Scholarlink

<p>Estimating cylinder pressure (CP) in internal combustion engines (ICEs) is essential for analyzing and optimizing engine performance. CP is a fundamental parameter affecting thermodynamic efficiency, emission levels, and overall engine performance. Since traditional CP measurement methods are time-consuming and costly, alternative methods capable of making fast and accurate estimations under different operating conditions and fuel mixtures are needed. In this context, this study aims to estimate CP using data from engine experiments with fuel mixtures that include varying ratios of biodiesel (from waste cooking oil) and fusel oil blended with diesel fuel, utilizing machine learning (ML) models. Engine experiments were conducted with the prepared fuel mixtures, and CP data were measured. Hyperparameter-optimized deep learning-based multilayer perceptrons (Deep-MLP), decision tree regression (DTR), and random forest regression (RFR) models were applied to estimate CP. The results indicate that all ML models performed strongly, demonstrating R² values ranging from 0.9946 to 0.9996 across all test cases. Based on its consistently high R² values and minimum MAPE value, the Deep-MLP model produces more sensitive results that better capture the specific characteristics of individual fuel types. The RFR model exhibited more balanced performance across all fuel varieties, demonstrating the most consistent results. Though showing comparatively higher error rates, the DTR model produced competitive results with high R² values. This study demonstrates that ML models can effectively predict cylinder pressure in ICEs under various operating conditions and have potential applications in anomaly detection and predictive maintenance strategies.</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 cylinder pressure estimation using newly developed biodiesel–fusel oil mixtures in diesel engines

  • 期刊论文

  • 发表日期：

    2025年9月2日

查看全文
  • Kürşat Mustafa Karaoğlan,
  • Burak Çiftçi,
  • Mustafa Karagöz,
  • Mustafa Bahattin Çelik

摘要

Estimating cylinder pressure (CP) in internal combustion engines (ICEs) is essential for analyzing and optimizing engine performance. CP is a fundamental parameter affecting thermodynamic efficiency, emission levels, and overall engine performance. Since traditional CP measurement methods are time-consuming and costly, alternative methods capable of making fast and accurate estimations under different operating conditions and fuel mixtures are needed. In this context, this study aims to estimate CP using data from engine experiments with fuel mixtures that include varying ratios of biodiesel (from waste cooking oil) and fusel oil blended with diesel fuel, utilizing machine learning (ML) models. Engine experiments were conducted with the prepared fuel mixtures, and CP data were measured. Hyperparameter-optimized deep learning-based multilayer perceptrons (Deep-MLP), decision tree regression (DTR), and random forest regression (RFR) models were applied to estimate CP. The results indicate that all ML models performed strongly, demonstrating R² values ranging from 0.9946 to 0.9996 across all test cases. Based on its consistently high R² values and minimum MAPE value, the Deep-MLP model produces more sensitive results that better capture the specific characteristics of individual fuel types. The RFR model exhibited more balanced performance across all fuel varieties, demonstrating the most consistent results. Though showing comparatively higher error rates, the DTR model produced competitive results with high R² values. This study demonstrates that ML models can effectively predict cylinder pressure in ICEs under various operating conditions and have potential applications in anomaly detection and predictive maintenance strategies.