chemical and process industries division...
TRANSCRIPT
CHAIR’S MESSAGEBy Kevin White([email protected])
Greetings CPID Members! What an honor to serve
as Chair of CPID. We have a lot of exciting things
planned for the ASQ year. But first, we need to thank
our Past-Chair, Jim Simpson, for all his contributions
to the Division as a member of the Executive
Committee over the last four plus years. Although
Jim now holds the most coveted title, Past-Chair, we
plan to keep him involved to continue his leadership
in other roles.
This year’s officers are Paula Reardon (Chair-Elect
and General Chair of the 2009 FTC), Marc Perry
(Secretary), and Wayne Wesley (Treasurer). Please
take a look at the “Meet you Officers” section of this
newsletter to learn a little more about who we are.
There are many other member leaders who are
gracious in their service to CPID and are recognized
on the last page of this newsletter.
The CPID roster currently lists approximately 1500
members. We believe our industries encounter unique
challenges. CPID exists to offer a support system of
fellow practitioners and resources to assist in the
solutions to problems you may be facing. Below are
just some of things you can expect to happen this
year.
The 53rd Fall Technical Conference will be held in
Indianapolis October 8-9, 2009. Please be sure to take a
look at the information in this newsletter and visit the
conference website (http://www.indyasq.org/stats/).
We also have just recently gotten a newly designed
website that will make it easier to share information.
There are already articles in the featured content
area. Take a look and let us know what you
think and what you would like to see
(http://www.asq.org/divisions-forums/cpi/).
Dotty Sempolinski, CPID Education Chair, leads
our Education and Training Sub-team. Look for some
great things out of this team this year. We’re very
close to having some webinars online and ready for
viewing.
With such a great support cast, I’m confident this
will be a great year. Finally, if you have questions,
comments, or suggestions for CPID, please feel free
to contact me or any other member leader.
CHEMICAL AND PROCESS INDUSTRIES DIVISIONOF THE AMERICAN SOCIETY FOR QUALITY
INSIDE THIS ISSUETopic Page
Chair Message. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Meet Your Officers . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
Fall Technical Conference 2009 . . . . . . . . . . . . . . . . . . 3
Mixture Design Case Study . . . . . . . . . . . . . . . . . . . . . 5
Process Capability Tidbit . . . . . . . . . . . . . . . . . . . . . . 11
CPID Leadership . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
CPID NEWSLETTERFALL 2009
Published by the Chemical and Process Industries Division of the American Society for QualityNewsletter Editor: Brenda Bishop, email ([email protected]), phone (314-577-7175)
CHEMICAL AND PROCESS INDUSTRIES DIVISIONOF THE AMERICAN SOCIETY FOR QUALITY
Fall 2009 CPID Newsletter
2
MEET YOUR OFFICERSKevin White, Chair
Employer: Eastman Chemical Company (Kingsport, TN)Job Title: Principal StatisticianEducation: B.S and M.S in Statistics, University of TennesseePersonal: Married to Lesley with two children, Kara and ZacharyInterests: Tennessee Athletics, Exercise, Softball, Picture FramingOther: Father, Jim White, was Chair of CPID (1986)
Paula Reardon, Chair-Elect
Employer: Eli Lilly (Clinton, IN)Job Title: Six Sigma BlackbeltEducation: B.S. in Secondary Math Education and M.S. in Statistics, Virginia TechPersonal: Married to Tim Reardon with three children, Meghann, Sean, and AllisonInterests: Shopping (not necessarily buying), visiting with friends…old and newOther: First job was at Eastman working for Jim White (Kevin’s father)
Marc Perry, Secretary
Employer: University of Alabama, TuscaloosaJob Title: Assistant Professor of StatisticsEducation: PhD, Florida State UniversityPersonal: Married to Barbie with one child, AshtonInterests: College football, traveling
Wayne Wesley, Treasurer
Employer: University of Technology (UTech), JamaicaJob Title: Program Director- Industrial EngineeringEducation: B.Ed. Industrial Technology, UTech, MSc. Manufacturing Systems, SIUC, and
Ph.D. Industrial Engineering, FSUPersonal: Married to Joy with three children Jowayne, Jowaynah and JowessanInterests: Playing the keyboard and BasketballOther: I am an ordained Evangelist minister
3
53rd AnnualFall Technical ConferenceStatistics & Quality - Accelerating To Higher Performance
October 7-10, 2009Hilton Indianapolis
www.indyasq.org/stats
SECTION ON PHYSICAL AND
ENGINEERING SCIENCES
QUALITY AND PRODUCTIVITY SECTION
Co-sponsored by:
You are invited to attend the 53rd Annual Fall Technical Conference
held this year in Indianapolis, Indiana. This conference is the premier
forum to discuss topics at the interface of statistics and quality. The
theme of this year’s conference is “Statistics and Quality: Accelerat-
ing to Higher Performance.” The goal is to engage researchers and
practitioners in a dialogue that leads to more effective use of statis-
tics to improve quality. The conference will serve to bring innovations
in statistical methodologies and quality tools to the forefront. You will
have the opportunity to meet informally and exchange views with
speakers and colleagues during breaks and in the hospitality suite.
