摘要
为获取建筑结构在地震过程中的位移响应,解决结构主动控制中的时滞问题,使用集成学习方法预测结构位移响应。提出了一套适用于结构地震动响应预测的集成学习模型,使用不同的神经网络和损失函数组合作为基学习器,在此基础上使用全连接网络构建二级学习器,得到最终预测模型。模型使用双向地震动加速度和结构位移响应作为输入,预测短期的结构位移响应;结构位移响应可实现多步预测,进行动态化输出。将该框架应用于某钢框架结构,进行数值实验,结果表明,集成后的模型预测结果要优于单一模型的预测结果,所有模型都表现出预测误差随着预测时间的增加而增加的规律。模型的预测随着地震过程在同步进行,当模型预测值第一次达到某一阈值时,则将预测结果反馈到主动控制系统中,提前进行结构振动控制,而不必等待位移传感器采集到真实的响应数据,从而减小甚至避免主动控制中时滞问题的影响。
In order to obtain the displacement response of building structures in the earthquake process,and solve the delay problem in the active control of structures,ensemble learning was used to predict the seismic structural response.In this paper,an ensemble learning model is proposed for the prediction of ground motion response of structures.Different neural networks and loss functions are combined as the base learners,then a Fully Connected Neural Network is used to construct a secondlevel learner to obtain the final prediction model.Bidirectional ground motion acceleration and structural displacement response are used as inputs to predict the short-term structural seismic response.This model could realize multi-step prediction of structural seismic response and dynamic outputs.Several numerical experiments are carried out on a steel frame structure,and the experimental results show that the prediction results of the integrated model is better than those of the single model,and all models display a trend of prediction errors increasing with the increase of the prediction time.When the model’s predicted value reaches a certain threshold for the first time,the prediction result will be sent to the active control system,so the structure vibration control will be carried out in advance without waiting for the real response data collected by displacement sensors,thereby reducing or even negating the influence of time delay in active control.
作者
安超
史健勇
潘泽宇
姜柳
王佳亮
AN Chao;SHI Jianyong;PAN Zeyu;JIANG Liu;WANG Jialiang(School of Naval Architecture,Ocean&Civil Engineering,Shanghai Jiao Tong University,Shanghai 200240,China;Shanghai Key Laboratory for Digital Maintenance of Buildings and Infrastructure,Shanghai 200240,China;Shanghai Urban Construction Design&Research Institute(Group)Co.,Ltd.,Shanghai 200001,China)
出处
《防灾减灾工程学报》
CSCD
北大核心
2023年第3期435-443,共9页
Journal of Disaster Prevention and Mitigation Engineering
基金
重庆市自然科学基金项目(CSTC2021JCYJ-MSXMX0986)
基于BIM+AI的智慧化建筑抗震管理系统关键技术研究项目(沪住建管科2020-009-005)资助。
关键词
集成学习
机器学习
时序预测
双向地震响应预测
ensemble learning
machine learning
time series forecasting
bidirectional seismic response prediction
