Comprehensive performance improvement of methanol/diesel dual-fuel engines using multi-physics field coupling simulation and surrogate-assisted many-objective optimization

Improving the comprehensive performance of methanol/diesel dual-fuel engines (MDDFEs) remains challenging due to the strong coupling of multi-physics fields and high computational cost of optimization. To address this issue, a dual-fidelity multi-physics coupling framework was developed for an MDDFE with Jet Controlled Compression Ignition (JCCI) in this study, combining a full-cylinder model for detailed physical representation with a one-sixth sector model for efficient data generation. Based on Latin Hypercube Sampling (LHS) and the dual-fidelity simulations, a Light Gradient Boosting Machine (LightGBM) surrogate was trained to predict engine performance precisely. Subsequently, the many-objective many-population hybrid genetic algorithm (MMHGA) was employed to simultaneously optimize economic and environmental indicators. The results indicate that LightGBM model achieves superior predictive performance, with coefficients of determination (R2) of 0.9691, 0.9632, 0.9579, 0.9772, 0.9808 and 0.9709 for indicated specific fuel consumption (ISFC), Soot, nitrogen oxides (NOx), hydrocarbon (HC), carbon monoxide (CO) and carbon dioxide (CO2), respectively. Furthermore, the Optimal Total solution delivers simultaneous reductions of 11.86 %, 15.06 %, 55.65 %, 12.71 %, 74.49 % and 11.88 % in ISFC, Soot, NOx, HC, CO and CO2. Flow-field diagnostics further reveals markedly more homogeneous in-cylinder temperature and equivalence ratio distributions, with suppressed hot spots and attenuated gradients, thereby enhancing combustion efficiency and stability while reducing various emissions. These findings demonstrate a computationally efficient and practically feasible pathway for many-objective optimization of MDDFE comprehensive performance.