Value from Data · 2014. 4. 10. · • More than 80% of Umetrics installation is based on the PI...

© Copyr i gh t 2014 O SIs o f t , LLC .

Presented by

Value from Data -Applying MVDA for real-time

monitoring, prediction & control

Petter Moree – Umetrics

Domenic Schimizzi – Genentech


• Enabled users in advanced

analysis, troubleshooting and

error detection

• Saved significant resources in

manual reviews

• Increased process knowledge

Solution Results and Benefits

OSIsoft PI System with SIMCA-online

in the Pilot Plant

Business Challenge

• Genentech PD traditionally

has relied on manual

reviews for troubleshooting

• Reviews are subjective,

reactive, time-consuming

and require expertise

• The PD group implemented

SIMCA-online using the

PI System and PI Batch for

online analysis of

in-process chromatography

at the pilot scale

“Access to model data and historical trends is invaluable as

a monitoring and troubleshooting tool”

William McGreevy

Genentech, Inc.

2


Agenda

• What is MVDA?

• Values using MVDA.

– PD & Pilot

– Manufacturing

• Case Study: MVDA at Genentech

– example from DSP in PD & Pilot

• Summary and Q&A

3


• Subsidiary of MKS Instruments

• MKS Instruments founded 1961 - Umetrics founded 1987

• 2400+ employees

• Global presence

Umetrics - MKS Instruments

4


• World leading user friendly solutions for PAT and QbD

– More than 700 leading companies & organizations

• World leading graphically driven software solutions

– More than 7000 users

• World leading consulting, support and training services

– More than 15000 individs educated

• Strong research cooperation with leading Chemometric research groups

Why Umetrics?

In-house training

Open courses

Explore, analyze and

interpret

For ensuring process

quality

For easier DOE and

QbD For embedded OEM

solutions

5


Why OSIsoft and Umetrics?

• Customer driven cooperation for more than 8+ years.

• More than 80% of Umetrics installation is based on the PI System as a fact the PI System offers an infrastructure fulfilling the needs and demands structuring batch data and metadata into one OTC solution.

• Large number of Use Cases and Success Stories presenting significant customer values by using the combined offering.

• Close collaboration in developing interfaces, documents, best practices and SOPs. Meetings on a regular basis between developers, product and marketing organizations from both parties.

• Joint Go-To-Customer approach when possible.

6

© Copyr i gh t 2014 O SIs o f t , LLC .7


Data

DATADATA = INFORMATION

•DOE – Generate informative data

•MVDA – Extraction of information in data

8

© Copyr i gh t 2014 O SIs o f t , LLC . 9

Building a capable process

DOE

MVDA

QFDQuality Function

Deployment

QRA:Quality Risk

Assessment

DOE Analysis

Design Space

Control Strategy

Manufacturing


DJIA = x1*Merck + x2*J&J + x3*Pfizer + x4*DuPont + ....

1

0

Is this chart familiar?


So this control chart is easy to understand....

-3

-2

-1

0

1

2

3

4

0 10 20 30 40 50 60 70 80 90 100 110 120 130

tPS

[1]

$Time (normalized)

PO_WST3433_EXJADE_Drying_V01.M3:3Predicted Scores [comp. 1]

+3 Std.Devt[1] (Avg)-3 Std.DevtPS[1] (Batch S0058_A_854826)

SIMCA-P+ 11 - 01.08.2009 14:42:24

t1= x1*Temperature + x2*Pressure + x3*Agitation speed + x4* pO2 ....

11


MSPC – Multivariate Statistical Process Control Evolution Level – Monitoring

• Example of a fermentation

-3

-2

-1

0

1

2

3

4

0 10 20 30 40 50 60 70 80 90 100 110 120 130

tPS

[1]

$Time (normalized)

PO_WST3433_EXJADE_Drying_V01.M3:3Predicted Scores [comp. 1]

+3 Std.Devt[1] (Avg)-3 Std.DevtPS[1] (Batch S0058_A_854826)

SIMCA-P+ 11 - 01.08.2009 14:42:24

Control limits

Average (signature)

of all batches

New batch assessed by the model

12


Statistical Process Control BATCH CONTROL CHART

-9

-8

-7

-6

-5

-4

-3

-2

-1

0

1

2

-100 0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000 2100 2200 2300 2400 2500 2600 2700

t[1]

