Model reference adaptive control of an electrohydraulic motor drive using an artificial neural network compensator

Abstract

An artificial neural network is used as the feedback compensation element of a servovalve/ motor speed control system. The network is established on a variable gain/acceleration feedback principle and trained using computer simulation techniques. The network is then emulated as a real-time controller to significantly improve the speed characteristic of the motor. It is then shown how a feedforward network utilizing the motor pressure differential and speed may be readily adapted, in response to changes in motor speed, supply pressure and load torque, to track the prescribed input/output dynamic model.