Weighted residual methods and the suboptimal control of distributed parameter systems†

Publisher Website
Google Scholar
Abstract
A suboptimal control algorithm for distributed parameter systems is developed in a framework which synthesizes weighted residual methods and mathematical programming. The heat exchanger example of Koppel et al. (1968) is employed for introducing the algorithm. First, the Galerkin procedure with polynomial modes is applied to obtain a lumped ODE model for the distributed parameter system. Then the state and control variables of the lumped control problem are approximated by cubic splines on a uniform mesh. Through collocation at the knots, the ODE model is reduced to a sot of linear algebraic equations and the suboptimal control is determined from the solution of a quadratic programming problem with sparse matrices. Numerical results for the heat exchanger example are presented and compared with those obtained by the authors (Neuman and Sen 1972) using the Ritz-Trefftz algorithm (Bosarge and Johnson 1970) for the lumped control problem. For this example, the two algorithms yield essentially identical results with comparable computational requirements. Application of the Ritz-Trefftz algorithm, however, is limited to lumped, linear-quadratic control problems without constraints on the state or control. The approach advocated in this paper, therefore, offers a viable approach to control problems in distributed parameter systems.