Scholarlink

<p>Laser drilling is a micromachining process that offers high precision and throughput, but results may vary depending on the material and laser characteristics. Moreover, the uniformity of results often depends on the operator’s experience, even when the same process conditions are applied. In this study, an artificial neural network algorithm was developed to accurately predict the depth of laser-drilled blind holes based on digital images despite various environmental changes such as brightness, gamma value, and exposure time by using scaled grayscale. Grayscale values were scaled from an arbitrary image condition to the actual hole depth using a regression model, and the scaling method utilized gradient data from three-point images under specific conditions. In addition, a method was proposed to improve the uniformity of hole depth such that the geometric deviation was analyzed and compensated via an additional laser post-treatment. Using the proposed model, the hole depth non-uniformity was improved from 2.3 to 1.9%. The present study is expected to improve the accuracy of laser drilling and minimize depth deviations during laser-assisted manufacturing.</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

Development of an Artificial Neural Network-Based Model for Prediction and Compensation of Hole Depth by Femtosecond Laser Drilling

  • 期刊论文

  • 发表日期：

    2025年3月12日

查看全文
  • Myeongjun Kim,
  • Pilgong Choi,
  • Kyunghan Kim,
  • Yun Young Kim

摘要

Laser drilling is a micromachining process that offers high precision and throughput, but results may vary depending on the material and laser characteristics. Moreover, the uniformity of results often depends on the operator’s experience, even when the same process conditions are applied. In this study, an artificial neural network algorithm was developed to accurately predict the depth of laser-drilled blind holes based on digital images despite various environmental changes such as brightness, gamma value, and exposure time by using scaled grayscale. Grayscale values were scaled from an arbitrary image condition to the actual hole depth using a regression model, and the scaling method utilized gradient data from three-point images under specific conditions. In addition, a method was proposed to improve the uniformity of hole depth such that the geometric deviation was analyzed and compensated via an additional laser post-treatment. Using the proposed model, the hole depth non-uniformity was improved from 2.3 to 1.9%. The present study is expected to improve the accuracy of laser drilling and minimize depth deviations during laser-assisted manufacturing.