基于独立分量分析和神经网络的钢结构损伤识别方法

Method of steel structural damage detection based on ICA and probabilistic neural network

为了有效剔除钢结构振动信号中的噪声,提取用于损伤识别的特征量,采用独立分量分析方法分离统计独立信号,同时得到表征结构损伤状态的混合矩阵,然后将混合矩阵作为特征量输入至神经网络进行训练,最后将训练好的神经网络作为分类器进行结构损伤识别。在冲击载荷作用下,针对钢框架结构模型进行了不同损伤部位的振动实验,结果表明:基于独立分量分析和神经网络的损伤识别方法具有较高的识别率和可重复性,而且实现简单,在结构损伤识别领域具有较大的应用潜力。

In order to effectively remove noises from vibration signals in the steel structure and extract the eigenvalue used for damage detection,the independent component analysis（ICA） was first used for separating and counting independent signals to obtain the mixed matrix representing the nature of structural damage.Then the artificial neural network（ANN） with the inputs of the mixed matrix was trained.Finally,the trained neural network as a classifier was applied to structural damage identification.Under the impact of the loads,the model of steel frame structure went through vibration tests at different damage positions.The experimental results demonstrate that the method of damage detection based on ICA-ANN not only has high identification rate and repeatability,but also is easy to implement and that the method has a great potential of application in the field of structural damage detection.