面向滚动轴承的自适应NA-MVMD融合GADF故障诊断方法  

Fault Diagnosis of Rolling Bearing Based on Adaptive NA-MVMD Fusion GADF

作　　者：瞿红春[1] 贾柏谊 郑剑青 韩松钰 马文博 QU Hong-chun;JIA Bai-yi;ZHENG Jian-qing;HAN Song-yu;MA Wen-bo(College of Aeronautical Engineering,Civil Aviation University of China,Tianjin 300300,China)

机构地区：[1]中国民航大学航空工程学院,天津300300

出　　处：《组合机床与自动化加工技术》2023年第3期99-103,108,共6页Modular Machine Tool & Automatic Manufacturing Technique

基　　金：中国民航大学科研基金项目(05yk08m);中央高校基本科研项目(ZXH2010D019)。

摘　　要：针对滚动轴承的故障特征易被湮没在噪声背景下,从而导致故障种类难以识别的问题,提出了一种基于自适应噪声辅助多元变分模态分解(NA-MVMD)降噪融合格拉姆角差场(GADF)特征提取的滚动轴承故障诊断方法。首先,使用鲸鱼优化算法(WOA)对NA-MVMD中的分解模态数K和惩罚因子α进行寻优;其次,利用NA-MVMD处理信号得到若干IMF分量,根据GADF将筛选重构后的一维数据转化为二维图片;随后,将故障特征图片输入LeNet-5卷积神经网络进行分类识别。采用某大学XJTU-ST轴承故障数据进行验证分析,分类准确率达到了97.5%,证明了该方法在较强噪声背景下具有较好的诊断性能。Aiming at the problem that the fault characteristics of rolling bearings are easily drowned in the noise background,which makes the fault types difficult to identify,a fault diagnosis method for rolling bearings based on adaptive noise-assisted multivariate variational mode decomposition(NA-MVMD)noise reduction combined with Gram angle difference field(GADF)feature extraction is proposed.First,the whale optimization algorithm(WOA)is used to optimize the decomposition mode number K and penalty factorαin NA-MVMD,and then NA-MVMD is used to process the signal to obtain several IMF components,and Convert the filtered and reconstructed one-dimensional data into two-dimensional images according to GADF,then input the fault feature image into the LeNet-5 convolutional neural network for classification and identification.The XJTU-ST bearing fault data of a university is used for verification and analysis,and the classification accuracy rate reaches 97.5%,which proves that the method has good diagnostic performance under strong noise background.

关 键 词：故障诊断 噪声辅助 多元变分模态分解 格拉姆角差场 卷积神经网络 

分 类 号：TH161[机械工程—机械制造及自动化] TG506[金属学及工艺—金属切削加工及机床]