RA
mean value variance and covariance which shows thetranslational invariance and invariance under nonsingulartransformation Mahalanobis distance can express the influ-encing factors of the correlation among different variablesand it is independent of the dimension of each characteristicparameter and the judgement basis is according to thecharacteristic distribution condition of variables in the wholespace [18 19]
Suppose that there are two variables 1198661 and 1198662 whichobey the state distribution 119909 isin 119877119901 is a new sample point andthe distance from 119909 to 1198661 and 1198662 is defined as follows
119889 (119909 1198661) = radic(119909 minus 119906(1)Σminus11 ) (119909 minus 119906(1)) 119889 (119909 1198662) = radic(119909 minus 119906(2)Σminus12 ) (119909 minus 119906(2))
(4)
in which 120583(1) 120583(2) are the mean vector of 1198661 and 1198662 Σ1 Σ2are covariance matrix of set 1198661 and 1198662
Therefore it can be determined according to the followingrules
The indices which can express the property of RA aredefined as a vector 119883 (11988311198832 119883119899) the mean vector of119883 is 120583 (1205831 1205832 120583119899) and the standard deviation vector is 120590(1205901 1205902 120590119899) The property standard is defined as vector 119884(1198841 1198842 119884119899) so the vector 119884 is degenerated into a pointof the multidimensional space
The global 1198661 is defined by the vector 119883 and the global1198662 as 119884 then the distance from a sample point of global 1198661to global 1198662 is equal to the distance from a point of vector119883to vector 119884119863 (119883 119884)= radic (1198841 minus 1198831)212059021 + (1198842 minus 1198832)212059022 + sdot sdot sdot + (119884119899 minus 119883119899)21205902119899 (6)
119863 is a normalized distance of nonuniform distributionand it can reflect the comprehensive index of RA propertywithout the need to consider the dimension of each index and
Table 6 Grading methods of RA
Grading of RA Meet the conditionsGrading 1 119863(11990110158401119901) lt 119863(11990110158402119901) cup 119863(11990110158401119901) lt 119863(11990110158403119901)Grading 2 119863(11990110158402119901) lt 119863(11990110158401119901) cup 119863(11990110158402119901) lt 119863(11990110158403119901)Grading 3 119863(11990110158403119901) lt 119863(11990110158401119901) cup 119863(11990110158403119901) lt 119863(11990110158402119901)the choice of unit Based on the characteristic distributioncondition in the whole space 119863 can express the effect ofcorrelation among indices of RA Therefore 119863 can betterdescribe the similarity between RA and the set so as toprovide a basis to determine the level of RA The smaller the119863 value the better the property and vice versa
52 Mahalanobis Distance Calculation and Grading Method
521 Mahalanobis Distance Calculation Water absorptioncrushing value and adhesion are represented by 119882 119862 and119873 respectively The evaluation index of RA comprehensiveproperty can be considered as a three-dimensional vector119875(119882119862119873) and assuming that the value of the three indi-cators is 1198751015840(1198821015840119899 1198621015840119899 1198731015840119899) we can calculate the distance 119863between any RA and NA and then distinguish the wholeproperty of RA
119863(1199011015840119899119901)= radic (119882 minus1198821015840119899)21205902119882 + (119876 minus 1198761015840119899)21205902119862 + sdot sdot sdot + (119873 minus 1198731015840119899)21205902119873 (7)
in which 1198821015840119899 1198761015840119899 1198731015840119899 are the water absorption crushingvalue and adhesion of different grades of NA 119882 119876 119873are the average value of water absorption crushing valueand adhesion of a certain RA 1205902119882 1205902119862 1205902119873 are the standarddeviation of water absorption crushing value and adhesionof a certain RA
522 Grading Method of Mahalanobis Distance When the119863(11990110158401119901) 119863(11990110158402119901) 119863(11990110158403119901) represent respectively Maha-lanobis distance grade 1 grade 2 and grade 3 the RA canbe graded by the methods of Table 6 therefore Mahalanobisdistance comprehensive evaluation method can be applied toevaluate some cases which are difficult to compare each other
8 Advances in Materials Science and Engineering
Table 7 Mahalanobis distance between RA and natural aggregate
523 The Comprehensive Property Evaluation of RA Basedon the RA property test results we can calculate the Maha-lanobis Distance of RA as is shown in Table 7
As can be seen from Table 6 119863(11990110158402119901) lt 119863(11990110158403119901) lt119863(11990110158401119901) and therefore this RA is nearest to the entirety ofsecond class and its level can be represented as two
Compared with the other grading standard Mahalanobisdistance grading method is more intuitive and it is con-venient for operations without requiring a large number ofproperty tests
6 Conclusions
(1) RA is influenced by many factors such as its source andcrushingmethodThe numerical value of the property exper-iments shows obvious variability and instability Comparedto the property of pavement RA building RA has a widervariability and lower stability
(2) The distance method of comparison is not affectedby the dimensions and it can be used to express the wholedifference of various sample sets simply The greater thedistance value the lower the similarity of the two samples
(3)The calculation results of Euclidean distance show thatthe comprehensive property of the pavement RA (119878119896 = 2208)is superior to that of building RA (119878119896 = 7341) And forthese two kinds of RA the Euclidean distance of the property(120588119896119895 = 052) and apparent relative density (120588119896119895 = 187) areslightly different whereas the change of water absorption(120588119896119895 = 1067) and crushing value (120588119896119895 = 2636) is quiteremarkable Thus the water absorption rate and crushingvalue can approximately reflect the property differences ofRA
(4) Euclidean distance calculations show that there is acertain correlation among apparent density water absorptionrate and crushing value The lower the apparent density thelarger the water absorption rate and crushing value Howeverthe correlation for flakiness content is low so the mostimportant property indices for the RA are crushing value andwater absorption and at same time the major improvementfor these two property indices is needed as well
(5) The Mahalanobis distance method can evaluate thecomprehensive property of RA directly and provide gradeevaluation guidance for RAs by considering the influencefactors of the correlation among variables such as parametersof mean variance and covariance
Additional Points
Highlights (i) It is clear that there has been a large variabilityabout the individual property of recycled aggregate (RA) andit has been difficult to express thewhole property in any singleproperty parameter (ii) A method is proposed to expressthe differences of the whole property of RAs using Euclidean
distance method (iii) A calculation model and evaluationmethod are provided for the whole property of RAs withMahalanobis distance method
Conflicts of Interest
The authors declare that they have no conflicts of interest
Acknowledgments
This study was supported by the National Natural ScienceFoundation of China (Project 51578027)
References
[1] F Xiao B Putman and S Amirkhanian ldquoRheological char-acteristics investigation of high percentage RAP binders withWMA technology at various aging statesrdquo Construction andBuilding Materials vol 98 pp 315ndash324 2015
[2] Y C Guo and K W Sun ldquoConsideration on new specificationsof road materials based on properties of recycled ag-gregaterdquoJournal of Changrsquoan University (Natural Science Edition) vol 3pp 1671ndash8879 2010
[3] S Nagataki A Gokce T Saeki and M Hisada ldquoAssessmentof recycling process induced damage sensitivity of recycledconcrete aggregatesrdquoCement and Concrete Research vol 34 no6 pp 965ndash971 2004
[4] Z Q Shi Jian guang ldquoEffects of different materials in theconstruction waste on the properties of concreterdquo Concrete vol6 pp 11ndash13 2008
[5] A R Chini S-S Kuo J M Armaghani and J P Duxbury ldquoTestof recycled concrete aggregate in accelerated test trackrdquo Journalof Transportation Engineering vol 127 no 6 pp 486ndash492 2001
[6] V W Y Tam and C M Tam ldquoA new approach in assessingcement mortar remains on recycled aggregaterdquo Magazine ofConcrete Research vol 59 no 6 pp 413ndash422 2007
[7] M Etxeberria A R Marı and E Vazquez ldquoRecycled aggregateconcrete as structural materialrdquo Materials and Structures vol40 no 5 pp 529ndash541 2007
[8] S YWang PH Zhu XWang J Zhou andH E Xia ldquoResearchon some key issues for applicat ion of recycled concreterdquoConcrete vol 4 pp 39ndash41 2008
[9] G R Robinson Jr W D Menzie and H Hyun ldquoRecycling ofconstruction debris as aggregate in the Mid-Atlantic RegionUSArdquo Resources Conservation amp Recycling vol 42 no 3 pp275ndash294 2004
[10] X Yuezhou The Numerical Simulation and Influence of Aggre-gate Gradation on the Compressive Strength of Recycled Aggre-gate Concrete Xiamen University 2007
[11] Z Zhang K Wang H Liu and Z Deng ldquoKey performanceproperties of asphalt mixtures with recycled concrete aggregatefrom low strength concreterdquo Construction and Building Materi-als vol 126 pp 711ndash719 2016
[12] H Sung Do P Hee Mun and R Suk keun ldquoA study onengineering characteristics of asphalt concrete using filler withrecycled waste limerdquoWaste Management vol 28 no 1 pp 191ndash199 2008
[13] Y D Wong D D Sun and D Lai ldquoValue-added utilisationof recycled concrete in hot-mix asphaltrdquo Journal of WasteManagement vol 27 no 2 pp 294ndash301 2007
Advances in Materials Science and Engineering 9
[14] S Wu J Zhong J Zhu and D Wang ldquoInfluence of demolitionwaste used as recycled aggregate on performance of asphaltmixturerdquo Road Materials and Pavement Design vol 14 no 3pp 679ndash688 2013
[15] L Huan The Research for Recycled Concrete Aggregate and ITSApplied in Asphalt Pavement Beijing University of TechnologyBeijing China 2011
