Research On An Intelligent System Of Vehicle Fault Diagnosis Based On Case Similarity

With the continuous development of social economy and the continuous improvement of people's living standards,the vehicle has become an essential travel means of transportation for people.And with the progress of vehicle technology,more and more functional modules are being integrated into the vehicle,making the structure of the vehicle increasingly complex.Accompanied by the larger probability of vehicle fault,the complexity of vehicle fault diagnosis increases as well.The traditional vehicle fault diagnosis method is: the maintenance personnel determines the coverage and the location of the fault according to the user's description and the fault code,and then overhauls the vehicle.If there is no fault code,then the maintenance personnel has to detect the components one by one according to the fault phenomenon.This diagnosis method depends mainly on the technical level and work experience of the maintenance personnel,if there is a misjudgment,then the parts faults need to be shot one by one,making the maintenance efficiency greatly reduced.How to locate the vehicle fault accurately and complete the maintenance efficiently has become one of the priorities of vehicle research.This paper introduces the natural language processing into the vehicle fault diagnosis.By studying and analyzing the related technology of natural language processing,we design and implement an intelligent system of vehiccle fault diagnosis based on case similarity.The main research contents are:1.The exact word segmentation of fault information.Collect and sort out the terminologies in vehicle field and the vehicle parts names,then establish the appropriate word segmentation dictionary.Solve the problem of ambiguous word segmentation and unlisted word recognition in Chinese word segmentation in the field of vehicle fault information.2.Reduction of the text information dimension.By adopting Hidden Markov Model,mark the words with their parts of speech,then remove the auxiliary words,conjunctions,modal particles and other words which are meaningless to the original information after word segmentation.3.Weight calculation.Calculate the weight of words through combination of parts of speech information and word frequency information,then figure out their important degrees of the original semantics.4.Normalization of synonyms.Construct the synonym table in the field of vehicle fault information,and the synonyms in fault description are normalized to solve the problem of synonym noise in the similarity calculation of text,so the accuracy of text similarity calculation will be improved.5.Improve the Simhash algorithm and extract the case library candidate sets,so the complexity of the system's space-time will be greatly reduced.Adopt BM25 model,and then design model parameters and weight calculation formula to achieve the case matching function.6.Relying on NET development platform,design system UI.Design and implement feedback mechanism,device library quickly add entries,device annotation,data management and other auxiliary functions.Experiments have shown that the proposed algorithm is feasible,and the system that has good accuracy and real-time capability can meet actual requirements with good application effect.