首页|基于WTMSE-AMCNN_1D的协作机器人故障诊断

基于WTMSE-AMCNN_1D的协作机器人故障诊断

扫码查看
原文链接
  • NETL
  • NSTL
  • 万方数据
  • 维普
六轴协作机器人在实际工作中难以采集到振动数据,且其故障诊断精度低,针对这一问题,提出一种基于多尺度小波分解、样本熵与一维注意力卷积神经网络(WTMSE-AMCNN_1D)的六轴协作机器人电流信号故障诊断方法.首先,对采集的原始故障数据进行随机采样;其次,采用多尺度小波分解后计算样本熵的方法来提取原始信号特征,将其作为引入注意力机制(AM)的一维卷积神经网络的输入并进行训练;最后,利用端到端训练后的模型实现故障诊断.通过实验采集某六轴协作机器人的电流数据进行诊断测试,并与其它模型对比,结果表明WTMSE-AMCNN_1D模型诊断精度达到99.21%,可以有效诊断协作机器人的故障.
Fault Diagnosis of Collaborative Robots Based on WTMSE-AMCNN_1D
Six-axis collaborative robots are difficult to collect vibration data in actual work,and the fault di-agnosis accuracy is low.To solve this problem,a fault diagnosis method of current signal of six-axis collab-orative robot based on multi-scale wavelet decomposition,sample entropy and one-dimensional attention convolutional neural network(WTMSE-AMCNN_1D)is proposed.In this study,the collected raw fault data was first randomly sampled;secondly,the method of calculating sample entropy after multi-scale wave-let decomposition is used to extract the original signal features,and they are as input to the one-dimensional convolutional neural network introduced into the attention mechanism(AM)and be trained;finally,trou-bleshoot the end-to-end trained model.Through experimental collection of current data of a domestic six-axis collaborative robot for diagnostic testing,and compared with other models,the results show that the di-agnostic accuracy of WTMSE-AMCNN_1D model reaches 99.21%,which can effectively diagnose the fault of the collaborative robot.

collaborative robotfault diagnosiswavelet decompositionmulti-scale sample entropyatten-tion mechanismone-dimensional convolutional neural network

戴天赐、王华、汪健、董凌浩、李帅康

展开 >

南京工业大学机械与动力工程学院,南京 211816

江苏省工业装备数字制造及控制技术重点实验室,南京 211899

协作机器人 故障诊断 小波分解 多尺度样本熵 注意力机制 一维卷积神经网络

江苏省研究生科研与实践创新计划资助项目江苏省重点研发计划项目

KYCX22_1282BE2019007-3

2024

10.13462/j.cnki.mmtamt.2024.01.026

组合机床与自动化加工技术
大连组合机床研究所 中国机械工程学会生产工程分会

组合机床与自动化加工技术

CSTPCD北大核心
影响因子：0.671
ISSN：1001-2265
年,卷(期)：2024.(1)
  • 5