$Time (smoothed)

Model Data Ciclo - Oct 2010 v5.M2:16

Scores [comp. 1] (Aligned)

+3 Std.Dev

t[1] (Avg)

-3 Std.Dev

t[1] (Aligned): 635

SIMCA-P+ 11 - 14.03.2011 17:53:19

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

8400 8500 8600 8700 8800 8900 9000 9100 9200 9300 9400 9500 9600 9700 9800 9900 10000 10100

Num

Variable, Batch: 617, Phase: 16-15.4122 * 543071TT--607 - 15.4122

-2.84936 * 543071TT--600 - 2.84936

0.00281207 * 543071PT--644 + 0.00767695

-1.10938 * 543071TT--602 - 1.10938

0.00826856 * 543071-Acetone_43x10e13

SIMCA-P+ 11 - 09.03.2011 19:17:38

Batch

Process

Signature

average of all batches control limits (± 3s from avg.)

-20

-10

0

10

20

-80 -70 -60 -50 -40 -30 -20 -10 0 10 20 30 40 50 60 70 80

t[2]

t[1]

Model Data Ciclo - Oct 2010 v5 - batch level (scores).M2 (PCA-X)

t[Comp. 1]/t[Comp. 2]

R2X[1] = 0.873728 R2X[2] = 0.0564179

Ellipse: Hotelling T2 (0.95)

612

614 615

616

617

618

619

620

621

622

623

624

625626 627

628

633

634

630

631

632

636637

638

635

639

641

SIMCA-P+ 11 - 10.03.2011 19:27:42

13


Role of Data analysis. Objectives for the pharmaceutical & biopharmaceutical industry

• Increase of process understanding

– Identification of influential process parameters

– Identification of correlation pattern among the process parameters

– Generation of process signatures

– Relationship between process parameters and quality attributes

• Increase of process control

– Efficient on-line tool for

• Multivariate statistical control (MSPC)

• Analysis of process variability

– Enabling on-line early fault detection

– Support for time resolved design space verification

• real time quality assurance

– Predicting quality attributes based on process data

– Excellent tool for root cause, trending analysis and visualization

– Fundament for Continued Process Verification (CPV)

Develo

pm

en

tP

rod

uctio

n

14


Work and Data flowFor Method Development

All Process

Parameters

Evolution

Level

Batch Level

Individual

Probes

Individual

Probes…

Reduction

of Dimensionality

0

1

2

3

4

5

6

7

8

9

10

11

12

13

14

320 330 340 350 360 370 380 390 400 410 420

Num

Recorded Process Parameter during granulationObsID(Obs ID ($PhaseID))

Mixer Power rate of change precss variable

0.01 * Mixer torqute process variable

0.1 * Mixer speed process variable

0.1 * Product temperature process variable

Mixer power process variavle (electrical)

Aims:

- Creation of batch signature

- Identify correlation patterns

- Fundament for CPV

15


Work and Data flowFor Routine Use in Production

Identification of

responsible

Parameter(s)

Evolution

Level

Batch Level

Investigation on

process data

Aims:

• Conformity check

• Real-time release testing

• Trend analysis

• Root cause analysis

• Early fault detection

-40

-20

0

20

40

60

80

100

120

0 10 20 30 40 50

$Time (normalized)

PO_WST10332_EXJADE_GRAN_Steintraining.M2:7

Predicted Liquid feed pump speed

+3 Std.Dev

XVar(Liquid feed pump speed) (Aligned) (Avg)

-3 Std.Dev

XVarPS(Liquid feed pump speed) (Batch S0007_B_854825)

SIMCA-P+ 11 - 08.02.2009 17:17:46

-180

-160

-140

-120

-100

-80

-60

-40

-20

0

20

40

-20 -10 0 10 20 30 40 50 60 70 80 90 100 110 120

tPS

[2]

tPS[1]

PO_WST10332_EXJADE_GRAN_Steintraining - batch level (scores).M1 (PCA-X), All Batches, PS-Complementory

tPS[Comp. 1]/tPS[Comp. 2]

R2X[1] = 0.402627 R2X[2] = 0.341738

Ellipse: Hotelling T2PS (0.95)

