中文摘要：

针对复杂系统发生故障时告警信号间的时序约束关系,提出一种时间贝叶斯Petri网模型（TBPN）,并基于该模型提出一种复杂系统的溯因故障诊断方法。该方法首先对观测到的告警信息建立时间Petri网,随后将其求逆并转换为TBPN。通过时间区间计算和溯因推理,分析告警信号的时序一致性并验证故障假说,最后对时序正确的故障事件链计算其故障概率并进一步分析干扰信息。仿真实验表明,该方法可对复杂故障进行快速诊断,并在告警信息存在丢失、虚警、时标差错时表现出较好的鲁棒性。与不考虑时间约束的同类方法相比,该方法具有搜索空间更小,抗干扰性更强的特点。

英文摘要：

For the temporal constraints among the alarms generated by complex system fault, this paper propose a kind of time Bayesian Pet net （TBPN） model, and give an abduetive fault diagnosis method for complex system based on this model. Firstly, a time Petri net （TPN） is established based on alarms which are observed. And then, inverse operation is done to the TPN, and the inverse TPN is converted to a TBPN. Through the time interval calculation and abductive reasoning, the temporal consistency among these alarms is analysed, and the calculation results are used to validate fault hypothesis. Finally, we get the fault event chains which have correct time sequences, calculate the probabilities of them, and further analyze interference information in these alarms. Simulation results show that the method can be used to make quick analysing on complex faults, and has good robustness when alarm is lost, false-alarm, and time mark mistake. Compared with other methods which not consider the temporal constraint, this method has the characteristics of smaller search space and stronger anti-jamming.