Fault Diagnosis Of Three-way Catalytic Converter Based On LVQ Neural Network

With the improvement of people's living standards, the car ownership is rising year by year. The problem caused by vehicle exhaust has become more and more serious. Thus, Three-Way Catalytic(TWC) Converter has received the unprecedented attention, since it is the most effective means to purify the automobile exhaust. However, the existing diagnosis models for TWC are mostly only associated with one function, poor fault tolerance and strong specification, so this article tries to put forward a new thought and method in order to solve all the problems mentioned above.Vehicle exhaust causing serious pollution to the environment contains numerous of fault features. Using automobile exhaust reasonably can diagnose the fault of TWC. Nevertheless, because of the isometric effect, it is not easy to distinguish the fault features, in order to solve this problem, this paper tries to use two separately kinds of feature extraction methods to generate the new features which are easy to be distinguished. The main work of this paper are stated as follows:1. This paper analyzes the TWC research situation of fault diagnosis in both domestic and foreign countries. The TWC's fault diagnosis technology based on preand post-converter oxygen model is summarized in this article. This paper also introduces the method of vehicle exhaust which based on the working priciple of TWC as well as deactivation mechanism, they can be further used for diagnoise the false caused by TWC.2. Using frequency spectrogram and linear canonical transform- Hilbert Huang Transform method to extract the fault features from the data of automobile exhaust can solve the isometric effect appeared in the original data which makes the fault features difficult to distinguish. What is more, it can also transform the abstract fault features into the intuitive image features, and have effect impacts on the feature contrasting and subsequent diagnostic work.3. In order to solve the problem of dead zone neural in the process of diagnosing, the algorithm is improved based on the original LVQ neural network weights adjustment algorithm, the distinguish rate in new network is improved. Comparing with BP neural network and PNN neural network and analyzing the experimental results, we can conclude that the method studied in this article is feasible.