Study Of Rough Set-based Knowledge Acquisition Of Fault Diagnosis

Engineer mechanical is more and more important in daily life. It's normal to confront fault in daily life. Using fault diagnosis technology can find out where the fault is in time, avoid the unnecessary missing, which is of higher importance in economics and society. The dissertation introduce Rough Set(RS) in the research of fault diagnosis. It focused on the knowledge acquisition in fault diagnosis technology, carried out deeply and meticulous research on the object of engine.The dissertation introduced some fault diagnosis methods, analyzed the current research and future development trend of these method. The RS and its relative knowledge was summarized, and the application of RS in knowledge acquisition was also deeply researched.After studying current attribute reduction algorithm, the dissertation put forward improved attribute reduction algorithm based discernible matrix. It used attribute importance as the importance, utilized a heuristic algorithm drawing second-rate reduction. A practical example indicated that this algorithm was superior than traditional algorithm in calculating time and steps.On the basis of current data reduction discussing and analyze, the dissertation put forward an improved heuristic data reduction algorithm. It used decision matrix drawing data core, used attribute importance as heuristic forming heuristic data reduction algorithm. After giving conception of decision quality and rule reliable, it revealed that the superior in all aspects by a practical example.The system structure of engine was detailed introduced. Based on analyzing the structure of engine, the dissertation put forward the idea of acquiring engine fault diagnosis knowledge in divided floors. The idea utilized artificial acquisition and RST acquisition respectively acquires one-rate and two-rate rule knowledge. Then it presented the system structure of fault diagnosis and the process of knowledge acquisition. Finally, it used oil-supply system as study object, utilized improved algorithm acquiring two-rate fault diagnosis knowledge.