Research On Damage Identification Of Bridge Structures Based On Deep Learning

Bridge is a very important large structure.In the process of operation,the environment is very complex and the bridge will inevitably damage.If the damage can be found and deal with the damage,this is very important for the structural safety of the bridge.Aiming at the problem of damage occurrence time of bridge structures,a method to identify damage occurrence time of bridge structures is proposed in this paper,which is based on Convolutional Autoencoder neural network.The convolutional autoencoder neural network is put forward in this article,processing the response signal.It can encode the original data of high dimension,then from low dimensional feature space by decoding operation data refactoring,the convolution coding neural network is its essence or coding neural networks,it is a kind of unsupervised learning model of deep learning,The data obtained from monitoring can be trained directly.In this paper,the identification of the damage occurrence time at different damages of the simple beam is performed by numerical simulation.Based on the position information contained in the measurement point,we can find the position of the damage occurs.The experiment is used to verified.The measured normal data of acceleration response are trained in the convolutional self-coding neural network,and the trained model is used to identify the time when damage occurs under different test conditions and to carry out preliminary localization of damage.Through the analysis of the numerical simulation and experimental results,it can be shown that the proposed method can identify the moment when the structure damage occurs under the testing conditions of different degrees of damage,and can roughly locate the damage,which is of great significance for the actual engineering.