Modified Feedback Error Learning Control of Shape Memory Alloys Actuator using Evolutionary Neural Network

Abstract

In this paper, a modified feedback error learning approach (called MFEL) is proposed for a nonlinear system. In MFEL, an inverse evolutionary neural (IEN) model that dynamically identifies offline all nonlinear features of the nonlinear system, provides the initial value of a feedforward compensator. A PID feedback control is combined with a feedforward compensator to eliminate the steady-state error and guarantee the global asymptotic stability of the overall system. The learning is based on the feedback error signal is employed. An adaptive backpropagation (aBP) with self-adaptive learning rate based Sugeno Fuzzy logic is developed and employed in the MFEL to adapt well to disturbances and dynamic variations in its operation. To prove the effectiveness of the proposed MFEL controller, first, the benchmark nonlinear SISO system is used to evaluate the controller. Then, the experimental shape memory alloys (SMA) actuator is set up to test the controller. The simulation and experimental results proved that the proposed controller provides better results compared to the feedback control.