基于冲突的离散事件系统诊断方法  

Conflict-based diagnosis of discrete event system

作　　者：王晓宇[1,2] 欧阳丹彤[1,2] 赵剑[1,2] 耿雪娜[1,2] 

机构地区：[1]吉林大学计算机科学与技术学院,长春130012 [2]吉林大学教育部符号计算与知识工程重点实验室,长春130012

出　　处：《吉林大学学报（工学版）》2013年第2期380-385,共6页Journal of Jilin University:Engineering and Technology Edition

基　　金：国家自然科学基金项目(61133011,60973089,60873148,61003101,61170092,60973088);浙江师范大学计算机软件与理论省级重中之重学科开放基金项目(ZSDZZZZXK12);浙江省自然科学基金项目(Y1100191)

摘　　要：提出了一种基于冲突的动态模型故障诊断方法。通过在离散事件系统的自动机模型中添加概率,离线计算模型中事件与状态的后验概率,从而处理不完全观测导致的不确定性,增强了离散事件系统处理不确定性问题的能力。在已经建立的离线模型上,在线计算观测与模型的冲突,在模型中提取符合观测的轨迹,推导系统运行状况,判断故障,并给出故障路径。将基于冲突的诊断方法扩展到离散事件系统的诊断上来,避免了对无关事件的搜索及判断,从而降低了诊断搜索空间,便于计算更大规模的系统模型。A dynamic model-based diagnosis method based on conflict computation is proposed. Discrete event system uses automaton to off-line add the probabilities to events and states. The method deals with the uncertainty lead by the incomplete observation. The ability of the discrete system in solving uncertain problem is enhanced. On the model built off-line, the conflict between model and observation is computed on-line, and the trajectory compatible with the observation is extracted. Based on the trajectory, the fault is estimated, the faulty trajectory is given and the state of the system is deduced. The conflict-based diagnosis is extended to the diagnosis in discrete event system. The search and judgment of unrelated events are avoided, thus, the research space is reduced. Therefore, the proposed diagnosis method is suitable for large scale systems.

关 键 词：人工智能 基于冲突诊断 离散事件系统 动态诊断 

分 类 号：TP312[自动化与计算机技术—计算机软件与理论]