文章摘要
杨汉博,赵飞,朱倪黎,高志聪,冯传锋.深度卷积神经网络在多工况下刀具磨损状态监测中的应用[J].机床与液压,2021,49(3):69-74.
YANG Hanb,ZHAO Fei,ZHU Nili,GAO Zhicong,FENG Chuanfeng.Application of Deep Convolutional Neural Network in Tool Wear Monitoring under Multiple Working Conditions[J].Machine Tool & Hydraulics,2021,49(3):69-74
深度卷积神经网络在多工况下刀具磨损状态监测中的应用
Application of Deep Convolutional Neural Network in Tool Wear Monitoring under Multiple Working Conditions
  
DOI:10.3969/j.issn.1001-3881.2021.03.014
中文关键词: 刀具磨损状态监测  多工况  敏感特征  深度卷积神经网络
英文关键词: Tool wear state monitoring  Multiple working conditions  Sensitive feature  Deep convolutional neural network
基金项目:国家科技重大专项(2017ZX04011017)
作者单位E-mail
杨汉博 西安交通大学机械工程学院 hanboyang@stu.xjtu.edu.cn 
赵飞 西安交通大学机械工程学院  
朱倪黎 西安交通大学机械工程学院  
高志聪 西安交通大学机械工程学院  
冯传锋 西安交通大学机械工程学院  
摘要点击次数: 76
全文下载次数: 0
中文摘要:
      为了解决复杂多工况下刀具磨损状态的监测问题,提出一种基于深度学习的刀具磨损状态监测方法,并构建敏感特征值提取函数。基于刀具磨损数据集,建立多种工况下刀具磨损状态的监测模型,进行多工况下刀具磨损状态监测研究。研究结果表明:当敏感值界限设置为0.3时,从声发射、振动和电流信号的特征值中可以提取出56个敏感特征值;以均方根误差作为评价函数,得到测试样本的评价函数均值为0.123;模型对严重磨损状态下的刀具磨损监测效果优于对正常磨损状态下的刀具磨损监测效果;多组重复性验证证明所提出的监测方法稳定有效。
英文摘要:
      In order to solve the problem of monitoring the tool wear state under multiple operating conditions,a tool wear state prediction method based on deep learning was proposed, and the function for extracting sensitive features was constructed. Based on the tool wear dataset, the tool wear state monitoring model under multiple working conditions was established, and the tool wear state monitoring method under multiple conditions was studied. The results show that when the sensitivity limit is set to 0.3, 56 sensitive features are extracted from the features of the acoustic emission, vibration and current signals; using root mean square error as the evaluation function, the average value of the test sample is 0.123; the prediction result under severe wear state of the tool is better than that of the normal wear state of the tool; multiple sets of repetitive verification experiments prove that the proposed method is stable and effective.
查看全文   查看/发表评论  下载PDF阅读器
关闭

分享按钮