结构系统识别不确定性分析的Bayes方法及其进展

受测试误差、建模误差、数值离散化以及环境变异等因素的影响，结构系统识别过程不可避免地存在不确定性，因此有必要引入概率统计方法来提高其鲁棒性，为工程结构安全监测提供更为可靠的结果.近年来，Bayes(贝叶斯)方法因为其诸多优势在系统识别领域受到了广泛关注.该文梳理了Bayes系统识别的历史脉络和研究进展.从Bayes系统识别的理论框架出发，分析了量化系统识别不确定性两类方法的适用条件与局限性.此外，文章综述了Bayes方法在模态参数识别、有限元模型修正以及结构损伤识别方面进行不确定性分析的理论、实现及其应用.最后对基于Bayes方法进行系统识别研究的发展趋势做出了展望.

Uncertainty Quantification for System Identification Utilizing the Bayesian Theory and Its Recent Advances

System identification is inevitably affected by various uncertainties involving measurement error, modeling error, numerical error as well as environmental variation, which indicates that it is of fundamental importance to explore statistical methods to improve the robustness in identification. The Bayesian approach has attracted widespread attention in the field of system identification due to a number of advantages. On the basis of the classic Bayesian theory, this paper systematically outlined the progress of the Bayesian system identification in the context of structural dynamics. In this study, the theoretical framework for the Bayesian system identification with special emphasis on applicable conditions and the limits on the two kinds of uncertainty quantification approaches were presented. In addition, this paper reviewed some theory, implementation and practice of the Bayesian approaches applied to uncertainty quantification for modal analysis, model updating and damage detection. Finally, the trends and challenges of the Bayesian system identification were prospected.