神经网络和模糊系统在故障诊断中的应用_宋彤.pdfVIP

23 2 计　算　技　术　与　自　动　化 V ol.23, No.2 2 0 0 4 6 Computing Technology and Automation 　Jun.2 0 0 4 :1003-6199(2004)02-0031-04 宋　彤, 孙增国,冯　冲 ( , 　116023) 　　　:本文提出了一种神经网络和模糊系统相结合的分级式故 诊断方法。神经网络通 过对部分测量数据的处理, 实现系统的回路级故 诊断,输出各回路故 出现的可信度。模糊 系统通过对神经网络得到的初步诊断结果和其他测量值的处理, 实现系统的元件级故 诊断, 并对最终诊断结果作出解释。该方法融合了神经网络自适应学习能力强和模糊系统知识表达 明确的优点,简化了神经网络学习数据获取及模糊推理规则建立的过程。通过对热硝酸冷却系 统故 诊断的仿真,证明了该故 诊断方法的有效性。 :故 诊断;神经网络;模糊系统 :TP277 　　　　　　　　:A Neural Network and Fuzzy System Applications to Fault Diagnosis SONG Tong ,SUN Zeng-guo,FENG Chong (School of Elect onic and Info mation Enginee ing, Dalian Unive sity of Technolgy, Dalian, 116023, China) 　　Abstract:In this pape , an a tificial neu al netwo k (ANN )is integ ated with fuzzy system fo fault diagnosis. ANN detects the loop faults sou ces th ough the pa tial data measu ed, and outputs the fault deg ees of co esponding loops.Fuzzy system detects the element faults th ough the p elimina y diagnosis esults obtained by A NN connected w ith othe co elative values measu ed, and inte p ets the final esults.The fault diagnosis system (FDS)combines the adaptive lea ning diagnosis p ocedu e of the ANN and the t anspa ent knowledge ep esentation of the fuzzy system, and simplifies the p ocess of obtaining the ANNs lea ning sample and establishment the fuzzy infe ence ules.Th ough the fault diagnosis simulation of a hot nit ic acid cooling system, it has been p oven that the fault diagnosis method is ve y valid. Key words:fault diagnosis;a tificial neu al netwo k;fuzzy system 。 1　 , , 、 ,、 , 。 。 ,