Scholarlink

<p>This study presents a vibration-based tool condition monitoring approach for face milling processes under varying cutting parameters. The analysis involved computing thirty features (ten statistical and twenty histogram features) from Continuous Wavelet Transform (CWT) coefficients using five wavelet families ‘Daubechies (db), Biorthogonal (bior), Reverse Biorthogonal (rbior), Symlet (sym), and Coiflet (coif)’ across sixty-four levels. The best levels of highly performing wavelet member from each wavelet family was identified using the J48 algorithm. They were rbior1.5 at level 14, bior5.5 at level 13, db8 at level 3, sym6 at level 3 and coif4 at level 13. The prominent features of the identified wavelet members at their respective levels were selected using decision tree and inputted to support vector machine (SVM) and K-star algorithms. The results show that the K-star outperformed the SVM with the CWT feature. The performance of other ML algorithms, namely, random forest, k-NN, Naïve Bayes, and Bayes net algorithms were also analysed. The findings highlight the potential adoption of this approach in manufacturing industries to optimize tool usage while ensuring product surface quality.</p> Graphical abstract

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

Continuous wavelet transform based face milling tool condition classification using support vector machine and K-star algorithm–a comparative study

  • 期刊论文

  • 发表日期：

    2025年1月16日

查看全文
  • D. Pradeep Kumar,
  • Syed Shaul Hameed,
  • V. Muralidharan,
  • S. Ravikumar,
  • Bernadatta Kwintiana

摘要

This study presents a vibration-based tool condition monitoring approach for face milling processes under varying cutting parameters. The analysis involved computing thirty features (ten statistical and twenty histogram features) from Continuous Wavelet Transform (CWT) coefficients using five wavelet families ‘Daubechies (db), Biorthogonal (bior), Reverse Biorthogonal (rbior), Symlet (sym), and Coiflet (coif)’ across sixty-four levels. The best levels of highly performing wavelet member from each wavelet family was identified using the J48 algorithm. They were rbior1.5 at level 14, bior5.5 at level 13, db8 at level 3, sym6 at level 3 and coif4 at level 13. The prominent features of the identified wavelet members at their respective levels were selected using decision tree and inputted to support vector machine (SVM) and K-star algorithms. The results show that the K-star outperformed the SVM with the CWT feature. The performance of other ML algorithms, namely, random forest, k-NN, Naïve Bayes, and Bayes net algorithms were also analysed. The findings highlight the potential adoption of this approach in manufacturing industries to optimize tool usage while ensuring product surface quality.

Graphical abstract