Scholarlink

<p>Monocrystalline silicon is widely applied as precision lenses in semi-conductor, optical, medical and electronic industries. To meet the service applications, precision grinding process is conducted to have excellent surface quality by generating massively striated ductile surfaces with shortened polishing time. However, predicting the process surface quality is difficult because of the stochastic distribution of abrasive grains, impacting its surfaces while inducing both ductility and brittleness. This necessitates characterizing the process surface quality by multivariate surface roughness parameters. In this study, the optimization of average roughness (<i>R</i><sub><i>a</i></sub>) and maximum heights (<i>R</i><sub><i>max</i></sub> and <i>R</i><sub><i>t</i></sub>) with respect to process parameters, depth of cut, feed rate and cutting speed is investigated. Taguchi’s L<sub>9</sub> orthogonal array is used for experimentation, while the principal component analysis-based unsupervised machine learning is applied for optimization. The optimum precision grinding condition, 20 µm depth of cut, 10 mm/min feed rate and 1696.50 m/mm cutting speed predict minimum values of <i>R</i><sub><i>a</i></sub>, <i>R</i><sub>max</sub> and <i>R</i><sub><i>t</i></sub> as 0.018, 0.312 and 0.934 µm, respectively. Cutting speed has the most effect on the surface roughness followed by the feed rate and then depth of cut. Validation experiments confirm <i>R</i><sub><i>a</i></sub>, <i>R</i><sub>max</sub> and <i>R</i><sub><i>t</i></sub> as 0.020, 0.350 and 0.850 µm, representing errors ranging from 9 to 12%. Finally, the optimum condition can be implemented in industries for cosmetic purposes because it meets the optical and ophthalmic surface quality requirements.</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

Simultaneous optimization of multivariate surface roughness parameters in precision grinding of silicon by unsupervised machine learning

  • 期刊论文

  • 发表日期：

    2025年7月19日

查看全文
  • Abdur-Rasheed Alao

摘要

Monocrystalline silicon is widely applied as precision lenses in semi-conductor, optical, medical and electronic industries. To meet the service applications, precision grinding process is conducted to have excellent surface quality by generating massively striated ductile surfaces with shortened polishing time. However, predicting the process surface quality is difficult because of the stochastic distribution of abrasive grains, impacting its surfaces while inducing both ductility and brittleness. This necessitates characterizing the process surface quality by multivariate surface roughness parameters. In this study, the optimization of average roughness (Ra) and maximum heights (Rmax and Rt) with respect to process parameters, depth of cut, feed rate and cutting speed is investigated. Taguchi’s L9 orthogonal array is used for experimentation, while the principal component analysis-based unsupervised machine learning is applied for optimization. The optimum precision grinding condition, 20 µm depth of cut, 10 mm/min feed rate and 1696.50 m/mm cutting speed predict minimum values of Ra, Rmax and Rt as 0.018, 0.312 and 0.934 µm, respectively. Cutting speed has the most effect on the surface roughness followed by the feed rate and then depth of cut. Validation experiments confirm Ra, Rmax and Rt as 0.020, 0.350 and 0.850 µm, representing errors ranging from 9 to 12%. Finally, the optimum condition can be implemented in industries for cosmetic purposes because it meets the optical and ophthalmic surface quality requirements.