Council Meetings
Three of the sponsoring organizations (Chemical & Process Indus-
tries & Statistics Divisions of ASQ and the Quality & Productivity Sec-
tion of ASA) will also hold council meetings during the conference
(days and times TBA). The council meetings are an opportunity for
those who wish to become involved in the activities of the societ-
ies to become better informed. Please check the conference website
(www.indyasq.org/stats) for more information on dates, times, room
locations, and other meetings of interest.
Hospitality Suite
-
soring organizations host a hospitality suite each year. This plays a
vital role in the strategic operations of the divisions. We welcome new
faces and new perspectives on division operations as well as share
technical insights with colleagues in a friendly, informal atmosphere.
Check at the registration desk for hospitality suite location and hours
of operation. Please come to meet us in Indianapolis.
Accommodations
A block of rooms is available at the Hilton Indianapolis through
September 15, 2009. Conference room rates are $129 for most
rooms, plus taxes. More deluxe suites are available at $179 and
$279. A link for registration will be available on the conference
website (www.indyasq.org/stats). REGISTER EARLY!
Travel Arrangements
Travel arrangements from the airport to the hotel are the responsibility
of the attendee. More information on options is available on the
website.
Cancellations and Refunds
A complete refund will be given if you cancel prior to September 24,
2009. Cancellations received on or after September 24, 2009 will
incur a $125 cancellation fee.
Short Courses will be offered on Wednesday, October 7 and on
Saturday, October 10 from 8:30 a.m. to 5:30 p.m. The fee for each
course includes coffee breaks and lunch. Registration is limited.
A Corporate Perspective:
Financial Savvy for Statisticians
by Frederick W. Faltin and Donna M. Faltin
Sponsored by ASA-Q&P
Wednesday, October 7, 2009 | $280
-
-
-
-
cial management. Along the way, we’ll acquaint attendees with the
them to better integrate into the corporate mainstream and to en-
hance their ability to contribute to, and prosper within, their respec-
tive organizations. Conceptual foundations will include: risk & return;
principles of borrowing/lending, including hedging & leasing; portfolio
theory; insurance (buying, selling, and pricing risk); Present Value;
from recent headlines on subprime lending, bank failures, etc.Specif-
-
ing practices; capital budgeting; dividend policy; and capital struc-
ture. Applications to recent events will include examples of corporate
fraud and the tenets of Sarbanes-Oxley.
Attendees will receive course notebooks with presentation contents;
A Short Course on Industrial Split Plot Experiments
by Geoff Vining
Sponsored by ASQ-STAT
Wednesday, October 7, 2009 | $250
Spilt-plot experiments involve two sets of factors. One set consists of
factors that are hard and/or expensive to change. The other set con-
sists of factors that are much easier and/or less expensive to change.
Most industrial processes involve both types of factors. The standard
designs and analyses taught in most Six-Sigma Black Belt courses
are not appropriate for such experimental situations. This course
illustrates the proper experimental protocol and analysis. Topics cov-
ered in this course include a historical background, hard-to-change
vs. easy-to-change factors, complete sub-plot experiments within
each whole-plot, fractional factorials within each whole-plot, proper
The course will use Minitab, JMP, and R for the analyses.
Expanding Your Factorial Two-Level Experimentation
Toolset
by Robert Mee
Sponsored by ASA-SPES
Saturday, October 10, 2009 | $320
($250 + $70 for textbook)
This workshop is intended to equip practitioners to select and ana-
lyze fractional factorial designs more knowledgeably. The workshop
will provide an introduction to the new book “A Comprehensive Guide
to Factorial Two-Level Experimentation.” Each student will receive a
copy of this textbook. The book contains 50 published experiments,
many of which will be utilized to illustrate the choice of design and
proper data analysis. Orthogonal array designs that are not regular
fractional factorials will receive particular emphasis. The importance
of run order restrictions will be explained, including blocking, split-
unit, and trend-free designs.
