摘要
通过对单自由度体系的统计矩理论分析,提出一种融合位移四阶矩和加速度八阶矩的特征指标,并结合L1正则化优化准则进行模型修正的框架结构损伤识别新方法。基于理论分析构建3种典型统计矩特征指标,考虑无噪音和信噪比40 dB下的不同损伤工况,对某12层标准框架模型进行损伤识别对比研究。在此基础上对常规残差优化准则和基于L1正则化优化准则进一步分析,结合12层标准框架振动台试验数据,选取3个典型振动工况中所有梁柱单元的损伤识别结果进行研究。结果表明:新方法下融合位移四阶矩和加速度八阶矩的特征指标较单一统计矩指标具有更高的敏感性和抗噪性,结合L1正则化优化准则的模型修正方法较常规残差优化准则的传统模型修正方法具有更高的稳定性及有效性。
By analyzing the statistical moment theory from the single degree of freedom system, a new frame structure damage identification method based on model updating is proposed. This method combines the characteristic fusion index of the fourth-order displacement and the eighth-order acceleration statistical moment with L1 regularization optimization criterion for model updating. Based on theoretical analysis, three typical statistical moment characteristic indicators for damage identification are formed. Then, a benchmark of 12-storey frame finite element model is established for comparative analysis with no measurement noise and the influence of SNR 40 dB on measurement noise. On this basis, the conventional residual optimization criterion and the L1 regularization optimization criterion are further analyzed. By combining the 12-storey benchmark frame structure shaking table test data, the damage identification results of all beam and column elements in 3 typical vibration cases are selected for research. Results show that the characteristic fusion index of the fourth-order displacement and the eighth-order acceleration statistical moment can achieve better sensitivity and anti-noise ability than the single statistical moment characteristic index. And the method based on L1 regularization optimization criterion is more stable and effective than the conventional residual optimization criterion.
作者
阳洋
陈熠昕
凌园
王松
王瑞琼
Yang Yang;Chen Yixin;Ling Yuan;Wang Song;Wang Ruiqiong(I.Key Laboratory of New Technology for Construction of Cities in Mountain Area,Ministry of Education,Chongqing University,Chongqing 400030,China;School of Civil Engineering,Chongqing University,Chongqing 400030,China;Pover China Chongqing Engineering Corporation Limited,Chongqing 400060,0 China)
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2020年第12期217-226,共10页
Chinese Journal of Scientific Instrument
基金
国家重点研发计划课题(2020YFF021780)
国家自然科学基金(51778090)
重庆市研究生科研创新项目(CYS20022,CYS19004)
重庆市科技局技术创新与引用发展专项(cstc2019jscx-msxmX0015)
中国电力建设股份有限公司科技项目(KJ-2020-117)
重庆市交通局科研项目(渝交科【2019】29号)资助。
关键词
统计矩理论
L1正则化优化准则
损伤识别
模型修正
statistical moment theory
L1 regularization optimization criterion
damage identification
model updating
