结构缺损的识别

Recognition of Structural Faults

利用小波分析与人工神经网络相结合的方法，对结构缺损进行了识别。敲击缺损试件后产生的振动信号由传感器拾取，经数据采集系统采集、适当处理后进行小波变换，形成人工神经网络的训练样本，并对所建网络进行训练，利用在训练样本中加入随机噪音的方法对网络的识别精度进行了讨论。本文用上述方法以悬臂梁试件为例，对结构缺损位置、缺损深度以及螺栓联接结构的紧固程度进行了试验。

In this paper,some structures with different faults are detected by wavelet analysis combined with artificial neural networks.The vibrating signals produced by striking the samples with different faults,are picked up by the data collection system,and then the wavelet transform is performed.The training samples of BP networks are extracted from the wavelet transform data.The neural networks are trained by the training samples.Finally,the recognition precisions of the neural networks,which we have trained above,are discussed by adding some of random noises to the training samples.When the controlling factor of the random noise is over 20%,some of them can not be detected correcty.Some examples of recognizing the structural faults are given.All of the machinery problems can be detected by methods mentioned above.