Speed Control of a Servo Hydraulic Actuator, Using Artificial Neural Networks and Feedback Error Learning Algorithm

In this article, speed control of a servo hydraulic rotary actuator is investigated, using flexible structure neural network (NN). The applied architecture of NN is a feedback error learning (FEL), whose underlying learning strategy is based on the inverse dynamics of the system under control. The classic control output was taken as the cost function to be minimized by the NN. A three-layer feedforward NN was applied and a flexible sigmoid activation function was used for the hidden layer nodes. The learning paradigm was online, making use of the back propagation of error with momentum. A simulation was performed and the results obtained indicated the high capability of the flexible NN in learning inverse dynamics in real time in controlling servo hydraulic systems.