基于BP网络识别框架结构损伤的研究

Study fo framed structural damage detection based on BP network

基于模态振型,提出一个新的参数——振型曲率比,作为BP网络的输入向量,然后用有限元分析计算结果作为网络的学习和预测样本,对不同的结构模拟损伤进行了识别,结果表明该方法简便有效,值得推广应用。

Based on mode shape,a new parameter is put forward-mode shape curvature ratio,for detecting structure damages.And it is also the input vector of the neural network.Then through finite element analysis and calculating the training and forecasting samples are got for the network.The neural network can identify the damage location and degree of the structure.It proves that this method is simple and valid,value to prompt and application.