Scholarlink

<p>With growing environmental concerns and stringent carbon dioxide (CO<sub>2</sub>) emissions regulations, reducing energy-intensive processes’ carbon footprint has become a top priority in sustainable manufacturing. Slicing silicon carbide (SiC) wafers using diamond wire sawing (DWS) presents notable challenges due to the exceptional hardness of SiC and the inherently low material removal rate, which together lead to elevated energy usage and significant CO<sub>2</sub> emissions. This study explores the influence of various cooling strategies on energy consumption and CO<sub>2</sub> emissions during the multi-DWS of the 4H-SiC wafer sawing process. Furthermore, advanced machine learning (ML) models are implemented to predict the CO<sub>2</sub> emissions, integrating variational mode decomposition (VMD) to disaggregate the CO<sub>2</sub> time series data into decomposed modes, and applied hyperparameter optimization based on Bayesian approach with regression algorithms such as SVR, LGBM, GBRT, and RF. The results showed significant improvements, including a 3.87% diminution in overall energy usage and a 6.80% cutback in CO<sub>2</sub> emissions during the sawing process. The integration of VMD significantly improves model accuracy by reducing noise and increasing the reliability of CO<sub>2</sub> emission forecasts. Among all cooling strategies, the electrophoretic-assisted method achieved the most pronounced reductions. Moreover, the hybrid VMD-ML models outperformed single-model approaches, lowering mean absolute error (MAE) by up to 31% and root mean square error (RMSE) by 35%. Single-step ahead prediction horizon demonstrates greater prediction accuracy. Overall, this study confirms the effectiveness of the ER-MDWS cooling strategy in reducing CO<sub>2</sub> emissions and demonstrates the potential of machine learning for data-driven decisions making aimed at lowering the environmental footprint of advanced 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

Sustainable multi-diamond wire sawing of 4H-SiC: cooling strategies and hybrid VMD-ML for CO2 reduction and prediction

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年11月11日

查看全文
  • Eyob Messele Sefene,
  • William Gomez,
  • Chao-Chang Arthur Chen

摘要

With growing environmental concerns and stringent carbon dioxide (CO2) emissions regulations, reducing energy-intensive processes’ carbon footprint has become a top priority in sustainable manufacturing. Slicing silicon carbide (SiC) wafers using diamond wire sawing (DWS) presents notable challenges due to the exceptional hardness of SiC and the inherently low material removal rate, which together lead to elevated energy usage and significant CO2 emissions. This study explores the influence of various cooling strategies on energy consumption and CO2 emissions during the multi-DWS of the 4H-SiC wafer sawing process. Furthermore, advanced machine learning (ML) models are implemented to predict the CO2 emissions, integrating variational mode decomposition (VMD) to disaggregate the CO2 time series data into decomposed modes, and applied hyperparameter optimization based on Bayesian approach with regression algorithms such as SVR, LGBM, GBRT, and RF. The results showed significant improvements, including a 3.87% diminution in overall energy usage and a 6.80% cutback in CO2 emissions during the sawing process. The integration of VMD significantly improves model accuracy by reducing noise and increasing the reliability of CO2 emission forecasts. Among all cooling strategies, the electrophoretic-assisted method achieved the most pronounced reductions. Moreover, the hybrid VMD-ML models outperformed single-model approaches, lowering mean absolute error (MAE) by up to 31% and root mean square error (RMSE) by 35%. Single-step ahead prediction horizon demonstrates greater prediction accuracy. Overall, this study confirms the effectiveness of the ER-MDWS cooling strategy in reducing CO2 emissions and demonstrates the potential of machine learning for data-driven decisions making aimed at lowering the environmental footprint of advanced manufacturing.