microsoft powerpoint - idrc_2008_cem6sigma
DESCRIPTION
TRANSCRIPT
![Page 1: Microsoft PowerPoint - IDRC_2008_CEM6sigma](https://reader033.vdocuments.net/reader033/viewer/2022042623/54bddcb34a7959bb608b4595/html5/thumbnails/1.jpg)
1
Chemometrics in Process Analytical Technology (PAT)-
A Six Sigma Perspective
Charles E. (Chuck) Miller, Ph.D.
Eigenvector Research, Inc.
©Copyright 2008Eigenvector Research, Inc.
No part of this material may be photocopied or
reproduced in any form without prior written
consent from Eigenvector Research, Inc.
![Page 2: Microsoft PowerPoint - IDRC_2008_CEM6sigma](https://reader033.vdocuments.net/reader033/viewer/2022042623/54bddcb34a7959bb608b4595/html5/thumbnails/2.jpg)
22
![Page 3: Microsoft PowerPoint - IDRC_2008_CEM6sigma](https://reader033.vdocuments.net/reader033/viewer/2022042623/54bddcb34a7959bb608b4595/html5/thumbnails/3.jpg)
33
Why this talk?
• I like to talk about process analytical chemistry
(PAT) and chemometrics a lot
• Concepts from Six Sigma training often “creep”
into the discussion
• Explore the relationship between Six Sigma (6σ),
PAT, and chemometrics
![Page 4: Microsoft PowerPoint - IDRC_2008_CEM6sigma](https://reader033.vdocuments.net/reader033/viewer/2022042623/54bddcb34a7959bb608b4595/html5/thumbnails/4.jpg)
44
Outline
• Introduction to Six-Sigma (6σ)
• Relationships: 6σ, PAT, Chemometrics
• What can Six-Sigma bring to PAT,
Chemometrics?
• Summary
(to be continued…)
![Page 5: Microsoft PowerPoint - IDRC_2008_CEM6sigma](https://reader033.vdocuments.net/reader033/viewer/2022042623/54bddcb34a7959bb608b4595/html5/thumbnails/5.jpg)
55
Six Sigma (6σσσσ)• A process improvement program
• Formulated in 1986, for Motorola
• Heavily influenced by teachings of
Shewhart, Deming, Taguchi, Ishikawa,
Juran, and others
• By 2000: ~2/3 of Fortune 500
• DuPont – since 1998
• Summarize into four components:
• A PROCESS,
• A TOOLKIT,
• AN INFRASTRUCTURE, and…
• A PHILOSOPHY!
Ref: http://www.sixsigmacompanies.com
Technical definition:
3.4 defects per million
opportunities
-50 0 500
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
+/- 6σ
+/- 1σ
![Page 6: Microsoft PowerPoint - IDRC_2008_CEM6sigma](https://reader033.vdocuments.net/reader033/viewer/2022042623/54bddcb34a7959bb608b4595/html5/thumbnails/6.jpg)
66
The Six-Sigma Process: “DMAIC”
• NOT LINEAR!
• Often backtrack
• Similar process
(“DMADV”), for R&D
projects
IImprove
AAnalyze
MMeasureDDefine
CControl
![Page 7: Microsoft PowerPoint - IDRC_2008_CEM6sigma](https://reader033.vdocuments.net/reader033/viewer/2022042623/54bddcb34a7959bb608b4595/html5/thumbnails/7.jpg)
77
The Six Sigma Toolkit
• Statistical tools
• Design of Experiments
• “Classical” Data analysis (i.e., ANOVA, linear regression)
• Organizational Tools
• Process mapping, brainstorming
• Templates/guides
• Risk Management
(punch line
forthcoming…)
![Page 8: Microsoft PowerPoint - IDRC_2008_CEM6sigma](https://reader033.vdocuments.net/reader033/viewer/2022042623/54bddcb34a7959bb608b4595/html5/thumbnails/8.jpg)
88
The Six Sigma Infrastructure
• NOT trivial, nor
cheap, to get
established!
• $15-25k BB
training costs
People Other support
Executive Management:
program vision, top-level support!
