桥(图论)
气动弹性
人工神经网络
风速
计算机科学
天气预报
机器学习
钥匙(锁)
辍学(神经网络)
弹道
数据挖掘
模式(计算机接口)
数据建模
气象学
人工智能
循环神经网络
预测建模
回归
风向
回归分析
工程类
风力发电
空气动力学
作者
Peng Liu,Wei Cui,Lin Zhao
摘要
Buffeting of long‐span bridges induced by strong winds has received attention recently due to traffic serviceability. An essential challenge is whether the bridge response can be estimated before the arrival of strong winds. In this work, a buffeting response early forecasting framework based on meteorological weather prediction is proposed. The initial input for the forecasting model is derived from the original low‐time‐resolution wind data at the meteorological station, as provided by the weather research and forecasting (WRF) system. High‐time‐resolution wind data at the bridge site are predicted using a combination of an LSTM neural network and a super‐resolution method. Then, the buffeting response regression model is built based on the LSTM seq2seq method for different types of wind. The proposed framework introduces two key innovations. First, a mapping method is devised to connect large‐scale wind speed prediction data with local wind conditions at the bridge site. Furthermore, a predictive model for the wind‐induced response of bridges, incorporating a dropout layer, has been formulated to address various types of wind disasters. This established mapping model considers sequence‐to‐sequence relationships and estimated uncertainty. This integrated approach establishes a forecasting framework that spans from long‐term future weather predictions to the prediction of wind‐induced responses in significant bridge projects. The proposed method is of significant value for optimizing bridge operation and improving traffic control.
科研通智能强力驱动
Strongly Powered by AbleSci AI