S0006_A_85S0007_A_85

S0007_B_85

S0006_A_85

S0006_B_85

S0007_A_85S0007_B_85

S0008_A_85

S0008-B_85S0009_A_85

S0009_B_85S0010_A_85S0010_B_85

S0011_A_85S0011_B_85

S0012_A_85S0012_B_85

S0014_A_85S0014_B_85S0018_A_85S0018_B_85S0019_A_85S0019_B_85S0021_A_85S0021-B_85

S0022_A_85S0022_B_85S0023-A_85S0023-B_85S0024-A_85S0006_B_85S0025_A_85S0025_B_85S0026_A_85S0026_B_85S0027_A_85S0027_B_85S0028_A_85

S0028-B_85

S0029-A_85S0029-B_85S0030-A_85S0030_TEILS0031_A_85S0031_B_85S0032_A_85S0032_B_85S0033_A_85S0033-B_85S0034_A_85S0034_B_85

S0035_A_85S0035_B_85S0037_A_85S0037_B_85S0039_A_85

S0039_B_85

S0040_A_85

S0040_B_85S0041_A_85S0041_B_85S0042_A_85S0042_B_85S0043_A_85

S0044_A_85

S0044_B_85S0045_A_85S0045_B_85S0046_A_85S0046_B_85S0047_A_85S0047_B_85S0048_A_85S0048_B_85

SIMCA-P+ 11 - 08.02.2009 17:15:01

Increased level of detailAnswers: What? When? How?

16


What makes Batch-SPC so powerful?

• The SIMCA product family uses a data

compression technique

– Multivariate data analysis

• Data from all relevant process parameters

are concentrated to a few highly informative

graphs

– Simplifies overview, analysis and

interpretation

– Enable use of data by increasing ease of use

• Simple drill-down functionality to transfer

compressed information back to raw data for

analysis

17


Drill-down for analysisFull transparency, perfect interpretation

18


Monitor• Early fault detection

– SIMCA-online technology is

acknowledged for its ability to

detect process issues before

they become critical

• Project dashboard

– Full drill-down to raw data for

cause analysis

• Knowledge building

– Instant analysis of process

changes improves

understanding

• Process visibility

– Easy-to-grasp graphics makes

the process status accessible to

colleagues at all levels

19


Various objectives applying MVDA

• Product quality information

– Indirect information based on

process behavior

– As long as a process behaves

well, product should be

according to specification

• Soft sensor modeling

– Predict hard-to-get process

properties from online process

data, spectral data etc.

• Predictive analytics

– Online prediction of product

quality and properties

• Continuoued Process

Verification

– Ongoing assurance is gained

during routine production that

the process remains in a state of

control.

20


Connecting process train

Raw Materials

-7

-6

-5

-4

-3

-2

-1

0

1

2

3

4

0 1 2 3 4 5 6 7 8 9 10 11

t[1]

Time ($Time)

-7.00

-6.00

-5.00

-4.00

-3.00

-2.00

-1.00

0.00

1.00

2.00

3.00

4.00

520095_GR_A.M4:5 - Scores [comp. 1] (Aligned) +3 Std.Dev

t[1] (Avg)

-3 Std.Dev

t[1] (Aligned): 1000016_A

-8

-6

-4

-2

0

2

4

6

8

-10 0 10

t[2]

t[1]

-8.00

-6.00

-4.00

-2.00

0.00

2.00

4.00

6.00

8.00

520095_GR_A - batch level.M1 (PCA-X), Untitled



-4

-3

-2

-1

0

1

2

3

4

5

6

-1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

t[1]

Time ($Time)

-4.00

-3.00

-2.00

-1.00

0.00

1.00

2.00

3.00

4.00

5.00

6.00

520095_DR_A.M1 - Scores [comp. 1] (Aligned) +3 Std.Dev

t[1] (Avg)

-3 Std.Dev

t[1] (Aligned): 665002T_A

-10

-5

0

5

10

-20 -10 0 10 20

t[2]

t[1]

-10.00

-5.00

0.00

5.00

10.00

520095_DR_A - batch level.M2 (PCA-X), Untitled



-20

-10

0

10

20

-40 -30 -20 -10 0 10 20 30 40

t[2]

t[1]

