A PATTERN-RECOGNITION APPROACH TO THE SOLUTION OF OPTIMAL SINGULAR CONTROL-PROBLEMS

A considerable number of algorithms have been proposed in the literatu re for the optimization of control policies of non-linear chemical sys tems. However, most algorithms do not seem to be flexible and robust e nough for their application to be a matter of routine. Rather, each of them presents typical shortcomings that restrict applications to spec ial classes of problems. In particular, bang-bang (or on-off) policies are difficult to optimize, due to control discontinuities. In this pa per we propose a strategy based on a combination of pattern recognitio n theory and non-linear mathematical programming for the computation o f optimal singular control problems. (C) 1997 Published by Elsevier Sc ience S.A.