基于特征自适应融合和集成学习的高性能铣削刀具状态监测  

Tool Condition Monitoring for High Performance Milling Based on Feature Adaptive Fusion and Ensemble Learning

作　　者：王润琼 宋清华[1,2,3] 彭业振 刘战强[1,2,3] WANG Runqiong;SONG Qinghua;PENG Yezhen;LIU Zhanqiang(School of Mechanical Engineering,Shandong University,Jinan 250061;Key Laboratory of High Efficiency and Clean Mechanical Manufacture of Ministry of Education,Shandong University,Jinan 250061;National Mechanical Engineering Experimental Teaching Demonstration Center,Shandong University,Jinan 250061)

机构地区：[1]山东大学机械工程学院,济南250061 [2]山东大学高效洁净机械制造教育部重点实验室,济南250061 [3]山东大学国家机械工程实验教学示范中心,济南250061

出　　处：《机械工程学报》2024年第1期149-158,共10页Journal of Mechanical Engineering

基　　金：国家自然科学基金(52275445);山东省重点研发计划(重大科技创新工程)(2020CXGC010204)资助项目。

摘　　要：通过人工智能、工业大数据实时感知切削加工中的刀具状态是实现面向性能的制造的重要技术途径,也是高性能制造的关键内涵。然而,在目前的切削刀具状态监测算法中,特征提取过程多依赖于人工经验,这无疑限制了刀具状态监测技术的在高性能制造中的推广应用。因此,针对高性能加工监测中的自主性和准确性要求,基于特征自适应融合和集成学习技术,提了出一种面向高性能铣削的刀具磨损监测方法。所提出的监测方法能够根据特征的表现自动为其赋予不同的权重从而实现特征的自适应融合,同时利用AdaBoost集成学习算法,在自动融合特征的同时保证了状态监测精度。薄壁件铣削实验表明,监测结果与真实磨损间的RMSE和MAE值最大为10.44,最小可达5.16。所提出的方法能够自主、准确地监测航空类薄壁件铣削加工中的刀具磨损状态,解决了高性能铣削加工刀具磨损监测中的人工经验依赖问题。Real-time sensing of tool conditions in cutting through artificial intelligence and industrial big data is an important technical way to realize performance-oriented manufacturing,which is also a key connotation of high-performance manufacturing.However,in the current tool condition monitoring(TCM)algorithm,the feature extraction still relies on manual experience,which limits the application of TCM in high-performance manufacturing(HPM).Therefore,a TCM method for HPM is proposed based on feature adaptive fusion and ensemble learning techniques to ensure autonomy and accuracy in the monitoring of HPM.The proposed fusion method automatically assigns weights to the extracted features based on their performance,thus achieving adaptive feature fusion.AdaBoost algorithm is also used to ensure monitoring accuracy while automatically fusing features.The milling experiments of thin-walled parts show that the maximum RMSE and MAE values between the monitoring results and the actual results are 10.44,and the minimum is 5.16.The proposed method monitors the tool wear in the machining of aerospace thin-walled parts autonomously and accurately,which solves the problem of manual experience dependence in the condition monitoring of HPM milling tools.

关 键 词：高性能制造 刀具磨损监测 特征融合 薄壁件 铣削 

分 类 号：TG68[金属学及工艺—金属切削加工及机床]