基于MTF与改进ResNeXt神经网络的齿轮箱故障诊断  

Gearbox Fault Diagnosis Method Based on MTF and Improved ResNeXt Neural Network

作　　者：郑心成 郝如江 孙汇宇 范亚飞 杨青松 ZHENG Xincheng;HAO Rujiang;SUN Huiyu;FAN Yafei;YANG Qingsong(School of Mechanical Engineering,Shijiazhuang Tiedao University,Shijiazhuang Hebei 050043,China;State Key Laboratory of Mechanical Behavior and System Safety of Traffic Engineering Structures,Shijiazhuang Tiedao University,Shijiazhuang Hebei 050043,China)

机构地区：[1]石家庄铁道大学机械工程学院,河北石家庄050043 [2]石家庄铁道大学,省部共建交通工程结构力学行为与系统安全国家重点实验室,河北石家庄050043

出　　处：《机床与液压》2025年第1期9-14,共6页Machine Tool & Hydraulics

基　　金：国家自然科学基金项目(12272243);河北省科技研发平台建设专项项目(21567622H);石家庄铁道大学创新项目(YC202430)。

摘　　要：齿轮箱的振动信号包含丰富的信息。通过将采集到的一维时序信号转换为二维图像,可进一步增强信号特征,从而更好地表征设备状态。基于此,提出一种基于马尔可夫变迁场(MTF)与改进ResNeXt神经网络相结合的齿轮箱故障诊断模型。通过MTF对采集到的一维信号进行转换,得到与时序相关的二维特征图。采用像素平均法对图像进行压缩,以更好地突显其特征信息。最后,将压缩后的图像送入改进ResNeXt神经网络中进行故障识别分类。通过使用动力传动故障诊断综合实验台齿轮箱数据,验证了模型的可行性,并确定了图像转换的最佳尺寸。此外,通过使用凯斯西储大学滚动轴承数据进行消融及抗噪性实验,验证了该模型的有效性与泛化性。The vibration signal of the gearbox contains a wealth of information.The acquired 1D timing signal converting to the 2D image can further enhance the signal characteristics and better characterize the state of the device.Based on this,a gearbox fault diagnosis model based on Markov transition fields(MTF)and improved ResNeXt neural network was proposed.The 1D signal was encoded to the 2D feature map related to the time series by MTF.The image was compressed by pixel averaging method to highlight the feature information of the picture.Finally,the compressed images were sent to the improved ResNeXt neural network to identify and classify faults.The feasibility of the model was verified by using the gearbox data of drivetrain dynamics simulator for power transmission fault diagnosis,and the optimal size of image conversion was determined.In addition,by using the rolling bearing data of Case Western Reserve University to perform ablation and noise resistance experiments,the validity and generalization of the model are verified.

关 键 词：马尔可夫变迁场 ResNeXt神经网络 故障诊断 图像压缩 

分 类 号：TH132[机械工程—机械制造及自动化]