基于EMD样本熵与改进DS证据理论的故障诊断方法

Fault Diagnosis Method Based on EMD Sample Entropy and Improved DS Evidence Theory

针对单一传感器故障诊断信息源简单、信息不完整的局限性,提出一种基于经验模态分解样本熵与改进DS证据理论的多传感器信息融合故障诊断方法。该方法对每路传感器采集的振动信号进行经验模态分解,并计算其固有模态函数IMF的样本熵作为故障特征变量;将故障特征变量输入事先训练好的各个随机森林分类器进行分类;以每个随机森林的分类结果作为证据体,采用改进的DS证据理论进行融合并输出最终分类结果,实现多个传感器信息有效融合,各传感器间形成信息互补,达到优化决策目的。实验结果表明:该方法故障诊断准确率达98.85%,且具有鲁棒性。

Aiming at the limitation of single information source and incomplete information for single sensor fault diagnosis,this paper proposed a fault diagnosis method ofmulti-sensor information fusion based on empirical mode decomposition(EMD)sample entropy and improved DS evidence theory.In this method,EMD is performed on the vibration signal collected by each sensor,and the sample entropy of its intrinsic mode function(IMF)is calculated as the fault characteristic variable.The fault characteristic variables are input into the previously trained random forest classifiers,and the classification results of each random forest are used as evidence bodies.Then the improved DS evidence theory is used to fuse and output the final classification results.The proposed method realizes the effective information fusion ofmultiple sensors,and the information between the various sensors is complementary,which achieves the purpose of optimization decision.Results of the experiment show that the accuracy of the method is 98.85%,and it is robust.