Neural Networks Compensation of Systems with Multi-segment Piecewise Linear Nonlinearities

Abstract: A neural networks (NN) compensator is designed for systems with multi-segment piecewise-linear nonlinearities. The compensator uses the back stepping technique with NN for inverting the multi-segment piecewise-linear nonlinearities in the feedforward path. This scheme provides a general procedure for determining the dynamic pre-inversion of an invertible dynamic system using NN. A tuning algorithm is presented for the NN compensator which yields a stable closed-loop system. In the case of nonlinear stability proofs, the tracking error is small. It is noted that PI controller without NN compensation requires much higher gain to achieve same performance. It is also difficult to ensure the stability of such highly nonlinear systems using only PI controllers. Using NN compensation, stability of the system is proven, and tracking errors can be arbitrarily kept small by increasing the gain. The NN weight errors are basically bounded in terms of input weight and hidden weight. Simulation results show the effectiveness of the piecewise linear NN compensator in the system. This scheme is applicable to xy table-like servo system and shows neural network stability proofs. In addition, the NN piecewise linear nonlinearity compensation can be further and applied to backlash, hysteresis, and another actuator nonlinear compensation.