Research Based On The Optimization Model Of Support Vector Machine For Car-engine Misfire Fault Diagnosis

Along with the rapid development of modern science and technology and the recovery of overall situation of economy, the ownership of car is rising The rapid development of automobile industry at the same time, effectively promote the process of precision and intelligent of the cars. On one hand, it makes the operation more convenient, which increases the number of consumer groups and leads to the popularity of the car. On the other hand, it causes some problems such as environment pollution, traffic congestion, energy shortage, noise pollution and other issues so on, especially the "fog" which is well-known now.Cars which have a lot of components and complex mechanism are prone to in failure when running. Because its working condition is not stable. Engine failure causes and failure phenomenon often present corresponding uncertainty The same form of the cause may lead to different failure, and the same failure phenomenon is not induced by the accurate cause of the problem, So it increases the difficulty of fault judgment and often causes unnecessary waste of maintenance, especially for the development of the intelligent car which increases incidence of failure and maintenance difficulty. The restriction of objective conditions make the demand of car engine diagnosis more urgent.Based on all the above problems, the fault diagnosis scheme for car engine misfire which is based on exhaust gas component analysis is proposed. Starting from the definition of misfire fault, this paper introduces the cause of the phenomenon of engine misfire, the types and hazard of engine misfire. Then, three kinds of the methods of misfire diagnosis which are in extensive research are introduced, Especially for its working mechanism the range of application. This paper focus on the method of misfire diagnosis from the Angle of exhaust gas composition. And the principle and influencing factors of the method are also introduced.The intelligence theory are used relatively commonly for engine misfire fault diagnosis such as expert system, neural network and support vector machine (SVM), etc. The three kinds of intelligent diagnosis methods can establish the corresponding relationship between fault characteristics and fault category and finally realize the pattern classification for misfire fault on the basis of fault samples learning. But the process of the training of the former two kinds of fault diagnosis method need too more samples. However, the misfire fault samples is not easy to get in practice. The focus of this article is the designing of the fault diagnosis classifier for small sample. Under the background of machine learning, this paper briefly introduces the several basic concepts in statistics. Then, it expounds basic theory of the support vector machine (SVM) such as the optimal hyperplane, the form of kernel function, classification model, the key parameters that determines the classifier performance and several typical optimization method for the key parameters, etc. Under this premise, the model for engine misfire diagnosis based on support vector machine (SVM) is established. Then the classifier which is based on the optimization of support vector machine based on the improved ABC algorithm is built for the samples in small scale. The classifier can be used for engine misfire fault diagnosis, through nonlinear corresponding between the five characteristics of automobile exhaust gases and the fault phenomena, with a higher accuracyThe progress of the research in theory is the improved ABC algorithm based on the combination of the adjustment in weight and the simulated annealing algorithm through the research which focus on the process of realization of artificial colony algorithm. The convergence of the improved ABC algorithm which is applied for the optimization of key parameters of the SVM classifier is good. And the simulation shows its advantage.