摘要
针对齿轮箱故障诊断时使用单一传感器进行信号获取过程中存在信息不完整的问题,导致故障特征信息及诊断推理方法具有随机性和模糊性。利用多传感器信息融合的二阶张量特征作为输入,构建了一个支持张量机和集成矩阵距离测度(Assembled Matrix Distance Metric,AMDM)的K最近邻分类器(k-nearest neighborhood classifier,KNN)决策融合故障诊断模型。首先,对多传感器信息时频域特征层进行融合,获得二阶张量的特征样本;其次,分别构建基于集成支持张量机、KNN-AMDM的故障诊断模型,并针对两类故障诊断模型的输入,设计了两种基本概率分配赋值的转化方法,通过不断调整参与的传感器数目获得6种不同的故障征兆张量集,进而得到12种不同的初步故障诊断结果;最后,采用D-S证据理论对12个证据体提供的基本概率分配值进行融合决策,得到最终的齿轮箱故障诊断结果。实验对比表明,该方法可提高齿轮故障诊断结果的可信度。
Aiming at the fault diagnosis of gearbox,the problem of incomplete information in the process of signal acquisition using a single sensor leads to the randomness and fuzziness of the fault feature information and diagnostic reasoning methods.Using the second-order tensor characteristics of the multi-sensor information fusion as input,a decision fusion method based on the support tensor machine and KNN-AMDM is proposed.First,the multi-sensor information time-frequency feature layer is fused to obtain second-order tensor feature samples.Secondly,fault diagnosis models based on the integrated support tensor and KNN-AMDM are constructed separately,and two types of fault diagnosis are targeted.Based on the output results of the two types of fault diagnosis models,two different methods for transforming the basic probability distribution assignments from the initial fault diagnosis results to the evidence body identification framework are designed.Six different fault symptom sets are obtained by adjusting the number of sensors involved.Twelve different preliminary fault diagnosis results are obtained.Finally,the D-S evidence theory is used to fuse the basic probability distribution values provided by the12evidence bodies,the more accurate gearbox decision fusion fault diagnosis result is obtained.The experimental comparison shows that this method can improve the credibility of the gear fault diagnosis results.
作者
葛江华
刘奇
王亚萍
许迪
卫芬
GE Jiang-hua;LIU Qi;WANG Ya-ping;XU Di;WEI Fen(School of Mechanical and Power Engineering, Harbin University of Science and Technology, Harbin 150080 , China;School of Mechanical Engineering, Hunan University of Science and Technology, Xiangtan 411201, China)
出处
《振动工程学报》
EI
CSCD
北大核心
2018年第6期1093-1101,共9页
Journal of Vibration Engineering
基金
国家自然科学基金资助项目(51575143)
黑龙江省自然科学基金资助项目(E2016046)
关键词
故障诊断
多传感器融合
支持张量机
集成矩阵距离测度
决策融合
fault diagnosis
multi-sensor fusion
support tensor machine
assembled matrix distance metric
decision fusion
