A Machine Learning Approach for Carbon di oxide and Other Emissions Characteristics Prediction in a Low Carbon Biofuel-Hydrogen Dual Fuel Engine

Abstract

To lower the carbon dioxide and other emissions from a single cylinder common rail direct injection (CRDI) engine, it is important to investigate the combinations of several methods. Lemon peel oil (LPO) and camphor oil (CMO), which are low carbon content biofuels, are the methods that are used and are induced by hydrogen in the intake manifold and zeolite-based after-treatment system. At full load, the injection of hydrogen decreased CO2 and smoke emissions by 39.7% and 49%, respectively. Even though the NO emission increases with hydrogen induction, it was decreased with zeolite after-treatment system. Predictions can be made using machine learning techniques, which will reduce the amount of time and money needed for engine trials. This work focuses on the prediction of engine emissions like CO2, Nitrogen Oxides (NO), Smoke, Brake Thermal Efficiency (BTE), Hydrocarbons (HC) using the ensemble learning techniques. The predictions are made using the ensemble learning methods like Extreme Gradient Boosting (XGBoost), Light Gradient Boosted Machine (LGBM), CatBoost, and Random Forest (RF). The CatBoost model has produced high accuracy predictions which was followed by XGBoost, RF and LightGBM models. The predicted and actual values are compared each other and the performance of the algorithms were analysed using the evaluation metrics like R-Square(R2), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE) and Root Mean Square Error (RMSE).