华中科技大学学报（自然科学版）2025，Vol.53Issue(12)：12-19,8.DOI:10.13245/j.hust.251202
混沌缺失数据下的高档数控加工刀具状态监测方法
Tool condition monitoring method for high-end CNC machining with chaotic data loss characteristics
摘要
Abstract
For the issue of data missing mechanism entanglement in high-end CNC machine tools during high-dynamic and deeply coupled machining processes,an intelligent tool condition monitoring method was proposed based on adaptive variational diffusion data recovery.First,a self-learning network with vectorized missing distribution features was designed.The network automatically identified complex missing patterns by unsupervised learning of the probability distribution structure of missing data,avoiding the limitations of manual identification and single matching in traditional methods.Then,based on the intrinsic temporal dependencies and multidimensional feature correlations within the data,an adaptive missing mechanism data recovery diffusion model was constructed.By combining multi-step reasoning mechanisms and dynamic strategy adjustments in the diffusion process,the model used the captured structured features of the data as supervisory signals to progressively explore the missing feature recovery path.Finally,a feature fusion-based tool wear state prediction model was constructed,and a knowledge distillation technique was designed to implement a cross-step sampling strategy,which significantly improved the training efficiency of the model while maintaining inference accuracy.Experimental results from actual machining cases show that the proposed method has distinct advantages under high missing rate chaotic missing data conditions,maintaining an accuracy of over 80%consistently.关键词
数控机床/高档数控加工/刀具状态监测/数据缺失/扩散模型Key words
CNC machine tool/high-end CNC machining/tool condition monitoring/data loss/diffusion model分类
信息技术与安全科学引用本文复制引用
XIAO Qinge,GU Yuntao,PENG Tao,YANG Zhile,XU Kai..混沌缺失数据下的高档数控加工刀具状态监测方法[J].华中科技大学学报（自然科学版）,2025,53(12):12-19,8.基金项目
国家自然科学基金青年基金资助项目(62403451) （62403451）
深圳市2024年度基础研究专项(深圳市自然科学基金)面上项目(JCYJ20240813154910014). （深圳市自然科学基金）
