Abstract
In recent years, linear multivariable control theory has been developed extensively. Today, it seems to be brought to near completion. However, advanced control technics such as optimal control, state estimation etc. have scarcely been applied to real industrial processes. To help bridge the gap between theory and practice, it is necessary to undertake experimental investigation on pilot model plants by applying such technics. In view of these, a boiler model plant is utilized in this paper to examine one of modern digital control technics.
The test plant is modeled by a two-dimensional discrete-time linear dynamical system which is obtained by linearizing lumped-parameter energy and mass balance equations and discretizing them through sampling and holding. The purpose of control is to maintain controlled variables (drum pressure and drum liquid level) at the normal set points under step-wise changes of steam-valve opening and random environmental disturbances. Then an FTSC algorithm (Finite Time Settling Control) recently proposed by Y. Takahashi and M. Tomizuka is applied to the model plant. In the beginning, a direct application of the algorithm yields an unsatisfactory system performance, resulting in excessively oscillatory responses. Therefore, two types of correction for control gain matrices are proposed, one of which is based on introducing staleness-weighting factors into the discrete transfer function matrix and the other adopts another gains designed by a single-loop parameter tuning method. After this correction, the FTSC algorithm gives rise to a satisfactory result.
