Intelligent prediction and multi-objective optimisation of CI engine performance and emissions using SiO2-enriched waste-derived biofuel

The present work investigates the performance and emission characteristics of a compression ignition (CI) engine fueled with a biofuel produced from Prosopis juliflora and enhanced with irregularly shaped SiO₂ nanoparticles, coupled with an intelligent machine learning–based optimization framework. The hybrid attention neuroevolutionary combustion intelligent network (HANCINet) developed in this study was employed to model the nonlinear combustion behavior of the engine. The input data were obtained from previously published open-access hybrid datasets of CI engines operating on waste-derived biofuels enriched with SiO₂ nanoparticles. The proposed model achieved high prediction accuracy, with coefficient of determination (R²) values of 0.910, 0.984, 0.986, 0.974, and 0.986 for brake thermal efficiency (BTE), carbon monoxide (CO), hydrocarbon (HC), nitrogen oxides (NOx), and smoke emissions, respectively. In addition, low prediction errors were obtained, including RMSE values of 0.863 for BTE and 0.005 for CO emissions. The experimental results indicate that the incorporation of SiO₂ nanoparticles enhances combustion efficiency while reducing CO and HC emissions, with only a moderate increase in NO x emissions. Furthermore, a comparative analysis with conventional machine learning models, including linear regression (LR), random forest (RF), and support vector regression (SVR), demonstrated that HANCINet provides competitive and balanced prediction performance. A Pareto-based multi-objective optimization framework was subsequently employed to identify optimal operating conditions that achieve an effective trade-off between engine efficiency and exhaust emissions. The findings demonstrate that the integration of nano-biofuel technology with intelligent combustion modeling offers a promising pathway toward sustainable and high-performance next-generation CI engine applications.