Process optimization and process robustness using experimental design: STAVEX

Thanks to user-friendly software and graphics-oriented techniques such as implemented in the expert system STAVEX developed at Ciba-Geigy, the powerful tools of statistical design of experiments (DoE) can now be used easily by non-statisticians. In an industrial context, the main applications of DoE are those of process optimization and process robustness. In both situations, a quantitative performance measure (response) of the process depends on measurable parameters or factors (process parameters, quality of raw materials...). The problem is to identify the "best" factor combination in order to achieve desirable values for the response. In the optimization case, "best" means that the response is as close as possible to the target value. In the robustness case, "best" means that at these factor settings, the response is as least as possible affected by noise perturbations. Two practical applications inspired from typical problems at Ciba-Geigy and that illustrate these two cases are discussed. The first one deals with the optimization of a dyestuff process with 11 a-priori important parameters using 29 experiments. The second example deals with the robustification of a pigment production process. Using adjustments of six easily controllable parameters, the impact of three noise factors on the color strength could be substantially reduced with a total of 32 experiments.