Scholarlink

<p>This study presents a machine learning-based approach for classifying the performance and emission balance of a dual-fuel (diesel–gasoline) internal combustion engine operating under various conditions. Four different machine learning algorithms were applied during the analysis process: Random Forest, Extreme Gradient Boosting (XGBoost), Support Vector Machines (SVM) and Artificial Neural Networks (ANN). During the data preprocessing stage, numerical variables were standardised and categorical variables were converted using one-hot encoding. Hyperparameter optimization of the models was performed using GridSearchCV. The findings revealed that all models achieved an acceptable level of classification success; however, the SVM model performed best with 99% accuracy and F1-score values. ROC–AUC analyses corroborated these results, demonstrating that the SVM model exhibited perfect discrimination across all three classes. Furthermore, statistical analyses revealed that the SVM model was significantly more successful than the others (<i>p</i> &lt; 0.05). As a result, it has been observed that the proposed approach can accurately classify the performance-emission balance in engine data and contribute to the development of sustainable engine technologies. In subsequent stages of the study, the method’s scope can be expanded to include different fuel types, hybrid combustion strategies and real-time engine data.</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

A Machine Learning Approach to Performance–Emission Analysis in Dual-Fuel Internal Combustion Engines

  • 期刊论文

  • 发表日期：

    2026年7月10日

查看全文
  • Güven Demirtaş,
  • Hakan Yurtkuran,
  • Feyyaz Alpsalaz,
  • Hasan Uzel

摘要

This study presents a machine learning-based approach for classifying the performance and emission balance of a dual-fuel (diesel–gasoline) internal combustion engine operating under various conditions. Four different machine learning algorithms were applied during the analysis process: Random Forest, Extreme Gradient Boosting (XGBoost), Support Vector Machines (SVM) and Artificial Neural Networks (ANN). During the data preprocessing stage, numerical variables were standardised and categorical variables were converted using one-hot encoding. Hyperparameter optimization of the models was performed using GridSearchCV. The findings revealed that all models achieved an acceptable level of classification success; however, the SVM model performed best with 99% accuracy and F1-score values. ROC–AUC analyses corroborated these results, demonstrating that the SVM model exhibited perfect discrimination across all three classes. Furthermore, statistical analyses revealed that the SVM model was significantly more successful than the others (p < 0.05). As a result, it has been observed that the proposed approach can accurately classify the performance-emission balance in engine data and contribute to the development of sustainable engine technologies. In subsequent stages of the study, the method’s scope can be expanded to include different fuel types, hybrid combustion strategies and real-time engine data.