exploring best practises in design of experiments
DESCRIPTION
TRANSCRIPT
Copyright © 2014, SAS Institute Inc. All rights reserved.
Exploring Best Practises inDesign of ExperimentsA Holistic Approach to DOE, Increasing Robustness, Efficiency and Effectiveness
Copyright © 2014, SAS Institute Inc. All rights reserved.
Contents
Background to DOE
Why Use DOE?
Tips for Effective DOE with Classical Designs
Definitive Screening
Case Studies 1-3
Role of Statistical Modelling and DOE in Learning
Holistic DOE
Case Study 4
Copyright © 2014, SAS Institute Inc. All rights reserved.
BACKGROUND TO DESIGN OF EXPERIMENTS (DOE)
Copyright © 2014, SAS Institute Inc. All rights reserved.
FATHER OF DOE RONALD A. FISHER
Rothamstead Experimental Station, England – Early 1920’s
Copyright © 2014, SAS Institute Inc. All rights reserved.
FISHER’S FOUR DESIGN PRINCIPLES
1. Factorial Concept - rather than one-factor-at-a-time
2. Randomization - to avoid bias from lurking variables
3. Blocking - to reduce noise from nuisance variables
4. Replication - to quantify noise within an experiment
Copyright © 2014, SAS Institute Inc. All rights reserved.
AGRICULTURAL IMPACT
US corn yields
Cornell University, http://usda.mannlib.cornell.edu/MannUsda
Copyright © 2014, SAS Institute Inc. All rights reserved.
WHY USE DOE?
Copyright © 2014, SAS Institute Inc. All rights reserved.
Typical ProcessThe properties of products and processes are often affected by many factors:
In order to build new or improve products and processes, we must understand the relationship between the factors (inputs) and the responses (outputs).
Typical
Process
Machine
Operator
Temperature
Pressure
Humidity
Yield
Cost
…
Inputs
Factors
Outputs
Responses
Copyright © 2014, SAS Institute Inc. All rights reserved.
Traditional One-Factor-at-a-Time A common approach is one-factor-at-a-time experimentation.
Consider experimenting one-factor-at-a-time to determine the values of temperature and time that optimise yield.
Copyright © 2014, SAS Institute Inc. All rights reserved.
Traditional One-Factor-at-a-Time
Copyright © 2014, SAS Institute Inc. All rights reserved.
Traditional One-Factor-at-a-Time
Copyright © 2014, SAS Institute Inc. All rights reserved.
Traditional One-Factor-at-a-Time
Copyright © 2014, SAS Institute Inc. All rights reserved.
Traditional One-Factor-at-a-Time
Copyright © 2014, SAS Institute Inc. All rights reserved.
Traditional One-Factor-at-a-Time
One-factor-at-a-time experimentation frequently leads to sub-optimal solutions.
Assumes the effect of one factor is the same at each level of the other factors, i.e. factors do not interact.
In practice, factors frequently interact.
Copyright © 2014, SAS Institute Inc. All rights reserved.
Interaction between factors
Copyright © 2014, SAS Institute Inc. All rights reserved.
Experimental Design Most efficient way of investigating relationships.
Runs (factor combinations) chosen to maximize the information
Ideally balanced for ease of analysis and interpretation
Copyright © 2014, SAS Institute Inc. All rights reserved.
ITERATIVE AND SEQUENTIAL NATURE OF CLASSICAL DOE
Copyright © 2014, SAS Institute Inc. All rights reserved.
TIPS FOR EFFECTIVE DOE WITH CLASSICAL DESIGNS
Copyright © 2014, SAS Institute Inc. All rights reserved.
Stages of Experimental Design
Designing an experiment involves much more than just selecting the sequence of experimental runs:
Historically, improper planning is the most common cause of failed experiments.
Plan Design Conduct Analyse Confirm
Copyright © 2014, SAS Institute Inc. All rights reserved.
Some Planning Steps
Review what we know
• Have peer discussions
Determine new questions to answer
Identify factors and ranges to investigate
Define responses
• Easy and precise to measure
Copyright © 2014, SAS Institute Inc. All rights reserved.
Common Experimental Objectives
Identify Important Factors
Screening Design
Classical Fractional Factorial
OptimiseProcess
RSM DesignClassical Central
Composite
OptimiseIngredients
Mixtures
Classical Simplex & Extreme Vertices
Copyright © 2014, SAS Institute Inc. All rights reserved.