Champions: organization-specific
leaders
Master Black Belts: 100% 6s, program uniformity efforts
Black Belts: 100% 6s, LEAD program
execution
Green Belts: part-time, program execution
INFRASTRUCTURE
IT: software, databases
Accountants
![Page 9: Microsoft PowerPoint - IDRC_2008_CEM6sigma](https://reader033.vdocuments.net/reader033/viewer/2022042623/54bddcb34a7959bb608b4595/html5/thumbnails/9.jpg)
99
The Six Sigma Philosophy
• Data-driven decision making
• vs. Folklore-driven
• Real financial verification of a project’s impact
• Validated by accountants (vs. “handwaving”)
• Comparable between functions, departments
• Strong, Top-down management support
![Page 10: Microsoft PowerPoint - IDRC_2008_CEM6sigma](https://reader033.vdocuments.net/reader033/viewer/2022042623/54bddcb34a7959bb608b4595/html5/thumbnails/10.jpg)
1010
Six-Sigma Detractors
• Another one of those
“bogus” programs?
• Charles Holland (Qualpro)
• 54 of 58 “large companies”
implementing it are “lagging
behind” in S&P 500
• Other criticisms
• Stifles creativity, innovation
• Offers nothing new
• Too “inward looking”
http://money.cnn.com/2006/07/10/magazines/fortune/r
ule4.fortune/index.htm
![Page 11: Microsoft PowerPoint - IDRC_2008_CEM6sigma](https://reader033.vdocuments.net/reader033/viewer/2022042623/54bddcb34a7959bb608b4595/html5/thumbnails/11.jpg)
1111
But, it’s NOT going away!
• Now, 100s of companies utilizing it• At least 100 in current Fortune 500
• US Military
• Wide diversity of business segments
• WHY the traction?• Toolkit rather similar to TQM, other previous efforts,
BUT…
Ref: http://www.sixsigmacompanies.com, and
http://www.isixsigma.com/library/content/c010723a.asp
• Better use of resources
• Different corporate attitude
• Easier project vs. project comparison
![Page 12: Microsoft PowerPoint - IDRC_2008_CEM6sigma](https://reader033.vdocuments.net/reader033/viewer/2022042623/54bddcb34a7959bb608b4595/html5/thumbnails/12.jpg)
1212
Relationship: PAT and Chemometrics
Chemometrics
Enabling technology for multivariate
PAT (NIR, Raman, FTIR..)
Exploratory tools (MCR) support PAT
scouting!
Supports process modeling/simulation-
to show PAT value!
High-value chemometrics deployment
opportunities!
PAT
![Page 13: Microsoft PowerPoint - IDRC_2008_CEM6sigma](https://reader033.vdocuments.net/reader033/viewer/2022042623/54bddcb34a7959bb608b4595/html5/thumbnails/13.jpg)
1313
Relationships: PAT, Chemometrics, 6σσσσ
Chemometrics
PAT
Multivariate data analysis toolset (cluster
analysis, variable selection, cross-validation)
Empirical modeling philosophy
High-frequency, high-
relevance DATA!
6σ
ChemometricsChemometrics
PAT
Chemometrics
![Page 14: Microsoft PowerPoint - IDRC_2008_CEM6sigma](https://reader033.vdocuments.net/reader033/viewer/2022042623/54bddcb34a7959bb608b4595/html5/thumbnails/14.jpg)
1414
Relationships: PAT, Chemometrics, 6σσσσ
PAT
Chemometrics
6σBUT…
What does Six-Sigma bring
to Chemometrics/PAT???...
