© 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
© Copyr i gh t 2014 O SIs o f t , LLC .
• 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
© Copyr i gh t 2014 O SIs o f t , LLC .
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
© Copyr i gh t 2014 O SIs o f t , LLC .
• Subsidiary of MKS Instruments
• MKS Instruments founded 1961 - Umetrics founded 1987
• 2400+ employees
• Global presence
Umetrics - MKS Instruments
4
© Copyr i gh t 2014 O SIs o f t , LLC .
• 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
© Copyr i gh t 2014 O SIs o f t , LLC .
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
© Copyr i gh t 2014 O SIs o f t , LLC .
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
© Copyr i gh t 2014 O SIs o f t , LLC .
DJIA = x1*Merck + x2*J&J + x3*Pfizer + x4*DuPont + ....
1
0
Is this chart familiar?
© Copyr i gh t 2014 O SIs o f t , LLC .
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
© Copyr i gh t 2014 O SIs o f t , LLC .
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
© Copyr i gh t 2014 O SIs o f t , LLC .
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
© Copyr i gh t 2014 O SIs o f t , LLC .
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
© Copyr i gh t 2014 O SIs o f t , LLC .
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
© Copyr i gh t 2014 O SIs o f t , LLC .
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
© Copyr i gh t 2014 O SIs o f t , LLC .
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
© Copyr i gh t 2014 O SIs o f t , LLC .
Drill-down for analysisFull transparency, perfect interpretation
18
© Copyr i gh t 2014 O SIs o f t , LLC .
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
© Copyr i gh t 2014 O SIs o f t , LLC .
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
© Copyr i gh t 2014 O SIs o f t , LLC .
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
t[Comp. 1]/t[Comp. 2]
Ellipse: Hotelling T2 (0.95)
-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 ($Time)
-4.00
-3.00
-2.00
-1.00
0.00
1.00
2.00
3.00
4.00
5.00
6.00
520095_DR_A.M1 - Scores [comp. 1] (Aligned) +3 Std.Dev
t[1] (Avg)
-3 Std.Dev
t[1] (Aligned): 665002T_A
-10
-5
0
5
10
-20 -10 0 10 20
t[2]
t[1]
-10.00
-5.00
0.00
5.00
10.00
520095_DR_A - batch level.M2 (PCA-X), Untitled
t[Comp. 1]/t[Comp. 2]
Ellipse: Hotelling T2 (0.95)
-20
-10
0
10
20
-40 -30 -20 -10 0 10 20 30 40
t[2]
t[1]
-20.00
-10.00
0.00
10.00
20.00
520095_CO_A - batch level.M2 (PCA-X), FinalMonitor
t[Comp. 1]/t[Comp. 2]
Ellipse: Hotelling T2 (0.95)
-7
-6
-5
-4
-3
-2
-1
0
1
2
3
4
5
-1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
t[1]
Time ($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
© Copyr i gh t 2014 O SIs o f t , LLC .
SIMCA-online with the PI System, PI Batch,
PI Event Frames and PI Asset Framework
22
© Copyr i gh t 2014 O SIs o f t , LLC .
• Member of the Roche Group
• Founded in 1976
• Research focused on oncology, immunology,
neuroscience and infectious disease
23
© Copyr i gh t 2014 O SIs o f t , LLC .
• 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
© Copyr i gh t 2014 O SIs o f t , LLC .
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
© Copyr i gh t 2014 O SIs o f t , LLC .
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
© Copyr i gh t 2014 O SIs o f t , LLC .
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
© Copyr i gh t 2014 O SIs o f t , LLC .
Project Objective
• Implement an automated
chromatography review
system to reduce labor,
accelerate deviation
detection and improve
process knowledge
28
© Copyr i gh t 2014 O SIs o f t , LLC .
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
© Copyr i gh t 2014 O SIs o f t , LLC .
Batch Evolution for Chrom Process
30
© Copyr i gh t 2014 O SIs o f t , LLC .
Batch Evolution in Single Chrom Step
31
© Copyr i gh t 2014 O SIs o f t , LLC .
Conductivity Deviation
32
© Copyr i gh t 2014 O SIs o f t , LLC .
Abnormal Elution Peak
33
© Copyr i gh t 2014 O SIs o f t , LLC .
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
© Copyr i gh t 2014 O SIs o f t , LLC .
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
© Copyr i gh t 2014 O SIs o f t , LLC .
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
© Copyr i gh t 2014 O SIs o f t , LLC .
QbD and PAT Strategies
• 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
© Copyr i gh t 2014 O SIs o f t , LLC .
QbD and PAT Strategies
• 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
© Copyr i gh t 2014 O SIs o f t , LLC .
• 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
© Copyr i gh t 2014 O SIs o f t , LLC .
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
© Copyr i gh t 2014 O SIs o f t , LLC .
• “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
© Copyr i gh t 2014 O SIs o f t , LLC .
Domenic Schimizzi• [email protected]
• Automation Engineer, Process Development
• Genentech42
Petter Moree• [email protected]
• Director Product and Marketing Management
• Umetrics
© Copyr i gh t 2014 O SIs o f t , LLC .
Brought to you by