-20.00

-10.00

0.00

10.00

20.00

520095_CO_A - batch level.M2 (PCA-X), FinalMonitor



-7

-6

-5

-4

-3

-2

-1

0

1

2

3

4

5

-1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23

t[1]

Time ($Time)

-7.00

-6.00

-5.00

-4.00

-3.00

-2.00

-1.00

0.00

1.00

2.00

3.00

4.00

5.00

520095_CO_A.M1 - Scores [comp. 1] (Aligned) +3 Std.Dev

t[1] (Avg)

-3 Std.Dev

t[1] (Aligned): 665001T_A

98

99

100

101

98 99 100 101

YV

ar(

A11

10-L

20-O

10

5-A

ssa

y_-_

Liq

uid

)

YPred[8](A1110-L20-O105-Assay_-_Liquid)

98.00

99.00

100.00

101.00

520095_CO_A - batch level.M3 (PLS), Assay

YPred[Comp. 8](YVar A1110-L20-O105-Assay_-_Liquid)/YVar(YVar A1110-L20-O105-Assay_-_Liquid)

RMSEE = 0.118192

y=1.001*x-0.127R2=0.9799

Used for characterizing operating

window of individual unit operations

Used for mapping interactive affects

and influence of process on quality

21


SIMCA-online with the PI System, PI Batch,

PI Event Frames and PI Asset Framework

22


• Member of the Roche Group

• Founded in 1976

• Research focused on oncology, immunology,

neuroscience and infectious disease

23


• Enabled users in advanced

analysis, troubleshooting and

error detection

• Saved significant resources in

manual reviews

• Increased process knowledge

Solution Results and Benefits

OSIsoft PI System with SIMCA-online

in the Pilot Plant

Business Challenge

• Genentech PD traditionally

has relied on manual

reviews for troubleshooting

• Reviews are subjective,

reactive, time-consuming

and require expertise

• The PD group implemented

SIMCA-online using the

PI System and PI Batch for

online analysis of

in-process chromatography

at the pilot scale

“Access to model data and historical trends is invaluable as

a monitoring and troubleshooting tool”

William McGreevy

Genentech, Inc.

24


Column Chromatography

• Used to purify compounds on the relative speed at which they travel through a medium

• Medium can select for ions, antibodies, certain sized particles, etc.

25


Traditional Chromatogram Review

26

MetMAb QSFF 3 2 x 19001:VP.12.002.29_UV1_280nm MetMAb QSFF 3 2 x 19001:VP.12.002.29_Cond MetMAb QSFF 3 2 x 19001:VP.12.002.29_pH MetMAb QSFF 3 2 x 19001:VP.12.002.29_Pressure MetMAb QSFF 3 2 x 19001:VP.12.002.29_Temp MetMAb QSFF 3 2 x 19001:VP.12.002.29_Logbook

0

500

1000

1500

2000

2500

3000

mAU

0 1000 2000 3000 4000 ml

Lo

ad

Wa

sh

Pre

Po

ol

Po

ol

Hig

h_

Sa

lt

Re

ge

n

Sto

rag

e

MetMAb QSFF 3 2 x 19001 31001:VP.12.002.31_UV1_280nm MetMAb QSFF 3 2 x 19001 31001:VP.12.002.31_Cond MetMAb QSFF 3 2 x 19001 31001:VP.12.002.31_pH MetMAb QSFF 3 2 x 19001 31001:VP.12.002.31_Pressure MetMAb QSFF 3 2 x 19001 31001:VP.12.002.31_Flow MetMAb QSFF 3 2 x 19001 31001:VP.12.002.31_Temp MetMAb QSFF 3 2 x 19001 31001:VP.12.002.31_Logbook

0

1000

2000

3000

mAU

0 1000 2000 3000 4000 ml

Lo

ad

Wa

sh

Pre

Po

ol

Po

ol

Hig

h_

Sa

lt

Re

ge

n

Sto

rag

e


Problem Statement

• Chromatography reviews require in-depthexpertise with many years of experience with the product family and the process

• Significant risk in a process that requires extensive on-the-job training with no quantifiable result

