Machine learning-based prediction of diesel engine performance fuelled with waste cooking oil biodiesel blends

The current research investigates the performance and emission characteristics of a single-cylinder diesel engine fueled with biodiesel blends of waste cooking oil in various proportions (WCBD10, WCBD20, and WCBD30). To improve the accuracy of predictions and to optimize engine settings, advanced machine learning algorithms like Random Forest Regressor, XGBoost and Support Vector Regressor were employed. The reliability of the model was assessed using the R-squared (R²) statistical method. Experiments were conducted at three different injection timings (20°, 23°, and 26° bTDC) at various engine loads. Engine Performance parameters and Emission characteristics were analysed at different load conditions and injection timings. The WCBD20 fuel sample exhibited superior performance, achieving a peak brake thermal efficiency of 28.16% (experimental) in contrast with 27.58% predicted by XGBoost, with deviations within ±1.5%. The lowest brake specific fuel consumption recorded was 0.325–0.350 kg/kWh at peak loads by WCBD20 at an injection timing of 26° bTDC. The emission tests indicated that hydrocarbon (13.1–15.5 ppm) and carbon monoxide (0.0117–0.0207%) decreased significantly with a moderate increase in oxides of Nitrogen emissions (1,875–2,070 ppm). Machine learning models optimized the engine performance across various blends and injection timings. Random Forest Regressor exhibited strong agreement with experiments (R² = 0.962), identifying WCBD20 at 26° BTDC and 89.5% load as effective. Support Vector Regressor excelled in emission prediction (R² = 0.981 for CO₂), reducing validation needs by 62%. XGBoost emphasized the importance of injection timing as a crucial factor and achieved 27.58% BTE with minimal HC emissions, demonstrating robustness in multi-objective optimization for performance and emissions. The convergence of experimental data and ML-driven prediction determined WCBD20 at 26° BTDC and 85–90% load as the optimal configuration, reducing fossil fuel dependence by 20% while ensuring adherence with emission standards.