Design of supervisor-based adaptive process fuzzy logic control

The industrial processes are nonlinear, subjected to disturbances, inertial, with time delay and parameters that change with the operating point, over the time or both. The adaptive fuzzy logic controllers (AFLCs) are promising in controlling of such plants ensuring a high system performance. Their complex structure and design and the computational load required restrict their wide application on industrial scale. An approach for the design of simple AFLCs is suggested on the basis of a main FLC and a FL supervisor (FLS) with input the online estimated current plant gain and output correction of selected scaling factors. The effect - a continuous adaptation of the control surface in response to plant changes, is equivalent to the sophisticated self-organising techniques. The needed input data are expert assessed ranges for the plant gain, plant delay and system error. The approach is applied for the design of process AFLCs for a chemical reactor and a two-tank plant using main FLCs of Mamdani type and parallel distributed compensation. The results from simulations show decreased overshoot, settling time, control effort and coupling in the whole range of operation and parameter changes. The comparison is made with a FLC system and a linear PI system.