基于深度迁移学习的Ti-6Al-4V合金微铣削毛刺尺寸预测  

Burr size prediction in micro-milling of Ti-6Al-4V alloy based on deep transfer learning

作　　者：吴凤和[1,2] 王宇 张会龙 张宁 马轩 王志勇 WU Fenghe;WANG Yu;ZHANG Huilong;ZHANG Ning;MA Xuan;WANG Zhiyong(School of Mechanical Engineering,Yanshan University,Qinhuangdao 066004,CHN;Hebei Heavy Intelligent Manufacturing Equipment Technology Innovation Center,Qinhuangdao 066004,CHN)

机构地区：[1]燕山大学机械工程学院,河北秦皇岛066004 [2]河北省重型智能制造装备技术创新中心,河北秦皇岛066004

出　　处：《制造技术与机床》2025年第4期63-69,共7页Manufacturing Technology & Machine Tool

基　　金：河北省中央引导地方科技发展基金项目(246Z1022G);石家庄市基础研究计划重点项目(241791077A)。

摘　　要：针对钛合金微铣削加工易产生毛刺缺陷影响使用的问题,提出一种基于深度迁移学习的Ti-6Al-4V微铣削顶部毛刺尺寸预测方法。首先,以工艺参数(主轴转速、轴向切深、径向切宽和每齿进给量)为网络输入,以顶部毛刺长度为预测目标,建立了微铣削毛刺尺寸的预测模型。其次,使用625个切削仿真样本进行预训练。最后,基于迁移学习机制,借助100个切削试验样本对预训练结果进行微调,从而将仿真规律迁移至试验规律。结果表明,迁移学习模型对顺、逆铣两侧毛刺尺寸的平均预测精度分别达到了95.77%、95.45%,为钛合金微铣削毛刺的预测及控制提供了一种有效方法。Aiming at the problem that the micro-milling of titanium alloy is prone to produce burr defects that affect the use,a prediction method of the burr size at the top of Ti-6Al-4V micro-milling based on depth transfer learning was proposed.Firstly,a prediction model of micro-milling burr size was established with the process parameters(spindle speed,axial cutting depth,radial cutting width and feed per tooth)as the network input and the burr length as the prediction target.Secondly,625 cutting simulation samples were used for pre-training.Finally,based on the transfer learning mechanism,100 cutting experimental samples were used to fine-tune the pre-training results,so as to transfer the simulation law to the experimental law.The results show that the average prediction accuracy of the transfer learning model for the burr size on both sides of forward and backward milling is 95.77%and 95.45%,which provides an effective method for the prediction and control of titanium alloy micro-milling burr.

关 键 词：微铣削毛刺 TI-6AL-4V合金 毛刺 尺寸预测 迁移学习 深度学习 

分 类 号：TH142.2[一般工业技术—材料科学与工程]