期刊文献+

GADF和改进CNN的齿轮箱复杂环境下故障诊断 被引量:1

GADF and Improved CNN Gearbox for Fault Diagnosis in Complex Environments
下载PDF
导出
摘要 针对齿轮箱同时处于变负载和含有噪声的复杂环境下的故障诊断难题,提出一种基于格拉姆角差场(Gram Angle Difference Field,简称GADF)和改进的具有多注意力机制的卷积神经网络(Multiple Attention Mechanism Convolutional Neural Network,简称MACNN)的齿轮箱故障诊断方法。首先将采集得到齿轮箱的一维振动信号的故障数据集进行预处理;然后通过格拉姆角差场将一维数据集转为二维图像数据;其次,将二维数据集数输入到改进的带有多注意力机制的卷积神经网络模型中进行训练;最后采用Softmax对齿轮箱故障数据集进行分类。通过试验验证,这里所提的方法在两个负载数据集上都可以达到99.80%以上,改进之后的模型训练效率更高、耗时更短,同时这里所提方法也优于一些已经发表的齿轮箱故障诊断的方法,此外本方法对齿轮箱在同时处于变负载和噪声的复杂环境中,依然有着较强的故障诊断能力。 Aiming at the problem of fault diagnosis of gear box under variable load and complex environment containing noise,a new method based on Gram Angle Difference Field was proposed(for short,GADF),and an improved method of gearbox fault diagnosis with Multiple Attention Mechanism Convolutional Neural Network(for short,MACNN).Firstly,the fault data set of one-dimensional vibration signal collected from the gearbox is preprocessed.Then the one-dimensional data set is converted into two-dimensional image data by Gram Angle difference field.Secondly,the number of two-dimensional data sets is input into the improved convolutional neural network model with multiple attention mechanism for training.Finally,Softmax was used to clas⁃sify the gearbox fault data set.Through experimental verification,the proposed method can reach more than 99.80%in two load data sets,and the improved model training efficiency is higher and the time is shorter.Meanwhile,the proposed method is better than some published gear box fault diagnosis methods.In addition,the proposed method can solve the problem of gear box in the complex environment of variable load and noise at the same time.Still has a strong fault diagnosis ability.
作者 刘成义 董绍江 唐倩 邓文亮 LIU Cheng-yi;DONG Shao-jiang;TANG Qian;DENG Wen-liang(Chongqing Kechuang Vocational College,Chongqing 402160,China;Chongqing Jiaotong University,Chongqing 400074,China;College of Mechanical and Vehicle Engineering,Chongqing University,Chongqing 400030,China)
出处 《机械设计与制造》 北大核心 2024年第9期92-96,103,共6页 Machinery Design & Manufacture
基金 国家自然科学基金(51775072) 重庆市科委基础与前沿项目(CSTC2017JCYJAX0279)。
关键词 齿轮箱故障诊断 格拉姆角差场 注意力机制 卷积神经网络 Gearbox Fault Diagnosis Gram Angle Difference Field Attention Mechanism Convolutional Neural Network
  • 引文网络
  • 相关文献

参考文献8

二级参考文献59

共引文献83

同被引文献11

引证文献1

;
使用帮助 返回顶部