摘要
刀具磨损监测是推动数控机床稳定运行的有效手段。为实现刀具健康实时监测,设计一种基于长短记忆网络的刀具磨损监测方法。将实时采集的切削力信号进行去噪、降维等预处理;用堆叠自编码器进行特征提取,获得影响刀具磨损的本质特征;用长短记忆网络构建刀具磨损监测模型,实现加工过程中刀具磨损监测。利用铣削实验数据进行实例验证,获得磨损量平均绝对误差为0.06080的铣刀磨损量监测模型,验证了该方法的有效性。
As tool wear monitoring is an effective means to promote the stable operation of CNC machine tools,a tool wear monitoring method based on long-short term network is designed to realize real-time monitoring of tool health.The cutting force signal collected in real time is preprocessed such as denoising and dimension reduction,and the stacked autoencoder is applied for feature extraction to obtain the essential features affecting tool wear.A long-short term network is used to build a tool wear monitoring model to realize tool wear monitoring during machining.And the milling experiment data is verified to obtain the milling cutter wear monitoring model with the average absolute error of wear as 0.06080,which verifies the effectiveness of the designed method.
作者
陈笑颖
许鹏
CHEN Xiaoying;XU Peng(AVIC Manufacturing Technology Institute,Beijing 100024,China)
出处
《机械制造与自动化》
2023年第3期106-111,共6页
Machine Building & Automation
基金
国防基础科学研究计划项目(JCKY2018205B013)。
关键词
机床健康监测
刀具磨损
长短记忆网络
machine tool health monitoring
tool wear
long-short term networks
