基于模态参数和神经网络对裂纹梁的损伤识别

为了更准确地判断简支梁中裂纹的损伤位置和损伤程度,使用模态固有频率的变化率判断出简支梁中裂纹的存在,通过模态振型的变化判别出其损伤位置,并利用基于模态振型的BP神经网络理论,对有限元数值分析得到的模态振型差作为输入向量进行网络训练。结果表明:通过固有频率的下降和振型的变化可以快速地判断裂纹的位置,BP神经网络预测值与理论值误差非常小,可以很准确地判断裂纹的损伤程度。

Damage Identification of Cracked Beam Based on Modal Parameters and Neural Network Technique

In order to determine damage location and level of cracks in simply supported beam more accurately.Using modal natural frequencies to determine presence of cracks in simply supported beam and using changes of modal shape to judge out location of cracks.Modal shapes got by finite element analysis were taken as inputting vector and trained through BP neural network theory based on modal shape.It was shown that location of cracks can be determined quickly through drop in natural frequencies and changes of modal shapes,the error values is very small between prediction and theoretical value by BP neural network which can determine damage degree of cracks.