Modeling inverse hysteresis using neural networks

In order to compensate the hysteresis nonlinearity and improve the precision of system with hysteresis, an inverse hysteresis model was constructed. As neural network can not be used to approximate the multi-valued mapping of inverse hysteresis directly, an inverse hysteretic operator is proposed to transform the multi-valued mapping into a one-to-one mapping which enables neural networks to approximate the behavior of inverse hysteresis. The advantage of the neural networks based inverse hysteresis model is that it has a rather simple architecture. Furthermore, it is convenient to tune the weights of neural networks for the identification of inverse hysteresis in different conditions. Finally, the approach was applied to model the inverse hysteresis in piezoelectric actuator.