A Research About Fuzzy Neural H Network And Its Application In Automobile Fault Diagnosis

The automobile has taken a more and more important part in moderntransportation. With the development of automobile engineering, the function andstructure of the automobile have improved rapidly. However, the development ofautomobile engineering also brings more and more complicated symptoms ofautomobile failure, which create more and more difficulties in the diagnosis andmaintenance of the automobile. For a quickly determining of the reason and the positionof automobile failure as well as an objective evaluation about the automobile and itstechnical condition by lowest price, it is urgent to give a research to study the advancedautomobile failure diagnosis technology, tool and theories.After an analysis of domestic and foreign literatures, this thesis comes up with anautomobile failure diagnosis method called fuzzy neural H net. It divides the FHNmodel into several layers according to its structual characteristics. This method aims atworking out all the identification numbers of the output H nodes on one layer throughthe operation of the fuzzy H nodes. Every fuzzy H node corresponds to a specific rule inFHN. The input of H nodes is the prerequisite of the rule while the output of it is theresult. The mapping function corresponds to its reliability as well as the successfulmatch of the rule corresponds to the ignition of the fuzzy nodes in FHN. Through themethod of error back propagation algorithm and parameter adjustment of each node, thevalue of each parameter is approching to its true value if BP is applied to FHN and aninitial value is given to the parameter. It can achieve a better diagnosis effect and proveits validity and accuracy by applying the fuzzy neural H net to the process of theautomobile failure diagnosis with the foundation of the inference model and simulationof the data sample.This thesis gives a general analysis of the condition of the automobile failure aswell as a focus on the phenomena and reasons of the engine failure with theeatablishment of the automobile engine failure diagnosing knowledge base model. The process of the automobile engine failure based on fuzzy neural H net is divided into twoparts: at first, the fuzzy neural H net model for automobile failure diagnosis should beconstructed; then the automobile failure is diagnosed through the reasoning algorithmand the fuzzy neural H net model which has been establised already. During thereasoning process of fuzzy neural H net, the neural net gives a preprocessing to theparameters provided by expertise. It also uses MATLAB to study the sample and givesan expected outcome. According to the reasoning algorithm and the reasoning processin the failure diagnosis, alone with the failure phenomenon in engine starting, theauthenticity and validity of the method used in this thesis are verified.