PI-controller parameters tuning method to reject disturbances acting on heating furnaces

On-line Kp, Ki gains adjustment of a Pi-controller to reject disturbances acting on a heating furnace is the main scope of this research. A neural tuner is used to resolve considered problem. it consists of two neural networks in order to ensure quality of control for heating and cooling processes. An additional network is integrated in its structure to enable disturbances attenuation. Network outputs are Kp, Ki values. Time moments when such networks need to be trained and learning rate values are determined by a rule base. A set of rules developed to reject step-like and impulse disturbances acting on the plant output is shown, as well as the tuner structure. The SNOL 40/1200 muffle electroheating furnace is used as a plant for experiments. Obtained results show the total amount of time spent on disturbance rejection may be reduced by 20% using the neural tuner in comparison with a control system with Pi-controller with fixed gains.