基于多模态的神经网络的结构损伤识别方法的研究

采用曲率模态和柔度曲率组合成多模态参数，针对连续梁结构在有限元模型基础上对结构进行了损伤识别研究．结果表明，以此多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可以准确地对结构损伤进行定位，而且对损伤的定量也取得了比较理想的效果，表明此网络还具备良好的容错性和鲁棒性．

Research on Damage Identification Based on Multi-modal Using Neural Networks

The multi-modal parameter which is composed of the curvature mode and the flexibility cur- vature is used to identify the damage on the base of finite element method model based on continuous beam. It has been proved that multi-mode parameter is considered as networks inputting parameter, and the location of the structure damage can be accurately determined and the quantitative of the struc- ture damage can be obtained by the neural networks. It has indicated that this neural network has a excellent identification ability with good ideal error tolerance and robustness.