Introduction to Regression Modeling
By Bob Brill
Sponsored by ASQ-CPID
Saturday, October 10, 2009 | $250
This course is designed for those who are interested in using and/or
creating regression models but have not had any formal training on
these methods. It is assumed that the student has had an introduc-
tory course on statistics or equivalent. After a brief review of some
of the relevant intro to statistics topics, the student will learn how
-
pret all the output from a typical software package (Minitab will be
used in this course), how to assess how good the model is and how
to intelligently use it. Some of the topics covered will be the sum
of squares and degrees of freedom decomposition, corresponding 2, adjusted R2, analysis
-
more. This is not a math course, we let the software worry about
doing that right. The emphasis will be on learning the concepts by
working with real world examples. If time permits some exploratory
and nonlinear regression methods will be presented.
photo from Hilton Indy webpage
CHEMICAL AND PROCESS INDUSTRIES DIVISIONOF THE AMERICAN SOCIETY FOR QUALITY
Fall 2009 CPID Newsletter
ASA-SPES: Cheryl Dingus, Battelle Memorial Institute
ASA-Q&P: Don McCormack, SAS Institute
ASQ-CPID: Flor Castillo, Dow Chemical
ASQ-STAT: Frank Rossi, Kraft Foods
Program Committee
Local Conference CommitteeRene Valverde-Ventura, Co-Chair
Tammy Forrester, Eli Lilly & Co., Co-Chair
Amy Komokovich, Eli Lilly & Co, Brochure Chair
Carmen Paniagua-Quiñones,Independent Consultant, Registrar
John Murphy, Registrar
Ron Robinette, Website
General Conference Chair Paula Reardon, Eli Lilly & Co.
Short Course Chair Bob Brill, ICL Performance Products
Session 6
1:30 -3:00
JQT Invited Session
Split Plot Designs:What, Why and How
Bradley Jones,JMP division of SAS
Christopher Nachtsheim,University of Minnesota
Multivariate SPC
Multivariate Poisson-distributed Control Charts
Busaba Laungrungrong,
Connie M. Borror,
Douglas C. Montgomery
Arizona State University
Screening Methodologies
Fractional FactorialBox-Behnken Designs
David J. Edwards
Virginia Commonwealth University
Robert W. Mee
University of Tennessee
Statistical Adjustmentto Engineering Models
Roshan Vengazhiyil
Shreyes Melkote,Georgia Institute
of Technology
A Bayesian Approachto Change Point Estimation
in Univariate andMultivariate SPC
Nicole Munden,
Lewis and Clark Community College
Rong Pan
Arizona State University
Steven E. Rigdon
Southern Illinois
University Edwardsville
A Screening Methodologyfor Mixture Experiments
Maria Weese,
Mary Leitnaker
University of Tennessee
Daniel Apley
Northwestern University
Kevin WhiteEastman Chemical Company
Moderators
11:45 - 1:15
L U N C H E O N“Both Hands on the Wheel: Statistics in the Journey from Evidence to Policy”
Speaker: Sally Morton, RTI International, ASA President
Session 5
10:00 -11:30
Sequential BayesianDesign of Experiments
for GeneralizedLinear Models
David M. Steinberg
Tel Aviv University
Prediction EmphasizedCovariate Selection for
Generalized Linear Models
Adam L. Pintar,
Huaiqing Wu
Iowa State
Christine M. Anderson-Cook
Los Alamos National Laboratory
Nonstandard GageR&R in ChromatographyColumn Manufacturing
Louis A. Johnson
Minitab, Inc.
Rick Crago
Restek Corporation
Connie Borror
Arizona State University
Cheryl Dingus
Battelle Memorial Institute
Gordon Clark
The Ohio State UniversityModerators
Generalized LinearModels
Indentifying Active Effectsin Screening Experiments
With Non-normal Response
Victor M. Aguirre
ITAM
Roman de la Vara
Centro de Investigacion en Matematicas
Measurement SystemsAnalysis
The Missing-IndexProblem in Process Control
Joseph G. Voelkel
RIT
9:30 - 10:00 B R E A K
Session 4
8:00 -9:30
CPID & Quality EngineeringInvited Session
Applying StatisticalThinking and Methodsin Quality Engineering:
A Look to the Future
Roger Hoerl
GE
Ron Snee
Snee Associates
TechnometricsInvited Session
Analysis of Reliabilityand Warranty Claimsin Products with Age
and Usage Scales
Jerry Lawless
University of Waterloo, Canada
Martin Crowder
Imperial College, London
Ker-Ai Lee
University of Waterloo, Canada
DOE Case Studies
Using Design ofExperiments to Develop a
Customer CorrelatedEngineering Test Procedure
Scott Sterbenz
Ford Motor Company
Panel Discussion
Using AcceleratedLife Test Results
to Predict ProductField Reliability
Hili Hong, William Meeker,
Iowa State University
Luis Escobar,
Louisiana State University
Top Reasons DesignedExperiments Fail
in a ManufacturingEnvironment
Louis Johnson,
Jim Colton,
Doug Gorman
Minitab, Inc.
