基于改进的RBFNN在线故障诊断专家系统设计

Design of On-Line Fault Diagnosis Expert System Based on Improved RBF Neural Network

针对在线故障诊断需求,通过分析神经网络和专家系统的各自特点和不足。为解决上述问题,构建了神经网络的在线故障诊断专家系统。通过对比分析选择径向基函数(RBF)神经网络,采用减聚类和k-means聚类混合算法求取RBF参数;建立了显示与隐式知识库,采用带可信度的产生式规则表示知识,运用ADO方法管理知识库和应用Visual C++、Matlab混合编程的方法,实现了诊断系统组建;最后在某随动装置在线故障诊断中仿真验证,构建系统很好地解决了在线快速故障诊断的问题。

According to requirements of on-line fault diagnosis, a on-line fault diagnosis expert system based on artificial neural network was designed by analyzing the characteristic and shortage of neural network and expert sys- tem. The paper chose RBF neural network through the comparative analysis of several neural networks, used a hybrid learning algorithm of subtractive clustering and k-means clustering to optimize the parameters of RBFNN, accessed to build explicit and implicit knowledge base, adopted production rules with credibility factor to represent knowledge, and utilized ADO method to manage knowledge base. And then the hybrid programming method of Visual C ＋ ＋ and Matlab were used to build fault diagnosis system. Finally, the system preferably resolved on-line fault diagnosis prob- lem of a servo device through fault diagnosis experiment.