Scholarlink

<p>Surface roughness of heat-treated and sanded wood is a crucial quality indicator in the furniture and woodworking industries. Predicting surface roughness in advance can reduce costs, improve production efficiency, and ensure better compatibility with subsequent material and coating applications, thereby improving overall quality and consistency while supporting more sustainable production. In this study, Siberian pine wood was heat-treated at different temperatures and subsequently sanded on a belt sander using various grit sizes. Taguchi design and Response Surface Methodology (RSM) were first employed to identify and quantify the significance of process parameters affecting surface roughness. Based on these results, several machine learning-based models, including Decision Tree (DT), Gaussian Process Regression (GPR), Multilayer Perceptron (MLP), Random Forest (RF), and Support Vector Machine (SVM), together with RSM, were developed to predict the roughness values of heat-treated and sanded wood. Heat treatment temperature, sandpaper grit size, and grain direction were considered as input variables for the prediction of Ra and Rz surface roughness parameters. Taguchi and RSM analyses revealed that all three factors have a statistically significant effect on surface roughness. Among the tested models, GPR achieved the highest prediction accuracy and was identified as the most suitable approach for modelling the roughness performance of sanded, heat-treated wood surfaces.</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-based prediction of surface roughness in heat-treated and sanded Siberian pine wood

  • 期刊论文

  • 发表日期：

    2026年1月16日

查看全文
  • Mehmet Güneş

摘要

Surface roughness of heat-treated and sanded wood is a crucial quality indicator in the furniture and woodworking industries. Predicting surface roughness in advance can reduce costs, improve production efficiency, and ensure better compatibility with subsequent material and coating applications, thereby improving overall quality and consistency while supporting more sustainable production. In this study, Siberian pine wood was heat-treated at different temperatures and subsequently sanded on a belt sander using various grit sizes. Taguchi design and Response Surface Methodology (RSM) were first employed to identify and quantify the significance of process parameters affecting surface roughness. Based on these results, several machine learning-based models, including Decision Tree (DT), Gaussian Process Regression (GPR), Multilayer Perceptron (MLP), Random Forest (RF), and Support Vector Machine (SVM), together with RSM, were developed to predict the roughness values of heat-treated and sanded wood. Heat treatment temperature, sandpaper grit size, and grain direction were considered as input variables for the prediction of Ra and Rz surface roughness parameters. Taguchi and RSM analyses revealed that all three factors have a statistically significant effect on surface roughness. Among the tested models, GPR achieved the highest prediction accuracy and was identified as the most suitable approach for modelling the roughness performance of sanded, heat-treated wood surfaces.