基于随机森林和最小角回归的结构地震需求重要性度量分析

An analysis of the importance of structural earthquake demand based on random forest and least angle regression

对带黏滞阻尼器的钢筋混凝土框架结构在El Centro地震波的作用下,采用OpenSEES软件进行了动力非线性时程分析,考虑了黏滞阻尼器的阻尼系数和刚度、钢筋的弹性模量和屈服强度、阻尼比、混凝土的抗压强度和弹性模量以及结构质量8个输入随机变量的影响,得到了框架结构的顶点位移、最大层间位移角以及基底剪力3种结构地震需求。提出将随机森林算法和最小角回归算法应用到结构地震需求的重要性度量分析中,得到了各个输入随机变量对3种结构地震需求的重要性性排序,并用Monte-Carlo数值模拟法进行了对比。结果表明,基于随机森林算法和最小角回归算法的结构地震需求重要性度量分析方法结果与Monte-Carlo数值模拟法基本一致,这两种方法是准确高效的方法,可以大大减少样本的数量。

The nonlinear time history analyses of a reinforced concrete frame structure with viscous dampers under El Centro seismic wave were performed by OpenSEES software.The effect of damping coefficient and stiffness of viscous dampers,elastic modulus and yield strength of steel bars,damping ratio,compressive strength and elastic modulus of concrete and structure quality were considered.Three kinds of earthquake demand for reinforced concrete frame structures with viscous dampers were obtained respectively,namely,top displacement,the most story drift angle,and base shear.The application of random forest algorithm and least angle regression algorithm to the analysis of the importance of structural earthquake demand was proposed.The importance ranking to the three kinds of earthquake demand of each input random variable was obtained.The results were compared with the Monte-Carlo numerical simulation method.The results of structural seismic demand importance measure method based on random forest algorithm and least angle regression algorithm were basically the same as the Monte-Carlo numerical simulation method.These two methods are accurate and efficient,which can greatly reduce the number of samples.