基于实际震害数据机器学习的构件损伤预测研究  

Strong ground motion characteristics of Moho-reflected phase(SmS)during M 5.5 Pingyuan earthquake in Shandong on August 6, 2023

作　　者：彭文 王江[1,2] 马强 解全才[1,2] 陶冬旺 PENG Wen;WANG Jiang;MA Qiang;XIE Quancai;TAO Dongwang(Key Laboratory of Earthquake Engineering and Engineering Vibration,Institute of Engineering Mechanics,China Earthquake Administration,Harbin 150080,China;Key Laboratory of Earthquake Disaster Mitigation,Ministry of Emergency Management,Harbin 150080,China)

机构地区：[1]中国地震局工程力学研究所地震工程与工程振动重点实验室,黑龙江哈尔滨150080 [2]地震灾害防治应急管理部重点实验室,黑龙江哈尔滨150080

出　　处：《世界地震工程》2025年第2期125-136,共12页World Earthquake Engineering

基　　金：国家自然科学基金项目(52378544)。

摘　　要：准确识别构件层面损伤是提升建筑物地震韧性的关键;但以往研究中缺乏震后详细损伤数据,构件损伤传统研究方法存在局限性,以及模拟真实复杂的地震情景较为困难。因此,根据新西兰建筑破坏记录,建立构件损伤数据库,筛选出影响构件损伤的18个关键特征,包括地震动参数(peak ground acceleration, PGA;peak ground velocity, PGV等)、建筑特性(建筑年代、建筑价值等)以及场地是否液化等。分别使用XGBoost(eXtreme Gradient Boosting)、随机森林(random forest, RF)、LightGBM(LightGradient Boosting Machine)和CatBoost(Categorical Boosting)四种机器学习模型;并采用贝叶斯超参数优化算法调整超参数,构建分类模型来预测结构构件与非结构构件的损伤等级;并使用SHAP算法分析易损性构件的敏感特征。研究结果表明:1)四种机器学习模型中XGBoost模型精度最高。2)结构构件中的桩基础和板式基础,以及非结构构件中的墙面饰品和墙面覆层为易损性最高的构件。3)结构破坏中最重要的五个特征参数分别为建筑价值、建筑面积、谱加速度(spectral acceleration, SA)、PGV和V_(S30),而非结构破坏中最重要的五个特征分别为:建筑价值、建筑面积、SA、PGV和PGA。研究结果为快速准确识别损伤构件和指导建筑物修复提供重要依据。On August 6,2023,an M 5.5 earthquake occurred in Pingyuan County,Dezhou City,Shandong Province.Based on three-component acceleration records obtained from the early warning network,this study identified an abnormal attenuation pattern within the 50~100 km epicentral distance range,which significantly deviates from the conventional geometric attenuation model and exhibits no attenuation with increasing distance.Through seismic phase arrival time analysis,the study confirms that this phenomenon is closely related to the development of the Moho-reflected phase(SmS).The analysis shows that the arrival of the SmS phase results in an average amplification of about 1.52 times in peak ground acceleration(PGA)within the 50~100 km range from the epicenter.Further analysis reveals that the SmS phase causes a slow increase in 90%significant duration of the earthquake within the 50~100 km range,and this seismic phase exhibits high-frequency characteristics,with dominant frequencies in the 3~10 Hz range.Finally,the theoretical critical epicentral distance for full reflection of the SmS phase is derived,providing valuable insights into the segmented intervals for complex path effects in ground motion prediction equations(GMPEs).The study highlights that the characteristics of the SmS phase have a significant impact on seismic attenuation and are crucial for improving ground motion prediction equations.

关 键 词：非结构构件 结构构件 损伤状态预测 机器学习 实际损失数据 

分 类 号：P315[天文地球—地震学]