基于曲率模态和BP神经网络的简支梁的损伤识别

Damage Identification Based on Curvature Mode and BP Neural Network for Simply Supported Beam

通过BP神经网络对简支梁的损伤位置和损伤程度进行了研究。文中首先采用有限元仿真软件ANSYS计算得到不同损伤情况下结构的前两阶固有频率并计算指定点的曲率模态值,并以此为输入参数建立用于识别简支梁损伤的BP网络,最后利用LM算法训练网络来进行损伤检测。结果表明:该方法能有效地对损伤位置及损伤程度进行识别,且对损伤程度进行识别的精确度较高。

BP neural network was used to study the damage position and degree for simply supported beam. Firstly, the finite element simulation software was used to calculate the first two natural frequencies of the structures with different damage and the curvature mode of the specially appointed points ; then they were made as the input of a BP network for simply supported beam＇s damage identification ; finally, the trained network was used to detect the damage. The results show that： this method can effectively identify the damage position and degree; there will be a high accuracy for the identification of damage degree.