Advanced control design for a high-precision heating furnace using combination of PI/Neural network

Abstract

Nowadays, thermal energy plays a very important role in manufacturing as well as daily activities, such as: steaming, drying, pasteurization, sterilization, steel refining, cooking, heating. Day by day, high-accuracy control is required to increase the performance of the thermal applications. However, nonlinear characteristics and high-inertial property are main obstacles in the development of precise controllers. In this paper, we propose an intelligent control scheme to improve the performance of a typical heat furnace. The mathematical model of the plant is first identified based on the experimental data acquired. The controller is then designed with a proportional-integral (PI) regulator and a neural-network-based adjustor. The linear control term is employed to ensure the excellent steady-state behavior while the intelligent control signal is worked as a feed-forward predictor to maintain the acceptable transient performance. The PI control gains are obtained using an auto-tuning toolbox. A nonlinear updating method is developed to increase the learning speed of the network. Effectiveness and feasibility of the proposed controller are confirmed both by simulation and real-time experimental results.