Physics-encoded machine learning for performance and emission prediction of nickel ferrite nanocatalyst and hydrogen-enriched biodiesel in diesel engines

Abstract

This work examined the combined effects of waste cooking oil biodiesel (WCOB), NiFe₂O₄ nanocatalysts, and hydrogen (H₂) enrichment on in-cylinder processes, engine performance, and emissions of a diesel engine. To accurately represent the physical processes and to derive thermodynamically consistent predictions, a physics-encoded multi-task machine learning (PE-MTMML) model was developed. Tests were carried out on a single-cylinder, four-stroke diesel engine at 1500 rpm. The engine was operated on blends of diesel, WCOB, and NiFe₂O₄ (50–150 ppm) with the addition of H₂ at 5 LPM. The analysis of combustion was done based on the measurement of in-cylinder pressure with respect to the crank angle from which peak pressure and heat release rate (HRR) were derived using a single-zone model. Performance parameters were obtained at various loads through the full load range. The PE-MTMML model incorporated not only the thermodynamic reciprocity but also catalytic oxidation trends and root-sum-square (RSS) uncertainty constraints. At full load, the NiFe₂O₄ nanoparticles improved catalytic oxidation and premixed combustion, thus, the peak pressure rose by 7.8% at most, and HRR by 14% compared to diesel. The ensemble of WCOB + NiFe₂O₄ 150 ppm + H₂ led to an increase in brake thermal efficiency (BTE) of 29.2% and a decrease in brake-specific energy consumption (BSEC) of 22.3% in comparison with diesel. Inspection of emissions revealed that the smoke opacity was lessened by 10.9%, carbon monoxide (CO) by 20%, and hydrocarbons (HC) by 25%, while nitrogen oxides (NO_x) were elevated by 7.5% relative to diesel. The PE-MTMML model had excellent prediction capabilities (mean R² = 0.9993) and at the same time adhered to thermodynamic constraints. The synergistic effect of NiFe₂O₄ nanocatalysis and H₂ enrichment makes WCOB both a high-efficiency and a low-pollution fuel. By allowing for optimization with minimal experimental input, the PE-MTMML scheme acts as a trustworthy digital twin, supporting circular-economy goals and the transition to sustainable, carbon-neutral diesel engines.