一种改进的基于RS与SVM的故障诊断方法

A Modified Fault Diagnosis Method Based on RS and SVM

传统的基于粗糙集与支持向量机的故障诊断方法在用支持向量机分类前用粗糙集进行数据约简,仅将粗糙集作为数据约简的工具,忽视了粗糙集所获取的决策规则对原有数据中所隐含知识的概括表达作用。本文提出了一种改进的基于粗糙集与支持向量机的故障诊断方法,首先基于粗糙集对样本数据进行约简和初步决策规则获取,然后将获取的规则作为先验知识集成到支持向量机中进行故障诊断。该方法结合了粗糙集的处理高维数据的优点和支持向量机具有较高推广能力的优势,并且在用支持向量机分类时有效地利用了粗糙集获取的决策规则,提高了故障诊断的准确率。使用该方法对柴油机常见故障进行诊断实验,结果表明了方法的有效性。

Traditional fault diagnosis method based on RS and SVM reduces the diagnosis samples using Rough Set before using SVM for classification. Rough Set is used only as a tool for data reduction, while it is ignored that the decision rules obtained by Rough Set is the brief description of the knowledge embedding in original data. This paper proposes a modified fault diagnosis method based on Rough Set and Support Vector Machine. Firstly, the diagnosis samples are reduced by Rough Set and the decision rules are ob- tained. Then the obtained rules are integrated into Support Vector Machine for fault diagnosis. This method combines the advantage of Rough Set in processing high dimensional data with the advantage of Support Vector Machine in better generalization capacity. And the obtained decision rules by Rough Set are used to assist Support Vector Machine with classification for improving the prediction accuracy. The diagnosis of a diesel shows a good fault diagnosis ability of the method.