从频域的角度分析,在Preisach类迟滞模型结构下,用具有相位超前和滞后的类迟滞元代替relay、stop和play迟滞算子,并将其作为神经网络的隐层及局部反馈层节点,构造了一种具有内部反馈的动态神经网络迟滞模型。该模型克服了Preisach类迟滞模型中迟滞算子不可微的缺点,使其能描述非单调迟滞特性。基于此模型,展开对高频响音圈电机非单调复杂迟滞特性建模的研究。实测数据的仿真实验结果表明所提出的建模方案能很好地逼近实际迟滞特性,并具有较好的预测精度。
The dynamic neural network hysteresis model hysteresis-like operators rather than relay, stop and play is constructed by using phase-lag and phase-lead hysteresis operators as the hidden nodes of neural network in the framework of Preisach model from the view point of frequency domain. This model overcomes the non-differentiable phenomenon of Preisach-class hysteresis, but it can describe the non-monotonic hyster- esis property. Based on this model, a research on non-monotonic complex hysteresis modeling for voice coil motor actuator under the high frequency response was carried out. Finally, the measured data-based simula- tion experimental results showed that the proposed model can approximate to the real hysteresis perfectly, and achieve an excellent prediction accuracy, which can illustrate the potential of proposed modeling tech- nique.