计算机集成制造系统 ›› 2017, Vol. 23 ›› Issue (第10): 2146-2155.DOI: 10.13196/j.cims.2017.10.008

• 产品创新开发技术 • 上一篇    下一篇

基于深度学习的刀具磨损监测方法

张存吉,姚锡凡+,张剑铭,刘二辉   

  1. 华南理工大学机械与汽车工程学院
  • 出版日期:2017-10-31 发布日期:2017-10-31
  • 基金资助:
    国家自然科学基金资助项目(51175187,51675186);广东省科技计划资助项目(2016A020228005,2016B090918035)。

ATool wear monitoring based on deep learning

  • Online:2017-10-31 Published:2017-10-31
  • Supported by:
    Project supported by the National Natural Science Foundation,China(No.51175187,51675186),and the Science & Technology Program of Guangdong Province,China(No.2016A020228005,2016B090918035).

摘要: 为监测制造车间机械加工设备刀具的磨损程度,提出应用深度学习方法实施刀具的磨损监测。深度学习理论作为人工智能领域的最新研究成果,以其中的深度卷积神经网络构建刀具磨损监测的模型,给出刀具磨损监测的流程,采用微型铣床与无线三轴加速度计搭建了数据采集实验平台。实验结果表明,与其他两种常用深度神经网络以及传统神经网络模型相比较,所提基于深度学习方法监测过程简单,不仅具有较高的准确度与较低的损失函数值,还实现了刀具磨损程度分类。

关键词: 刀具磨损监测, 数据采集, 深度学习, 卷积神经网络, 无线三轴加速度计

Abstract: To monitor the tool wear for machining equipment in manufacturing workshops,deep learning was proposed to realize the tool wear monitoring.As the latest research result in Artificial Intelligence (AI) field,the Convolutional Neural Network (CNN) was adopted to build the model of tool wear monitoring.A flow chart of tool wear monitoring was given,and a micro milling machine and a wireless triaxial accelerometer were used to build the experimental setup to acquire measurement data.The experimental results showed that the proposed approach was simple to realize the tool wear monitoring with higher accuracy and lower loss during the learning process by comparing with other two common models that were deep CNNs and traditional Neural Network (NN),and a classification of tool wear degree was realized.

Key words: tool wear monitoring, data acquisition, deep learning, convolutional neural network, wireless triaxial accelerometer

中图分类号: