摘要
针对现有的永磁同步电机(PMSM)轴承故障检测方法准确度低的问题,对PMSM轴承故障表征方法和基于神经网络的检测方法进行了研究,提出了一种PMSM轴承故障归一化表征指标集合的构建方法,和一种基于VMD和MLP的PMSM轴承故障检测方法。首先,采用融合PMSM轴承故障频域特征进行归一化处理的方法,构建了一个PMSM轴承故障表征指标集合;然后,利用优化后的变分模态分解(VMD)方法,对振动信号进行了降噪重构,提取了故障频域特征,并计算出了归一化指标集合;利用基于多层感知器(MLP)的神经网络模型对获取的归一化指标集合进行了训练,得到了一种高准确度PMSM轴承故障检测器;最后,采用了一套可以模拟数控机床进给传动系统的试验测试装置,对基于VMD和MLP的PMSM轴承故障检测方法的有效性和先进性进行了验证。研究结果表明:PMSM轴承故障表征指标集合比现有的指标具有更强的故障表征能力,基于VMD和MLP的PMSM检测方法的平均检测准确度高达95.4%;该结果验证了归一化PMSM轴承故障表征指标集合的先进性,以及基于VMD与MLP的PMSM轴承故障检测方法的有效性。
Aiming at the problem of the low detection accuracy of the existing permanent magnet synchronous motor(PMSM)bearing fault detection methods,the study of the representation method of the PMSM bearing fault and the detection method based on neural network were executed.A method for constructing a normalized characterization index set for PMSM bearing faults,and a PMSM bearing fault detection method based on VMD and MLP were proposed.Firstly,a set of PMSM bearing fault characterization indices was constructed by a method of fusing the bearing fault frequency domain characteristic parameters of PMSM and normalizing them.Then,the optimized variational mode decomposition(VMD)method was used to denoise and reconstruct the vibration signals,extract the bearing fault frequency domain characteristic parameters,and the normalized set of PMSM bearing fault characterization indices was calculated.The neural network model based on the multi-layer perceptron(MLP)was used to train the normalized set of PMSM bearing fault characterization indices,and a model of PMSM bearing fault detection with high detection accuracy was obtained.Finally,a set of test and test device which can simulate the feed drive system of NC machine tool was adopted.The validity and advanced nature of the PMSM bearing fault detection method based on VMD and MLP were verified.The experimental results show that the proposed normalized set of PMSM bearing fault characterization indices has stronger fault characterization ability than the existing ones,and the average detection accuracy of the proposed method is up to 95.4%.The experimental results verify the advanced nature of the normalized set of PMSM bearing fault characterization indices,and the proposed bearing fault detection method of PMSM based on the VMD,and the effectiveness of PMSM bearing fault detection method based on VMD and MLP.
作者
黄晓诚
贺青川
陈文华
HUANG Xiao-cheng;HE Qing-chuan;CHEN Wen-hua(National and Local Joint Engineering Research Center of Reliability Analysis and Testing for Mechanical and Electrical Products,Zhejiang Sci-Tech University,Hangzhou 310018,China)
出处
《机电工程》
CAS
北大核心
2022年第7期911-918,共8页
Journal of Mechanical & Electrical Engineering
基金
国家自然科学基金资助项目(51875529)
装备预先研究领域基金资助项目(80902010302)
NSFC-浙江两化融合资助项目(U1709210)
浙江省万人计划杰出人才资助项目(2018R51008)。
关键词
轴承故障特征提取
永磁同步电机
故障表征
神经网络
变分模态分解
多层感知器
归一化处理
bearing fault feature extraction
permanent magnet synchronous motor(PMSM)
fault characterization
neural network
variational mode decomposition(VMD)
multi-layer perceptron(MLP)
normalization processing