基于自然语言理解的故障诊断方法研究

Research on Fault Diagnosis Method Based on Nature Language Understanding

为解决知识规则推理故障诊断系统对用户输入限制的问题,提出了一种基于自然语言理解的二次回溯语义分词故障诊断方法,采用词频作为分词主要依据对用户自然语言输入进行分词切分,同时按邻近关系组词进行二次回溯处理,计算出故障征兆与规则的语义相似度,从而获得与故障字典的最佳匹配结果,使诊断系统不仅具备对用户自然语言输入处理的能力,而且具有自学习和诊断多故障现象的功能,将该方法应用于某大型设备故障诊断专家系统,取得了较好的诊断效果。

In order to solve the problems of the knowledge--rule--based reasoning diagnosis system restrictions on user input, the paper proposed a fault diagnosis method of two times backtracking semantic word segmentation based on nature language understanding. The method uses word frequency as the main basis for segmentation of the natural language word input by user, and deals with the word segmentation by the second backtracking on close relationship phrases; By calculation of the semantic similarity between fault symptoms and rules, the re- sults matching with fault dictionary are obtained. The system has not only the ability to handle natural language input by user, but also the function of self--learning and multiple faults diagnosis. Finally, the method is applied to a large fault diagnosis expert system, and achieves good diagnosis results.