摘要
针对现有数控机床主轴故障检测算法存在的检测精度低、故障分类效果差等不足,提出一种基于激光多普勒振动谱的在线针对方案。先构建以激光器和探测器为核心部件的激光多普勒故障检测系统,采集故障数据集并对数据进行降维处理;然后针对机床主轴多故障模态的情况,基于可变经验模态分解(VMD)算法分解故障的本征模态分量,形成新的故障特征集;最后利用经过改进的深度卷积神经网络训练故障特征集,实现对多故障模态的分类识别和诊断。实验结果显示,所提算法信号源频率误差值较低,在故障点的诊断和分类识别过程中,效果显著优于传统故障在线诊断算法。
Aiming at the shortcomings of existing CNC machine tool spindle fault detection such as low detection accuracy and poor fault classification,an online targeting scheme based on Doppler laser vibration spectrum is proposed.Firstly,a Doppler laser fault detection system with laser and detector as the core components is constructed,and the fault data set are collected and processed by dimensionality reduction.In view of the multi-fault modes of machine tool spindle,a new fault feature set is formed based on the decomposition of the fault mode classification based on VMD algorithm.An improved deep convolutional neural network is used to train the fault feature set to realize the classification and diagnosis of multiple fault modes.The experimental results show that the proposed algorithm has lower frequency error value,and the diagnosis effect is significantly better than the traditional online fault diagnosis algorithm in the process of fault targeting and classification recognition.
作者
邓宇翔
李正红
Deng Yuxiang;Li Zhenghong(Faculty of Electrical and Mechanical Engineering,Kunming Metallurgy College,Yunnan Kunming,650300,China;College of Information Engineering,Kunming University,Yunnan Kunming,650214,China)
出处
《机械设计与制造工程》
2024年第4期61-66,共6页
Machine Design and Manufacturing Engineering
基金
云南省教育厅科学研究基金(2023J1525)。
关键词
激光多普勒
振动谱
机床主轴
深度卷积神经网络
Doppler laser
vibration spectrum
machine tool spindle
deep convolutional neural networks
