摘要
针对存在抗力退化结构的时变可靠性问题,提出一种动态贝叶斯网络(dynamic Bayesian network,DBN)模型,以gamma过程作为抗力退化模型,并离散为Bayes网络,同时建立观测模型、可靠性模型,组合为动态Bayes网络,通过连续节点消除与离散得到仅含离散变量的动态Bayes网络;给出精确推理的3种情况,评估现在(滤波)、未来(预测)以及过去时刻(平滑)结构的状态.当测量信息出现时,对退化模型参数重新估计,利用精确推理来更新结构时变可靠性.以存在抗力退化的一跨刚架作为研究对象,验证了模型的合理性.
A dynamic Bayesian network （DBN） model was proposed for timedependent relia bility analysis of structures in deterioration. The structural resistance deterioration was modeled as a gamma process while the loads as random variables. The stochastic deterioration process was discretized in time domain as deterioration models. A DBN was established and comprised of the reliability model, deterioration model and observation model. Node elimination algorithm and discretization were applied to modify the DBN into a network with only discrete variables. Exact inferences with the DBN were presented to estimate the 3 structural states at present （ ill tering）, in the future （prediction） and in the past （smoothing）, respectively. The structural timedependent reliability was updated with the reestimated deterioration model when meas urements were available. The proposed model was validated through the timedependent relia bility analysis of a onebay example frame in resistance deterioration.
出处
《应用数学和力学》
CSCD
北大核心
2014年第1期102-110,共9页
Applied Mathematics and Mechanics
基金
国家自然科学基金(51278420)~~
关键词
时变可靠性
动态Bayes网络
抗力退化
精确推理
更新
time-dependent reliability
dynamic Bayesian network
resistance deterioration
exact inference
updating