Jim Lucas
L.M. Lucas and Associates
David M. Steinberg
Tel Aviv University
Frank Rossi
Kraft FoodsModerators
Friday, October 9, 2009
53rd AnnualFall Technical ConferenceStatistics & Quality - Accelerating to Higher Performance
REGISTRATION DESK OPENS7:30
Q&P Invited Session
Orthogonal LatinHypercube Design for
Computer Experiments
Dennis Lin
Penn State
Christine Anderson-CookLos Alamos National Laboratory
Presentation of WILLIAM G. HUNTER AWARDW. J. YOUDEN MEMORIAL ADDRESS
“Perspectives on Bias and Variance in Developing and Assessing Experimental Designs”
Speaker: Greg Piepel, PhD, Laboratory Fellow Statistics and Sensor AnalyticsPacific Northwest National Laboratory
4:00 -5:00
Session 3
2:00 -3:30
FutureVijay Nair
University of Michigan
MultiresponseExchange Algorithms
for Model RobustExperimental Design
Byran Smucker,
Enrique del Castillo,
James Rosenberger,
Penn State University
Profile Control Chart:A Nonparametric L-1Regression Approach
Dennis K.J. Lin,
Zhibiao Zhao,Penn State University
Ying Wei,Columbia University
Christine Anderson-Cook
Los Alamos National Laboratory
Brenda Bishop Steven Schuelka
Calumet College of St. JosephModerators
STAT Invited Session
ASQ Statistics Division:Past, Present, and Future.
Past
Bob Mitchell, 3M
PresentDaksha Choksha
Pratt & WhitneyDoug Hlavacek
ECOLAB
Design Algorithms
Designing FractionalFactorial Split PlotExperiments Using
Integer Programming
Shay R. Capehart
Air Force Institute of Technology
Murat Kulahci
Technical University of Denmark
Ahmet Keha,
Douglass C. Montgomery,
Arizona State University
Advanced Process Control
Statistical ProcessControl and ModelingUsing Homogeneous
Finite Mixtures
Dario Nappa
Texas Instruments
12:15 - 1:45L U N C H E O N
“The Centennial Year of the Indianapolis Motor Speedway”
Speaker: Joie Chitwood, President & COO Indianapolis Motor Speedway
Session 2
10:30 -12:00
SPES Invited Session
Feature Selection with aBayesian Ensemble Model
Hugh Chipman,
Acadia University
Reliability
Inference With CensoredDegradation Data
Yang Yang, Vijay Nair
University of Michigan
Strip or SplitPlot Case Studies
Designed Experimentsfor BiopharmaceuticalProcess Development:
Challenges and Opportunities
Nelson Lee Afanador,
Julia C. Oneill
Merck and Company Inc.Experimental Design and
Analysis Strategies forFactor Screening
Christopher Nachtsheim,
University of Minnesota
Engineering Applicationsfor the Eternal Survivors
Model for Limited FailurePopulations
Mindy Hotchkiss
Pratt and Whitney Rocketdyne
Use of Strip-Strip BlockDesign for Multistage
Processes to Reduce theCost of Experimentation
Ma. del Carmen
Paniagua-Quiñones
Independent Consultant
Thomas M. Loughin
Simon Fraser University Surrey
Daksha Chokshi
Pratt and Whitney
Flor Castillo
Dow ChemicalModerators
10:00 - 10:30 B R E A K
WELCOME / PLENARY SESSION“Insights into Statistics in An R&D Culture”
Janet Buckingham, Staff Analyst Statistical Analysis Section, Southwest Research Institute
8:00 -9:00
Session 1
9:15 -10:00
Advanced Visualization
Exploratory Data AnalysisMethods for Robust
EngineeringTheodore Allen,
Ravishankar Rajagopalan
The Ohio State University,
Mikhail Bernshteyn, Sagata Ltd.
