Abstract: To improve the accuracy and generalization ability of Atomic Force Microscope (AFM) tip wear online monitoring in nano-machining,a detection method based on an Incremental Adaptive Support Vector Machine (IASVM) was proposed.The peak-to-peak value and variance of the lateral force were calculated as feature variables,and the data set was updated via a moving window.Based on the current Support Vector Machine (SVM) structure,the regularization parameter C and kernel parameter σ were adaptively modified through maintaining the optimization conditions required by Kuhn-Tucher theorem to obtain the updated SVM solution,then the incremental data and the original data disturbed by it were performed classification,the trend of the tip failure number reflected the degree of tip wear.Experiments showed that the algorithm could meet the requirements of online detection in recognition accuracy and time.Compared with the directional acyclic graph support vector classifier,IASVM had stronger robustness and higher generalization ability.

Key words: nano machining, online recognition of tip wear, lateral force characteristics, incremental adaptive support vector machine, statistical pattern damage detection

摘要： 为了提高纳米加工刀具磨损状态在线监测的精度与泛化能力,提出一种基于增量自适应支持向量机的基于原子力显微镜(AFM)尖端磨损识别方法。该方法以横向力的峰-峰值和方差作为特征变量,通过移动视窗获取增量数据;以维持Kuhn-Tucher定理所要求的最优化条件为准则,在当前支持向量机解结构基础上自适应修改正则化参数C和核参数σ,以获得更新支持向量机结构,并对增量数据及受其扰动的原数据进行分类;根据尖端失效点数量走势,判定尖端磨损程度。实验证明该算法在识别精度与时间上可满足在线检测要求。与定向非循环图支持向量分类器对比,该算法具有更强的鲁棒性与更高的泛化能力。

关键词: 纳米加工, 尖端磨损在线识别, 横向力特征, 增量自适应支持向量机, 统计模式损伤检测