[16] R HuiThe Evaluation for Property of Recycled Coarse ConcreteAggregate Basing on Distance Method Beijing University ofTechnology Beijing China 2012
[17] A Stimilli G Ferrotti A Graziani and F Canestrari ldquoPerfor-mance evaluation of a cold-recycled mixture containing highpercentage of reclaimed asphaltrdquo Road Materials and PavementDesign vol 14 no 1 pp 149ndash161 2013
[18] K L Vasconcelos A Bhasin and D N Little ldquoInfluence ofreduced production temperatures on the adhesive propertiesof aggregates and laboratory performance of fine aggregate-asphalt mixturesrdquo Road Materials and Pavement Design vol 11no 1 pp 47ndash64 2011
[19] Z-G Zhao and F Liu ldquoApplication research of statistical mon-itoring index based on mahalanobis distancerdquo Acta AutomaticaSinica vol 34 no 4 pp 493ndash495 2008
31 Property Data of RA Tested and Investigated RA prop-erties can be evaluated by some experimental indices whichinclude property density water absorption needle and flaki-ness content and crushing value [12ndash14]
Sixteen groups of RAs were selected from pavements andbuildings respectively [15 16]Most of the property data wereinvestigated except for groups 119864 119865 and NAs which weretested by ourselves in the laboratory All of the RAs werecrushed by the jaw-crush method (see Tables 2 and 3)
As shown in Tables 2 and 3 the basic property data ofthe NA are generally property density at 263 gcm3 waterabsorption at 16 needle and flakiness content at 52 andcrushing value at 96
32 Variability Analysis of the NA Property Data As canbe seen from Tables 1 and 2 there is a little distinction inapparent relative density between RA and NA Most of theRArsquos apparent relative density is 88sim97 of the NAs about231ndash261 kgm3The shapes of RA fromdifferent sources havea high degree of similarity some of them are even better thanthe NA The shapes of different sources of RA are similarand for some the quality is better than the NA the apparentrelative density and needle-flakiness content are relativelystable while the water absorption and crushing values arerelatively higher The water absorption rate of RA is between2sim10 the highest value even is up to 12 which is 2 to 7times that of the NAThe crushing value of RA varies greatlyand its crush resistance ability is significantly lower than theNA as is shown in Figure 4
After being crushed the RA might produce a largenumber of microcracks on the aggregate surface and theinterior between aggregate and adhesive cement mortar
Besides the RA would be cracked more easily when beingpressed and it contains a variety of impurities on the surfacetherefore the RA strength would be lower than NA As aresult the water absorption and crushing value of RA weresignificantly larger [7 17] and the experimental data showeda large variability as well Since the content of impurities andmicrocrack is influenced by many factors such as the typesof concrete waste engineering construction and crushingmethod and the RA from building contains more impuritiesthan that from pavement its variability would be greater
33 Pavement Property with RA The conclusions of ourprevious study about RArsquos instability are as follows [11]
(1) Compared with NA the RA recycled from low gradeconcrete had lower apparent relative density higher waterabsorption and poorer adhesion to asphalt but had probablylower crushing and wearing value
(2) The rutting deformation of the HMAmixes generallyincreased with the level of RA replacement but the lowtemperature cracking resistance test results showed that thefailure load tensile stress and modulus of the HMA mixeswith RA are all significantly higher than those of themix withNA
(3) The water stability measured by immersion compres-sion tests showed a similar trend to that measured by freeze-thaw splitting strength tests The water stability of the HMAconcrete decreased with increasing RA replacement (up to50) When the RA content reached 75 the water resis-tance of the HMA concrete increased rather than decreasedfurther
Due to instability a large variability has been indicatedby experimental data of RA property index Therefore it isinaccurate to illustrate the whole property through a singleproperty index It is possible that the whole property is better
Advances in Materials Science and Engineering 5
22
23
24
25
26
27
28
aA bB cC dD eE fF gG hH
RA sample number
Building RAPavement RANA
Perfo
rman
ce d
ensit
y (g
cＧ
3)
(a) Contrast chart of density of NA and RA
0
2
4
6
8
10
aA bB cC dD eE fF gG hH
Abso
rptio
n (
)
RA sample number
Building RAPavement RANA(b) Contrast chart of absorption of NA and RA
Building RAPavement RANA
2
25
3
35
4
45
5
55
6
65
7
aA bB cC dD eE fF gG hH
Flak
ines
s con
tent
()
RA sample number
(c) Contrast chart of flakiness of NA and RA
4
8
12
16
20
24
28
32
aA bB cC dD eE fF gG hH
Crus
hing
val
ue (
)
RA sample number
Building RAPavement RANA
(d) Contrast chart of crushing value of NA and RA
Figure 4 Contrast charts of NA and RA properties
in the asphalt mixture while a single property index exhibitspoor results However the distance method is not affectedby the dimension and directly expresses the difference ofmultiple variables or sample set of indices According to thedistance method it is feasible to illustrate the whole propertyof RA
4 Euclidean Distance ComparisonMethod of RA
41 Euclidean Distance Calculation of RA Property (1) Thesingle property Euclidean distance 120588119896119895 is generally
(119909119895 minus 119910119896119894119895)2 (1)
in which 119909119895 is property test value of the NA 119910119896119894119895 is propertytest values of RA 119896 is pavement RAor building RAwhere 119896 =I (pavement RA) and 119896 = II (building RA) 119894 is the numberof test data groups when 119896 = I 119894 = 119860 119861 119862119863 119864 119865 119866119867119896 = II 119894 = 119886 119887 119888 119889 119890 119891 119892 ℎ 119895 is the different property test119895 = 1 (apparent density) 2 (water absorption) 3 (needle andflakiness content) and 4 (crushing value)
(2) The whole property Euclidean distance of RA 120588119896119894 isgenerally
120588119896119894 = radic 4sum119895=1
(119909119895 minus 119910119896119894119895)2 (2)
in which 119909119895 119910119896119894119895 119896 119894 and 119895 are the same as formulation(1)
6 Advances in Materials Science and Engineering
Table 4 Euclidean distance of single property of building and pavement RA
Property index Building RA (119860) Pavement RA (119861) 119860 minus 119861Property density 0516 0241 0275Absorption () 10669 8595 2074Flakiness content () 187 0943 0927Crushing value () 26355 1072 15635
052
1067
187
2636
024
860
094
1072
0
10
20
30
Performance Absorption Flakiness Crushing value
Eucli
dean
dist
ance
Single performances of building and pavement RA
Building RAPavement RA
density content
Figure 5 Single property of RA
(3) The whole property Euclidean distance standardvariance of RA 119878119896 is generally
119878119896 = radicsum119866119892119894=119860119886 (120588119896119894 minus 120588119896119894)27 (3)
in which type 119878119896 is the standard variance of whole propertyEuclidean distance for building RA and pavement RA 120588119896119894is the arithmetic mean value of whole property Euclideandistance for the different number of RA 120588119896119894 119896 and 119894 are thesame as formulation (1)
42 The Comparison and Analysis of Euclidean Distance ofRA Property According to the Euclidean distance formula(1)ndash(3) for the two types of RA the Euclidean distances ofsingle property andwhole property are shown in Tables 4 and5 and Figures 5 and 6 respectively
The greater the value of the properties the greater thedifference between the RA and theNA otherwise the smallerthe value the closer the difference The standard varianceof the whole property Euclidean distance of building RAand pavement RA was calculated to be 7341 and 2208respectively
Table 4 and Figure 5 show slightly differences of theproperty density and flakiness content of RA compared withthe NA while the crushing value and water absorption areobvious differences In addition pavement RA was signifi-cantly better than building RA in terms of single propertyindexAs shown inTable 5 andFigure 6 thewhole property of
2051
1590
450 440 438 388
091 020
789
506 400 372
247 199 150 130
02468
10121416182022
G F h e C A B f H g b c a d D EEu
clide
an d
istan
ce
Whole performance of building and pavement RA
Building RAPavement RA
Figure 6 Whole property of RA
RA with better single property may be poor while the wholeproperty of RA with poor property may be better
From the size of Euclidean distance it can be seen that thecomprehensive properties of group 119864 group 119863 and group 119889are the best whereas group 119866 shows the worst results TheEuclidean distance of crushing value and water absorptionis the largest which shows a primary need to improve theproperty index of the RA
The standard variances of the whole property Euclideandistances are 7341 and 2208 respectively which shows thatthe property of building RA has great fluctuation whilepavement RA is more stable
5 Mahalanobis Distance Evaluation andGrading Methods of RA
The analysis above indicates that the property evaluation ofRA contains multiple indicators Due to the instability of theproperties of RA its experimental data has a large variabilityFor different evaluation indices different evaluation resultsare achieved Since there exists certain dependence amongproperties of RA andEuclidean distance could not express theinfluence factors of the correlation among various propertiesof the samples another index is proposed to solve thisproblem named Mahalanobis distance
51 Mahalanobis Distance Evaluation Model Mahalanobisdistance can discriminate the spatial distribution of a pointrelative to the total by calculating the distance to a certainpoint Mahalanobis distance considers three parameters
Advances in Materials Science and Engineering 7
Table 5 Euclidean distance of whole property of building and pavement RA
