摘要
对刚架结构，利用传递矩阵方法建立含裂纹单元的刚度矩阵，用有限元正分析得到在裂纹的不同位置和深度时结构的固有频率．用已知（计算）的裂纹与结构频率所相应的一定数量的样本，对三层前馈式神经网络进行训练，自学习后的网络将用于裂纹位置和深度的识别．以悬臂梁为算例，计算实验结果表明，此方法对裂纹的识别有一定的精度．
Stiffness matrices of beam elements with crack are established by transfer matrix method.Natural frequencies of structure with crack at different position are obtained by FEM.The neural network is to be used to identify the position and the size of crack after training of feedforward neural network with three layers using some samples.An example of the clamped free beam is provided to indicate that this method is of feasible for crack identification.
出处
《武汉交通科技大学学报》
1998年第6期567-570,共4页
Journal of Wuhan University of Technology(Transportation Science & Engineering)
基金
国家自然科学基金
关键词
裂纹
识别
反问题
神经网络
刚架结构
crack
identification
inverse problem
neural network
