机器故障诊断模糊分类规则的归纳学习

机器状态的有效监控是机器安全运行,提高系统生产效率的重要途径,由于环境的复杂性与不确定性,现有机器故障诊断系统的可靠性还不能满足应用要求,为此近来年发展了模糊诊断技术,对于模糊系统而言,其关键问题在于如何获得解决具体问题的模糊知识或规则,为此提出了一种被称为模糊归纳学习的学习方法,并利用模糊信息熵概念建立了从不确定信息数据中归纳模糊规则的学习准则,实验结果表明,利用所提出的模糊归纳学习方法所获得的模糊规则能不确定怀数据进行有效分类,从而能有效地提高机器故障诊断系统的与可靠性。

Inductive Learning of Fuzzy Classification Rules for Machine Fault Diagnosis

For ensuring the machines operating in a safe situation and improving the productivity, reliably monitoring the machine condition is becoming necessary and important in automatic ma-nufacturing systems. Due to high uncertainties inherent in industrial environment, the reliability of the developed fault diagnostic systems is still unsatisfactory. Extensive research has been performed to develop fuzzy diagnosis system for handling the problems related to the uncertainties. In the design of a fuzzy classification system, the most important task is to extract a set of fuzzy rules related to a particular problem. This paper presents a new inductive learning method, termed as fuzzy inductive learning, for extracting fuzzy classification rules from a set of numerical training data. In this paper, an alternative criterion is proposed for the induction of fuzzy decision trees by measuring the information gain in the possibility domain. The effectiveness of the proposed fuzzy inductive learning method has been illustrated in extracting a set of fuzzy rules for a machine diagnostic system that is worked in a highly uncertain situation. The experimental results show that the generated fuzzy rules can classify the machine conditions with high accuracy.