Bridge Health Monitoring Data Analysis Based On PSO-VMD And LSTM Neural Networ

At present,many bridges are equipped with bridge structural health monitoring systems.Due to the timing and unrest of the bridge monitoring data,the data obtained by the monitoring lacks practicality.Therefore,this paper proposes a combined PSOVMD-LSTM model based on particle swarm optimization variational modal parameter decomposition-long short-term memory neural network data noise reduction-prediction,so as to realize the whole process of bridge monitoring data from collection noise reduction to prediction.The main contents of this paper include :(1)A data noise reduction combination algorithm for particle swarm(PSO)optimization variational mode decomposition(VMD)parameters is proposed to reduce and reconstruct the data.A continuous rigid bridge with a span of 220 m was used as the background bridge to simulate and collect the deflection changes during bridge operation,and the decomposition mode number K and penalty factor α parameters of the PSO optimization VMD algorithm were used to decompose the data and reduce noise.At the same time,the wavelet threshold noise reduction method and the empirical mode decomposition(EMD)method are used to compare the noise reduction effect,which verifies the effectiveness and superiority of the proposed PSO-VMD noise reduction method in removing the noise contained in the original data series.(2)Construction of multi-source bridge dataset and construction and implementation of long short-term memory neural network(LSTM)model.The input features are divided and combined,the combined dataset A-D is established and the experiment is trained.The input combination B(strain + acceleration)with the best prediction effect of the model is used as the input feature combination for subsequent research to select the final model optimal hyperparameter combination.Based on this combination parameter,the LSTM model is built and trained.(3)The combined model of PSO-VMD-LSTM bridge deflection noise reductionprediction is verified by example.The PSO-VMD method is used to predict the deflection of the monitoring data,and the predicted value of the PSO-VMD-LSTM combined model is obtained.Through model comparison,it is concluded that data noise reduction can improve the prediction accuracy of the prediction model,and the accuracy is the highest after PSO-VMD decomposition.The comparison and verification of LSTM deep neural network can improve the prediction accuracy of the model and reduce the error between the predicted value and the true value in the deflection prediction.Based on the comprehensive results,it shows that the PSO-VMD-LSTM model proposed in this paper can effectively eliminate the noise and interference in the monitoring data sequence,and apply the deep learning technology to the field of bridge health monitoring.Compared with other prediction models,it has great advantages in performance and the highest prediction accuracy in data noise reduction and prediction bridge deflection.On the basis of historical monitoring data,it can provide a basis for the subsequent reinforcement,repair and maintenance of bridges,ensure the safe operation of bridges during service,and have positive contribution value to the development of China’s infrastructure industry...