Research And Implementation Of Knowledge Acquisition Methods For Fault Diagnosis Expert System

The intelligent fault diagnosis expert system has been applied in every field, and becomes a new direction of the technology of fault diagnosis. Expert system is a complicated subject. In the development of expert system, there are three problems must be solved. They are the methods of knowledge expression, reasoning machine and knowledge acquisition. Knowledge acquisition is a bottleneck of expert system. It is a topic of the day that how to acquire knowledge of high quality and achieve automatic knowledge acquisition.Knowledge acquisition and rule maintenance is studied in this paper. The research content and accomplishment is as followed:(1) Semiautomatic knowledge acquisition based on fault treeAfter the research on the expression, storage of the nodes information and design of the structure of fault tree, a semiautomatic knowledge acquisition model based on fault tree is finished. The fault tree could be transformed into rules in this model. In this way, the knowledge is derived. The rules contain complicated relations between knowledge and they are available for reasoning. At last, the model is verified by plane fault knowledge.(2) Automatic knowledge acquisition based on rough setFirst, a series of key problems are solved such as discretization of continuous attribute, reducing redundancy, extracting rules and so on. Then an automatic knowledge acquisition model based on rough set is finished. The knowledge could be extracted from many samples by this model. Besides, a new incremental learning algorithm is also studied. The efficiency of acquisition could be improved greatly by it. Finally, it is proved to be validated by the iris data. With this model, the class of the sample can be achieved.(3) Automatic maintenance of rule setThe technology of automatic maintenance of rule set is studied. Based on the theories of word package and rule containing, a model of automatic maintenance of rule set is achieved. The redundant and circular rules can be checked out and disposed. In this way, the consistency of the rule set could be insured. In the end, it is verified by plane fault rules.(4) Software platform of knowledge acquisitionUnder VC++6.0 and DB2 developing environment, integrating the above technologies, a software platform of knowledge acquisition is developed in this paper. Knowledge can be acquired through two ways of fault tree and rough set. It can also be used to maintain rule set. This paper makes positive contribution to deep research in the project. Besides, it could be applied to other fields.A software platform is developed through the research on knowledge acquisition and rule maintenance. The algorithms are verified according to it. Finally, three aims are accomplished that knowledge acquisition based on fault tree and rough set, automatic maintenance of rule set.