mean value variance and covariance which shows thetranslational invariance and invariance under nonsingulartransformation Mahalanobis distance can express the influ-encing factors of the correlation among different variablesand it is independent of the dimension of each characteristicparameter and the judgement basis is according to thecharacteristic distribution condition of variables in the wholespace [18 19]
Suppose that there are two variables 1198661 and 1198662 whichobey the state distribution 119909 isin 119877119901 is a new sample point andthe distance from 119909 to 1198661 and 1198662 is defined as follows
119889 (119909 1198661) = radic(119909 minus 119906(1)Σminus11 ) (119909 minus 119906(1)) 119889 (119909 1198662) = radic(119909 minus 119906(2)Σminus12 ) (119909 minus 119906(2))
(4)
in which 120583(1) 120583(2) are the mean vector of 1198661 and 1198662 Σ1 Σ2are covariance matrix of set 1198661 and 1198662
Therefore it can be determined according to the followingrules
The indices which can express the property of RA aredefined as a vector 119883 (11988311198832 119883119899) the mean vector of119883 is 120583 (1205831 1205832 120583119899) and the standard deviation vector is 120590(1205901 1205902 120590119899) The property standard is defined as vector 119884(1198841 1198842 119884119899) so the vector 119884 is degenerated into a pointof the multidimensional space
The global 1198661 is defined by the vector 119883 and the global1198662 as 119884 then the distance from a sample point of global 1198661to global 1198662 is equal to the distance from a point of vector119883to vector 119884119863 (119883 119884)= radic (1198841 minus 1198831)212059021 + (1198842 minus 1198832)212059022 + sdot sdot sdot + (119884119899 minus 119883119899)21205902119899 (6)
119863 is a normalized distance of nonuniform distributionand it can reflect the comprehensive index of RA propertywithout the need to consider the dimension of each index and
Table 6 Grading methods of RA
Grading of RA Meet the conditionsGrading 1 119863(11990110158401119901) lt 119863(11990110158402119901) cup 119863(11990110158401119901) lt 119863(11990110158403119901)Grading 2 119863(11990110158402119901) lt 119863(11990110158401119901) cup 119863(11990110158402119901) lt 119863(11990110158403119901)Grading 3 119863(11990110158403119901) lt 119863(11990110158401119901) cup 119863(11990110158403119901) lt 119863(11990110158402119901)the choice of unit Based on the characteristic distributioncondition in the whole space 119863 can express the effect ofcorrelation among indices of RA Therefore 119863 can betterdescribe the similarity between RA and the set so as toprovide a basis to determine the level of RA The smaller the119863 value the better the property and vice versa
52 Mahalanobis Distance Calculation and Grading Method
521 Mahalanobis Distance Calculation Water absorptioncrushing value and adhesion are represented by 119882 119862 and119873 respectively The evaluation index of RA comprehensiveproperty can be considered as a three-dimensional vector119875(119882119862119873) and assuming that the value of the three indi-cators is 1198751015840(1198821015840119899 1198621015840119899 1198731015840119899) we can calculate the distance 119863between any RA and NA and then distinguish the wholeproperty of RA
119863(1199011015840119899119901)= radic (119882 minus1198821015840119899)21205902119882 + (119876 minus 1198761015840119899)21205902119862 + sdot sdot sdot + (119873 minus 1198731015840119899)21205902119873 (7)
in which 1198821015840119899 1198761015840119899 1198731015840119899 are the water absorption crushingvalue and adhesion of different grades of NA 119882 119876 119873are the average value of water absorption crushing valueand adhesion of a certain RA 1205902119882 1205902119862 1205902119873 are the standarddeviation of water absorption crushing value and adhesionof a certain RA
522 Grading Method of Mahalanobis Distance When the119863(11990110158401119901) 119863(11990110158402119901) 119863(11990110158403119901) represent respectively Maha-lanobis distance grade 1 grade 2 and grade 3 the RA canbe graded by the methods of Table 6 therefore Mahalanobisdistance comprehensive evaluation method can be applied toevaluate some cases which are difficult to compare each other
8 Advances in Materials Science and Engineering
Table 7 Mahalanobis distance between RA and natural aggregate
523 The Comprehensive Property Evaluation of RA Basedon the RA property test results we can calculate the Maha-lanobis Distance of RA as is shown in Table 7
As can be seen from Table 6 119863(11990110158402119901) lt 119863(11990110158403119901) lt119863(11990110158401119901) and therefore this RA is nearest to the entirety ofsecond class and its level can be represented as two
Compared with the other grading standard Mahalanobisdistance grading method is more intuitive and it is con-venient for operations without requiring a large number ofproperty tests
6 Conclusions
(1) RA is influenced by many factors such as its source andcrushingmethodThe numerical value of the property exper-iments shows obvious variability and instability Comparedto the property of pavement RA building RA has a widervariability and lower stability
(2) The distance method of comparison is not affectedby the dimensions and it can be used to express the wholedifference of various sample sets simply The greater thedistance value the lower the similarity of the two samples
(3)The calculation results of Euclidean distance show thatthe comprehensive property of the pavement RA (119878119896 = 2208)is superior to that of building RA (119878119896 = 7341) And forthese two kinds of RA the Euclidean distance of the property(120588119896119895 = 052) and apparent relative density (120588119896119895 = 187) areslightly different whereas the change of water absorption(120588119896119895 = 1067) and crushing value (120588119896119895 = 2636) is quiteremarkable Thus the water absorption rate and crushingvalue can approximately reflect the property differences ofRA
(4) Euclidean distance calculations show that there is acertain correlation among apparent density water absorptionrate and crushing value The lower the apparent density thelarger the water absorption rate and crushing value Howeverthe correlation for flakiness content is low so the mostimportant property indices for the RA are crushing value andwater absorption and at same time the major improvementfor these two property indices is needed as well
(5) The Mahalanobis distance method can evaluate thecomprehensive property of RA directly and provide gradeevaluation guidance for RAs by considering the influencefactors of the correlation among variables such as parametersof mean variance and covariance
Additional Points
Highlights (i) It is clear that there has been a large variabilityabout the individual property of recycled aggregate (RA) andit has been difficult to express thewhole property in any singleproperty parameter (ii) A method is proposed to expressthe differences of the whole property of RAs using Euclidean
distance method (iii) A calculation model and evaluationmethod are provided for the whole property of RAs withMahalanobis distance method
Conflicts of Interest
The authors declare that they have no conflicts of interest
Acknowledgments
This study was supported by the National Natural ScienceFoundation of China (Project 51578027)
References
[1] F Xiao B Putman and S Amirkhanian ldquoRheological char-acteristics investigation of high percentage RAP binders withWMA technology at various aging statesrdquo Construction andBuilding Materials vol 98 pp 315ndash324 2015
[2] Y C Guo and K W Sun ldquoConsideration on new specificationsof road materials based on properties of recycled ag-gregaterdquoJournal of Changrsquoan University (Natural Science Edition) vol 3pp 1671ndash8879 2010
[3] S Nagataki A Gokce T Saeki and M Hisada ldquoAssessmentof recycling process induced damage sensitivity of recycledconcrete aggregatesrdquoCement and Concrete Research vol 34 no6 pp 965ndash971 2004
[4] Z Q Shi Jian guang ldquoEffects of different materials in theconstruction waste on the properties of concreterdquo Concrete vol6 pp 11ndash13 2008
[5] A R Chini S-S Kuo J M Armaghani and J P Duxbury ldquoTestof recycled concrete aggregate in accelerated test trackrdquo Journalof Transportation Engineering vol 127 no 6 pp 486ndash492 2001
[6] V W Y Tam and C M Tam ldquoA new approach in assessingcement mortar remains on recycled aggregaterdquo Magazine ofConcrete Research vol 59 no 6 pp 413ndash422 2007
[7] M Etxeberria A R Marı and E Vazquez ldquoRecycled aggregateconcrete as structural materialrdquo Materials and Structures vol40 no 5 pp 529ndash541 2007
[8] S YWang PH Zhu XWang J Zhou andH E Xia ldquoResearchon some key issues for applicat ion of recycled concreterdquoConcrete vol 4 pp 39ndash41 2008
[9] G R Robinson Jr W D Menzie and H Hyun ldquoRecycling ofconstruction debris as aggregate in the Mid-Atlantic RegionUSArdquo Resources Conservation amp Recycling vol 42 no 3 pp275ndash294 2004
[10] X Yuezhou The Numerical Simulation and Influence of Aggre-gate Gradation on the Compressive Strength of Recycled Aggre-gate Concrete Xiamen University 2007
[11] Z Zhang K Wang H Liu and Z Deng ldquoKey performanceproperties of asphalt mixtures with recycled concrete aggregatefrom low strength concreterdquo Construction and Building Materi-als vol 126 pp 711ndash719 2016
[12] H Sung Do P Hee Mun and R Suk keun ldquoA study onengineering characteristics of asphalt concrete using filler withrecycled waste limerdquoWaste Management vol 28 no 1 pp 191ndash199 2008
[13] Y D Wong D D Sun and D Lai ldquoValue-added utilisationof recycled concrete in hot-mix asphaltrdquo Journal of WasteManagement vol 27 no 2 pp 294ndash301 2007
Advances in Materials Science and Engineering 9
[14] S Wu J Zhong J Zhu and D Wang ldquoInfluence of demolitionwaste used as recycled aggregate on performance of asphaltmixturerdquo Road Materials and Pavement Design vol 14 no 3pp 679ndash688 2013
[15] L Huan The Research for Recycled Concrete Aggregate and ITSApplied in Asphalt Pavement Beijing University of TechnologyBeijing China 2011
