This study proposes an optimization method that combines a stacking generalization (SG)-based machine learning model with swarm intelligence to enhance the performance of low-speed engines fueled with diesel/biodiesel blends. First, a one-dimensional working process simulation model of the engine fueled by diesel/biodiesel blends was developed and calibrated against experimental data. A systematic analysis was then performed to investigate the influence mechanisms of four combustion control parameters, including intake temperature, injection timing, injection duration, and exhaust valve closing timing, on engine performance. Based on the simulation results, an extensive dataset encompassing a wide range of operating conditions was constructed to support the training of the machine learning model. Subsequently, a high-precision predictive model for the engine performance was established using the SG framework. Following hyperparameter optimization, the model achieved a coefficient of determination of 0.9980 on an independent validation set, with maximum relative errors below 10% across all performance indicators. Finally, the optimized predictive model was coupled with the Harris Hawks Optimization algorithm to perform automated multi-objective optimization. Without modifying the cyclic fuel injection quantity, the optimization yielded a 1.70% increase in brake power, a 2.22% reduction in brake specific fuel consumption, a 2.26% improvement in indicated thermal efficiency, and decreases in NO x and soot emissions by 2.31% and over 90%, respectively. These results demonstrate that the proposed method effectively achieves coordinated optimization of engine power, fuel economy, and emission characteristics, thereby offering a novel technical pathway toward the intelligent and sustainable development of marine power systems.
Sun et al. (Mon,) studied this question.