A NEURAL NETWORK BASED THICKNESS CONTROL FOR TANDEM ROLLING MILL SYSTEMS

摘要

Since the last decade, artificial neural networks have been successfully applied to the control of several industrial processes. In this work an interesting application of neural network control of tandem systems is presented. One of the main problems in the control design for rolling systems is the difficulty to eliminate time delays from the measurement of the strip output thickness. The time delay is a consequence of the output sensor placement that is usually located at some distance from the roll gap. The proposed technique uses the gap direct measurement to feed an output thickness predictor algorithm (OTPA). The OTPA is based on sensitivity functions whose parameters are previously calculated from the differentiation of the outputs of a properly trained artificial neural network (ANN). The main aspect of this ANN-based control scheme is that it allows to overcome the undesired consequences of the existing time delays. Simulation results are presented to illustrate the proposed technique performance.