Sequential Experimentation Reduces Total Cost
Common Experimental Objectives
Identify Important Factors
Screening Design
Classical Fractional Factorial
OptimiseProcess
RSM DesignClassical Central
Composite
OptimiseIngredients
Mixtures
Classical Simplex & Extreme Vertices
Copyright © 2014, SAS Institute Inc. All rights reserved.
Sequential Experimentation
Common Experimental Objectives
Identify Important Factors
Screening Design
Classical Fractional Factorial
OptimiseProcess
RSM DesignClassical Central
Composite
OptimiseIngredients
Mixtures
Classical Simplex & Extreme Vertices
Definitive Screening Design Simplifies Experimental Workflow
Copyright © 2014, SAS Institute Inc. All rights reserved.
Sequential Experimentation
Common Experimental Objectives
Identify Important Factors
Screening Design
Classical Fractional Factorial
OptimiseProcess
RSM DesignClassical Central
Composite
OptimiseIngredients
Mixtures
Classical Simplex & Extreme Vertices
Definitive Screening Design
Optimal Design Manages Experimental Constraints
Copyright © 2014, SAS Institute Inc. All rights reserved.
Determining the Appropriate Factors
Determining the factors to be included in your experiment is a critical part of planning.
• Exploring too many factors may be costly and time consuming.
• Exploring too few may limit the success of your experiment.
Prior knowledge and analysis of existing data are useful aids to identifying and prioritising factors for study. Other methods may include:
• Brainstorming
• Ishikawa
Copyright © 2014, SAS Institute Inc. All rights reserved.
Selection of Factor Range is Critical With Two Level Designs …
Copyright © 2014, SAS Institute Inc. All rights reserved.
Selection of Factor Range is Critical With Two Level Designs …
By experimenting at the two settings in
yellow, X would be declared unimportant
Copyright © 2014, SAS Institute Inc. All rights reserved.
Selection of Factor Range is Critical With Two Level Designs …
By using half and often times much less than than
half the factor range X is declared important
Copyright © 2014, SAS Institute Inc. All rights reserved.
Selection of Factor Range is Critical With Two Level Designs …
By using half and often times much less than than
half the factor range X is declared important
Often leads to narrow factor ranges
to force linear relationships but
consequence is high risk of
determining sub-optimal solution
Copyright © 2014, SAS Institute Inc. All rights reserved.
Determining the Appropriate Responses
Selection of your responses will also be critical to the success of your experiment. Whenever possible:
• Choose variables that correlate to internal or external customer requirements
• Find responses that are easy to measure
• Make sure your measurement systems are precise, accurate, and stable
Analysis of current data, prior knowledge, measurement systems analysis are useful aids.
Copyright © 2014, SAS Institute Inc. All rights reserved.
DEFINITIVE SCREENING
Copyright © 2014, SAS Institute Inc. All rights reserved.
Fractional Factorials: Complex workflow from many factors to optimum settings
Tempting to miss out
middle step which can
result in selection of
wrong factors and decisions
Copyright © 2014, SAS Institute Inc. All rights reserved.
Definitive Screening Design
Identifies active main effects, uncorrelated with other effects.
May identify significant quadratic effects, uncorrelated with main effects and at worst weakly correlated with other quadratic effects.
If few factors turn out to be important, can identify significant two-way interactions uncorrelated with main effects and weakly correlated with other higher order effects.
One stage experiment if three or fewer factors important:
• progress straight to full quadratic model
• optimise process with no further experimentation
• otherwise augment DSD for optimization goals
Copyright © 2014, SAS Institute Inc. All rights reserved.
New Class of Screening Design
Three-level screening design
• 2m + 1 runs based on m fold-over pairs and an overall center point, where m is number of factors
• the values of the ±1 entries in the odd-numbered runs are determined using optimal design.
Copyright © 2014, SAS Institute Inc. All rights reserved.
Use of Three Level DesignsAdvantageous
Scientists and engineers are uncomfortable using two-level designs
• Restricting factor ranges may result in sub-optimal solutions
• Scientific/engineering judgment suggests relationships nonlinear over wide ranges
Investigators frequently have an opinion regarding the “best” levels of each factor for optimizing a response
• Experimental region centered at these levels.
• Two-level design might screen out an important factor when experimental region centred at “best”
• Adding centre points allows test for curvature
• However ambiguity over factors causing curvature
• DSD avoids ambiguity by making it possible to uniquely identify the source(s) of curvature.
Copyright © 2014, SAS Institute Inc. All rights reserved.
CASE STUDIES
Copyright © 2014, SAS Institute Inc. All rights reserved.
