基于平均多粒度决策粗糙集和NNBC的滚动轴承故障诊断

采用求同排异思想的悲观多粒度粗糙集是一种规避风险的决策策略,其限制条件过于苛刻,导致约简后的征兆属性集维数过低,难于对滚动轴承的状态做出准确判断。为此,提出一种基于平均多粒度决策粗糙集和非朴素贝叶斯分类器(Non-Naive Bayesian Classifier, NNBC)的滚动轴承故障诊断方法。该方法提取训练样本中滚动轴承的故障特征,用于构建平均多粒度决策粗糙集;采用基于平均多粒度决策粗糙集的属性约简算法,降低训练样本中征兆属性集的维数;根据约简后的训练样本构建NNBC,用于判断待诊样本中滚动轴承状态。实验结果表明该方法能够准确地判断滚动轴承的故障类型及故障程度。

Rolling bearing fault diagnosis based on mean multi-granularity decision rough set and NNBC

Pessimistic multi-granularity rough set using thought of seeking common ground and excluding differences is a risk-aversion decision strategy. Due to its harsh constraint conditions, the dimensionality of a reduced symptom attribute set is too low and difficult to correctly diagnose rolling bearings’ state. Here, a fault diagnosis method of rolling bearings based on the mean multi-granularity decision rough set and the non-naive Bayesian classifier(NNBC) was proposed. Firstly, fault features of rolling bearings in training samples were extracted to construct a mean multi-granularity decision rough set. Then, an attribute reduction algorithm based on the mean multi-granularity decision rough set was applied to reduce the dimensionality of symptom attribute set in training samples. Finally, NNBC was constructed according to reduced training samples to classify rolling bearing states in samples to be diagnosed. Test results showed that the proposed method can be used to correctly diagnose fault type and fault level of rolling bearings.