Process identification from closed-loop response using optimization methods

Existing methods employ different approaches and analyses for closed-loop i dentification of a suitable model such as a second order plus dead time (SO PDT) model for a single-input single-output (SISO) process. In this study, an optimization approach for identifying a SOPDT model from closed-loop res ponse is proposed and its performance is extensively evaluated mainly via s imulation. Typical examples, test conditions, controller and response types , step and relay responses as well as measurement noise are considered. Thr ee optimization methods, namely, the multi-pass LJ method (which finds the global minimum), modified Simplex method and a quasi-Newton method are trie d. The results show that the recovered model is more accurate than those ob tained by other methods, and that a reliable and consistent model can be ob tained even from the response with measurement noise, At least two minima a re likely when step response is used and several minima are probable when r elay response is employed. The optimization approach is independent of the nature of the closed-loop response and controller used in the test, be it a PSD controller or a relay. importantly, the test duration can be considera bly shortened because the step test need not be conducted until steady stat e, and in the relay test, steady oscillations are not required.