Method For On Line Structural Health Monitoring Based On Convolutional Neural Network

For most equipment and structures,on-line health monitoring is the key method to ensure their normal operation and increase their working life.There are two main steps in the process of on-line health monitoring: 1.feature extraction;2.feature classification.Convolutional neural network integrates these two steps into a learning model,and improves the processing speed and accuracy of these two steps,so as to obtain more accurate classification results.In this paper,by studying the principle of convolutional neural network,a convolutional neural network model is established,and then the model is applied to the fault diagnosis of two structures,and finally the results show that the convolutional neural network model can be used in weak fault information and complex structure.The two tests are: the first is to simulate the fatigue damage of the cantilever beam,and the second is the fault diagnosis test of the mill main reducer.Firstly,a convolutional neural network model is established.Then the signals collected in the experiment are divided into two groups of training data and test data.The model is trained through the training data,and then the test data is input into the trained model and output.The result is to optimize the parameters of the model,and finally obtain a convolutional neural network model that can automatically diagnose the fault.At present,there are related papers at home and abroad to apply convolutional neural networks to fault diagnosis,and the effectiveness of the method is preliminarily proved.However,most of them currently apply the convolutional neural network model to a structure with simple structure or obvious fault information,which is difficult to be applied to practical complex structures or weak fault signals.Therefore,the above test simultaneously verified the convolutional neural network.The fault can still be automatically identified if the fault signal is weak or contains a lot of noise.