[16] R HuiThe Evaluation for Property of Recycled Coarse ConcreteAggregate Basing on Distance Method Beijing University ofTechnology Beijing China 2012
[17] A Stimilli G Ferrotti A Graziani and F Canestrari ldquoPerfor-mance evaluation of a cold-recycled mixture containing highpercentage of reclaimed asphaltrdquo Road Materials and PavementDesign vol 14 no 1 pp 149ndash161 2013
[18] K L Vasconcelos A Bhasin and D N Little ldquoInfluence ofreduced production temperatures on the adhesive propertiesof aggregates and laboratory performance of fine aggregate-asphalt mixturesrdquo Road Materials and Pavement Design vol 11no 1 pp 47ndash64 2011
[19] Z-G Zhao and F Liu ldquoApplication research of statistical mon-itoring index based on mahalanobis distancerdquo Acta AutomaticaSinica vol 34 no 4 pp 493ndash495 2008
Building RAPavement RANA(b) Contrast chart of absorption of NA and RA
Building RAPavement RANA
2
25
3
35
4
45
5
55
6
65
7
aA bB cC dD eE fF gG hH
Flak
ines
s con
tent
()
RA sample number
(c) Contrast chart of flakiness of NA and RA
4
8
12
16
20
24
28
32
aA bB cC dD eE fF gG hH
Crus
hing
val
ue (
)
RA sample number
Building RAPavement RANA
(d) Contrast chart of crushing value of NA and RA
Figure 4 Contrast charts of NA and RA properties
in the asphalt mixture while a single property index exhibitspoor results However the distance method is not affectedby the dimension and directly expresses the difference ofmultiple variables or sample set of indices According to thedistance method it is feasible to illustrate the whole propertyof RA
4 Euclidean Distance ComparisonMethod of RA
41 Euclidean Distance Calculation of RA Property (1) Thesingle property Euclidean distance 120588119896119895 is generally
(119909119895 minus 119910119896119894119895)2 (1)
in which 119909119895 is property test value of the NA 119910119896119894119895 is propertytest values of RA 119896 is pavement RAor building RAwhere 119896 =I (pavement RA) and 119896 = II (building RA) 119894 is the numberof test data groups when 119896 = I 119894 = 119860 119861 119862119863 119864 119865 119866119867119896 = II 119894 = 119886 119887 119888 119889 119890 119891 119892 ℎ 119895 is the different property test119895 = 1 (apparent density) 2 (water absorption) 3 (needle andflakiness content) and 4 (crushing value)
(2) The whole property Euclidean distance of RA 120588119896119894 isgenerally
120588119896119894 = radic 4sum119895=1
(119909119895 minus 119910119896119894119895)2 (2)
in which 119909119895 119910119896119894119895 119896 119894 and 119895 are the same as formulation(1)
6 Advances in Materials Science and Engineering
Table 4 Euclidean distance of single property of building and pavement RA
Property index Building RA (119860) Pavement RA (119861) 119860 minus 119861Property density 0516 0241 0275Absorption () 10669 8595 2074Flakiness content () 187 0943 0927Crushing value () 26355 1072 15635
052
1067
187
2636
024
860
094
1072
0
10
20
30
Performance Absorption Flakiness Crushing value
Eucli
dean
dist
ance
Single performances of building and pavement RA
Building RAPavement RA
density content
Figure 5 Single property of RA
(3) The whole property Euclidean distance standardvariance of RA 119878119896 is generally
119878119896 = radicsum119866119892119894=119860119886 (120588119896119894 minus 120588119896119894)27 (3)
in which type 119878119896 is the standard variance of whole propertyEuclidean distance for building RA and pavement RA 120588119896119894is the arithmetic mean value of whole property Euclideandistance for the different number of RA 120588119896119894 119896 and 119894 are thesame as formulation (1)
42 The Comparison and Analysis of Euclidean Distance ofRA Property According to the Euclidean distance formula(1)ndash(3) for the two types of RA the Euclidean distances ofsingle property andwhole property are shown in Tables 4 and5 and Figures 5 and 6 respectively
The greater the value of the properties the greater thedifference between the RA and theNA otherwise the smallerthe value the closer the difference The standard varianceof the whole property Euclidean distance of building RAand pavement RA was calculated to be 7341 and 2208respectively
Table 4 and Figure 5 show slightly differences of theproperty density and flakiness content of RA compared withthe NA while the crushing value and water absorption areobvious differences In addition pavement RA was signifi-cantly better than building RA in terms of single propertyindexAs shown inTable 5 andFigure 6 thewhole property of
2051
1590
450 440 438 388
091 020
789
506 400 372
247 199 150 130
02468
10121416182022
G F h e C A B f H g b c a d D EEu
clide
an d
istan
ce
Whole performance of building and pavement RA
Building RAPavement RA
Figure 6 Whole property of RA
RA with better single property may be poor while the wholeproperty of RA with poor property may be better
From the size of Euclidean distance it can be seen that thecomprehensive properties of group 119864 group 119863 and group 119889are the best whereas group 119866 shows the worst results TheEuclidean distance of crushing value and water absorptionis the largest which shows a primary need to improve theproperty index of the RA
The standard variances of the whole property Euclideandistances are 7341 and 2208 respectively which shows thatthe property of building RA has great fluctuation whilepavement RA is more stable
5 Mahalanobis Distance Evaluation andGrading Methods of RA
The analysis above indicates that the property evaluation ofRA contains multiple indicators Due to the instability of theproperties of RA its experimental data has a large variabilityFor different evaluation indices different evaluation resultsare achieved Since there exists certain dependence amongproperties of RA andEuclidean distance could not express theinfluence factors of the correlation among various propertiesof the samples another index is proposed to solve thisproblem named Mahalanobis distance
51 Mahalanobis Distance Evaluation Model Mahalanobisdistance can discriminate the spatial distribution of a pointrelative to the total by calculating the distance to a certainpoint Mahalanobis distance considers three parameters
Advances in Materials Science and Engineering 7
Table 5 Euclidean distance of whole property of building and pavement RA
mean value variance and covariance which shows thetranslational invariance and invariance under nonsingulartransformation Mahalanobis distance can express the influ-encing factors of the correlation among different variablesand it is independent of the dimension of each characteristicparameter and the judgement basis is according to thecharacteristic distribution condition of variables in the wholespace [18 19]
Suppose that there are two variables 1198661 and 1198662 whichobey the state distribution 119909 isin 119877119901 is a new sample point andthe distance from 119909 to 1198661 and 1198662 is defined as follows
119889 (119909 1198661) = radic(119909 minus 119906(1)Σminus11 ) (119909 minus 119906(1)) 119889 (119909 1198662) = radic(119909 minus 119906(2)Σminus12 ) (119909 minus 119906(2))
(4)
in which 120583(1) 120583(2) are the mean vector of 1198661 and 1198662 Σ1 Σ2are covariance matrix of set 1198661 and 1198662
Therefore it can be determined according to the followingrules
The indices which can express the property of RA aredefined as a vector 119883 (11988311198832 119883119899) the mean vector of119883 is 120583 (1205831 1205832 120583119899) and the standard deviation vector is 120590(1205901 1205902 120590119899) The property standard is defined as vector 119884(1198841 1198842 119884119899) so the vector 119884 is degenerated into a pointof the multidimensional space
The global 1198661 is defined by the vector 119883 and the global1198662 as 119884 then the distance from a sample point of global 1198661to global 1198662 is equal to the distance from a point of vector119883to vector 119884119863 (119883 119884)= radic (1198841 minus 1198831)212059021 + (1198842 minus 1198832)212059022 + sdot sdot sdot + (119884119899 minus 119883119899)21205902119899 (6)
119863 is a normalized distance of nonuniform distributionand it can reflect the comprehensive index of RA propertywithout the need to consider the dimension of each index and
Table 6 Grading methods of RA
Grading of RA Meet the conditionsGrading 1 119863(11990110158401119901) lt 119863(11990110158402119901) cup 119863(11990110158401119901) lt 119863(11990110158403119901)Grading 2 119863(11990110158402119901) lt 119863(11990110158401119901) cup 119863(11990110158402119901) lt 119863(11990110158403119901)Grading 3 119863(11990110158403119901) lt 119863(11990110158401119901) cup 119863(11990110158403119901) lt 119863(11990110158402119901)the choice of unit Based on the characteristic distributioncondition in the whole space 119863 can express the effect ofcorrelation among indices of RA Therefore 119863 can betterdescribe the similarity between RA and the set so as toprovide a basis to determine the level of RA The smaller the119863 value the better the property and vice versa
52 Mahalanobis Distance Calculation and Grading Method
521 Mahalanobis Distance Calculation Water absorptioncrushing value and adhesion are represented by 119882 119862 and119873 respectively The evaluation index of RA comprehensiveproperty can be considered as a three-dimensional vector119875(119882119862119873) and assuming that the value of the three indi-cators is 1198751015840(1198821015840119899 1198621015840119899 1198731015840119899) we can calculate the distance 119863between any RA and NA and then distinguish the wholeproperty of RA
119863(1199011015840119899119901)= radic (119882 minus1198821015840119899)21205902119882 + (119876 minus 1198761015840119899)21205902119862 + sdot sdot sdot + (119873 minus 1198731015840119899)21205902119873 (7)
in which 1198821015840119899 1198761015840119899 1198731015840119899 are the water absorption crushingvalue and adhesion of different grades of NA 119882 119876 119873are the average value of water absorption crushing valueand adhesion of a certain RA 1205902119882 1205902119862 1205902119873 are the standarddeviation of water absorption crushing value and adhesionof a certain RA
