摘要
近来以深度学习算法为代表的滚动轴承特征智能提取和故障辨识技术被广泛研究,但目前研究大多局限于无强干扰的轴承故障。在齿轮箱存在较强齿轮振动干扰条件下,基于此类算法的轴承故障辨识率将显著降低。为提高在较强齿轮振动信号干扰下齿轮箱轴承故障智能辨识的准确率,提出了一种基于自参考自适应噪声消除技术(SANC)和一维卷积神经网络(1D-CNN)的齿轮箱轴承故障诊断方法。首先利用SANC将齿轮箱振动信号分离为周期性信号分量成分和随机信号分量,抑制齿轮等周期强干扰成分,再通过1D-CNN对包含轴承故障特征的随机信号成分进行智能特征提取和识别,实现在齿轮振动干扰下齿轮箱轴承故障辨识率的提高。通过与不同方法的对比验证了本文所提方法的优势和有效性。
Recently,feature intelligent extraction and fault recognition techniques of rolling bearing based on deep learning algorithm are widely studied,but most of studies are limited to bearing faults without strong interference.Under the condition of stronger gear vibration interference existing in gearbox,the bearing fault recognition rate based on this type algorithm significantly drops.Here,in order to improve the accuracy rate of gearbox bearing fault intelligent recognition under stronger gear vibration signals interference,a gearbox bearing fault diagnosis method based on self-reference adaptive noise cancellation(SANC)technique and one-dimensional convolution neural network(1-DCNN)was proposed.Firstly,SANC was used to decompose gear vibration signals into periodic signal components and random signal ones,and suppress gear periodic strong interference components.Then,1-D CNN was used to do intelligent feature extractionand recognition of random signal components containing bearing fault features,and realize improving gearbox bearing fault recognition rate under strong gear vibration interference.The advantages and effectiveness of the proposed method were verified with comparison to different methods.
作者
高佳豪
郭瑜
伍星
GAO Jiahao;GUO Yu;WU Xing(Faculty of Mechanical and Electrical Engineering,Kunming University of Science and Technology,Kunming 650500,China)
出处
《振动与冲击》
EI
CSCD
北大核心
2020年第19期204-209,257,共7页
Journal of Vibration and Shock
基金
国家自然科学基金(51675251)。
关键词
齿轮箱
自参考自适应噪声消除技术
一维卷积神经网络
故障诊断
gearbox
self-reference adaptive noise cancellation(SANC)technique
1-D convolution neural network(CNN)
fault diagnosis
