【作者】 李杰峰； 王国锋； 丛君宇； 马凯乐；

【Author】 Jiefeng Li;Guofeng Wang;Junyu Cong;Kaile Ma;School of Mechanical Engineering,Tianjin University;

【机构】 天津大学机械工程学院；

【摘要】 加工过程监测是智能制造技术的关键部分,对于刀具磨损和颤振的监测能够有效提高加工质量,降低生产成本。本文针对型腔件加工中存在的路径复杂多变特点,建立了一套多模型融合的加工过程监测系统。本系统使用VB.net作为平台开发语言,结合Python的深度学习算法库进行模型搭建,并集成了多套采集方案。针对颤振问题,通过变分模态分解寻找敏感特征,并且和深度残差网络结合,构建了早期颤振的识别模型;针对刀具磨损问题,建立了基于熵特征以及长短时记忆网络的刀具磨损识别模型,并进行了有效的融合,实现了从监测方案建立、信号采集、模型导入到在线实时加工监测的全过程。更多还原

【Abstract】 Machining process monitoring is a key part of intelligent manufacturing technology.Monitoring tool wear and chatter can effectively improve machining quality and reduce production cost.In view of the complexity and changeability of the machining path of cavity parts,a machining process monitoring system based on multi-model fusion is established in this paper.The system uses VB.net as the platform development language,combines Python deep learning algorithm library to build the model,and integrates several sets of acquisition schemes.For the flutter problem,the sensitive features were found through variational mode decomposition,and combined with deep residual network,the identification model of early flutter was constructed.Aiming at the problem of tool wear,a tool wear recognition model based on entropy feature and short and long time memory network was established,and the effective fusion was carried out to realize the whole process from monitoring scheme establishment,signal acquisition,model import to online real-time machining monitoring.更多还原