DOE UnderModel Uncertainty
A General Criterionfor Multifactor Designs
Under Model UncertaintySteven Gilmour, Queen Mary
University of London,
Pi-wen Tsai, National
Taiwan Normal University
Statistics for Biometric Identification
Performance Modeling,Prediction, and Validation for
a Continuously Evolving,Large Operating Biometrics
Identification SystemTeddy Ko,
Raytheon
Don McCormack
SAS Institute
Bob Brill
ICL Performance Products
Mindy Hotchkiss
Pratt and Whitney RocketdyneModerators
Thursday, October 8, 2009
53rd AnnualFall Technical ConferenceStatistics & Quality - Accelerating to Higher Performance
REGISTRATION DESK OPENS7:00
4
CHEMICAL AND PROCESS INDUSTRIES DIVISIONOF THE AMERICAN SOCIETY FOR QUALITY
Fall 2009 CPID Newsletter
5
MAKING USE OF MIXTURE DESIGN TOOPTIMIZE OLIVE OIL
– A CASE STUDYMark J. Anderson ([email protected]) and Patrick J. Whitcomb
Olive oil, an important commodity of the Mediterranean region and a main ingredient of their world-renowned diet (see sidebar), mustmeet stringent European guidelines to achieve the coveted status of “extra virgin.” Oils made from single cultivars (a particularcultivated variety of the olive tree) will at times fall into the lower “virgin” category due to seasonal variation. Then it becomesadvantageous to blend in one or more superior oils based on a mixture design for optimal formulation. For example, a team offormulators* experimented on four Croatian olive oils – Buza, Bianchera (pronounced “be an kay ra”), Leccino (pronounced “la cheeno”) and Karbonaca – to achieve an overall sensory rating of at least 6.5 on a 9-point hedonic scale, thus easily exceeding the cut off for“extra virgin” (5.5 considered to be “virgin”). The ratings were done by ten assessors trained on fundamental tastes (sweet, salt, sourand bitter) and defects of virgin olive oils, such as rancidity.
FIGURE 1: OLIVE OIL (SOURCE: ROYALTY-FREE INTERNET POST 670878 BY STOCKXPERT)
We’ve adapted the original study a bit to simplify it and make it more educational, while capturing the essence of how these formulatorsmade use of mixture design and what they discovered as a result.
The assessors can discern very tiny differences in sensory attributes that may depend on subtle non-linear blending of two or more oils.Therefore the four component (q = 4) simplex lattice is set up to the third degree (m = 3). That produces 20 unique blends. To augmentthis lattice, the formulators add 4 axial check blends and the overall centroid. They then specify that the four vertices (chosen for theirhigh leverage) and centroid be replicated (for added pure error measure) at random intervals. (Always randomize!) Assume that theformulators use a 1 liter blender to mix the oils – 30 blends in total after the augmentation. This ASL (augmented simplex lattice)design and the end results for overall sensory ratings are shown in Table 1. (Note that, for the sake of simplicity, the one-third and two-third levels are rounded to 0.333 and 0.667; respectively – thus adding to the total of 1.)
CHEMICAL AND PROCESS INDUSTRIES DIVISIONOF THE AMERICAN SOCIETY FOR QUALITY
Fall 2009 CPID Newsletter
# PointType A:Buza B:Bianchera C:Leccino D:Karbonaca Sensory Rating1 Vertex 1 0 0 0 6.982 “ 1 0 0 0 6.843 Vertex 0 1 0 0 6.494 “ 0 1 0 0 6.455 Vertex 0 0 1 0 7.256 “ 0 0 1 0 7.307 Vertex 0 0 0 1 5.888 “ 0 0 0 1 5.959 Third Edge 0.667 0.333 0 0 7.3810 Third Edge 0.333 0.667 0 0 7.1211 Third Edge 0.667 0 0.333 0 6.8712 Third Edge 0 0.667 0.333 0 6.8413 Third Edge 0.333 0 0.667 0 6.9514 Third Edge 0 0.333 0.667 0 7.1715 Third Edge 0.667 0 0 0.333 7.3616 Third Edge 0 0.667 0 0.333 7.1417 Third Edge 0 0 0.667 0.333 7.5018 Third Edge 0.333 0 0 0.667 7.1619 Third Edge 0 0.333 0 0.667 6.9520 Third Edge 0 0 0.333 0.667 7.0021 Triple Blend 0.333 0.333 0.333 0 7.5622 Triple Blend 0.333 0.333 0 0.333 7.5323 Triple Blend 0.333 0 0.333 0.333 7.2924 Triple Blend 0 0.333 0.333 0.333 7.2825 Axial CB 0.625 0.125 0.125 0.125 7.4126 Axial CB 0.125 0.625 0.125 0.125 7.3727 Axial CB 0.125 0.125 0.625 0.125 7.5028 Axial CB 0.125 0.125 0.125 0.625 7.1929 Centroid 0.25 0.25 0.25 0.25 7.5830 “ 0.25 0.25 0.25 0.25 7.55
The chosen model is a reduced special cubic:
Sensory Rating = 6.91 A + 6.47 B + 7.29 C + 5.93 D + 2.51 AB - 0.91 AC + 3.70 AD + 0.54 BC + 3.75 BD + 2.78 CD + 11.65 ABC
The presence of the ABC non-linear blending term supports the choice of a third-degree lattice design. The other three special-cubicterms (ABD, ACD, BCD) were insignificant (p>0.1) so we chose to remove them from the final model. Rather than laboriouslydissecting the model by its remaining terms, let’s focus on the response surface graphics: The pictures will tell the story.Unfortunately, now that we’ve gone to the third dimension the imaging gets trickier – only three out of the four components can bedepicted on a contour plot, for example. This complication provides the perfect opportunity to present the “trace” plot – a way to viewthe relative effects of any number of components. A trace plot for the olive-oil mixture experiment is shown in Figure 2.
