基于DBN的多状态时变系统可靠性建模与评估

The failure of most stucture or parts is a process for many engineering structure which have remarkable multi-state gradual change character. Partial failure of part often leads to working performance degradation of it and the failure cumulative effect causes the system function deterioration to accomplish the required task. Reliability method based on two-state hypothesis neglects the failure cumulative effect so the reliablility model based on two-state is different with actual situation.To solve this problem, DBN(Dynamic Bayesian Network)is introduced into the generized structure system reliability research. DBN is a key tool in indetermined research field which can inference and graphic description of stochastic indetermined probolem. By using the advantages of dealing the uncertainty reasoning and graphical expression of stochastic process, the relationship between dynamic sequential system state and each element state can be built conveniently.Thus the logic relationship of various factors causing the failure of multi-state system can be described and accurate modeling of complex system can be realized. At the same time, it providing scientific basis for reliability assessment,fault diagnosis, finding weak link and improve system reliability.

对于多数工程结构系统，绝大部分结构或零件的失效都是一个过程，呈现明显的多态渐变特征。元件部分失效往往会导致元件工作性能的退化，由此产生的失效累积效应会导致系统完成规定工作任务的能力退化。基于2态假设的可靠性研究方法忽略了系统在渐变失效过程中呈现出的失效累积效应，由此建立的可靠性模型与实际情况存在较大差异。针对这一问题,本项目拟应用不确定性研究领域的核心工具 - 动态贝叶斯网络(Dynamic Bayesian Networks，DBN)对时变多状态系统可靠性展开研究。利用DBN的不确定性推理和图形化表达的优势，研究时变多状态系统状态（退化或失效）与系统中各单元状态之间的动态关系，从而系统地描述产生多状态系统失效的各种因素间的关系，实现复杂多状态系统可靠性的精确建模；同时，还可以进一步对系统可靠性进行故障诊断，查寻系统的薄弱环节，为提高系统可靠性提供依据。

本项目拟应用不确定性研究领域的核心工具 - 动态贝叶斯网络(Dynamic Bayesian Networks，DBN)对时变多状态系统可靠性展开研究。利用DBN的不确定性推理和图形化表达的优势，研究时变多状态系统状态（退化或失效）与系统中各单元状态之间的动态关系，从而系统地描述产生多状态系统失效的各种因素间的关系，实现复杂多状态系统可靠性的精确建模；同时，对系统可靠性进行故障诊断，查寻系统的薄弱环节，为提高系统可靠性提供依据。基于贝叶斯网络随机模拟推理算法结合可靠性评估中采用的时变模拟法，开发计算机辅助可靠性工程软件平台，用于复杂系统可靠性建模、可靠性设计与可靠性评估。制定详细的试验方案，设计加工了部分试验件，进行初步的实验室试验，并运用有限元与Monte-Carlo仿真方法进行系统可靠性仿真分析，实验结果正确有效，可以在复杂系统可靠性评估中应用。