GENERATION OF REDUCED STATISTICAL-MODELS FOR NLP AND MINLP OPTIMIZATION

A general strategy for the generation of statistical models with a red uced complexity of the mathematical formulation, in order to handle NL P and MINLP optimization problems efficiently, is proposed. Resulting from increasing complexity of structural and parameter optimization pr oblems in chemical engineering, we are very often, especially when dea ling with highly nonlinear problems, confronted with the situation tha t the currently available solvers are not able to cope with the proble m size or require very good starting points. In many cases the utiliza tion of reduced statistical models can help to overcome these difficul ties by enabling a better insight into the optimization problem and pr oviding sufficient initial guesses, with respect to the process struct ure and the unit parameters. In order to fulfill the entire optimizati on task it is desirable that the reduced models also consider the infl uence of significant parameters. The generation of such reduced statis tical models on the basis of associated rigorous models is described a nd examined by using the commercial flowsheeting package SPEEDUP.