6
TABLE 1: ASL DESIGN FOR BLENDING FOUR OLIVE OILS AND THEIR SENSORY RESULTS
CHEMICAL AND PROCESS INDUSTRIES DIVISIONOF THE AMERICAN SOCIETY FOR QUALITY
Fall 2009 CPID Newsletter
Continued on page 8
The traces are drawn from the overall centroid – all components at equal volume within the 1 liter vessel. This is called the “referenceblend.” Each component alone is then mathematically varied while holding all others in constant proportion. This reveals, for example,that the predicted sensory evaluation falls off dramatically as the Karbonaca oil (D) is increased relative to the three alternatives.To give you a better feel for how the trace plot is produced, consider the simpler case of only three components. Figure 3 shows thepaths of the three traces.
FIGURE 2: TRACE PLOT FOR OLIVE-OIL MIXTURE EXPERIMENT
FIGURE 3: TRACES FOR THREE COMPONENTS ONLY
CHEMICAL AND PROCESS INDUSTRIES DIVISIONOF THE AMERICAN SOCIETY FOR QUALITY
Fall 2009 CPID Newsletter
7
The trace for x1 starts at the overall centroid where it amounts to one-third of the three-component blend. The other two componentsare also at one-third, thus their ratio is one-to-one. Tracing x1 from the centroid down to the base of the ternary diagram reduces theamount of this individual component to zero. At this point the amounts of the other two components become one half each – thus theirratio remains one-to-one. In fact if you pick any point along any of the three traces, the other two components remain at constantproportions! Try working this out for yourself – it will be good practice for reading off coordinates on the ternary diagram.
TRACE VERSUS PERTURBATION
Those of you who are knowledgeable about response surface methods (RSM) for process optimization may be familiar with the“perturbation” plot – the RSM equivalent of the mixture trace. This plot predicts what will happen if you perturb your process bychanging only one factor at a time, for example by first varying time and then temperature of a chemical reaction. The perturbationplot generally emanates from the center point – all process factors at their middle level. It looks the same as a trace but thedifference is that, given a fixed total on amount, no single component can change without one or more of the others taking up theslack. Creating a trace as we’ve detailed is a work-around that provides the same benefits as a perturbation plot for RSM, that is,graphically depicting the relative effects of individual components as they become more or less concentrated in your formulation.However, keep in mind the one-dimensional nature of the trace, which cannot substitute for contour plots or 3D views of thesurface as a function of any two components. Only then will you see an accurate picture of non-linear blending effects.
PS. Warning: The trace (and perturbation) plot can change dramatically with change in the reference point. Suggestion: Once yousettle on the optimum draw the trace (or perturbation) plot from that location – this provides perspective on how robust thesolution may be to undesirable changes caused by variations in the inputs.
Now that we’ve been provided with clues on the non-linear blending behavior of the four olive oils, it seems sensible to study theresponse surfaces of the three good components ‘sliced’ at varying levels of the inferior fourth component. For example, Figures 4a and4b show the sensory results at the overall centroid (all components, including D, at 0.25 concentration) versus no Karbonaca oil (D =0). If anything, it’s the Bianchera (B) oil that creates the greatest effect on taste – very noticeably on these slices with D held fixed attwo specific levels (0 and 0.25).
FIGURES 4A AND 4B: SENSORY RESULTS AT THE OVERALL CENTROID VERSUS NO KARBONACA OIL (D = 0)
CHEMICAL AND PROCESS INDUSTRIES DIVISIONOF THE AMERICAN SOCIETY FOR QUALITY
Fall 2009 CPID Newsletter
8Continued on page 9
Continued on page 10
These response surface graphs are very illuminating! It appears as if the complete four-part blend at the centroid, shown on the left(Figure 4a), will be most robust to variations in olive oil concentrations and deliver a superior sensory rating for the most part. A morecomprehensive computer-aided search of the entire tetrahedral formulation space produced the optimal blend (best tasting) depicted inFigure 5:
A. 0.333 Buza
B. 0.299 Bianchera
C. 0.189 Leccino
D. 0.179 Karbonaca
This is predicted to produce a sensory rating of 7.63 – higher than any of the actual test results. However, any such prediction must besubject to verification.
