Scholarlink

<p>One ton of sugarcane yields approximately 130&#xa0;kg of dry bagasse, a fibrous byproduct. It serves as an effective feedstock for microbial bioethanol production. To enhance its usability, the chemical pretreatment with 1&#xa0;M sodium hydroxide achieves a hydrolysis rate of 35.68%. For ethanol fermentation, 30&#xa0;g of the pretreated sugarcane bagasse was used per 100 mL of medium, along with cellulase enzymes immobilized in 6.5% calcium alginate beads. The fermentation medium was enriched with yeast extract at a concentration of 10&#xa0;g/L to serve as a nitrogen source, while fermentation was done using <i>Saccharomyces cerevisiae</i>. Following 24&#xa0;h of fermentation, alcohol production commenced and reached its peak after 72&#xa0;h. At this point, the alcohol concentration was determined to be 8.1%, with a productivity rate of 1.14&#xa0;ml/l/h. The impact of key process parameters – including pH (5.0), incubation duration (72&#xa0;h), inoculum volume (10 mL/L), and substrate concentration (40&#xa0;g/100 mL)- and their interactive effects on ethanol yield was assessed. The experimental setup followed a central composite design (CCD) under the principles of response surface methodology (RSM). The robustness and validity of the developed model was evaluated through an F-test, which yielded notably high value of 92.362, confirming the model’s strong statistical significance. The accuracy of the experimental model was further confirmed through validation with laboratory results. Three machine learning algorithm models, viz., SVM, KNN and RF, for the prediction of ethanol yield were developed. These models were tested for their accuracy and adequacy. The simulation results obtained from these models indicated the value of R<sup>2</sup> and RMSE for training data as 0.97 and 0.1815 for SVM model, 0.98 and 0.1446 for KNN model, and 0.98 and 0.1429, respectively. The findings indicated that the KNN model is capable of predicting ethanol yield with a minimal margin of error.</p> Graphical Abstract <p></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

Statistical Machine Learning Models for Carbohydrate-Based Bioethanol Production from Sugarcane Bagasse: Optimization and Experimental Validation

  • 期刊论文

  • 发表日期：

    2025年10月21日

查看全文
  • Fouziya Parveen,
  • Ayush Saxena,
  • Akhtar Hussain,
  • Lamya Ahmed Al-Keridis,
  • Mohd Saeed,
  • Nadiyah M. Alabdallah,
  • Safia Obaidur Rab,
  • Nawaf Alshammari,
  • Mohammad Ashfaque

摘要

One ton of sugarcane yields approximately 130 kg of dry bagasse, a fibrous byproduct. It serves as an effective feedstock for microbial bioethanol production. To enhance its usability, the chemical pretreatment with 1 M sodium hydroxide achieves a hydrolysis rate of 35.68%. For ethanol fermentation, 30 g of the pretreated sugarcane bagasse was used per 100 mL of medium, along with cellulase enzymes immobilized in 6.5% calcium alginate beads. The fermentation medium was enriched with yeast extract at a concentration of 10 g/L to serve as a nitrogen source, while fermentation was done using Saccharomyces cerevisiae. Following 24 h of fermentation, alcohol production commenced and reached its peak after 72 h. At this point, the alcohol concentration was determined to be 8.1%, with a productivity rate of 1.14 ml/l/h. The impact of key process parameters – including pH (5.0), incubation duration (72 h), inoculum volume (10 mL/L), and substrate concentration (40 g/100 mL)- and their interactive effects on ethanol yield was assessed. The experimental setup followed a central composite design (CCD) under the principles of response surface methodology (RSM). The robustness and validity of the developed model was evaluated through an F-test, which yielded notably high value of 92.362, confirming the model’s strong statistical significance. The accuracy of the experimental model was further confirmed through validation with laboratory results. Three machine learning algorithm models, viz., SVM, KNN and RF, for the prediction of ethanol yield were developed. These models were tested for their accuracy and adequacy. The simulation results obtained from these models indicated the value of R2 and RMSE for training data as 0.97 and 0.1815 for SVM model, 0.98 and 0.1446 for KNN model, and 0.98 and 0.1429, respectively. The findings indicated that the KNN model is capable of predicting ethanol yield with a minimal margin of error.

Graphical Abstract