Scholarlink

<p>Conventional roughness parameters capture only a limited portion of a surface’s multiscale structure, which restricts their ability to describe how machined surfaces evolve under different conditions. This study introduces a continuous wavelet-based descriptor, obtained through reconstruction driven multiscale decomposition to characterize surface features from fine roughness to coarse waviness. High resolution chromatic confocal scans of milled stainless steel surfaces were collected under controlled variations in tool condition and two feed rates, providing a dataset with diverse multiscale content for validating the method. The SMa(a) curves reveal clear scale dependent changes in surface structure that are not detected by single valued roughness metrics. To address the computational cost of wavelet reconstruction, a deep learning surrogate model is proposed to estimate the SMa spectrum directly from surface patches. The model captures the global trend of the multiscale response with good fidelity, offering a practical and efficient alternative to the proposed wavelet analysis. The study offers both a multiscale analysis framework and a step-by-step methodological basis, from the fundamentals to culminating in an efficient CWT implementation for surface characterization. Overall, the proposed framework enables scalable and interpretable multiscale surface characterization suitable for tribological studies, where a multiscale characterization of surface topography is critical to understanding contact, lubrication, and wear behavior.</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

Accelerated 2D continuous wavelet transform for scale-resolved surface evolution across tool wear states in stainless steel milling

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年3月16日

查看全文
  • Sourish Ghosh,
  • Ricardo Knoblauch,
  • Mohamed El Mansori,
  • Cosimi Corleto

摘要

Conventional roughness parameters capture only a limited portion of a surface’s multiscale structure, which restricts their ability to describe how machined surfaces evolve under different conditions. This study introduces a continuous wavelet-based descriptor, obtained through reconstruction driven multiscale decomposition to characterize surface features from fine roughness to coarse waviness. High resolution chromatic confocal scans of milled stainless steel surfaces were collected under controlled variations in tool condition and two feed rates, providing a dataset with diverse multiscale content for validating the method. The SMa(a) curves reveal clear scale dependent changes in surface structure that are not detected by single valued roughness metrics. To address the computational cost of wavelet reconstruction, a deep learning surrogate model is proposed to estimate the SMa spectrum directly from surface patches. The model captures the global trend of the multiscale response with good fidelity, offering a practical and efficient alternative to the proposed wavelet analysis. The study offers both a multiscale analysis framework and a step-by-step methodological basis, from the fundamentals to culminating in an efficient CWT implementation for surface characterization. Overall, the proposed framework enables scalable and interpretable multiscale surface characterization suitable for tribological studies, where a multiscale characterization of surface topography is critical to understanding contact, lubrication, and wear behavior.