The upward ramp for the sensory response shows how a rating below 6.5 will be completely undesirable, whereas a rating of 9represents the peak of desirability (prima!). The predicted optimum falls 0.454 up the scale from zero to one on desirability. This maybe the best the blenders can do with these particular varietals. A sensory outcome of 9 on the hedonic scale remains in the province ofthe gods who enjoy only the best of the best.
Figure 6 flags the attainable optimal blend of these four olive oils. It displays the prediction interval (7.51 to 7.76) based on 95 percentconfidence. Ideally the verification blend will be rated by the sensory panel within this range. A result outside of the prediction intervalwould cast doubt on the validity of the model.
FIGURE 5: MOST DESIRABLE BLEND VIA COMPUTER-AIDED NUMERICAL SEARCH
CHEMICAL AND PROCESS INDUSTRIES DIVISIONOF THE AMERICAN SOCIETY FOR QUALITY
Fall 2009 CPID Newsletter
9
Remember that these plots are derived from the Scheffé-polynomial predictive-model fitted to the actual experimental data andvalidated statistically. However, only by producing a confirmatory blend and subjecting it to sensory evaluation will this be verified forall practical purposes.
FIGURE 6: MOST DESIRABLE BLEND FLAGGED ON CONTOUR PLOT (D SLICED AT 0.179)
MEDITERRANEAN DIETAccording to the various sources the key components of the Mediterranean diet include:• Olive oil as an important source of monounsaturated fat • Generous amount of fruits and vegetables • Red wine in moderation• Fish on a regular basis• Very little red meat Many benefits have been attributed to this diet, including reduced rate of coronary events and weigh loss. See the American HeartAssociation’s internet post on the “Lyon Diet Heart Study” for details on a randomized, controlled trial with free-living subjects.“Two themes characterize people who have lost a significant amount of weight and kept it off long-term: 1) they don't eat as muchas in the past, and 2) they exercise more. Look for these when you search for effective weight-loss programs.”
- Dr. Steve Parker, author of The Advanced Mediterranean Diet (Vanguard Press, 2008).
The optimum sensory comes from a blend that goes light on the Karbonaca oil. But what if this inferior oil can be bought very cheaply?Then a blend with it being maximized could be made at a sensory just good enough to pass the panel at Extra Virgin level. This mightbe worth a try!
By employing mixture design these sophisticated olive oil formulators obtained “a knowledge of the whole experimental dominion withthe advantage to be able to find various mixtures bearing the same qualities.” Furthermore, “in this way, in spite of the presence ofexternal limits such as olive oil availability or other economic aspects, a variety of optimal blendings can be selected according tomarket preference.”
– Mark & Pat
* Vojnovic, D., et al. “Experimental Mixture Design to Ameliorate the Sensory Quality Evaluation of Extra Virgin Oils. Chemometricsand Intelligent Laboratory Systems, 1995, Vol. 27, pp. 205-210.
CHEMICAL AND PROCESS INDUSTRIES DIVISIONOF THE AMERICAN SOCIETY FOR QUALITY
Fall 2009 CPID Newsletter
10
Upgrading to Become an ASQ Senior MemberSince July 2005 our membership has changed from having 24% Senior members to 56%. In fact, since Jan 2005 we have had 590members become Senior members, furthermore 79% of all our current Senior members achieved this status after Jan 2005. This impliesmore and more CPID members understand the benefits to becoming a Senior member. It is a necessary step if you ever plan to becomean ASQ Fellow. You must be a Senior member for a minimum of 5 years before submitting your application for Fellow. If you becomea Senior member in addition to the ASQ member benefits you currently receive, as a Senior member you will also receive:• A Senior member certificate and card.• Recognition of your achievement through an announcement in ASQ Weekly, ASQ’s member e-newsletter.• Special Senior member name badge at ASQ events you attend.• Your choice of one extra benefit journal, or two Forums or Divisions, or one additional Section, or choose to waive additional benefits
as part of your Senior member benefit package.
Qualifying to become a Senior member is not as difficult as some people may think but it’s no cake walk either. See the informationbelow and for more information see ASQ policy G-02-01 and the Application for Senior Membership on the website www.asq1106.orgSee membership and upgrades.
