Optimal Adaptive Control of an Ash Stabilization Batch Mixing Process using Change Detection

We present an optimal adaptive controller that is used to regulate the chemical process of wood ash stabilization (WAS). The model parameters of the time-varying process dynamics are estimated using recursive least squares (RLS). At each batch an auto-tuning sequence produced with the controller disabled is carried through in order to obtain a good estimate of the process dynamics. After the auto-tuning sequence is completed, a generalized predictive controller is enabled to control the WAS process. The control objective is to regulate the normalized effective power Pe(t) to the level Pe crit that represents the critical rate of useful work being performed by the three-phase asynchronous machine used for the stirrer drive. Hence, Pecrit also represents the desired mixture viscosity. If more water is added to the stabilization process after Pecrit has been reached, one will obtain a mixture useless for granular material. To cope with this problem, change detection is used to reach the desired level Pecrit without any pre-determined set-point. Two methods are evaluated; a probing strategy and the geometric moving average test, both adequate for successful implementation. The used control strategies are presented and off-line simulations with a model of the physical process evaluate the control performance