Robust control of nonlinear systems using pattern recognition

Abstract

Pattern-recognition based methodology implemented in the architecture of artificial neural networks can be used to model knowledge-intensive feedback control systems. The procedure for development and practical design of a neural-net based control system is described and demonstrated by the example of a nonlinear hydraulic system. The neural-net control system has been trained to provide stabilizing controls and to detect and identify malfunctioning sensors. The results of computer simulations and experiments are presented to illustrate the proposed approach. They show that the neural-net control system is capable of maintaining any desired level of liquid in a tank, for tanks of various shapes; of stabilizing the system in the presence of large disturbances; of detecting and identifying defective level sensors; and of estimating parameters in nonlinear processes.