Achieving Senior membership in ASQ is an indication of professional growth and accomplishments in quality or the allied arts andsciences. To be eligible for advancement, a member shall demonstrate professional growth and significant achievements in his/herprofession as indicated by meeting all of the following requirements:1. Ten years of active professional experience. Up to four years of this vocational requirement may be satisfied by graduation from an
accredited college, university, or similar institution.2. ASQ Full (formerly Regular) member in good standing for at least one calendar year prior to the date of application for
advancement.3. Qualified in one of the following ways described below:a. Conducting quality-related engineering, inspection or audit, or statistical work, or applying the methods and principles of quality on
the job for at least two years.b. Teaching quality or related arts or sciences at an accredited institution for at least two years.c. Being a Senior member or comparable type in a recognized professional organization.d. Currently holding an ASQ certification that requires recertification. If you have questions about becoming a Senior member or filling out the application, feel free to contact the CPID Examining Chair,Arved Harding, [email protected]
The Difference between Cp and CpkHave you ever wondered about the difference between Cp and Cpk? No, I'm not referring to the fact that Cp looks at the ratio of thespecification range to the process variation and that Cpk is similar but uses the average to impose an off-target penalty. What does itreally say if your Cp=1.1 and Cpk=0.8? Yes, this is an indication that your process is off target, but how far off-target? Going throughsome algebra, it can be shown that 3*(Cp-Cpk) equals the distance from target in standard deviation units. In the above example, 3*(1.1-0.8) indicates that the process is 0.9 standard deviations off-target. So, if you ever stuck in a meeting and someone is throwing Cp's andCpk's at you with no charts, maybe this will help you better visualize the situation. My general rule of thumb is that if a process is morethan 0.5 standard deviations from target, then it might be worth taking a look to see if it can be improved. - Kevin White
ARE YOU A POTENTIAL ASQ FELLOW?The Chemical & Process Industries Division wants high achieving members to receive the recognition due them bynominating qualified candidates to be ASQ Fellows. Our CPID membership currently has 46 or 3.6% Fellow members.We know that we other qualified members out there; we just need help in finding them. In order to determine if you or acolleague may be qualified, see ASQ Policy G-02-02.
Please notify the CPID Examining Chair, Arved Harding, [email protected] if you are a candidate or you know of apossible candidate, so that the complete Fellow nomination form can be provided. Ideally, we need to begin to processnomination forms in January to have them completed by the 1st Monday in May deadline each year. It’s never too early tostart working on this.
CHEMICAL AND PROCESS INDUSTRIES DIVISIONOF THE AMERICAN SOCIETY FOR QUALITY
Fall 2009 CPID Newsletter
11
CPID Position Member Leader E-mail Address Phone Number (s)
Executive Committee – Officers
Chair Kevin White [email protected] 423-229-4939 (Work) or 423-963-7733 (Cell)
Chair-Elect/2009 FTC GeneralChair/ Strategic Planning Chair Paula Reardon [email protected] 765-832-4231
Secretary/2010 FTC General Chair Marc Perry [email protected] 937-238-9297 (Cell) or 205-348-9864 (Work)
Treasurer/2011 FTC General Chair Wayne Wesley [email protected] Jamaica 876-927-1680 x3223 (Work) or 876-890-3350
Examining Committee Chair Arved Harding [email protected] 423-229-4957
Membership/Voice of the Customer Chair Diane Kulisek [email protected] 805-522-5005
Newsletter Chair/Internet Liaison Brenda Bishop [email protected] 618-567-6327
Education/BOK Project ChairFTC Short Course Rep Dotty Sempolinski [email protected] 607-962-5387
Webinars Jack Herman [email protected] 615-824-0988
WCQI Session Chair VACANT
ASQ HQ Education Liaison Jim Lucas [email protected]
Shewell Awards Chair/FTCSteering Committee Rep Malcolm Hazel [email protected] 914-378-2167
Wilcoxon/Youden Awards Chair Bill Notz [email protected] 614-292-3154
Standards Chair Rudy Kittlitz [email protected] 432-837-9937
ASQ Community Development Leta Thrasher [email protected] 800-248-1946 x7423
FTC Technical Program Rep Flor Castillo [email protected] 979-665-5447 (Cell) or 979-238-7554 (Work)
Past Chair/ Nominating Chair Jim Simpson [email protected] 850-882-0619
Past Chair 2/ Audit Chair Connie Borror [email protected] 602-543-5637
Past Chair 3 Lori Pfahler [email protected] 215-652-2967 (Work) or 215-799-9837 (Home)
Past Chair 4 Julia O'Neill [email protected] 215-652-4548
Past Chair 5 Dean Neubauer [email protected] 607-974-6777
CPID 2008-2009 Leadership Committee
CHEMICAL AND PROCESS INDUSTRIES DIVISIONOF THE AMERICAN SOCIETY FOR QUALITY
Fall 2009 CPID Newsletter