522 Grading Method of Mahalanobis Distance When the119863(11990110158401119901) 119863(11990110158402119901) 119863(11990110158403119901) represent respectively Maha-lanobis distance grade 1 grade 2 and grade 3 the RA canbe graded by the methods of Table 6 therefore Mahalanobisdistance comprehensive evaluation method can be applied toevaluate some cases which are difficult to compare each other
8 Advances in Materials Science and Engineering
Table 7 Mahalanobis distance between RA and natural aggregate
523 The Comprehensive Property Evaluation of RA Basedon the RA property test results we can calculate the Maha-lanobis Distance of RA as is shown in Table 7
As can be seen from Table 6 119863(11990110158402119901) lt 119863(11990110158403119901) lt119863(11990110158401119901) and therefore this RA is nearest to the entirety ofsecond class and its level can be represented as two
Compared with the other grading standard Mahalanobisdistance grading method is more intuitive and it is con-venient for operations without requiring a large number ofproperty tests
6 Conclusions
(1) RA is influenced by many factors such as its source andcrushingmethodThe numerical value of the property exper-iments shows obvious variability and instability Comparedto the property of pavement RA building RA has a widervariability and lower stability
(2) The distance method of comparison is not affectedby the dimensions and it can be used to express the wholedifference of various sample sets simply The greater thedistance value the lower the similarity of the two samples
(3)The calculation results of Euclidean distance show thatthe comprehensive property of the pavement RA (119878119896 = 2208)is superior to that of building RA (119878119896 = 7341) And forthese two kinds of RA the Euclidean distance of the property(120588119896119895 = 052) and apparent relative density (120588119896119895 = 187) areslightly different whereas the change of water absorption(120588119896119895 = 1067) and crushing value (120588119896119895 = 2636) is quiteremarkable Thus the water absorption rate and crushingvalue can approximately reflect the property differences ofRA
(4) Euclidean distance calculations show that there is acertain correlation among apparent density water absorptionrate and crushing value The lower the apparent density thelarger the water absorption rate and crushing value Howeverthe correlation for flakiness content is low so the mostimportant property indices for the RA are crushing value andwater absorption and at same time the major improvementfor these two property indices is needed as well
(5) The Mahalanobis distance method can evaluate thecomprehensive property of RA directly and provide gradeevaluation guidance for RAs by considering the influencefactors of the correlation among variables such as parametersof mean variance and covariance
Additional Points
Highlights (i) It is clear that there has been a large variabilityabout the individual property of recycled aggregate (RA) andit has been difficult to express thewhole property in any singleproperty parameter (ii) A method is proposed to expressthe differences of the whole property of RAs using Euclidean
distance method (iii) A calculation model and evaluationmethod are provided for the whole property of RAs withMahalanobis distance method
Conflicts of Interest
The authors declare that they have no conflicts of interest
Acknowledgments
This study was supported by the National Natural ScienceFoundation of China (Project 51578027)
References
[1] F Xiao B Putman and S Amirkhanian ldquoRheological char-acteristics investigation of high percentage RAP binders withWMA technology at various aging statesrdquo Construction andBuilding Materials vol 98 pp 315ndash324 2015
[2] Y C Guo and K W Sun ldquoConsideration on new specificationsof road materials based on properties of recycled ag-gregaterdquoJournal of Changrsquoan University (Natural Science Edition) vol 3pp 1671ndash8879 2010
[3] S Nagataki A Gokce T Saeki and M Hisada ldquoAssessmentof recycling process induced damage sensitivity of recycledconcrete aggregatesrdquoCement and Concrete Research vol 34 no6 pp 965ndash971 2004
[4] Z Q Shi Jian guang ldquoEffects of different materials in theconstruction waste on the properties of concreterdquo Concrete vol6 pp 11ndash13 2008
[5] A R Chini S-S Kuo J M Armaghani and J P Duxbury ldquoTestof recycled concrete aggregate in accelerated test trackrdquo Journalof Transportation Engineering vol 127 no 6 pp 486ndash492 2001
[6] V W Y Tam and C M Tam ldquoA new approach in assessingcement mortar remains on recycled aggregaterdquo Magazine ofConcrete Research vol 59 no 6 pp 413ndash422 2007
[7] M Etxeberria A R Marı and E Vazquez ldquoRecycled aggregateconcrete as structural materialrdquo Materials and Structures vol40 no 5 pp 529ndash541 2007
[8] S YWang PH Zhu XWang J Zhou andH E Xia ldquoResearchon some key issues for applicat ion of recycled concreterdquoConcrete vol 4 pp 39ndash41 2008
[9] G R Robinson Jr W D Menzie and H Hyun ldquoRecycling ofconstruction debris as aggregate in the Mid-Atlantic RegionUSArdquo Resources Conservation amp Recycling vol 42 no 3 pp275ndash294 2004
[10] X Yuezhou The Numerical Simulation and Influence of Aggre-gate Gradation on the Compressive Strength of Recycled Aggre-gate Concrete Xiamen University 2007
[11] Z Zhang K Wang H Liu and Z Deng ldquoKey performanceproperties of asphalt mixtures with recycled concrete aggregatefrom low strength concreterdquo Construction and Building Materi-als vol 126 pp 711ndash719 2016
[12] H Sung Do P Hee Mun and R Suk keun ldquoA study onengineering characteristics of asphalt concrete using filler withrecycled waste limerdquoWaste Management vol 28 no 1 pp 191ndash199 2008
[13] Y D Wong D D Sun and D Lai ldquoValue-added utilisationof recycled concrete in hot-mix asphaltrdquo Journal of WasteManagement vol 27 no 2 pp 294ndash301 2007
Advances in Materials Science and Engineering 9
[14] S Wu J Zhong J Zhu and D Wang ldquoInfluence of demolitionwaste used as recycled aggregate on performance of asphaltmixturerdquo Road Materials and Pavement Design vol 14 no 3pp 679ndash688 2013
[15] L Huan The Research for Recycled Concrete Aggregate and ITSApplied in Asphalt Pavement Beijing University of TechnologyBeijing China 2011
[16] R HuiThe Evaluation for Property of Recycled Coarse ConcreteAggregate Basing on Distance Method Beijing University ofTechnology Beijing China 2012
[17] A Stimilli G Ferrotti A Graziani and F Canestrari ldquoPerfor-mance evaluation of a cold-recycled mixture containing highpercentage of reclaimed asphaltrdquo Road Materials and PavementDesign vol 14 no 1 pp 149ndash161 2013
[18] K L Vasconcelos A Bhasin and D N Little ldquoInfluence ofreduced production temperatures on the adhesive propertiesof aggregates and laboratory performance of fine aggregate-asphalt mixturesrdquo Road Materials and Pavement Design vol 11no 1 pp 47ndash64 2011
[19] Z-G Zhao and F Liu ldquoApplication research of statistical mon-itoring index based on mahalanobis distancerdquo Acta AutomaticaSinica vol 34 no 4 pp 493ndash495 2008
Table 4 Euclidean distance of single property of building and pavement RA
Property index Building RA (119860) Pavement RA (119861) 119860 minus 119861Property density 0516 0241 0275Absorption () 10669 8595 2074Flakiness content () 187 0943 0927Crushing value () 26355 1072 15635
052
1067
187
2636
024
860
094
1072
0
10
20
30
Performance Absorption Flakiness Crushing value
Eucli
dean
dist
ance
Single performances of building and pavement RA
Building RAPavement RA
density content
Figure 5 Single property of RA
(3) The whole property Euclidean distance standardvariance of RA 119878119896 is generally
119878119896 = radicsum119866119892119894=119860119886 (120588119896119894 minus 120588119896119894)27 (3)
in which type 119878119896 is the standard variance of whole propertyEuclidean distance for building RA and pavement RA 120588119896119894is the arithmetic mean value of whole property Euclideandistance for the different number of RA 120588119896119894 119896 and 119894 are thesame as formulation (1)
42 The Comparison and Analysis of Euclidean Distance ofRA Property According to the Euclidean distance formula(1)ndash(3) for the two types of RA the Euclidean distances ofsingle property andwhole property are shown in Tables 4 and5 and Figures 5 and 6 respectively
The greater the value of the properties the greater thedifference between the RA and theNA otherwise the smallerthe value the closer the difference The standard varianceof the whole property Euclidean distance of building RAand pavement RA was calculated to be 7341 and 2208respectively
Table 4 and Figure 5 show slightly differences of theproperty density and flakiness content of RA compared withthe NA while the crushing value and water absorption areobvious differences In addition pavement RA was signifi-cantly better than building RA in terms of single propertyindexAs shown inTable 5 andFigure 6 thewhole property of
2051
1590
450 440 438 388
091 020
789
506 400 372
247 199 150 130
02468
10121416182022
G F h e C A B f H g b c a d D EEu
clide
an d
istan
ce
Whole performance of building and pavement RA
Building RAPavement RA
Figure 6 Whole property of RA
RA with better single property may be poor while the wholeproperty of RA with poor property may be better
From the size of Euclidean distance it can be seen that thecomprehensive properties of group 119864 group 119863 and group 119889are the best whereas group 119866 shows the worst results TheEuclidean distance of crushing value and water absorptionis the largest which shows a primary need to improve theproperty index of the RA
The standard variances of the whole property Euclideandistances are 7341 and 2208 respectively which shows thatthe property of building RA has great fluctuation whilepavement RA is more stable
5 Mahalanobis Distance Evaluation andGrading Methods of RA
