Research On Bridge Health Condition Detection Method Based On Deep Learning

Bridges are expensive and closely related to people’s livelihood.If the bridge is repaired after the collapse,it will cause huge losses to the country.In order to reduce the future costs related to bridge management,it is necessary to inspect health condition of the bridges for a long time and repair and reinforce bridges timely.At present,it is one of the main challenges in engineering practice to improve the accuracy and efficiency of bridge structural state detection.Aiming at the above problems,this paper proposes a model based on residual network and attention mechanism algorithm,which can realize a more efficient detection of bridge health state,and can help improve the operation and maintenance efficiency of bridges.Firstly,according to the characteristics of the bridge’s original data,an appropriate data preprocessing scheme is designed.Then,by analyzing the characteristics of various deep learning models,we combine the residual network with efficient feature extraction ability and the LSTM network that can capture the time series data information.Moreover,the attention mechanism is used to realize the weighting of features in the time dimension.Therefore,a new model structure is designed based on the data of bridge structure to realize the detection of bridge health condition.Finally,the bridge simulation data provided by the cooperative research group and the measured data of Tianjin Yonghe Bridge structure provided by the Structure Monitoring and Control Research Center(SMC)of Harbin Institute of Technology are used to verify the model algorithm.Compared with other four existing algorithm models(SVM,CNN,LSTM,and Res Net),the model is more accurate and more stable in two data sets,which proves that the algorithm proposed in this paper can better meet the needs of practical engineering applications than the existing algorithm technology.The effectiveness of the improved algorithm is verified.In the process of solving the problem,in order to obtain bridge deflection data more conveniently,an image-based deflection measurement method was proposed.Through image processing,distortion correction,calibration of central coordinates and other operations,the deflection data of dynamic and static loads of the bridge with higher accuracy can be obtained.