基于多数据融合的电机故障诊断方法研究  

Research on Motor Fault Diagnosis Method Based on Multi Data Fusion

作　　者：袁媛[1] 方红彬[1] 殷忠敏[1] YUAN Yuan;FANG Hongbin;YIN Zhongmin(Department Electrical Engineering,Hebei Institute of Mechanical and Electroninc Technology,Xingtai 054000,Hebei,China)

机构地区：[1]河北机电职业技术学院电气工程系,河北邢台054000

出　　处：《电气传动》2021年第9期75-80,共6页Electric Drive

基　　金：河北省高等学校科学研究计划项目青年基金(QN2019222);河北省高等学校科学研究计划项目(QN2015014)。

摘　　要：电机作为各类电驱设备的主要动力装置,具有结构简单、控制方便、能效高、无污染等优点。在电机运行过程中,受载荷多变、零部件老化、散热条件差等影响,故障频发,进而降低电驱装置的工作效率和稳定性。此外,电机故障种类繁多,各故障的征兆与表现又极其相似,不同故障产生的原因也错综复杂,这极大地提高了电机故障诊断的难度。传统的电机故障诊断过程中多是基于单一传感器信号,存在不确定性大、诊断精度差等问题,为克服上述缺点,提出一种基于多传感器参数融合的电机故障诊断方法,基于振动加速度计和电流传感器信号,结合BP神经网络算法和D-S证据理论对电机故障进行准确辨识,提高电机故障诊断的准确性。简要介绍了多传感器数据融合技术的结构框架,在分析异步电机典型故障机理的基础上,对基于BP神经网络学习算法和D-S证据理论的多传感器数据融合电机故障诊断系统进行详细分析,并通过实例对所提出故障诊断方法的有效性进行验证。研究结果表明,采用所提出的多数据融合电机故障诊断方法可以高置信度地诊断出电机的故障类型。As the main power equipment of all kinds of electric drive devices,motor has the advantages of simple structure,convenient control,high energy efficiency and no pollution.In the process of motor operation,affected by variable load,aging of parts and poor heat dissipation conditions,faults occur frequently,thus the working efficiency and stability of the electric drive device are reduced.In addition,there are many kinds of motor faults,the symptoms and performance of each fault are very similar,the causes of different faults are also complex,which greatly improve the difficulty of motor fault diagnosis.The traditional motor fault diagnosis process is mostly based on a single sensor signal,which has the problems such as large uncertainty and poor diagnosis accuracy.In order to overcome the above shortcomings,a motor fault diagnosis method based on multi-sensor parameter fusion was proposed.Based on vibration accelerometer and current sensor signal,combined with BP neural network algorithm and D-S evidence theory,the motor fault was accurately identified and the accuracy of motor fault diagnosis was improved.The structure framework of multi-sensor data fusion technology was briefly introduced.Based on the analysis of the typical fault mechanism of asynchronous motor,the multi-sensor data fusion motor fault diagnosis system based on BP neural network learning algorithm and D-S evidence theory was analyzed in detail,and the effectiveness of the proposed fault diagnosis method was verified by an example.The results show that the proposed multi data fusion motor fault diagnosis method can diagnose the motor fault type with high confidence.

关 键 词：电机故障诊断 数据融合 神经网络 证据理论 

分 类 号：TM28[一般工业技术—材料科学与工程]