基于KNN算法的数控机床加工过程异常检测方法研究  

Research on Anomaly Detection Method for Machining Process of NC Machine Tools Based on KNN

作　　者：刘福民 凌思庆 于音 冯子豪 董琦 高诚 LIU Fumin;LING Siqing;YU Yin;FENG Zihao;DONG Qi;GAO Cheng(Intelligent Manufacturing Innovation Center,AVIC Manufacturing Technology Institute,Beijing 100024,China)

机构地区：[1]中国航空制造技术研究院智能制造装备中心,北京100024

出　　处：《机床与液压》2024年第21期168-172,共5页Machine Tool & Hydraulics

摘　　要：针对数控机床加工过程异常检测问题,提出一种基于KNN算法的数控机床加工过程异常检测方法。该方法利用机床加工过程信号,通过时、频域分析提取信号特征,利用KNN算法进行决策判断,可检测并识别出数控机床加工过程中存在的异常情况。利用某生产线上的实验案例,在数控机床上完成了多组正常零件和常见异常零件的加工实验,采集了加工过程各轴的高频电流数据,对信号进行处理,完成了加工过程信号的特征提取并从中选取了对异常检测有效的特征,经过交叉实验,确定了KNN算法合适的K值。最后,通过训练,得到了异常检测模型,并利用验证集对模型进行了验证,证明了该异常检测模型具有较高的准确率。A KNN based anomaly detection method was proposed for the machining process of CNC machine tools.In this method,CNC processing signals were utilized to extract signal features through time-domain and frequency-domain analysis,KNN algorithm was used for decision making,so abnormal situations in CNC processing process could be detected.Using an experimental case on a certain production line,multiple sets of normal and common abnormal parts were processed on a CNC machine tool,high-frequency current datasets of each axis during the machining processes were collected and processed.Feature extraction of the signals was implemented and effective features for anomaly detection were selected.Through cross experiments,the appropriate K value for the KNN algorithm was determined.Finally,through training,an anomaly detection model was obtained and validated using a validation set,proving that the anomaly detection model has high accuracy.

关 键 词：数控机床加工过程 异常检测 KNN算法 特征提取 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]