摘要
电机轴承运行状态及时诊断及分级预警可有效保障电机的运用安全可靠。针对轴承电信号故障特征微弱、随机干扰多以及现有方法可解释性较差和自适应性能力不强的问题,提出一种基于线性预测消噪谱估计及谱平均中心峭度指标相结合的轴承电信号故障诊断方法。首先,采用结合线性预测消噪和谱估计方法,消除电流信号中电源基波、固有偏心谐波等无关轴承故障的可预测噪声分量,提取轴承故障特征;其次,通过同一工况不同数据集进行标准化和频谱平均,以消除电流频谱诸多随机干扰的影响;最后,以轴承不同阶次故障特征频率为中心构建中心峭度指标,以此作为定量评估轴承状态及其严重程度指标,实现电机轴承故障诊断与分级预警。为验证所提方法的正确性和有效性,在牵引电机故障模拟对托试验台的基础上,选用3种不同程度的内圈缺陷轴承和1块正常轴承为试验对象,并选用常用的其他3钟特征提取方法进行对比验证分析。试验结果表明:相较于其他3钟特征提取方法,提出的结合线性预测消噪和谱估计特征提取方法能有效提取微弱的轴承故障信息,具有更强的特征提取的能力。提出的结合频谱平均和中心峭度的诊断评估方法,在消除电流频谱随机干扰影响的同时还能定量评估轴承健康状态,实现轴承诊断与分级预警,具有可解释性强、评估效果好及自适应性能力强的优点。
Timely diagnosis and layered warning of motor bearing operational state can effectively ensure safe and reliable running of motors.A bearing fault diagnosis method by electrical signals based on linear prediction denoising spectrum estimation and spectral average central kurtosis was proposed to address the problems of weak fault characteristics,multiple random interferences,poor interpretability,and adaptability of existing methods.First,a combination of linear prediction denoising and spectral estimation method was adopted to eliminate the predictable irrelevant bearing fault noise,such as power supply fundamental waves and inherent eccentricity harmonics in the current signal,to extract bearing fault characteristics.Second,standardization and spectral averaging were carried out on different datasets under the same working condition to eliminate the influence of many random interferences in the current spectrum.Finally,a central kurtosis index was constructed based on the different order characteristic frequencies for quantitative assessment of bearing status and its severity,to achieve motor bearing fault diagnosis and layered warning.To verify the correctness and effectiveness of the proposed method,three different degrees of inner ring defect bearings and one normal bearing were selected on the traction motor fault simulation test bench,and other three commonly used feature extraction methods were selected for comparative verification and analysis.The experimental results show that the proposed method combining linear prediction denoising and spectral estimation feature extraction can effectively extract weak bearing fault information and has stronger feature extraction ability compared to the other three feature extraction methods.The proposed diagnostic evaluation method combining spectrum averaging and center kurtosis can eliminate the random interference of the current spectrum and quantitatively evaluate the health status of bearings,achieving bearing diagnosis and layered warning.It has the advantages of strong interpretability,good evaluation effect and strong adaptability.
作者
戴计生
丁荣军
付勇
刘欣
于天剑
DAI Jisheng;DING Rongjun;FU Yong;LIU Xin;YU Tianjian(College of Mechanical and Vechicle Engineering,Hunan University,Changsha 410083,China;Zhuzhou CRRC Times Electric Co.,Ltd.,Zhuzhou 412001,China;School of Traffic and Transportation Engineering,Central South University,Changsha 410075,China)
出处
《铁道科学与工程学报》
EI
CAS
CSCD
北大核心
2024年第10期4344-4356,共13页
Journal of Railway Science and Engineering
基金
国家重点研发计划项目(2021YFB3400702-04)。
关键词
电机轴承
诊断与分级预警
线性预测消噪
谱平均
中心峭度
motor bearing
diagnosis and layered warning
linear prediction denoising
spectrum averaging
central kurtosis