![Page 15: Microsoft PowerPoint - IDRC_2008_CEM6sigma](https://reader033.vdocuments.net/reader033/viewer/2022042623/54bddcb34a7959bb608b4595/html5/thumbnails/15.jpg)
1515
How Six Sigma can help Chemometrics/PAT
IImprove
AAnalyze
MMeasureDDefine
CControl
“Voice of the
Customer” tools
Measurement
System Analysis
Impact/Control
Piloting, Risk
Tools
Control Plan
![Page 16: Microsoft PowerPoint - IDRC_2008_CEM6sigma](https://reader033.vdocuments.net/reader033/viewer/2022042623/54bddcb34a7959bb608b4595/html5/thumbnails/16.jpg)
1616
PAT/Chemometrics Scenario
Model
Interpreter
Real-Time Analyzer Data
Field Analyzer
Model Application
Math
Model Outputs
DCS
On-line model
parameters
Analyzer
Control Software
Analyzer
Control Parameters
Modbus,
4-20mA, etc...
Field PC Final Model
Package
Model Deployment
Software
Existing
PAT/Chemometrics
application
“Not performing
well-enough” to be
useful
The plant might
“scrap it”
![Page 17: Microsoft PowerPoint - IDRC_2008_CEM6sigma](https://reader033.vdocuments.net/reader033/viewer/2022042623/54bddcb34a7959bb608b4595/html5/thumbnails/17.jpg)
1717
Define: Voice of the Customer
• Define Phase:
• Time-saving project phase
• Problem definition/information-gathering
• Get the “Voice of the Customer”
• Seems trivial, but• WHO are the customers?
• HOW do you get their “voice”?
• WHAT questions do you ask them?
• HOW do you “process” their responses?
• Interpersonal skills• BE ANNOYING, PERSISTANT!
IImprove
AAnalyze
MMeasureDDefine
CControl
![Page 18: Microsoft PowerPoint - IDRC_2008_CEM6sigma](https://reader033.vdocuments.net/reader033/viewer/2022042623/54bddcb34a7959bb608b4595/html5/thumbnails/18.jpg)
1818
WHO are the customers?
Process engineers
Operations Management
Process operators
Maintenance Technicians Project Engineers
Chemometrics Software Users?
WHAT do I ask them?:
• what precision is required of the analyzer?
• can calibration samples be extracted from the process?
• how often does the sampling system “foul”?
• are analyzer outputs useful during product transitions?
![Page 19: Microsoft PowerPoint - IDRC_2008_CEM6sigma](https://reader033.vdocuments.net/reader033/viewer/2022042623/54bddcb34a7959bb608b4595/html5/thumbnails/19.jpg)
1919
What to do with the answers?
Helps to focus on the REAL problems!
Affinity diagram“sampling
system fouls too much”
“outputs for analytes a&bare too noisy for control”
“transitions from product
D to E are problematic”
“there are too many
analyzer outputs”
“model deployment S/W is not
stable”
“output for analyte c is unstable at
certain times”
“we really need a NEW method, for analyte z”
“instrument hardware is very hard to
service”
methods
infrastructure
“sampling system fouls too much” “model
deployment S/W is not
stable”
“transitions from product
D to E are problematic”
“outputs for analytes a&bare too noisy for control”
Method usage
“there are too many
analyzer outputs”“instrument
hardware is very hard to
service”
“we really need a NEW method, for analyte z”
“output for analyte c is unstable at
certain times”
![Page 20: Microsoft PowerPoint - IDRC_2008_CEM6sigma](https://reader033.vdocuments.net/reader033/viewer/2022042623/54bddcb34a7959bb608b4595/html5/thumbnails/20.jpg)
2020
![Page 21: Microsoft PowerPoint - IDRC_2008_CEM6sigma](https://reader033.vdocuments.net/reader033/viewer/2022042623/54bddcb34a7959bb608b4595/html5/thumbnails/21.jpg)
2121
Measure: Measurement System Analysis
• Measure Phase:
• Identify “The Project Y”• Ex. Effectiveness of Analyzer output “B”
• Establish “baseline” performance
• Assess, validate existing measurements!
• Measurement System Analysis (MSA)
• Is an existing measurement capable of
assessing analyzer output quality?
