Research On The Fault Diagnosis Expert System Of Engines Based On Artificial Intelligence And Virtual Instrument Technology

Automotive engine is a very complex system, and there is more than 40% vehicle faults resulted from it. For avoiding the casualty, the work performance and conditions of engine need to be known in time. So it is very important to research the engine fault diagnosis technology theory and system program. However, the universal platform of an engine fault diagnosis system , which is used for the collection of many parameters acquisition, signal processing and diagnosis reasoning methods, is not very perfect and can not be suited to the demand of modern diagnosis system. At the same time, the fault diagnosis reasoning methods in existence can't be applied into fault system without any changes, it needs to be improved.So, the technologies of signal acquisition, signal processing, neural network, rough sets theory and virtual instrument were introduced in this paper. Also the principle, peculiarities and application in fault diagnosis of these technologies were discussed in detail. Then a new engine fault diagnosis expert system, which is based on a virtual instrument universal-platform and with an integration of many artificial intelligent technologies, was proposed too.The main contents of this dissertation is:1 .On the basis of analyzing the basic principle, reasoning methods and function peculiarities of some popular fault diagnosis technologies, and their applications in fault diagnosis field, the design view is given. We take the artificial neural network reasoning as a main method of system diagnosis reasoning method, the expert system method as the basis of fault system, primarily with additional methods. Then three typical neural models, which are BP neural network, RBF neural network and SOM neural network, their concepts, model structures and learning algorithms are analyzed. Last, by comparing these different BP improved algorithms, the guidance view of how to choose these algorithms in reason is given.2. The basic conception of Rough sets, the discrete methods of continuous attribute value and attribute parameter reduction method were analyzed in detail. And a new algorithm for obtaining rules is suggested. Then a new engine fault diagnosis model based on Rough Sets theory and neural network technology is established in this paper. The model can reduce the sample size, optimize the neural network, decrease the computation, and increase the diagnosis correctness. It has been testified satisfying by the experimentation of EQ6102 engine.3. The artificial neural network may not train easily and correctly while the number of its sample is very large. For solving this problem, The basic contents, neural network structure, classified method and data-fusion methods of an integrated neural network technology were analyzed clearly. Then with the help of some pretreatment methods, two models were established in this paper. One model is based on the principal component analysis and Neural Network integration for analyzing the relation between engine exhaust emission and fault. Another model is based on clustering analysis method, pattern recognizing and BP neural network for analyzing the relationship between technical states and performance parameters of diesel engines.4. The principle of signal acquisition and signal processing methods, and its applications in engine fault symptom inspection are studied carefully. Then inspection system of the engine working temperature, oil pressure and other steadily signal is designed. Also an engine fault symptom signal acquisition instrument on the basis of the power spectral analysis is given too. For improving the precision of measurement data, the data-fusion based on parameter estimation is applied into the measuring exhaust emission contents.5. The concept, structure and peculiarities of the virtual instrument, and the LabVIEW programming technology are discussed deeply. Also, the main parameters of engine working performance are analyzed too. Then we designed a new universal platform of engine fault diagnosis measurement, which is based on a virtual instrument and an...