Machine Learning Approaches for Predicting Diesel engine Emissions Using Waste tire Pyrolysis Oil – Hydrotreated Vegetable Oil Blends

Abstract

This experimental study explores the use of blended Tire Pyrolysis Oil (TPO) with Hydrotreated Vegetable Oil (HVO) as potential substitutes for diesel fuel in compression ignition engines. The assessment of the investigation on the three blends of TPO with HVO as the varying %vol addition as 15 %, 30 % and 60 % under various engine load conditions. The performance and emission characteristics are compared with the neat fuels like diesel oil, HVO and neat TPO for conclusive results. Further to enhance the analysis and reduce dependency on extensive physical testing, machine learning (ML) techniques were employed to model and predict engine out emissions. Four machine learning models including Linear Regression (LR), k-Nearest Neighbors (KNN), Random Forest (RF), and Gradient Boosting (GB) were developed to estimate these outputs. The performance of the models was evaluated using R², Root Mean Square Error (RMSE), and Mean Absolute Error (MAE). Among the models, Random Forest achieved the lowest RMSE, MAE values and highest R² across the target variables, followed by Gradient Boosting, while KNN and Linear Regression demonstrated relatively lower R² and higher errors. The findings emphasize the strength of ensemble-based models in accurately predicting engine behaviour under varying fuel conditions. The experimental results shows that the engine operations with HVO as the working fuel has improved brake thermal efficiency of 30.8 % with reduced emission formation. The addition of 15 % vol of TPO with HVO also has the improved thermal efficiency of 28.2 % and with the consistent increase of TPO with HVO as 30 and 60 %vol the brake thermal efficiency tends to decrease. The integration of experimental data with machine learning provides a valuable framework for optimizing alternative fuel usage in diesel engines, contributing to more sustainable energy systems.