IImprove
AAnalyze
MMeasureDDefine
CControl
![Page 22: Microsoft PowerPoint - IDRC_2008_CEM6sigma](https://reader033.vdocuments.net/reader033/viewer/2022042623/54bddcb34a7959bb608b4595/html5/thumbnails/22.jpg)
2222
Measurement System Analysis
1. Identify all possible
variation sources
4. Compare error(s)
to Performance Specs
2. Design and execute “variability
experiment” on system
3. Analyze data with
appropriate tool (Gage
R&R, ANOVA,..) to get
measurement error, and
contributions to error
Operator
Instrument ID Reagent
supplier
Ref. Method Measurements
Sample ID Operator A Operator B
1 2.650 2.698
2 2.096 2.115
3 3.033 3.015
4 2.712 2.698
5 2.111 2.125
Gage Error (GRR): 0.0987
Performance
Spec (+/-) 2.000
![Page 23: Microsoft PowerPoint - IDRC_2008_CEM6sigma](https://reader033.vdocuments.net/reader033/viewer/2022042623/54bddcb34a7959bb608b4595/html5/thumbnails/23.jpg)
2323
Analyze: Impact/Control
• Analyze Phase:
• Identify ALL sources of variation in The
Project Y (called “X”s)
• Explore X/Y relationships
• Start with many X’s, reduce to a few
critical X’s
• Useful Tools
• Cause and Effect (“fishbone”, Ishikawa)
• Process Map
• Impact/Control Matrix
IImprove
AAnalyze
MMeasureDDefine
CControl
![Page 24: Microsoft PowerPoint - IDRC_2008_CEM6sigma](https://reader033.vdocuments.net/reader033/viewer/2022042623/54bddcb34a7959bb608b4595/html5/thumbnails/24.jpg)
2424
Process Map
Model Data
Data Import
Data
“cleaning”
Exploratory
Modeling
Model Assessment (cross validation RMSECV, test
set RMSEP, NAS,..)
Final Model GenerationData Pre-
processing (2nd d., MSC, SNV,…)
Final Model “Packaging”
Final Model Package
DOE
Lab Work
Prior Knowledge
Process Sampling
Modeling Tools
(PCA, PLS, MLR, SIMCA, KNN,…)
Lab Analyzer
Field Analyzer
Visualization
Tools
Commercial
Model Development
software
Model development
process
Model
Interpreter
Real-Time
Analyzer Data
Field Analyzer
Model
Application Math
Model Outputs
DCS
On-line model
parameters
Analyzer Control
Software
Analyzer
Control Parameters
Modbus,
4-20mA, etc...
Field PC Final Model Package
Model
Deployment Software
Field PC
Model development Model deployment
Model data
collection
Model
development
process
The model itself
(parameters and
“recipe”)
Model
development
S/W
WHAT impacts analyzer effectiveness?
Deployment S/W
reliability
![Page 25: Microsoft PowerPoint - IDRC_2008_CEM6sigma](https://reader033.vdocuments.net/reader033/viewer/2022042623/54bddcb34a7959bb608b4595/html5/thumbnails/25.jpg)
2525
Impact/Control Matrix
Focus on high impact and high controllability!
If high impact and low controllability- must address!
High Medium Low
Co
ntr
olla
ble
• • •
Unco
ntr
olla
ble
• • •
IMPACT / CONTROL MATRIX
IMPACT
Model
development
S/W
Model development
process (experienced
user)
The model itself
(parameters and
“recipe”)
Model data
collectionField PC
Deployment S/W
reliability
Model development
process
(INexperienced user)
![Page 26: Microsoft PowerPoint - IDRC_2008_CEM6sigma](https://reader033.vdocuments.net/reader033/viewer/2022042623/54bddcb34a7959bb608b4595/html5/thumbnails/26.jpg)
2626
Improve: Model Piloting
• Improve Phase:
• Quantify changes needed to optimize
improvement
• Demonstrate that these changes will improve
the process
• What are tolerances in new settings?
• Useful Tools:
• Risk assessment: FMEA!
