A Direct Prediction Method of Bridge Wind-induced Response Based on Long-term Measured Data

Abstract

The invention discloses a method for directly predicting the wind-induced vibration response of a bridge based on long-term actually measured data. The method comprises the following steps: extracting primary information of long-term actually measured data of a structural health monitoring system (SHMS) of the bridge by utilizing a principal component analysis method at first, and constructing an SHMS actually measured data-bridge wind-induced vibration response model by adopting a BP neural network in combination with a parameter influencing the wind-induced vibration response of a bridge structure; then, inputting data, such as actually measured wind speeds and wind directions, into the model to obtain a wind-induced vibration response predication value, and comparing and checking the wind-induced vibration response predication value with an SHMS actually measured value of the bridge so as to obtain reliable wind-induced vibration response of the bridge; and finally, storing the latest actually measured data in an SHMS long-term actually measured database, and continuously optimizing a BP neural network prediction model so as to further improve the prediction precision of the wind-induced vibration response. With the help of the SHMS long-term actually measured data, the traditional finite element model is replaced by the neural network; and thus, an effective means and reliable analysis basis are provided for monitoring and safety assessment of wind-induced vibration of the bridge.

Description

technical field [0001] The invention relates to the field of wind resistance of long-span bridges in civil engineering, uses a BP neural network model established based on long-term measured data to predict the wind-induced vibration response of bridges, and is applicable to the wind-induced vibration response prediction of long-span bridges in strong / typhoon-prone areas. Background technique [0002] In recent years, many long-span bridge projects across rivers and seas have been built at home and abroad, and their structural forms are mainly suspension bridges and cable-stayed bridges, such as the Akashi Kaikyo Bridge, Sutong Bridge, and Xihoumen Bridge. For long-span cable-supported bridges, as the span increases, the structural stiffness decreases significantly, and the wind sensitivity increases significantly, making wind-induced vibration one of the main factors threatening the safety of bridges. [0003] The Bridge Structural Health Monitoring System (SHMS) is a bridg...

AI Technical Summary

Problems solved by technology

However, due to assumption or measurement errors between the model and the actual structure in terms of materials, geometric properties, and boundary conditions, as well as the discretization error of the finite element model itself, the bridge finite element model based on the structural design drawings is often in static dynamics. There is a difference between the response and the actual structure. In addition, the finite element modeling process is cumbersome and complicated, the calculation efficiency is low, and the model update is slow. Therefore, the present invention breaks away from the finite element model innovatively, and uses BP neural network to establish a long-term measured data-based The direct prediction method of bridge wind-induced response provides a reliable and effective means for real-time and accurate prediction of bridge wind safety in strong / typhoon areas in the future

Images