Establishment of a Hybrid Model for Bridge Structure Risk Prediction Based on Long Short-Term Memory Model

The aim of this study is to establish a hybrid model based on wavelet denoising and long short-term memory (LSTM) network for the prediction of bridge structural health monitoring risk. Firstly, 7-day bridge structural health monitoring indicators of a bridge were obtained from a bridge platform in Hubei Province; secondly, a hybrid prediction model was established after wavelet denoising and LSTM network analysis; finally, the accuracy of the prediction model was evaluated. The results show that the hybrid model with wavelet denoising of the time series combined with LSTM exhibits lower test errors compared with the single LSTM model, indicating that the model has higher accuracy in predicting bridge structural health monitoring indicators. The establishment of this hybrid model provides reference and technical support for the prediction of bridge structural health monitoring risks and provides a research basis for deep exploration of the interdisciplinary field of combining deep learning and bridge structural health monitoring indicators.