Research On Vane Compressor Fault Diagnosis Expert System

This paper airs vane-type compressor for diagnostic target, discusses exposition of expert system and neural network theory and the basic expert system structure. Expert system can simulate the logic process of human brain, the needing for complex reasoning to resolve the complexity of the problem. It can store and promote expert's valuable experience and knowledge, play a more effective role in a variety of specialized personnel, general maintenance personnel can be the expert's level in the fault diagnosis.Expert system continues to be added with knowledge, to revise the existing knowledge, and continuously updated. However, the poor capability to process uncertain information is a very important bottleneck for fault diagnosis expert system. This paper is based by production, in dealing with uncertain information, using the fuzzy method. Using knowledge of the database, making the learning expert system becomes very simple and convenient.Artificial neural network which simulates the human brain is a new and promising method of fault diagnosis. In the knowledge acquisition,knowledge of the neural network does not need finishing by the knowledge engineers, experts in the field to digest and summarize the knowledge, only need experts in the field or examples of problem-solving examples to train the neural network. Neural network systems knowledge acquisition and expert system compared, as well as with more time efficient, can ensure that the higher the quality. The paper use the three-layer feed-forward neural network and BP algorithm, vane compressor experimental platform tests the data samples for neural network training and diagnostic input. Compressor experimental platform tests various parts of the compressor performance data, through neural network calculation, identifying the compressor's fault. This paper also adopted the experimental method for initial weight, the hidden layer neurons, momentum coefficient and the error level on the impact of network training.In this paper, a reasonable selection of the traditional expert system based on neural network and expert system integration and reasonable use Visual C++6.0 development language for the system to Microsoft Access as a data source, the use of database technology ADO visit. System is designed aesthetic, practical, powerfully and realized the using of two different methods of the vane compressor's fault diagnosis purposes.