Why should engineers, scientists and anyone who experiments take this course?

Some of the most important functions of engineers in any discipline, R&D or manufacturing, is to come up with new products and make them work, improve the existing ones, optimize processes, improve quality and reliability, maximize productivity, etc. To do this they need to run experiments, collect and analyze data and, based on the results, make important decisions. Unfortunately most engineering schools do not prepare engineers for this important job. They do not teach modern techniques for experimentation.

Lack of knowledge of design of experiments (DOE) and data analysis leaves engineers and their employers at a great disadvantage. It hurts the companies' productivity and may even threaten their survival. Fairchild Semiconductor, in the early 1980's attempted to produce 64k DRAMs. By the time the product and process problems were solved and the yield was at a point that was feasible to go to market, the window of opportunity had passed, DRAM prices had fallen and the product had to be scrapped. Fairchild had very knowledgeable engineers but they did not know how to experiment. They would have been successful had they introduced the product sooner.

All examples in the seminar are real industrial applications.

OBJECTIVES

One of the objectives of this seminar is to introduce modern design of experiments (DOE) techniques to individuals who need to run experiments, collect and analyze data. However experimentation is more than using DOE techniques. These techniques are one of the building blocks of efficient experimentation. Equally important are strategies for both experimentation and empirical model building. An important objective of the course is to discuss these strategies. Another objective of the course is to introduce the use of computer for design and analysis of experiments. JMP, an interactive software package produced by SAS® will be used.

COURSE SCHEDULE

• Elements of objective investigation
• Design vs. analysis
• WS: Optimizing yield of a process
• Innovation, invention and commercialization
• Scientific learning process and the role of design of experiments and statistical analysis
• Data vs. information vs. knowledge
• Strategies for experimentation
• Getting more information for your money
• One factor at a time vs. simultaneous multifactor experimentation
• Types of factors/variables in experimentation
• How to start the planning experimentation
• Factorial experiments
• 22 factorial experiment
• Main effects and interaction plots
• 23 and 2k factorial experiments
• Analysis of 2k factorial experiments
• Graphical analysis
• Introduction and use of JMP software
• Specifying variables
• Choosing and generating designs
• Analysis of data from designed experiments
• Probability plots
• Reference distribution / Confidence intervals
• Is process A better than Process B?
• Independent samples
• Confidence interval analysis
• Graphical analysis
• Paired samples
• Confidence interval analysis
• Graphical analysis
• Analysis of Variance
• JMP analysis of 25 factorial analysis
• Bayes Plot
• Fractional factorial experiments
• Choosing 1/2, 1/4, 1/8, 1/(2k) of a factorial design
• Confounding effect
• Construction of fractional factorial designs
• Generators of a design
• Defining relation of a design
• Resolution of a design
• Using JMP to design experiments
• Analysis of fractional factorial experiments
• JMP analysis of 2(5-1) and 2(5-2) factorial analysis
• De-confounding: Foldover designs
• More fractional factorial designs
• Orthogonal array for 2 & 3 level factors, resolution III factorial designs
• Plackett-Burman designs
• Introduction to regression analysis
• Simple regression model
• The least squares estimation method
• Multiple regression models
• Polynomial regression models
• Regression analysis Using JMP
• Introduction to response surface methodology
• Central composite designs
• Construction of central composite designs
• Box-Behnken designs
• Response surface methodology analysis Using JMP
• Strategy for response surface model building
• What to do:Before the experiment, during the experiment, after the experiment
• How to choose a matrix for an experiment
• Application in system reliability prediction
• WS: Optimizing yield of a process, strategy and methodology
• Review of the workshop and participants' feedback