Nonlinear process modeling and optimization based on Multiway Kernel Partial Least Squares model

MKPLS (multiway kernel partial least squares) methods are used to model the batch processes from process operational data. To improve the optimization performance, a batch-to-batch optimization strategy is proposed based on the idea of the similarity between the iterations during numerical optimization and successive batch runs. SQP (Sequential Quadratic Programming) coupling with MKPLS model is used to solve the optimization problem, and the plant data, instead of the MKPLS model predictions, are used in gradient calculation. The proposed strategy is illustrated on a simulated bulk polymerization of styrene. The results demonstrate that the optimization performance has been improved in spite of the model-plant mismatches.