Analysis of combustion characteristics of a diesel engine run on ternary blends using machine learning algorithms

Diesel engines are vital in industries, such as transportation, agriculture, and power generation. Enhancing fuel efficiency and reducing emissions in these engines are critical goals, and machine learning (ML) techniques offer novel solutions for achieving them. This study investigates the use of ML models, specifically random forest regression (RFR) and polynomial regression (PR) to predict key combustion characteristics, namely cylinder pressure and heat release rate (HRR), in a dual-fuel diesel engine powered by a ternary blend (TB) and acetylene. The experimental setup involved modifying a conventional diesel engine to operate in dual-fuel mode, using a TB fuel comprising 70% diesel, 20% waste cooking oil biodiesel (WCOB), and 10% methanol by volume. The cylinder head, piston crown, and intake/exhaust valves were coated with partially stabilized zirconia (PSZ) to improve combustion efficiency. The RFR model achieved an impressive R² score of 0.9987 for cylinder pressure and 0.9878 for HRR predictions, with corresponding mean absolute error (MAE) values of 0.124 and 0.021, indicating high predictive accuracy and minimal deviation from experimental values. The PR model, while capturing some nonlinear trends, performed less reliably, with R² scores of 0.7689 for cylinder pressure and 0.6720 for HRR. These results underscore the RFR model's robustness in predicting complex combustion behavior, offering a reliable, cost-effective approach to optimize fuel efficiency and emission reduction in diesel engines. The findings suggest that implementing ML models, particularly RFR, can aid in tuning engine parameters for sustainable fuel blends, contributing to reduced greenhouse gas emissions and improved fuel efficiency.