基于SVM和最大熵模型的桥梁极值风速预测研究

Extreme wind speed forecasting of bridge based on SVM and maximum entropy model

大跨度桥梁对极值风荷载作用十分敏感,基于桥址处长期实测风速数据得到合理的极值风速至关重要。由于在桥址处进行长期风速监测较难实现,研究极值风速预测方法对于大跨度桥梁抗风研究具有重要意义。以泉州湾高速铁路斜拉桥为工程背景,提出一种基于支持向量机(SVM)和最大熵模型的极值风速预测方法。通过临近风速观测塔与桥面风速仪同步实测短期风速数据建立SVM风速预测模型,预测得到桥面长达3 a的平均风速数据,进而采用最大熵方法计算得到桥梁高度处不同重现期内极值风速,最后采用时域分析方法进行桥梁抖振响应分析,探讨不同风速下桥梁抖振响应的差异。研究结果表明:SVM模型应用于风速预测效果比较理想,测试集预测风速与实测风速均方根误差(RMSE)和平均绝对误差(MAE)分别为0.115和0.252。基于桥面预测平均风速样本,采用最大熵理论计算得到100 a重现期极值风速为21.52 m/s,小于泉州湾高铁斜拉桥抗风设计时采用的极值风速49.4 m/s。SVM模型预测得到的极值风速对应的主梁跨中位置横向、竖向和Rotx扭转抖振位移响应峰值与采用设计风速为49.4 m/s的桥梁抖振响应比较,降幅分别为80.3%,79.6%和78.7%。桥梁抖振响应受平均风速影响较大,选择准确的桥址处极值风速对桥梁抖振响应研究至关重要。研究结果可应用于仅有短期实测风速数据的桥梁得到合理的极值风速,为大跨度桥梁抗风研究提供参考。

Long-span bridges are highly sensitive to extreme wind load effects.It is essential to obtain reasonable extreme wind speed based on long-term measured wind speed data at the bridge site.Since long-term wind speed monitoring at bridge sites is difficult to achieve,it is of great significance to study the extreme wind speed forecasting method for wind resistance studies of long-span bridges.Taking the Quanzhou Bay high-speed railway cable-stayed bridge as the engineering background,the wind speed forecasting method based on support vector machine(SVM)and maximum entropy model was proposed.Establishing SVM wind speed prediction model by synchronizing short-term wind speed data measured by adjacent wind speed observation tower and bridge deck anemometer.Thus,the average wind speed data of the bridge deck was predicted for up to three years.The maximum entropy method was used to derive the extreme wind speed at the height of the bridge for different recurrence periods.Finally,the time domain analysis method was used for bridge buffeting response analysis,and the differences in bridge buffeting response under different wind speeds were explored.The results of the study are showed.The SVM model is applied to wind speed prediction with satisfactory results,and the root mean square error(RMSE)and mean absolute error(MAE)of the predicted wind speed and the measured wind speed in the test set are 0.115 and 0.252,respectively.Based on the predicted average wind speed samples of the bridge deck,the maximum entropy theory is used to calculate the 100-year recurrence period extreme wind speed of 21.52 m/s,which is smaller than the extreme wind speed of 49.4 m/s used in the wind resistance design of the Quanzhou Bay high-speed railway cable-stayed bridge.The peak values of the transverse,vertical and Rotx torsional buffeting responses at the mid-span position of the main girders predicted by the SVM model are 80.3%,79.6%,and 78.7%respectively when compared with the buffeting response of the bridge using 49.4 m/s as the design wind speed.The bridge buffeting response is greatly influenced by the average wind speed,so selecting he accurate extreme wind speed at the bridge site is essential for the bridge buffeting response study.The results of the study can be applied to obtain reasonable extreme wind speed for bridges where only short-term measured wind speed data are available,and provide a reference for the wind resistance research of long-span bridges.