基于改进DSmT的高速电主轴早期振动故障诊断方法

Diagnosis of a High-Speed Motorized Spindle with Early Vibration Fault Symptom Based on Improved DSmT

摘要: 提出在高速电主轴关键部位安装多个振动传感器组成多传感器数据采集系统采用多源信息融合方法对早期振动故障进行融合诊断的方法。构建高速电主轴早期振动故障诊断系统框架：包括组建多传感器采集系统、基于本征模态函数的信息熵特征提取、基于BP神经网络的广义基本置信配置以及决策级融合等各个环节。针对Dezert-Smarandache理论(DSmT)在解决高冲突证据融合时存在的不足综合考虑融合证据的数量、证据本身的信任度以及各个证据之间的相互支持度，提出对各个证据进行加权赋值然后进行融合的方法。算例表明此种方法增加可信度，更合理地分配了冲突信息，对决策结果更为有利。

Abstract: Multi-sensor network was built on different positions of the high-speed motorized spindle. Multisensor network information fusion method was used to diagnose the high-speed motorized spindle with early vibration fault symtoms. Early vibration fault diagnosis framework was proposed in this paper. A method based on the intrinsic mode functions (IMF) energy entropy was used to extract the signal’s feature. On the basis of the output of (BP) neural network basic belief assignment function was constructed. Information fusion was used to combine the different evidences and make the final decision. There is something illogical in dealing with multi evidences. A weight assignment method is adopted considering number conflict and support of different evidences. The results indicate this method advanced the reliability and solved the conflict among evidences much reasonably. This method is much advantageous to the fusion result.