基于偏最小二乘回归的地震动复合强度参数构造与统计性分析

为解决某些地震动强度参数对某些结构有效性较好，而对其他结构有效性较差的问题，基于偏最小二乘回归提出一种地震动复合强度参数的构造方法。对构造的地震动复合强度参数与工程需求参数进行概率地震需求分析，从充分性、有效性、有益性和鲁棒性四个角度对地震动强度复合参数开展统计性分析。为校验所提方法的准确性，选择具有不同高度和不同抗震设防水平的16个钢筋混凝土框架结构为例，并遴选100条真实地震动记录作为地震输入。选取8个地震动强度参数作为基本参数，以结构最大层间位移角作为工程需求参数，采用偏最小二乘回归方法，构造地震动复合强度参数，并对其进行统计性分析。分析结果表明：偏最小二乘回归方法是一种较为合理构造地震动强度复合参数的方法，所构造的地震动复合强度参数具有比单个强度参数更优良的统计性，具有更好的有效性、有益性、充分性和鲁棒性。

Development and statistical analysis of compound intensity measure using partial least-squares regression

Some ground motion intensity measures are significantly effective for some structures while they are not effective for other structures. To solve this problem, a method for developing compound ground motion intensity measures was proposed based on the partial least-squares regression. Probabilistic seismic demand analysis was then conducted using the developed compound ground motion intensity measures and engineering demand parameters. The statistical properties of the developed hybrid ground motion intensity measure were examined from four aspects in terms of effectiveness, sufficiency, proficiency and robustness. To illustrate the proposed method, a total of 16 reinforced concrete frame structures with different heights and different seismic fortification levels were used as the study cases. A set of 100 real ground motion records were selected as earthquake inputs. Eight ground motion intensity measures were taken as the basic parameters. The maximum inter-story drift was used as the engineering demand parameter. Based on the partial least-squares regression, a compound ground motion intensity measure was developed and its statistical properties were investigated. The results show that the partial least-squares regression is a reasonable way to develop compound ground motion intensity measures. The developed compound intensity measure has better statistical properties, effectiveness, proficiency, sufficiency and robustness than the single intensity measure.