• Design of Experiments
• Piloting the solution
IImprove
AAnalyze
MMeasureDDefine
CControl
![Page 27: Microsoft PowerPoint - IDRC_2008_CEM6sigma](https://reader033.vdocuments.net/reader033/viewer/2022042623/54bddcb34a7959bb608b4595/html5/thumbnails/27.jpg)
2727
Model Piloting Tool
B. Wise, “Tools for Multivariate Calibration Robustness Testing
with Observations on Effects of Data Preprocessing”, CAC 2008
1000 1500 2000 2500-0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
WAVELENGTH (nm)
NIR DIFFUSE REFLECTANCE SPECTRA OF CORN
reference
moisture
data
+MODEL
(by BMW)
How is model
affected by an
additional
component?
![Page 28: Microsoft PowerPoint - IDRC_2008_CEM6sigma](https://reader033.vdocuments.net/reader033/viewer/2022042623/54bddcb34a7959bb608b4595/html5/thumbnails/28.jpg)
2828
Control: Control Plan• Control Phase:
• Implement “sustainable” solution
• Assess REAL impact
• Documentation, Translation
• Useful tools• Control charts
• Control Plan: “lock in the gains”!
IImprove
AAnalyze
MMeasureDDefine
CControl
Dept. or
Indiv.Property
CTQ
or XSpec.
Meas.
Tech.
Sample
SizeFreq.
Who
Measures
Where
Recd.
Action Timing Owner
Response Plan
![Page 29: Microsoft PowerPoint - IDRC_2008_CEM6sigma](https://reader033.vdocuments.net/reader033/viewer/2022042623/54bddcb34a7959bb608b4595/html5/thumbnails/29.jpg)
2929
Control Plan- PAT/Chemometrics
Model Interpreter
Real-Time Analyzer Data
Field Analyzer
Model Application
Math
Model Outputs
DCS
On-line model parameters
Analyzer
Control Software
Analyzer Control
Parameters
Modbus, 4-20mA,
etc...
Field PC Final Model
Package
Model Deployment
Software
Periodic auditing of
DEPLOYMENT
software
Monitor model
performance:
Hotelling’s T2, Q
Residuals
Remote Access
(PCAnywhere/Tim
buktu): Enabling
Technology!
Quick model
updating
capability
![Page 30: Microsoft PowerPoint - IDRC_2008_CEM6sigma](https://reader033.vdocuments.net/reader033/viewer/2022042623/54bddcb34a7959bb608b4595/html5/thumbnails/30.jpg)
3030
Six Sigma Questionnaire
• Current & former DuPont colleagues
• Which element(s) of Six-Sigma do you feel were
the most useful to your work?
• Did your "Six-Sigma experience" affect the way
you operate?
• 8 respondents
• 1 Master BB, 1 BB, 6 GB
• 3 process analytical, 1 PAT management, 2
statisticians, process modeler, project engineer
![Page 31: Microsoft PowerPoint - IDRC_2008_CEM6sigma](https://reader033.vdocuments.net/reader033/viewer/2022042623/54bddcb34a7959bb608b4595/html5/thumbnails/31.jpg)
3131
Questionnaire results• Most useful elements:
• Statistical toolset (3)
• Voice of Customer tools (2)
• Data-driven decision making (2)
• Gage R&R (1)
• Management Top-Down approach, “Bite-sized” projects,
Documentation discipline, “Locking in” solutions, REAL
validation of benefits, more data for process modeling
• Affect the way you operate?
• 3 Yes: REAL value of new measurements, less “hard
selling” of statistics!
• 2 “Not Much” : Aware of tools already
![Page 32: Microsoft PowerPoint - IDRC_2008_CEM6sigma](https://reader033.vdocuments.net/reader033/viewer/2022042623/54bddcb34a7959bb608b4595/html5/thumbnails/32.jpg)
3232
Summary
• PAT, Chemometrics have much to contribute to
Six Sigma, but..
• Six Sigma has much to contribute to PAT,
Chemometrics
![Page 33: Microsoft PowerPoint - IDRC_2008_CEM6sigma](https://reader033.vdocuments.net/reader033/viewer/2022042623/54bddcb34a7959bb608b4595/html5/thumbnails/33.jpg)
3333
Acknowledgements
• Eigenvector Research Colleagues
• DuPont Colleagues
• DuPont Six Sigma