Optimization of process performance



The growing availability of fast, large memory digital computers has made it practical to consider the physical implementation of control system designs incorporating appropriate strategies for automatic process optimization. The control system is taken to consist of the process to be optimized together with the interconnected digital computer. The control strategy is realized in the program of the digital computer. In the present paper a number of such programs or algorithms are discussed for carrying out a search of the possible settings of the process input (independent) variables in such a way as to locate an extremal of the possible values of a chosen objective function. The magnitude of these variables is determined from measurements taken of the dependent variables in the process. It is shown that for the particular process used as an example it is desirable to alter the search strategy as the optimization proceeds in order to locate the extremal in a minimum amount of time. The emphasis at the beginning of the search is on speed in moving towards the optimum and at the end on accuracy.

Further, a computational technique is described whereby the dynamic response of the process to the various search steps (or settings) is under time-optimal control. This procedure is important for the fast execution of the search programs and consequent rapid location of the extremal of the chosen objective function.