基于改进决策树的故障诊断方法研究

Research on Fault Diagnosis Method based on Improved Decision Tree

为解决设备故障无法定位和无法及时预测的问题,提出一种基于决策树的故障诊断方法。该方法采用基尼系数进行分类树的无偏节点分裂,按照最小代价复杂度剪枝法对生成的决策树进行剪枝,并采用袋装技术建立分类回归树的组合预测模型。最后对空调智能远程控制器故障数据进行分类研究,结果证明了该方法的有效性与可行性。

In order to solve the problems that equipment failure can’t be located and the fault of equipment can’t be predicted in time,a fault diagnosis method based on decision tree is proposed.The Gini coefficient is used to classify the unbiased node of the classification tree.The decision tree is pruned according to the minimum cost complexity pruning method,and the combination forecasting model of classification and regression tree is established by bagging technology.At last,the fault data of air conditioning intelligent remote controller are classified and studied.The results prove the effectiveness and feasibility of the method.