April Meeting -- Speaker: Dr. Stanley N. Deming

Dr. Stanley N. Deming

Professor Emeritus, Department of Chemistry, University of Houston

and

President, Statistical Designs

Sequential Simplex Optimization
and its Application to the Improvement of Industrial Processes and Consumer Products

In industry, when improving old processes and products or developing new ones, it is important to make changes that rapidly achieve optimal performance. To that end, Spendley, Hext, and Himsworth introduced the sequential simplex method of experimental design in 1962 as a more efficient substitute for the original factorial-based evolutionary operation (EVOP) presented in 1957 by George Box. Later, in 1965, Nelder and Mead modified the original fixed-size simplex method to make it a variable-size method and applied it to the mathematical minimization of the sum of squares of residuals when fitting non-linear models to data. The advantage of the variable-size simplex to accelerate when moving in good directions and decelerate when moving in poor directions was quickly adopted by industrial researchers working in the real world. Today, the sequential simplex methods are often used in the early stages of process and product development or improvement to rapidly find a region of suitable operating conditions, after which more complex experimental designs can be used to map and understand the region in more detail.