基于神经网络的故障LRU的检测与定位-控制理论与控制工程专业论文.docxVIP

Classified Index: TP206 U.D.C: 621 Dissertation for the Master Degree of Engineering DETECTION AND ISOLATION OF FAULTY LRU BASED ON NEURAL NETWORK Candidate: Xiaodong Du Supervisor: Co- Supervisor: Prof. Michael Y.Wang Baoan Li, Howard Ip Academic Degree Applied for: Master of Engineering Specialty: Control Theory and Engineering Affiliation: Shenzhen Graduate School Date of Defence: December, 2006 Degree-Conferring-Institution: Harbin Institute of Technology To my family and friends! ABSTRACT Machine diagnosis helps machines to work more efficiently in a wide variety of industrial fields. The efficiency of maintenance work is enhanced by detecting machine abnormalities, analyzing their cause, and predicting machine deterioration. Therefore, it is meaningful to be able to diagnose machine faults in the most effective manner. Electronic packaging machine is important equipment widely used in IC manufacturing, while the motion control system is the core part of the machine. We may take it as our study object. Nowadays, machine fault diagnosis technology has been a new interdiscipline covering mathematics, physics, mechanics, electronics, chemistry, computer science, signal processing, artificial intelligence etc. Some theories and methods in common use are: spectrum analysis, time sequence analysis, fuzzy mathematics analysis, artificial neural network analysis, wavelet analysis and so on. In accordance with our object and our goal, we proposed the method based on neural network. In this project, Focus is given to detect and isolate the faulty Line Replaceable Unit (LRU). To begin with, a database is established for the storage of machine operation data which include both numerical and categorical data. But the measurement data are not suitable for immediate usage, therefore, we applied some method to extract some features to represent the symptom of faults. Then a model is developed based on the features extracted from measurement signal/data and the types of fault.