因果链解耦的时间—概率模型  

Time-probability model for causal chains decoupling

作　　者：佘维[1,2] 宋伟[1,2] 叶阳东[1] 

机构地区：[1]郑州大学信息工程学院,河南郑州450052 [2]北京交通大学轨道交通控制与安全国家重点实验室,北京100044

出　　处：《计算机集成制造系统》2013年第10期2536-2549,共14页Computer Integrated Manufacturing Systems

基　　金：国家863计划资助项目(2011AA110501);国家重点实验室开放课题基金资助项目(RCS2009K003);河南省重点科技攻关计划资助项目(122102210004);湖南省科技攻关计划资助项目(2012FJ3092)~~

摘　　要：为了对离散事件系统中的复合故障进行快速准确的诊断,提出一种时间—概率Petri网(TPPN)模型及复合故障因果链的解耦算法。该方法首先根据故障的观测事件集和关系集建立扩展时间Petri网(ETPN)以描述事件间的时序和因果关系;随后将ETPN求逆并转换为TPPN,使其能够进一步描述事件发生的概率信息,并赋予TPPN初始状态使其运行;最后对TPPN终止状态下的各类令牌进行分析及故障源诊断,并析出各单纯故障因果链。仿真实验和对比分析表明,TPPN在观测事件集不完备和存在干扰的情况下,仍能准确地辨识故障源,解耦因果链,比同类方法的诊断精度更高。TPPN作为一种可有效描述离散事件之间的时序特征、概率信息和因果关系的模型,还可进一步应用于离散事件系统的行为预测和过程分析。To analyze the composite fault in Discrete Event System （DES）, a Time-Probability Petri Net （TPPN） and a decoupling algorithm were proposed. According to the observation event set and faults relationship set, an Extend- ed Time Petri Net （ETPN） was constructed to describe the time sequence of cause-and-effect relationship, and ET- PN was inversed and converted into TPPN, so that it could further represent the probability of event occurrence. An initial state was allocated to TPPN for its operation. The tokens were analyzed to find the fault origins when the TP- PN was terminated, and the causal chains were extracted from ETPN. The simulation experiment showed that TP- PN could accurately identify fault origins and decouple causal chains, even if the observation event set was incom- plete with disturbance. Compared with other similar methods, TPPN had higher diagnosis precision. As a kind of DES model which could describe the timing sequence, probability and causal relationship effectively, TPPN could be further applied to behavior prediction and process analysis.

关 键 词：复合故障 因果链 解耦 故障诊断 时间—概率Petri网 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程] TP391.4[自动化与计算机技术—控制科学与工程]