Scholarlink

<p>Tool anomalies in computer numerical control (CNC) milling processes are unpredictable, hindering the promotion of fully automated machining. Traditional detection systems often struggle with stability due to the irregular and non-parametric nature of signals generated during dynamic milling. This study proposes an improved probability and statistics-based model for constructing tool anomaly thresholds in the time domain, with the decision-making strategy seamlessly integrated into CNC milling systems. A robust data acquisition and preprocessing framework was developed to improve the accuracy and reliability of real-time monitoring data. A Gaussian process model was employed to construct an anomaly detection threshold data set from irregular signals. Anomalies were identified when monitoring indicators surpassed the established threshold. The proposed method was validated through milling experiments of a turbine blisk, demonstrating an overall anomaly detection accuracy of 91.45%, which exceeds those of four other typical anomaly detection methods. These results confirm the effectiveness of the proposed method and its potential applicability in industry.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

An innovative strategy for real-time tool anomaly detection in CNC milling processes using time series monitoring

  • 期刊论文

  • 发表日期：

    2025年6月20日

查看全文
  • Yao Li,
  • Zhengcai Zhao,
  • Zhigao Peng,
  • Lei Zhang,
  • Wenfeng Ding,
  • Yucan Fu

摘要

Tool anomalies in computer numerical control (CNC) milling processes are unpredictable, hindering the promotion of fully automated machining. Traditional detection systems often struggle with stability due to the irregular and non-parametric nature of signals generated during dynamic milling. This study proposes an improved probability and statistics-based model for constructing tool anomaly thresholds in the time domain, with the decision-making strategy seamlessly integrated into CNC milling systems. A robust data acquisition and preprocessing framework was developed to improve the accuracy and reliability of real-time monitoring data. A Gaussian process model was employed to construct an anomaly detection threshold data set from irregular signals. Anomalies were identified when monitoring indicators surpassed the established threshold. The proposed method was validated through milling experiments of a turbine blisk, demonstrating an overall anomaly detection accuracy of 91.45%, which exceeds those of four other typical anomaly detection methods. These results confirm the effectiveness of the proposed method and its potential applicability in industry.