The analysis above indicates that the property evaluation ofRA contains multiple indicators Due to the instability of theproperties of RA its experimental data has a large variabilityFor different evaluation indices different evaluation resultsare achieved Since there exists certain dependence amongproperties of RA andEuclidean distance could not express theinfluence factors of the correlation among various propertiesof the samples another index is proposed to solve thisproblem named Mahalanobis distance
51 Mahalanobis Distance Evaluation Model Mahalanobisdistance can discriminate the spatial distribution of a pointrelative to the total by calculating the distance to a certainpoint Mahalanobis distance considers three parameters
Advances in Materials Science and Engineering 7
Table 5 Euclidean distance of whole property of building and pavement RA
mean value variance and covariance which shows thetranslational invariance and invariance under nonsingulartransformation Mahalanobis distance can express the influ-encing factors of the correlation among different variablesand it is independent of the dimension of each characteristicparameter and the judgement basis is according to thecharacteristic distribution condition of variables in the wholespace [18 19]
Suppose that there are two variables 1198661 and 1198662 whichobey the state distribution 119909 isin 119877119901 is a new sample point andthe distance from 119909 to 1198661 and 1198662 is defined as follows
119889 (119909 1198661) = radic(119909 minus 119906(1)Σminus11 ) (119909 minus 119906(1)) 119889 (119909 1198662) = radic(119909 minus 119906(2)Σminus12 ) (119909 minus 119906(2))
(4)
in which 120583(1) 120583(2) are the mean vector of 1198661 and 1198662 Σ1 Σ2are covariance matrix of set 1198661 and 1198662
Therefore it can be determined according to the followingrules
The indices which can express the property of RA aredefined as a vector 119883 (11988311198832 119883119899) the mean vector of119883 is 120583 (1205831 1205832 120583119899) and the standard deviation vector is 120590(1205901 1205902 120590119899) The property standard is defined as vector 119884(1198841 1198842 119884119899) so the vector 119884 is degenerated into a pointof the multidimensional space
The global 1198661 is defined by the vector 119883 and the global1198662 as 119884 then the distance from a sample point of global 1198661to global 1198662 is equal to the distance from a point of vector119883to vector 119884119863 (119883 119884)= radic (1198841 minus 1198831)212059021 + (1198842 minus 1198832)212059022 + sdot sdot sdot + (119884119899 minus 119883119899)21205902119899 (6)
119863 is a normalized distance of nonuniform distributionand it can reflect the comprehensive index of RA propertywithout the need to consider the dimension of each index and
Table 6 Grading methods of RA
Grading of RA Meet the conditionsGrading 1 119863(11990110158401119901) lt 119863(11990110158402119901) cup 119863(11990110158401119901) lt 119863(11990110158403119901)Grading 2 119863(11990110158402119901) lt 119863(11990110158401119901) cup 119863(11990110158402119901) lt 119863(11990110158403119901)Grading 3 119863(11990110158403119901) lt 119863(11990110158401119901) cup 119863(11990110158403119901) lt 119863(11990110158402119901)the choice of unit Based on the characteristic distributioncondition in the whole space 119863 can express the effect ofcorrelation among indices of RA Therefore 119863 can betterdescribe the similarity between RA and the set so as toprovide a basis to determine the level of RA The smaller the119863 value the better the property and vice versa
52 Mahalanobis Distance Calculation and Grading Method
521 Mahalanobis Distance Calculation Water absorptioncrushing value and adhesion are represented by 119882 119862 and119873 respectively The evaluation index of RA comprehensiveproperty can be considered as a three-dimensional vector119875(119882119862119873) and assuming that the value of the three indi-cators is 1198751015840(1198821015840119899 1198621015840119899 1198731015840119899) we can calculate the distance 119863between any RA and NA and then distinguish the wholeproperty of RA
119863(1199011015840119899119901)= radic (119882 minus1198821015840119899)21205902119882 + (119876 minus 1198761015840119899)21205902119862 + sdot sdot sdot + (119873 minus 1198731015840119899)21205902119873 (7)
in which 1198821015840119899 1198761015840119899 1198731015840119899 are the water absorption crushingvalue and adhesion of different grades of NA 119882 119876 119873are the average value of water absorption crushing valueand adhesion of a certain RA 1205902119882 1205902119862 1205902119873 are the standarddeviation of water absorption crushing value and adhesionof a certain RA
522 Grading Method of Mahalanobis Distance When the119863(11990110158401119901) 119863(11990110158402119901) 119863(11990110158403119901) represent respectively Maha-lanobis distance grade 1 grade 2 and grade 3 the RA canbe graded by the methods of Table 6 therefore Mahalanobisdistance comprehensive evaluation method can be applied toevaluate some cases which are difficult to compare each other
8 Advances in Materials Science and Engineering
Table 7 Mahalanobis distance between RA and natural aggregate
523 The Comprehensive Property Evaluation of RA Basedon the RA property test results we can calculate the Maha-lanobis Distance of RA as is shown in Table 7
As can be seen from Table 6 119863(11990110158402119901) lt 119863(11990110158403119901) lt119863(11990110158401119901) and therefore this RA is nearest to the entirety ofsecond class and its level can be represented as two
Compared with the other grading standard Mahalanobisdistance grading method is more intuitive and it is con-venient for operations without requiring a large number ofproperty tests
6 Conclusions
(1) RA is influenced by many factors such as its source andcrushingmethodThe numerical value of the property exper-iments shows obvious variability and instability Comparedto the property of pavement RA building RA has a widervariability and lower stability
(2) The distance method of comparison is not affectedby the dimensions and it can be used to express the wholedifference of various sample sets simply The greater thedistance value the lower the similarity of the two samples
(3)The calculation results of Euclidean distance show thatthe comprehensive property of the pavement RA (119878119896 = 2208)is superior to that of building RA (119878119896 = 7341) And forthese two kinds of RA the Euclidean distance of the property(120588119896119895 = 052) and apparent relative density (120588119896119895 = 187) areslightly different whereas the change of water absorption(120588119896119895 = 1067) and crushing value (120588119896119895 = 2636) is quiteremarkable Thus the water absorption rate and crushingvalue can approximately reflect the property differences ofRA
(4) Euclidean distance calculations show that there is acertain correlation among apparent density water absorptionrate and crushing value The lower the apparent density thelarger the water absorption rate and crushing value Howeverthe correlation for flakiness content is low so the mostimportant property indices for the RA are crushing value andwater absorption and at same time the major improvementfor these two property indices is needed as well
(5) The Mahalanobis distance method can evaluate thecomprehensive property of RA directly and provide gradeevaluation guidance for RAs by considering the influencefactors of the correlation among variables such as parametersof mean variance and covariance
Additional Points
Highlights (i) It is clear that there has been a large variabilityabout the individual property of recycled aggregate (RA) andit has been difficult to express thewhole property in any singleproperty parameter (ii) A method is proposed to expressthe differences of the whole property of RAs using Euclidean
distance method (iii) A calculation model and evaluationmethod are provided for the whole property of RAs withMahalanobis distance method
Conflicts of Interest
The authors declare that they have no conflicts of interest
Acknowledgments
This study was supported by the National Natural ScienceFoundation of China (Project 51578027)
References
[1] F Xiao B Putman and S Amirkhanian ldquoRheological char-acteristics investigation of high percentage RAP binders withWMA technology at various aging statesrdquo Construction andBuilding Materials vol 98 pp 315ndash324 2015
[2] Y C Guo and K W Sun ldquoConsideration on new specificationsof road materials based on properties of recycled ag-gregaterdquoJournal of Changrsquoan University (Natural Science Edition) vol 3pp 1671ndash8879 2010
[3] S Nagataki A Gokce T Saeki and M Hisada ldquoAssessmentof recycling process induced damage sensitivity of recycledconcrete aggregatesrdquoCement and Concrete Research vol 34 no6 pp 965ndash971 2004
[4] Z Q Shi Jian guang ldquoEffects of different materials in theconstruction waste on the properties of concreterdquo Concrete vol6 pp 11ndash13 2008
[5] A R Chini S-S Kuo J M Armaghani and J P Duxbury ldquoTestof recycled concrete aggregate in accelerated test trackrdquo Journalof Transportation Engineering vol 127 no 6 pp 486ndash492 2001
[6] V W Y Tam and C M Tam ldquoA new approach in assessingcement mortar remains on recycled aggregaterdquo Magazine ofConcrete Research vol 59 no 6 pp 413ndash422 2007
[7] M Etxeberria A R Marı and E Vazquez ldquoRecycled aggregateconcrete as structural materialrdquo Materials and Structures vol40 no 5 pp 529ndash541 2007
[8] S YWang PH Zhu XWang J Zhou andH E Xia ldquoResearchon some key issues for applicat ion of recycled concreterdquoConcrete vol 4 pp 39ndash41 2008
[9] G R Robinson Jr W D Menzie and H Hyun ldquoRecycling ofconstruction debris as aggregate in the Mid-Atlantic RegionUSArdquo Resources Conservation amp Recycling vol 42 no 3 pp275ndash294 2004
[10] X Yuezhou The Numerical Simulation and Influence of Aggre-gate Gradation on the Compressive Strength of Recycled Aggre-gate Concrete Xiamen University 2007
[11] Z Zhang K Wang H Liu and Z Deng ldquoKey performanceproperties of asphalt mixtures with recycled concrete aggregatefrom low strength concreterdquo Construction and Building Materi-als vol 126 pp 711ndash719 2016
