故障诊断系统中的冗余特征处理

特征参数是故障诊断的基础和关键,然而特征参数中总是存在很多冗余特征,影响故障诊断的准确率;根据设备的结构特点提出了基于故障树分解的冗余特征处理方法;首先,采用基于分类的粗糙集算法对不同部件的冗余特征进行约简,根据故障树结构分层处理;进而采用Apriori算法挖掘隶属于同一父节点的部件的频繁特征,降低不同部件特征参数的相关性;仿真实验证明,对原始特征进行冗余处理后,故障诊断系统的性能有较大提高。

Redundant Feature Disposal in Fault Diagnosis System

A Characteristic parametersetis fundamental and key to design of fault diagnosis.However there are always a lot of redundant featuresin characteristic parameters,affecting the performance of fault diagnosis.According to the structural characteristics of equipment,a redundant features approach is put forward based on fault tree decomposition.Firstly,redundant features of different parts are reduced by rough set algorithms based on the classification,in accordance with hierarchical structure of the fault tree.Secondly,frequent features of the parts under the same parent node are found out using Apriori algorithm to lower the relativity of characteristic parameters of different parts. Simulation experiments results show that performance of fault diagnosis system proves a lot with redundant feature disposal.