Study to Apply Artificial Neural Network for Establishing Displacement Models of a Cable-Stayed Bridge

Establishing a displacement model of a cable-stayed bridge plays an important role in the operating process of structures in order to assess the structural health conditions, detect damages to structures, and suggest recommendations for maintenance. This paper studies to application of Artificial Neural Network (ANN) for establishing the displacement models along to X, Y, and Z directions of the center main span point of a cable-stayed bridge. The strategy of the building displacement model includes three steps. Firstly, Structural Health Monitoring systems (SHMs) data of a cable-stayed bridge were acquired in long-term monitoring which includes Global Navigation Satellite System (GNSS) displacement data in 3D of the center main span point, air-temperature data, windspeed, and stress data. The acquired data were de-noised and then assessed the correlation between the GNSS displacement data and other influencing factors such as air-temperature, windspeed, and stress data. Secondly, the displacement models were established by applying the ANN method for the long-term monitoring data. Finally, the establishing models were then assessed for their precision by some criterions such as Root Mean Square Error (RMSE), the determination coefficient (R2), and making comparisons between the real GNSS measurement data and the predicting data. The results show that the displacement models have high precision and reliability with the RMSE less than ± 3 mm and R2 over 0.96.