• Chromatogram reviews take resources that could be spent elsewhere

27


Project Objective

• Implement an automated

chromatography review

system to reduce labor,

accelerate deviation

detection and improve

process knowledge

28


Batch Context and Model Development

• PI Batch configured in less than two hours on all

units

• Other data historians have taken weeks of time

to develop and configure batch context

• After initial MVDA model put online, batch data

extracted from the PI System via SIMCA-online,

accelerating model development

29


Batch Evolution for Chrom Process

30


Batch Evolution in Single Chrom Step

31


Conductivity Deviation

32


Abnormal Elution Peak

33


Motivation for QbD

• Reducing process variability is not necessarily desirable

00.20.40.60.8

1

Input

00.20.40.60.8

1

Process

00.20.40.60.8

1

Output

With variation in inputs• Initial material qualities

• Environment

• Equipment

Static process Results in variability in outputs

34


Motivation for QbD

• Reducing process variability is not necessarily desirable

00.20.40.60.8

1

Input

00.20.40.60.8

1

Process

00.20.40.60.8

1

Output

With variation in inputs• Initial material qualities

• Environment

• Equipment

Static process Results in variability in outputs

35


QbD and PAT Strategies

• Control strategy a) batch to batch control

00.20.40.60.8

1

Input

00.20.40.60.8

1

Output

00.20.40.60.8

1

Process

00.20.40.60.8

1

OutputAdjusting the process based on variations in the output

• Used in Semiconductor and other high throughput industries

• Affective if input variations are slow relative to the rate of production

36



• Control strategy b) feedforward control

00.20.40.60.8

1

Input

00.20.40.60.8

1

Process

00.20.40.60.8

1

Output

Adjusting the process based on variations in the input

• Media and feed composition

• Used in pulp and paper and other industries with natural products with high variability

37



• Control strategy c) PAT control

00.20.40.60.8

1

Input

00.20.40.60.8

1

Process

Adjusting the process based on measurement of quality in the process

• Used in many processing industries using various methods

• Direct measurement of material qualityInferential control – estimation of quality from process measurementsSpectral calibration

00.20.40.60.8

1

Output

38


• Final day VCD improved on average 23% with SIMCA-control– 7 confirmation runs in parallel - open vs closed loop

• Improved robustness e.g. reduced variation in VCD, time etc.

• Harvest time decreased 20%

39

PAT control / SIMCA-control / SIMCA-onlinePresented at IFPAC 2014


Lonza: Multivariate Online Batch modeling

• Lonza is a global company serving the needs

of the pharmaceutical and specialty

ingredients markets.

• Presented by Christine Bernegger /

Head Program Management,

February - Workshop der ISPE Affiliate D/A/CH

Solution

Six sigma approach variability

analysis for Yield optimization

And Time Based MVA of On-

line Process Parameters.

Customer Results / BenefitsCustomer Business Challenge

• Average Yield was lower

than expected

• Variation in Yield gave a

more difficult situation to

plan work and delivery to

end customer

40


• “The new automated analytics system saves time and allows the team to focus on the trends instead of gathering and charting data. In one case, a yield issue was rectified in less than a day thanks to the new system; with Amgen’s old setup, it would not have become apparent for weeks.”

• “In one recent example, Amgen identified the cause of a cell culture problem about a month earlier than it otherwise might have. For that biologic product, making the fix early -and not losing that month- saved $2.4 million, she says. Depending on the product, manufacturing a lot can cost more than $1 million”, CIO Diana McKenzie says. "This is Amgen's taming of big data.“

• “Using statistical analysis of data points collected in real-time during the process, Amgen identifies "weak signals" that could indicate brewing problems in the manufacturing cycle. Scientists then delve deeper, taking corrective steps if necessary. The analytics system includes virtualized data warehousing tools from Denodo, multivariate analysis tools from Umetrics and various software modules from SAP.”

• Source: http://www.cio-asia.com/print-article/40347/

Amgen receive CIO 100 Award 2013

41


Domenic Schimizzi• [email protected]

• Automation Engineer, Process Development

• Genentech42

Petter Moree• [email protected]

• Director Product and Marketing Management

• Umetrics

mailto:[email protected]

mailto:[email protected]


Brought to you by

Date post:	21-Aug-2020
Category:	Documents
Upload:	others
View:	0 times
Download:	0 times

Value from Data · 2014. 4. 10. · • More than 80% of Umetrics installation is based on the PI...

Documents