KL-Bayes方法在故障模式识别中的应用

Application of KL-Bayes Method to the Fault Pattern Diagnosis

有效特征向量的提取和状态识别是设备状态监测与故障诊断领域中的关键技术。近年来,国内外很多学者都非常重视自动特征向量选择与提取方法的研究和模式识别方法的探讨。文中提出的KL-Bayes 方法是KL变换特征提取方法与Bayes 逐步判别分析方法的结合,前者可在不改变原始样本空间分布特点的基础上降低特征空间的维数[4],后者是一种集“有效特征选择与状态识别”功能于一身的方法[1]。KL-Bayes方法用于不太复杂的系统故障诊断,如轴承、齿轮箱故障诊断中是非常简单有效的。文中给出了应用实例及分类器自学习前后的分类结果。

The extraction of effective feature vectors and pattern recognition are the key technique in the subject of mechanical condition monitoring and diagnosis. In recent years, people pay much attention to the research of effectiveness of feature vectors and the discussion of methods for pattern recognition. The KL Bayes method presented in this paper is a combination of KL transform method and Bayes discrimination function. The former can lower the dimension of feature vectors without changing the distribution of sample data, the latter is a good combination method which can get the discrimination function while accomplishing the selection of feature vectors. The method of KL Bayes can be applied to the analysis of state signals obtained from bearings or gearboxes in simple machinery. The pattern recognition results of the classifier designed with KL Bayes method and also the results after self learning are given in this paper.