Reconstruction Method for Lateral and Longitudinal Bridge Displacements Based on Temporal Splicing

The lateral and longitudinal displacements on the bridge are essential parameters in identifying the structure's health. However, due to the error accumulation and data noise, existing methods struggle to resolve issues of accuracy degradation and fluctuations. To address these challenges, this study proposes a novel approach for monitoring bridge displacement using environmental factors. The main innovation of this study lies in utilizing the sustainably precise observability of environmental factors to reconstruct bridge displacements. The proposed method incorporates a dual-channel module and two temporal splicing steps. One channel of the dual-channel module processes a full day of data using a single-layer LSTM, ensuring the effective application of long-range features. The other channel of dual-channel module utilizes a single-head attention mechanism to capture sensitive points of real-time changes, aiming to improve the adaptability of local features. These features are processed through two temporal splicing steps to produce the final output, enhancing sensitivity under different environmental conditions. The performance of the proposed method is validated using one year of displacement data from the Shengli Yellow River Bridge. Compared to state-of-the-art models, the proposed method reduces root mean square error (RMSE) and mean absolute error (MAE) by ≥53.5% and ≥57.4%. It is indicated that the proposed method improves the accuracy of lateral and longitudinal displacements reconstruction of the bridge and enhances stability.