Sequential Simplex Optimization

Who Should Attend

Technicians, scientists, engineers, research supervisors, project managers, and vice presidents who need to learn and understand a rapid way to optimize many products and processes.

The course is aimed at both beginning and experienced workers. The course assumes no previous knowledge of statistics.

Key Topics You Will Learn About

  • Systems theory approaches to defining optimization problems
  • Resonse surface concepts
  • Both fixed-size and variable-size simplex algorithms
  • Desirability functions for dealing with multiple responses

How You Will Benefit From This Course

  • You will master an effective optimization strategy that will allow you to tune-up many processes
  • You will learn a unified way of thinking about your processes
  • You will achieve your real research goals much more rapidly
  • You will communicate more easily with statisticians about what you really want
  • You will avoid "Yabuts" when developing new processes and products
  • You will begin to think in terms of response surface concepts
  • You will achieve confidence in working with multifactor systems
  • You will know how to avoid becoming stranded on ridges on response surfaces
  • You will know when to stop using the simplex and switch to other types of experimentation
  • You will master the simplex worksheets and feel confident about using statistical software

Day 1, Morning

  • Learn the intimate relationship between optimization and good statistical process control
  • See why classical experimental strategies must be turned "head over heels" for more effective optimization
  • Review the fundamentals of systems theory
  • Learn common response surface concepts and definitions
  • Review other common optimization methods and learn their weaknesses
  • Understand the historical importance of evolutionary operation (EVOP)

Day 1, Afternoon

  • Learn the rules and calculations of the fixed-size sequential simplex of Spendley, Hext, and Himsworth
  • Learn the rules and calculations of the variable-size sequential simplex of Nelder and Mead
  • Discuss five ways to design the initial simplex
  • Learn about mathematical convergence criteria, and discover two practical criteria that always work
  • Discuss the possibility of multiple optima and two ways to discover them
  • Understand why the "k+1" rule is necessary

Day 2, Morning

  • Understand why some factors should not be included in optimization studies
  • Realize why it is important to know all of the important responses before optimization is started
  • See how optimization can be done with discrete factors
  • Learn how the simplex can handle boundary conditions and other constraints
  • Work through a four-variable example to see how the simplex calculations generalize

Day 2, Afternoon

  • Review modifications of original simplex methods — some that work, and some that don't
  • Understand how the sequential simplex method is different from the linear programming simplex
  • Learn about Harrington's desirability functions and how they are related to Zadeh's fuzzy logic
  • Discuss when to make the transition from running the simplex to carrying out a classical design
  • Conclusion of course

Return Arrow