基于云模型和模型修正的不确定性损伤识别
Uncertainty-Based Damage Identification Using Cloud Model and Model Updating Techniques
郑金铃 1骆勇鹏 2齐林 3陈鑫 2廖飞宇2
作者信息
- 1. 福建农林大学交通与土木工程学院 福州,350002
- 2. 福建农林大学交通与土木工程学院 福州,350002;数字福建智能交通技术物联网实验室 福州,350002
- 3. 中铁上海设计院集团有限公司 上海,200070
- 折叠
摘要
结构损伤识别中存在的不确定性因素相互渗透、相互耦合,对识别结果有较大影响.针对此问题,提出基于随机模型修正和云模型漂移性度量(Kullback-Leibler divergence based on cloud model,简称KLDCM)的不确定性损伤识别方法.首先,采用云模型数字特征参数量化不同状态下实测数据的不确定性,并通过云发生器扩充实测数据;其次,基于改进随机模型修正方法计算与扩充后数据所对应的结构物理参数,根据云模型外包络曲线计算未知工况与健康工况下结构各单元物理参数的漂移度,并以归一化后各单元漂移度指标均值为阈值判别损伤单元位置;然后,在损伤位置判断的基础上,以损伤单元物理参数期望值来衡量损伤单元的损伤程度;最后,采用数值模拟和实际结构验证所提方法的可行性及可靠性,探讨了噪声水平、原始样本个数对损伤识别结果的影响.研究结果表明,与基于最大边界曲线法(maximum-boundary curve method,简称MCM)相比,所提方法受不确定性因素影响较小,具有更好的损伤识别精度.
Abstract
Uncertainty factors in damage identification infiltrate and interact with each other,significantly influ-encing the outcomes of structural damage identification.Therefore,a method for uncertainty-based damage iden-tification using stochastic model updating and the Kullback-Leibler divergence based cloud model(KLDCM)is proposed.First,the uncertainty of the measured data in different scenarios is quantified by using the numerical characteristic parameter of the cloud model.The measured data are extended using cloud generator.Second,the structural physical parameters corresponding to the expanded data are calculated based on the improved stochas-tic model updating process.Based on the outer envelope curve of the cloud model,the excursion degree of the physical parameters for each structure element under both unknown scenarios and healthy scenarios is calculated.The mean value of the normalized excursion degree index for each element is then used as the threshold to iden-tify the location of the damage element.Once the damage location is identified,the extent of damage is assessed by the expected values of the physical parameters of the damaged element.The feasibility and reliability of the proposed method are verified through numerical simulation and actual structure tests.Then,the effects of the noise level and number of original samples on the damage identification results are discussed.The results indi-cate that the proposed method is less susceptible to uncertain factors and achieves higher accuracy in damage identification compared to the method of maximum-boundary curve method(MCM).
关键词
损伤识别/响应面/模型修正/云模型/不确定性Key words
damage identification/response surface/model updating/cloud model/uncertainty引用本文复制引用
基金项目
国家自然科学基金资助项目(51808122)
福建省自然科学基金面上资助项目(2020J01580)
福建省结构工程与防灾重点实验室(华侨大学)开放研究课题资助项目(SEDPFJ-2018-01)
出版年
2024
NETL
NSTL
万方数据