对Preisach类的迟滞非线性神经网络建模

Modeling Preisach-type hysteresis nonlinearity using neural networks

为了消除迟滞非线性对系统的不良影响,本文利用神经网络对Preisach类的迟滞非线性进行建模．通过引入一个特殊的迟滞因子,将多映射的迟滞非线性转换成一一映射,然后建立了基于神经网络的迟滞非线性模型．该模型结构简单,简化了辨识过程,可以调整神经网络权值以适应不同条件下的迟滞辨识．最后．应用该方法对压电执行器中的迟滞非线性建模,并与KP模型进行了比较．

In order to approximate the behavior of hysteresis nonlinerity which often severely limits the performance of the system, a neural-network-based hysteresis model is presented in this paper. A novel hysteretic operator is firstly proposed to transform the multi-valued mapping of Preisach-type hysteresis into a one-to-one mapping so that the neural networks are capable of implementing identification for hysteresis. The proposed model has a simple structure and simplifies identification procedure. Moreover, it is convenient to tune the weights of neural networks for the identification of hysteresis in different conditions. Finally the approach is applied to model the hysteresis in piezoelectric actuator and compared with the well-known KP model.