基于知识工程的图形化网络拓扑诊断系统

Development of Intelligent Fault Diagnosis System Based on KBE Graph

针对目前智能诊断系统缺乏有效的知识库开发工具、推理功能单一等问题,开发了一套基于知识工程(KBE)的图形化网络拓扑诊断系统。KBE图形化网络拓扑专家知识表达使专家知识的建立、修改、扩充和维护于一体,不需编写专门代码,节省了时间,减少了人为出错的可能性。集专家规则、模糊逻辑和神经网络于一体的组合智能推理机使系统适应于多变量、多参数、多目标及多过程的复杂系统。在Bently模拟转子试验台上对系统的有效性进行了验证,取得较好的效果。该系统已获国家发明专利。

This paper aimed at solving the insufficiency problem in the current intelligent fault diagnosis system,such as lack of development tools,inference oversimplification,and developed a kind of diagnosis system based on network and KBE graph.The system is simple and easy to establish, modify,expand and maintain and does not need further coding.In addition,it can significantly save time and reduce the chance of failure.The integration of expert rules,fuzzy logic and nerve network enables the system to adapt to those complicated systems which involve in multi-variable,multi-pa- rameter,multi-object and multi-process.As an example,an on-line detection system for simula- tion rotor test bench had been established and the typical faults were successfully diagnosed through it,which proved the validity and reliability of this system.Furthermore,the system had been de- clared national invention patent.