Case Study 1: Optimising a Chemical Process
Why Consider Definitive Screening Designs?
Copyright © 2014, SAS Institute Inc. All rights reserved.
Background
Five factors
One response yield
Goal optimise yield
Keep total cost of experimentation to minimum
Contrast traditional approach of main effect screening design plus augmentation to RSM with DSD
Copyright © 2014, SAS Institute Inc. All rights reserved.
Traditional screening approach correlates main effects with two factor interaction effects
Cost constraint and inexperience with such designs can lead to missed DOE steps
Investigator missed step of augmenting main effect design to separate correlated interaction effects from assumed important main effects
Resulted in wrong set of factors selected for RSM design which results in wrong solution
Background
Copyright © 2014, SAS Institute Inc. All rights reserved.
Traditional Approach with Missed Step
Copyright © 2014, SAS Institute Inc. All rights reserved.
Resolution III Design Perfectly Correlates Main Effects With Interaction Effects
Copyright © 2014, SAS Institute Inc. All rights reserved.
Model Interpretation
Fitted Model
Y = b0 + b1*X1 + b2*X2 + b3*X3 + Error
Correct Interpretation of Fitted Model
Y = b0 + b1*(X1+X2X3) + b2*(X2+X1X3) + b3*(X3+X1X2) + Error
Copyright © 2014, SAS Institute Inc. All rights reserved.
Missed Step Augments Initial Design to Separate Main Effects From Interactions
Copyright © 2014, SAS Institute Inc. All rights reserved.
Model Interpretation of Augmented Design
Correct Interpretation of Model Fitted to Augmented design
Y = b0 + b1*X1 + b2*X2 + b3*X3 + b12*X1X2 + b13*X1X3 + b23*X2X3 + Error
Allows clear separation of main and interaction effects
This step was missed in case study prior to modelling curvature
Copyright © 2014, SAS Institute Inc. All rights reserved.
DSD results in correct identification of important factors due to non correlated main and two factor interaction effects
Because just three factors are important DSD results in one step design:
• In addition to correctly identifying correct factors
• DSD requires no augmentation to identify optimal settings of important factors
Background
Copyright © 2014, SAS Institute Inc. All rights reserved.
CASE STUDY 1
Copyright © 2014, SAS Institute Inc. All rights reserved.
Conclusions
Fractional factorial designs can lead to selection of wrong factor set
Complex workflow for avoiding this risk which may be misunderstood or not applied by users new to DOE
May lead to conclusion that DOE does not work for us!
DSD simplifies DOE process and removes risk of selecting wrong factor set
Provides one step DOE when three or fewer important factors
• Sufficient to identify correct factor set and determine best settings of selected factors
Copyright © 2014, SAS Institute Inc. All rights reserved.
Case Study 2: Optimising Reaction Conditions for Chemical Methods
Augmenting Definitive Screening Designs
Copyright © 2014, SAS Institute Inc. All rights reserved.
Background
How effective are DSD when more than three factors are important?
Use example from literature
• Response-Surface Co-optimization of Reaction Conditions in Clinical Chemical Methods, Gopal S. Rautela, Ronald 0. Snee,’ and Warren K. Miller, CLINICAL CHEMISTRY, Vol. 25, No. 11, 1979
• CCF RSM in six factors
• Five factors are important
• Use model from this experiment to contrast CCF with augmented DSD
Copyright © 2014, SAS Institute Inc. All rights reserved.
Aspartate Aminotransferase Assay: http://www.chem.qmul.ac.uk/iubmb/enzyme/EC2/6/1/1.html
Six factors (reagent conditions): tris(hydroxymethyl)aminomethane, pH, L-aspartic acid, pyridoxal-5’-phosphate, 2-oxoglutarate, and malate dehydrogenase
Response: aspartate aminotransferase activity measured for human serum with above normal activity at 30C
Goal: select reagent conditions that maximise aspartate aminotransferase activity
Example selected to stress DSD when >3 factors are important, in this case 5 factors are important.
Background
Copyright © 2014, SAS Institute Inc. All rights reserved.
CASE STUDY 2
Copyright © 2014, SAS Institute Inc. All rights reserved.
Conclusions
When >3 factors are important, augmenting DSD works
When >3 factors are important, an augmented DSD approach is more efficient than classical Response Surface Designs
Copyright © 2014, SAS Institute Inc. All rights reserved.
CASE STUDY 3
Copyright © 2014, SAS Institute Inc. All rights reserved.
