Closed-loop identification of TITO processes using time-domain curve fitting and genetic algorithms

Abstract

Closed-loop identification of two-input two-output (TITO) processes by time-domain curve fitting to minimize the sum of the squared errors between the actual and calculated closed-loop responses is studied. The perceived difficulties with this methodology are the need for global optimization because of the possible existence of multiple minima, the extensive computations, the evaluation of numerical derivatives for efficient optimization methods, and the accuracy and reliability of the results. In this study, a genetic algorithm (GA) is employed to locate reliably the global minimum of the least-squares problem. Results show that time-domain curve fitting via GA, followed by the BFGS method, recovers accurate and reliable transfer function models of TITO processes from closed-loop responses under a range of test conditions including different controller settings in the closed-loop test, different durations of the test, underdamped or overdamped closed-loop responses, and measurement noise. Because closed-loop tests need not be conducted until steady state is reached, the test duration can be reduced significantly.