Scholarlink

<p>Corona plasma treatment is increasingly adopted in advanced manufacturing as a scalable and environmentally benign surface engineering technique for functional textiles. However, the strong nonlinearity between plasma processing parameters and surface modification outcomes remains a critical challenge for reliable process control. In this study, a machine-learning-assisted framework is proposed to optimize low-temperature corona plasma treatment of ramie nonwoven fabrics as a model natural fiber system. Plasma exposure time and electrode distance were systematically varied, and plasma-induced weight loss was employed as a quantitative indicator of surface activation intensity. Experimental characterization using scanning electron microscopy and Fourier transform infrared spectroscopy confirmed that weight loss correlates with surface roughening and the formation of oxygen-containing functional groups. Artificial neural network (ANN), support vector regression (SVR), and random forest (RF) models were developed to predict plasma-induced weight loss and to identify optimal processing conditions. Among the evaluated models, RF exhibited the highest agreement with the experimental data (R² = 0.85) and demonstrated the closest correspondence between predicted and experimentally validated weight loss values under the optimized processing conditions, outperforming ANN (R² = 0.84) and SVR (R² = 0.80), owing to its superior capability in capturing interaction-driven and nonlinear plasma-material behavior. The optimized plasma conditions were subsequently applied to facilitate the immobilization of a natural antibacterial agent, serving as a functional validation of the surface activation process. Fabrics treated under the machine-learning-assisted optimized conditions exhibited enhanced antibacterial activity against <i>Escherichia coli</i>, with an inhibition zone diameter of 3.40 ± 0.50&#xa0;cm. The results demonstrate that integrating machine learning with corona plasma processing provides a practical framework for identifying promising operating regions for surface modification in nonwoven manufacturing. The experimentally validated optimization results highlight the potential of data-driven approaches to support intelligent plasma process design under constrained experimental conditions.</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

Machine-learning-assisted optimization of corona plasma surface engineering for functional nonwoven manufacturing

  • 期刊论文

  • 发表日期：

    2026年6月30日

查看全文
  • Markus Paramahasti,
  • Eri Widianto,
  • Valentinus Galih Vidia Putra,
  • Yusril Yusuf

摘要

Corona plasma treatment is increasingly adopted in advanced manufacturing as a scalable and environmentally benign surface engineering technique for functional textiles. However, the strong nonlinearity between plasma processing parameters and surface modification outcomes remains a critical challenge for reliable process control. In this study, a machine-learning-assisted framework is proposed to optimize low-temperature corona plasma treatment of ramie nonwoven fabrics as a model natural fiber system. Plasma exposure time and electrode distance were systematically varied, and plasma-induced weight loss was employed as a quantitative indicator of surface activation intensity. Experimental characterization using scanning electron microscopy and Fourier transform infrared spectroscopy confirmed that weight loss correlates with surface roughening and the formation of oxygen-containing functional groups. Artificial neural network (ANN), support vector regression (SVR), and random forest (RF) models were developed to predict plasma-induced weight loss and to identify optimal processing conditions. Among the evaluated models, RF exhibited the highest agreement with the experimental data (R² = 0.85) and demonstrated the closest correspondence between predicted and experimentally validated weight loss values under the optimized processing conditions, outperforming ANN (R² = 0.84) and SVR (R² = 0.80), owing to its superior capability in capturing interaction-driven and nonlinear plasma-material behavior. The optimized plasma conditions were subsequently applied to facilitate the immobilization of a natural antibacterial agent, serving as a functional validation of the surface activation process. Fabrics treated under the machine-learning-assisted optimized conditions exhibited enhanced antibacterial activity against Escherichia coli, with an inhibition zone diameter of 3.40 ± 0.50 cm. The results demonstrate that integrating machine learning with corona plasma processing provides a practical framework for identifying promising operating regions for surface modification in nonwoven manufacturing. The experimentally validated optimization results highlight the potential of data-driven approaches to support intelligent plasma process design under constrained experimental conditions.