基于BP神经网络的压电智能结构振动系统辨识

Identification of Vibration System for Piezoelectric Smart Structures Based on BP Neural Network

针对神经网络在压电智能结构振动控制中的关键问题之一——系统模型的神经网络辨识,用引入时延的多层前馈BP神经网络串—并联型结构对表面粘贴压电片的柔性悬臂梁进行非线性动态系统模型辨识。考虑压电片对梁的质量和刚度矩阵的影响和实验提取数据的繁琐问题,用有限元分析软件ANSYS对智能梁进行模态和瞬态响应分析,利用获取的系统动力响应时间序列对神经网络进行离线训练,通过MATLAB神经网络工具箱对算例进行仿真显示。

Neural network identification of the system is a key issue in the vibration control of piezoelectric smart structures. A kind of BP neural network with delay time is employed to identify the nonlinear dynamic system of piezoelectric smart cantilever beam, which surface is bonded several piezoelectric patches. Considering the effect of piezoelectric patches on the mass and stiffness of beam and the difficulty of extracting the experimental data, the modals and the transient response of beam is analysed by use of ANSYS. A BP neural network with delay time is built and trained offline in basis of the data using MATLAB Neural Network Toolbox.