基于模糊区间优化的模糊推理故障诊断方法  

Fault Diagnosis Method Based on Fuzzy Interval Optimization Using Fuzzy Inference

作　　者：董炜[1] 陈卫征[1,2] 徐晓滨[2] 吉吟东[1] 

机构地区：[1]清华大学清华信息科学与技术国家实验室,北京100084 [2]清华大学自动化系,北京100084

出　　处：《北京工业大学学报》2012年第12期1905-1912,共8页Journal of Beijing University of Technology

基　　金：国家自然科学基金资助项目(61104019);国家科技支撑计划项目(2009BAG12A08)

摘　　要：提出一种基于模糊区间优化三角均分的模糊规则生成方法.在获取故障样本之后,通过基于故障识别率最佳准则的寻优过程,确定模糊区间均匀划分数的下确界.在此基础上,利用所提出的改进型DM算法提取模糊规则,最终构建出模糊推理故障诊断系统.与模糊区间划分数低于该下确界的模糊推理系统相比,利用该方法能保证在所生成模糊规则数较少的情况下,使构建的推理系统具有更高的故障识别率和诊断准确率,同时避免了人为确定模糊区间划分数时所带来的盲目性.最后,在轨道电路该类复杂模拟电路常见硬故障的诊断实例中,说明了新方法的有效性.This paper presents a fuzzy rule extraction method based on optimal triangular equipartition. An optimization process based on the optimal criterion of fault identification rate is implemented to adaptively obtain the infimum of the number of fuzzy intervals, into which the input and output spaces of the given fault sample data should be divided. Subsequently, an improved DM algorithm is proposed to generate fuzzy rules from the given sample data based on the fuzzy intervals that compose the fuzzy inference system. Compared with factitious or random partition that the number of fuzzy intervals is lower than that of the infimum, the proposed method can generate more reasonable fuzzy inference system with higher fault identification rate and diagnostic accuracy. Moreover, the calculation burden in the process of fuzzy inference is significantly reduced because of smaller number of generated rules. As a result, the blindness in factitious or random partition is avoided. Application to fault diagnosis of track circuit shows the effectiveness of the new method.

关 键 词：故障诊断 模糊区间 模糊推理 模糊专家系统 轨道电路 

分 类 号：U461[机械工程—车辆工程] TP308[交通运输工程—载运工具运用工程]