抄録
In general, conventional automatic optimizing techniques can be classified into two categories, the exploratory method and the predictive method using process models. These methods both have disadvantages. Namely, the former can not follow up and compensate for fast outer disturbances in the controlled system with large time lag, while the latter may settle into a nonoptimum point when the model does not picture the real characteristics of the controlled system. Presented in this paper are two new optimizing control schemes which use combination of the above two methods so that they compensate for the disadvantages each other. Especially, an application of model modification method, one of the two methods, to a hypothetical process is analyzed on an analog computer as to the setting time to an optimum point, relation between deviation of the settled point from the true optimum point and an accuracy of model, and the effect of execution time of computer. Also described is an experimentally-built optimizer that operates on the model modification scheme.
