基于模态分析和BP神经网络的复合材料脱层损伤监测研究

Investigation of Delamination Detection for Composite Structures Based on Modal Analysis and BP Neural Network

基于假定自然应变法建立了一个新的八节点压电固体单元,并采用具有相同坐标但不同节点号的节点对模拟脱层,分析了含不同脱层损伤梁的模态特性;进而提出了一种将计算力学、神经网络和实验模态分析相结合的复合材料结构脱层损伤检测的新方法。该方法通过数值模拟的手段为神经网络提供充足的训练样本,以实验模态结果作为神经网络的输入来预测复合材料结构的脱层损伤,实验结果证明了这一方法的可行性。

An novel Assumed Natural Strain(ANS) piezoelectric solid element formulation was developed to analyze composite beam with different delamination size and location by using pairs of nodes with the same coordinates but different node numbers. Furthermore, a new method combining computational mechanics, neural network and experimental modal analysis was demonstrated for composite health monitoring. The numerical results obtained by FEM were used to train the neural network and the experimental modal frequencies were input to the neural network to predict the demalination location and extent.