[12] H Sung Do P Hee Mun and R Suk keun ldquoA study onengineering characteristics of asphalt concrete using filler withrecycled waste limerdquoWaste Management vol 28 no 1 pp 191ndash199 2008
[13] Y D Wong D D Sun and D Lai ldquoValue-added utilisationof recycled concrete in hot-mix asphaltrdquo Journal of WasteManagement vol 27 no 2 pp 294ndash301 2007
Advances in Materials Science and Engineering 9
[14] S Wu J Zhong J Zhu and D Wang ldquoInfluence of demolitionwaste used as recycled aggregate on performance of asphaltmixturerdquo Road Materials and Pavement Design vol 14 no 3pp 679ndash688 2013
[15] L Huan The Research for Recycled Concrete Aggregate and ITSApplied in Asphalt Pavement Beijing University of TechnologyBeijing China 2011
[16] R HuiThe Evaluation for Property of Recycled Coarse ConcreteAggregate Basing on Distance Method Beijing University ofTechnology Beijing China 2012
[17] A Stimilli G Ferrotti A Graziani and F Canestrari ldquoPerfor-mance evaluation of a cold-recycled mixture containing highpercentage of reclaimed asphaltrdquo Road Materials and PavementDesign vol 14 no 1 pp 149ndash161 2013
[18] K L Vasconcelos A Bhasin and D N Little ldquoInfluence ofreduced production temperatures on the adhesive propertiesof aggregates and laboratory performance of fine aggregate-asphalt mixturesrdquo Road Materials and Pavement Design vol 11no 1 pp 47ndash64 2011
[19] Z-G Zhao and F Liu ldquoApplication research of statistical mon-itoring index based on mahalanobis distancerdquo Acta AutomaticaSinica vol 34 no 4 pp 493ndash495 2008
mean value variance and covariance which shows thetranslational invariance and invariance under nonsingulartransformation Mahalanobis distance can express the influ-encing factors of the correlation among different variablesand it is independent of the dimension of each characteristicparameter and the judgement basis is according to thecharacteristic distribution condition of variables in the wholespace [18 19]
Suppose that there are two variables 1198661 and 1198662 whichobey the state distribution 119909 isin 119877119901 is a new sample point andthe distance from 119909 to 1198661 and 1198662 is defined as follows
119889 (119909 1198661) = radic(119909 minus 119906(1)Σminus11 ) (119909 minus 119906(1)) 119889 (119909 1198662) = radic(119909 minus 119906(2)Σminus12 ) (119909 minus 119906(2))
(4)
in which 120583(1) 120583(2) are the mean vector of 1198661 and 1198662 Σ1 Σ2are covariance matrix of set 1198661 and 1198662
Therefore it can be determined according to the followingrules
The indices which can express the property of RA aredefined as a vector 119883 (11988311198832 119883119899) the mean vector of119883 is 120583 (1205831 1205832 120583119899) and the standard deviation vector is 120590(1205901 1205902 120590119899) The property standard is defined as vector 119884(1198841 1198842 119884119899) so the vector 119884 is degenerated into a pointof the multidimensional space
The global 1198661 is defined by the vector 119883 and the global1198662 as 119884 then the distance from a sample point of global 1198661to global 1198662 is equal to the distance from a point of vector119883to vector 119884119863 (119883 119884)= radic (1198841 minus 1198831)212059021 + (1198842 minus 1198832)212059022 + sdot sdot sdot + (119884119899 minus 119883119899)21205902119899 (6)
119863 is a normalized distance of nonuniform distributionand it can reflect the comprehensive index of RA propertywithout the need to consider the dimension of each index and
Table 6 Grading methods of RA
Grading of RA Meet the conditionsGrading 1 119863(11990110158401119901) lt 119863(11990110158402119901) cup 119863(11990110158401119901) lt 119863(11990110158403119901)Grading 2 119863(11990110158402119901) lt 119863(11990110158401119901) cup 119863(11990110158402119901) lt 119863(11990110158403119901)Grading 3 119863(11990110158403119901) lt 119863(11990110158401119901) cup 119863(11990110158403119901) lt 119863(11990110158402119901)the choice of unit Based on the characteristic distributioncondition in the whole space 119863 can express the effect ofcorrelation among indices of RA Therefore 119863 can betterdescribe the similarity between RA and the set so as toprovide a basis to determine the level of RA The smaller the119863 value the better the property and vice versa
52 Mahalanobis Distance Calculation and Grading Method
521 Mahalanobis Distance Calculation Water absorptioncrushing value and adhesion are represented by 119882 119862 and119873 respectively The evaluation index of RA comprehensiveproperty can be considered as a three-dimensional vector119875(119882119862119873) and assuming that the value of the three indi-cators is 1198751015840(1198821015840119899 1198621015840119899 1198731015840119899) we can calculate the distance 119863between any RA and NA and then distinguish the wholeproperty of RA
119863(1199011015840119899119901)= radic (119882 minus1198821015840119899)21205902119882 + (119876 minus 1198761015840119899)21205902119862 + sdot sdot sdot + (119873 minus 1198731015840119899)21205902119873 (7)
in which 1198821015840119899 1198761015840119899 1198731015840119899 are the water absorption crushingvalue and adhesion of different grades of NA 119882 119876 119873are the average value of water absorption crushing valueand adhesion of a certain RA 1205902119882 1205902119862 1205902119873 are the standarddeviation of water absorption crushing value and adhesionof a certain RA
522 Grading Method of Mahalanobis Distance When the119863(11990110158401119901) 119863(11990110158402119901) 119863(11990110158403119901) represent respectively Maha-lanobis distance grade 1 grade 2 and grade 3 the RA canbe graded by the methods of Table 6 therefore Mahalanobisdistance comprehensive evaluation method can be applied toevaluate some cases which are difficult to compare each other
8 Advances in Materials Science and Engineering
Table 7 Mahalanobis distance between RA and natural aggregate
523 The Comprehensive Property Evaluation of RA Basedon the RA property test results we can calculate the Maha-lanobis Distance of RA as is shown in Table 7
As can be seen from Table 6 119863(11990110158402119901) lt 119863(11990110158403119901) lt119863(11990110158401119901) and therefore this RA is nearest to the entirety ofsecond class and its level can be represented as two
Compared with the other grading standard Mahalanobisdistance grading method is more intuitive and it is con-venient for operations without requiring a large number ofproperty tests
6 Conclusions
(1) RA is influenced by many factors such as its source andcrushingmethodThe numerical value of the property exper-iments shows obvious variability and instability Comparedto the property of pavement RA building RA has a widervariability and lower stability
(2) The distance method of comparison is not affectedby the dimensions and it can be used to express the wholedifference of various sample sets simply The greater thedistance value the lower the similarity of the two samples
(3)The calculation results of Euclidean distance show thatthe comprehensive property of the pavement RA (119878119896 = 2208)is superior to that of building RA (119878119896 = 7341) And forthese two kinds of RA the Euclidean distance of the property(120588119896119895 = 052) and apparent relative density (120588119896119895 = 187) areslightly different whereas the change of water absorption(120588119896119895 = 1067) and crushing value (120588119896119895 = 2636) is quiteremarkable Thus the water absorption rate and crushingvalue can approximately reflect the property differences ofRA
(4) Euclidean distance calculations show that there is acertain correlation among apparent density water absorptionrate and crushing value The lower the apparent density thelarger the water absorption rate and crushing value Howeverthe correlation for flakiness content is low so the mostimportant property indices for the RA are crushing value andwater absorption and at same time the major improvementfor these two property indices is needed as well
(5) The Mahalanobis distance method can evaluate thecomprehensive property of RA directly and provide gradeevaluation guidance for RAs by considering the influencefactors of the correlation among variables such as parametersof mean variance and covariance
Additional Points
Highlights (i) It is clear that there has been a large variabilityabout the individual property of recycled aggregate (RA) andit has been difficult to express thewhole property in any singleproperty parameter (ii) A method is proposed to expressthe differences of the whole property of RAs using Euclidean
distance method (iii) A calculation model and evaluationmethod are provided for the whole property of RAs withMahalanobis distance method
Conflicts of Interest
The authors declare that they have no conflicts of interest
Acknowledgments
This study was supported by the National Natural ScienceFoundation of China (Project 51578027)
References
[1] F Xiao B Putman and S Amirkhanian ldquoRheological char-acteristics investigation of high percentage RAP binders withWMA technology at various aging statesrdquo Construction andBuilding Materials vol 98 pp 315ndash324 2015
[2] Y C Guo and K W Sun ldquoConsideration on new specificationsof road materials based on properties of recycled ag-gregaterdquoJournal of Changrsquoan University (Natural Science Edition) vol 3pp 1671ndash8879 2010
[3] S Nagataki A Gokce T Saeki and M Hisada ldquoAssessmentof recycling process induced damage sensitivity of recycledconcrete aggregatesrdquoCement and Concrete Research vol 34 no6 pp 965ndash971 2004
[4] Z Q Shi Jian guang ldquoEffects of different materials in theconstruction waste on the properties of concreterdquo Concrete vol6 pp 11ndash13 2008
[5] A R Chini S-S Kuo J M Armaghani and J P Duxbury ldquoTestof recycled concrete aggregate in accelerated test trackrdquo Journalof Transportation Engineering vol 127 no 6 pp 486ndash492 2001
[6] V W Y Tam and C M Tam ldquoA new approach in assessingcement mortar remains on recycled aggregaterdquo Magazine ofConcrete Research vol 59 no 6 pp 413ndash422 2007
