机器学习辅助隔震支座优化布置  

Machine learning-aided optimization for seismic isolation bearings layout

作　　者：党育[1] 刘全明 马小科 DANG Yu;LIU Quanming;MA Xiaoke(School of Civil Engineering,Lanzhou University of Technology,Lanzhou 730050,China)

机构地区：[1]兰州理工大学土木工程学院,甘肃兰州730050

出　　处：《地震工程与工程振动》2023年第4期26-36,共11页Earthquake Engineering and Engineering Dynamics

基　　金：国家自然科学基金项目(62166025,51668043);甘肃省重点研发计划(21YF5GA073)。

摘　　要：针对如何快速确定隔震结构的最优隔震支座布置问题,使用机器学习辅助结构优化方法,采用人工神经网络构建了预测隔震结构设计参数与结构响应之间复杂关系的机器学习模型,以隔震结构响应最小为目标函数,用粗粒度并行遗传算法,得到隔震结构的最优支座布置。用一个实际隔震工程验证,结果表明:构建的人工神经网络模型对各结构响应参数的预测准确率均达到了92%以上,平均预测准确率达到93%,与精确计算的优化结果相比,用机器学习辅助隔震支座布置的优化结果平均误差为3%,加速比约为300,说明该方法具有很高的预测精度和计算效率。A machine learning aided structural optimization method is used to quickly determine the optimal layout of isolation bearings for isolated buildings.An artificial neural networks(ANN)model is used to construct a machine learning model for predicting the complex relationship between design parameters and structural response of isolated buildings.With the objective function of minimum the responses of isolated buildings,a coarse-grained parallel genetic algorithm is used to obtain the optimal layout of isolation bearings for isolated buildings.Verified by an actual isolated building,the results show that the prediction accuracy of the constructed ANN model for each structural response is more than 92%,and the average prediction accuracy is 93%.Compared with the optimization results of accurate calculation,the average error of the optimization results of the isolation bearings layout with machine learning is 3%,and the acceleration ratio is about 300.It is indicated that this method has high prediction accuracy and computational efficiency.

关 键 词：隔震结构 粗粒度并行遗传算法 优化 机器学习 

分 类 号：TU352.1[建筑科学—结构工程]