Application of Self-Optimizing Controllers to Variable Time-Delay Processes

A comparison of model-free and model-based controllers is made for a stochastic process with a variable time-delay. A process is controlled by a modified Nelder-Mead evolutionary optimization technique suitable for noisy systems, the self-tuning controller of Vogel-Edgar and a developed method combining a self-tuning controller with a Smith predictor. An algorithmic approach to explicit dead-time estimation is presented. Controller performance is studied using a laboratory scale thermal process unit. Of the three controllers, the developed self-tuning controller was found to respond the best in the variable time-delay environment.