利用组合模态参数识别结构损伤的神经网络法

Structural Damage Identification Based on Combination of Modal Parameters Using Neural Networks

采用组合模态参数在有限元模型基础上对结构损伤进行了识别.同时考虑了噪声输入情况下,即存在数据误差时神经网络的损伤识别能力.结果表明,以组合模态参数作为网络输入参数,并通过学习训练所得网络不仅具有理想的损伤识别能力,还具备良好的容错性和鲁棒性.

The combination of modal parameters is used to identify the damage of a FEM（finite element method） model using neural networks. The identification ability with different levels of noise and incomplete mode shapes are also investigated. It has been proved that the neural network using combination of modal parameters as input has a excellent identification ability with ideal error tolerance and robustness