A simplex approach to evolutionary operation - methods and software
Öberg, T.
Lecture at Workshop on process analysis and process optimisation, Mariehamn, Åland, June 14-19, 2001.

Abstract
Analyses of historical data seldom give enough information to improve a manufacturing process and find the optimal set points. Information must instead be gathered by imposing external disturbances, i.e. by making practical trials. The information content from such process studies will be high if we use systematic methods like design of experiments (DOE). There are many benefits using DOE, but also limitations to what can be achieved when trying to optimize a typical technical process. One major problem is that the focus very soon has to be set on a limited number of variables; else the number of experiments will grow out of hand¹. Another problem is the validity of the results when the process age or the feedstock changes. Both the benefits and problems with traditional DOE and model-based optimization served as catalysts and challenges when the MultiSimplex-project was initiated, resulting in the two software: MultiSimplex® for desktop optimization and OMM™ for real-time process optimization.

The methods employed in the MultiSimplex® and the OMM™ software are different variants of the sequential simplex approach to evolutionary operation (EVOP)^2-3, and fuzzy set concepts for definition of multiple optimization criteria⁴. Practical experience has now been gained over several years. The MultiSimplex® package was introduced in 1997 and has been successfully employed in both research and production, with most published applications from the field of analytical chemistry. This should of course not come as a surprise since the sequential simplex method has been very popular among analytical chemists dating back to Stan Deming's pioneering work in the 70s and 80s. The OMM™ package was launched in 1998 and has not been sold as a stand-alone product, but as a part in real-time optimization projects. The first of these installations was commissioned in late 1998 and I will review results from 2½ years of operation in different industrial boilers and energy producing plants.

1. Öberg, T. G., and Deming, S. N. Find optimum operating conditions fast, Chem. Engr. Progr., 96, 53-59 (2000).
2. Box, G. E. P.Evolutionary operation: a method for increasing industrial productivity, Appl. Stat., 6, 81-101 (1957).
3. Spendley, W., Hext, G. R. and Himsworth, F. R. Sequential application of simplex designs in optimisation and evolutionary operation, Technometrics, 4, 441-461 (1962).
4. Zadeh, L. A. Fuzzy sets, Information and Control 8, 338-363 (1965).