[7] M Etxeberria A R Marı and E Vazquez ldquoRecycled aggregateconcrete as structural materialrdquo Materials and Structures vol40 no 5 pp 529ndash541 2007
[8] S YWang PH Zhu XWang J Zhou andH E Xia ldquoResearchon some key issues for applicat ion of recycled concreterdquoConcrete vol 4 pp 39ndash41 2008
[9] G R Robinson Jr W D Menzie and H Hyun ldquoRecycling ofconstruction debris as aggregate in the Mid-Atlantic RegionUSArdquo Resources Conservation amp Recycling vol 42 no 3 pp275ndash294 2004
[10] X Yuezhou The Numerical Simulation and Influence of Aggre-gate Gradation on the Compressive Strength of Recycled Aggre-gate Concrete Xiamen University 2007
[11] Z Zhang K Wang H Liu and Z Deng ldquoKey performanceproperties of asphalt mixtures with recycled concrete aggregatefrom low strength concreterdquo Construction and Building Materi-als vol 126 pp 711ndash719 2016
[12] H Sung Do P Hee Mun and R Suk keun ldquoA study onengineering characteristics of asphalt concrete using filler withrecycled waste limerdquoWaste Management vol 28 no 1 pp 191ndash199 2008
[13] Y D Wong D D Sun and D Lai ldquoValue-added utilisationof recycled concrete in hot-mix asphaltrdquo Journal of WasteManagement vol 27 no 2 pp 294ndash301 2007
Advances in Materials Science and Engineering 9
[14] S Wu J Zhong J Zhu and D Wang ldquoInfluence of demolitionwaste used as recycled aggregate on performance of asphaltmixturerdquo Road Materials and Pavement Design vol 14 no 3pp 679ndash688 2013
[15] L Huan The Research for Recycled Concrete Aggregate and ITSApplied in Asphalt Pavement Beijing University of TechnologyBeijing China 2011
[16] R HuiThe Evaluation for Property of Recycled Coarse ConcreteAggregate Basing on Distance Method Beijing University ofTechnology Beijing China 2012
[17] A Stimilli G Ferrotti A Graziani and F Canestrari ldquoPerfor-mance evaluation of a cold-recycled mixture containing highpercentage of reclaimed asphaltrdquo Road Materials and PavementDesign vol 14 no 1 pp 149ndash161 2013
[18] K L Vasconcelos A Bhasin and D N Little ldquoInfluence ofreduced production temperatures on the adhesive propertiesof aggregates and laboratory performance of fine aggregate-asphalt mixturesrdquo Road Materials and Pavement Design vol 11no 1 pp 47ndash64 2011
[19] Z-G Zhao and F Liu ldquoApplication research of statistical mon-itoring index based on mahalanobis distancerdquo Acta AutomaticaSinica vol 34 no 4 pp 493ndash495 2008
523 The Comprehensive Property Evaluation of RA Basedon the RA property test results we can calculate the Maha-lanobis Distance of RA as is shown in Table 7
As can be seen from Table 6 119863(11990110158402119901) lt 119863(11990110158403119901) lt119863(11990110158401119901) and therefore this RA is nearest to the entirety ofsecond class and its level can be represented as two
Compared with the other grading standard Mahalanobisdistance grading method is more intuitive and it is con-venient for operations without requiring a large number ofproperty tests
6 Conclusions
(1) RA is influenced by many factors such as its source andcrushingmethodThe numerical value of the property exper-iments shows obvious variability and instability Comparedto the property of pavement RA building RA has a widervariability and lower stability
(2) The distance method of comparison is not affectedby the dimensions and it can be used to express the wholedifference of various sample sets simply The greater thedistance value the lower the similarity of the two samples
(3)The calculation results of Euclidean distance show thatthe comprehensive property of the pavement RA (119878119896 = 2208)is superior to that of building RA (119878119896 = 7341) And forthese two kinds of RA the Euclidean distance of the property(120588119896119895 = 052) and apparent relative density (120588119896119895 = 187) areslightly different whereas the change of water absorption(120588119896119895 = 1067) and crushing value (120588119896119895 = 2636) is quiteremarkable Thus the water absorption rate and crushingvalue can approximately reflect the property differences ofRA
(4) Euclidean distance calculations show that there is acertain correlation among apparent density water absorptionrate and crushing value The lower the apparent density thelarger the water absorption rate and crushing value Howeverthe correlation for flakiness content is low so the mostimportant property indices for the RA are crushing value andwater absorption and at same time the major improvementfor these two property indices is needed as well
(5) The Mahalanobis distance method can evaluate thecomprehensive property of RA directly and provide gradeevaluation guidance for RAs by considering the influencefactors of the correlation among variables such as parametersof mean variance and covariance
Additional Points
Highlights (i) It is clear that there has been a large variabilityabout the individual property of recycled aggregate (RA) andit has been difficult to express thewhole property in any singleproperty parameter (ii) A method is proposed to expressthe differences of the whole property of RAs using Euclidean
distance method (iii) A calculation model and evaluationmethod are provided for the whole property of RAs withMahalanobis distance method
Conflicts of Interest
The authors declare that they have no conflicts of interest
Acknowledgments
This study was supported by the National Natural ScienceFoundation of China (Project 51578027)
References
[1] F Xiao B Putman and S Amirkhanian ldquoRheological char-acteristics investigation of high percentage RAP binders withWMA technology at various aging statesrdquo Construction andBuilding Materials vol 98 pp 315ndash324 2015
[2] Y C Guo and K W Sun ldquoConsideration on new specificationsof road materials based on properties of recycled ag-gregaterdquoJournal of Changrsquoan University (Natural Science Edition) vol 3pp 1671ndash8879 2010
[3] S Nagataki A Gokce T Saeki and M Hisada ldquoAssessmentof recycling process induced damage sensitivity of recycledconcrete aggregatesrdquoCement and Concrete Research vol 34 no6 pp 965ndash971 2004
[4] Z Q Shi Jian guang ldquoEffects of different materials in theconstruction waste on the properties of concreterdquo Concrete vol6 pp 11ndash13 2008
[5] A R Chini S-S Kuo J M Armaghani and J P Duxbury ldquoTestof recycled concrete aggregate in accelerated test trackrdquo Journalof Transportation Engineering vol 127 no 6 pp 486ndash492 2001
[6] V W Y Tam and C M Tam ldquoA new approach in assessingcement mortar remains on recycled aggregaterdquo Magazine ofConcrete Research vol 59 no 6 pp 413ndash422 2007
[7] M Etxeberria A R Marı and E Vazquez ldquoRecycled aggregateconcrete as structural materialrdquo Materials and Structures vol40 no 5 pp 529ndash541 2007
[8] S YWang PH Zhu XWang J Zhou andH E Xia ldquoResearchon some key issues for applicat ion of recycled concreterdquoConcrete vol 4 pp 39ndash41 2008
[9] G R Robinson Jr W D Menzie and H Hyun ldquoRecycling ofconstruction debris as aggregate in the Mid-Atlantic RegionUSArdquo Resources Conservation amp Recycling vol 42 no 3 pp275ndash294 2004
[10] X Yuezhou The Numerical Simulation and Influence of Aggre-gate Gradation on the Compressive Strength of Recycled Aggre-gate Concrete Xiamen University 2007
[11] Z Zhang K Wang H Liu and Z Deng ldquoKey performanceproperties of asphalt mixtures with recycled concrete aggregatefrom low strength concreterdquo Construction and Building Materi-als vol 126 pp 711ndash719 2016
[12] H Sung Do P Hee Mun and R Suk keun ldquoA study onengineering characteristics of asphalt concrete using filler withrecycled waste limerdquoWaste Management vol 28 no 1 pp 191ndash199 2008
[13] Y D Wong D D Sun and D Lai ldquoValue-added utilisationof recycled concrete in hot-mix asphaltrdquo Journal of WasteManagement vol 27 no 2 pp 294ndash301 2007
Advances in Materials Science and Engineering 9
[14] S Wu J Zhong J Zhu and D Wang ldquoInfluence of demolitionwaste used as recycled aggregate on performance of asphaltmixturerdquo Road Materials and Pavement Design vol 14 no 3pp 679ndash688 2013
[15] L Huan The Research for Recycled Concrete Aggregate and ITSApplied in Asphalt Pavement Beijing University of TechnologyBeijing China 2011
[16] R HuiThe Evaluation for Property of Recycled Coarse ConcreteAggregate Basing on Distance Method Beijing University ofTechnology Beijing China 2012
[17] A Stimilli G Ferrotti A Graziani and F Canestrari ldquoPerfor-mance evaluation of a cold-recycled mixture containing highpercentage of reclaimed asphaltrdquo Road Materials and PavementDesign vol 14 no 1 pp 149ndash161 2013
[18] K L Vasconcelos A Bhasin and D N Little ldquoInfluence ofreduced production temperatures on the adhesive propertiesof aggregates and laboratory performance of fine aggregate-asphalt mixturesrdquo Road Materials and Pavement Design vol 11no 1 pp 47ndash64 2011
[19] Z-G Zhao and F Liu ldquoApplication research of statistical mon-itoring index based on mahalanobis distancerdquo Acta AutomaticaSinica vol 34 no 4 pp 493ndash495 2008
[14] S Wu J Zhong J Zhu and D Wang ldquoInfluence of demolitionwaste used as recycled aggregate on performance of asphaltmixturerdquo Road Materials and Pavement Design vol 14 no 3pp 679ndash688 2013
[15] L Huan The Research for Recycled Concrete Aggregate and ITSApplied in Asphalt Pavement Beijing University of TechnologyBeijing China 2011
[16] R HuiThe Evaluation for Property of Recycled Coarse ConcreteAggregate Basing on Distance Method Beijing University ofTechnology Beijing China 2012
[17] A Stimilli G Ferrotti A Graziani and F Canestrari ldquoPerfor-mance evaluation of a cold-recycled mixture containing highpercentage of reclaimed asphaltrdquo Road Materials and PavementDesign vol 14 no 1 pp 149ndash161 2013
[18] K L Vasconcelos A Bhasin and D N Little ldquoInfluence ofreduced production temperatures on the adhesive propertiesof aggregates and laboratory performance of fine aggregate-asphalt mixturesrdquo Road Materials and Pavement Design vol 11no 1 pp 47ndash64 2011
[19] Z-G Zhao and F Liu ldquoApplication research of statistical mon-itoring index based on mahalanobis distancerdquo Acta AutomaticaSinica vol 34 no 4 pp 493ndash495 2008
