Scholarlink

<p>This study designed a system for diagnosing the wear condition of ball screws and identifying improper preloading of the end cap in a single-axis slant-type ball screw machine. This system detects ball wear or part misalignment according to decreases in the thickness of a preload spacer. Experiments were performed in this study by using a single-axis slant-type ball screw machine, and an accelerometer was used to collect vibration data. Subsequently, feature engineering was conducted to enhance signal features, following which principal component analysis was performed to filter out data features with low importance. The important features were then incorporated into a Gaussian mixture model to develop an unsupervised model for bearing fault detection. The Kullback–Leibler (KL) and Jensen–Shannon (JS) divergence analysis methods were used to calculate the differences between data corresponding to normal and abnormal operation conditions, enabling the monitoring of abnormal preload conditions at the bearing end of a single-axis machine. Experimental results indicated that reductions of &gt; 10&#xa0;μm in the preload spacer thickness could be identified with accuracy values of over 60% and 88% through KL divergence analysis and JS divergence analysis, respectively. The unsupervised prognostic diagnosis for the preload loss of a ball screw drive system of a machine tool is validated successfully for practical application.</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

Prognostic diagnosis system based on feature engineering and a Gaussian mixture model for detecting preload failure at the bearing end of a single-axis slant bed

  • 期刊论文

  • 发表日期：

    2025年5月10日

查看全文
  • Yi-Cheng Huang,
  • Chi-Hsien Wang,
  • Tzu-Teng Peng

摘要

This study designed a system for diagnosing the wear condition of ball screws and identifying improper preloading of the end cap in a single-axis slant-type ball screw machine. This system detects ball wear or part misalignment according to decreases in the thickness of a preload spacer. Experiments were performed in this study by using a single-axis slant-type ball screw machine, and an accelerometer was used to collect vibration data. Subsequently, feature engineering was conducted to enhance signal features, following which principal component analysis was performed to filter out data features with low importance. The important features were then incorporated into a Gaussian mixture model to develop an unsupervised model for bearing fault detection. The Kullback–Leibler (KL) and Jensen–Shannon (JS) divergence analysis methods were used to calculate the differences between data corresponding to normal and abnormal operation conditions, enabling the monitoring of abnormal preload conditions at the bearing end of a single-axis machine. Experimental results indicated that reductions of > 10 μm in the preload spacer thickness could be identified with accuracy values of over 60% and 88% through KL divergence analysis and JS divergence analysis, respectively. The unsupervised prognostic diagnosis for the preload loss of a ball screw drive system of a machine tool is validated successfully for practical application.