Case Study 3: Optimising Yield
What About Constrained Factor Spaces?
Copyright © 2014, SAS Institute Inc. All rights reserved.
Background
From chapter 5 of Goos & Jones
Chemical reaction
Goal: maximise yield
2 factors: Temperature and Time
Copyright © 2014, SAS Institute Inc. All rights reserved.
Background
Expert knowledge tells us
• Certain conditions will give poor results (hence, constraints)
• Behaviour very non-linear
We will show
• Design where prior knowledge is ignored.
• Fitting the design to the problem
Copyright © 2014, SAS Institute Inc. All rights reserved.
Example of Process Constraint
Copyright © 2014, SAS Institute Inc. All rights reserved.
Shrink Experimental Range to Factorial
Copyright © 2014, SAS Institute Inc. All rights reserved.
Shrink Experimental Range to Factorial
Copyright © 2014, SAS Institute Inc. All rights reserved.
Shrink Experimental Range to Factorial
Copyright © 2014, SAS Institute Inc. All rights reserved.
Optimal Design: Use Actual Factor Range
Copyright © 2014, SAS Institute Inc. All rights reserved.
… optimal designs allow investigation of complete factor space properly adjusted for constraints
Typical
Process
Machine
Operator
Temperature
Pressure
Humidity
Yield
Cost
…
Inputs
Factors
Outputs
Responses
Optimal Design: Fit to Model
Model
Y = f(X)
The process is not seen as a black box anymore…
Copyright © 2014, SAS Institute Inc. All rights reserved.
CASE STUDY 3
Copyright © 2014, SAS Institute Inc. All rights reserved.
Conclusions
Custom Design permits studying any:
• combination of factors with or without constraints,
• number of factor levels,
• blocking structure.
Build your design to suit the problem instead of fitting the problem into a design
Copyright © 2014, SAS Institute Inc. All rights reserved.
Case Study 4: Designing Products People Want to Buy
Holistic DOE
Copyright © 2014, SAS Institute Inc. All rights reserved.
ROLE OF STATISTICAL MODELLING AND DOE IN LEARNING
Copyright © 2014, SAS Institute Inc. All rights reserved.
Data Sources
DOE and/or observational (historical)
Potential problems with observational data:
• X’s are correlated – identification of “best” model difficult
• Outliers (potential or real) - bias model estimation
• Missing data cells – result in loss of whole data rows with traditional least squares based analysis
• Range over which X’s varied may be limited –restricting model usefulness
• May not have measured all relevant X’s
In some situations these can also be issues with DOE datasets
Copyright © 2014, SAS Institute Inc. All rights reserved.
WHAT IS HOLISTIC DOE?
Copyright © 2014, SAS Institute Inc. All rights reserved.
Holistic DOE Approach: Integrating Statistical Modelling and DOE
Learning is incremental and effective statistical modelling of observational data aids design of next experiment.
Analysis approach needs to manage real (messy) data simply
• Correlated X’s, outliers, missing cells
• Quickly deliver “best” current model to revise with new DOE data
• Aid better analysis of new experimental data when unexpected occurs
• Build models based on individual datasets and aggregated data
Good statistical modelling integrated with DOE helps reduce total learning time, effort and cost
It would be a shame to not use pre-existing data that comes for free
Copyright © 2014, SAS Institute Inc. All rights reserved.
Holistic DOE Example
Copyright © 2014, SAS Institute Inc. All rights reserved.
Background
PC retailer is observing appreciably retail price variation in its laptop computer line.
Goals:
• Investigate factors associated with retail price variation.
• Perform further experimentation in key factors to optimise and standardise pricing across stores.
Copyright © 2014, SAS Institute Inc. All rights reserved.
CASE STUDY 4
Copyright © 2014, SAS Institute Inc. All rights reserved.
Conclusions
Analysis of prior data helps identify factors and ranges to use in next DOE.
Analysis of prior data helps reduce risk and increase efficiency and effectiveness of future experiments.
DOE is not just for science and engineering.
Copyright © 2014, SAS Institute Inc. All rights reserved.
Holistic DOE: Integrated Statistical Modelling and DOE
Supports wide range of user skills
Exploratory analysis and statistical modelling of historical messy data simplifies and shortens whole DOE process.
Next generation DOE enables more staff to apply DOE with reduced learning and implementation effort
Interact with model predictions to build consensus
Integrated simulation capabilities enables rapid progression from models to decisions
Drag and drop charts help monitor processes and identify potential causes of issues
Manage risk better by correctly identifying signal from noise