Waste tyre pyrolysis oil-hemp biodiesel ternary blends with graphene oxide and thermal barrier coating: CI engine performance and machine learning prediction.

This research investigates the synergistic effects of the 10% waste tyre pyrolysis oil, 10% hemp seed biodiesel, and 80% diesel blend in addition to graphene oxide nano additive (25, 50, 75, 100 ppm) and a partially stabilised zirconia thermal barrier coating on a Kirloskar TV1 single-cylinder, four-stroke CI engine (5.2 kW, CR 17.5:1, 1500 rpm). Six response variables were investigated: brake thermal efficiency (BTE), brake-specific fuel consumption (BSFC), unburned hydrocarbons (HC), carbon monoxide (CO), nitrogen oxides (NOx), and smoke opacity at a load of 0% to 100%. The B20GO50 has peaked in BTE at 35.36% with an increase of 4.12% against diesel and a BSFC of 0.25 kg/kWh, which is lesser by 7.4%. The B20GO50 + TBC configuration achieved a BTE of 36.22% (+ 6.65%), CO of 0.15% (- 40%), HC of 32 ppm (- 25.6%) and around a 20% reduction in smoke. NOx emission was maintained at 1010 ppm, which shows an increment of 2.4% compared to the neat diesel at full load. The dataset was partitioned into 80% training (n = 58) and 20% testing (n = 14) subsets using stratified random sampling. Nine regression algorithms were evaluated under five-fold cross-validation with systematic hyperparameter tuning. Nine regression algorithms in ML were trained on a dataset that contained ten distinctive features. The algorithm set includes XGBoost, Gradient Boosting (GB), Random Forest (RF), Linear Regression, Ridge, Lasso, Elastic Net, Support Vector Regression (SVR) and K-Nearest Neighbours (KNN). The highest R2 across all models was 0.9905 (RMSE = 7.05) for XGBoost, followed by GB with R2 = 0.9888 and RF with R2 = 0.9714. SHAP analysis indicated that engine load was by far the most predominant feature affecting BTE, BSFC and the CO and NOx emissions. GO and TBC were the main determining variables for HC and smoke. Overall, the results prove XGBoost as the most efficient multi-output model for characterisation of CI engines, particularly in terms of complex interactions involved between fuel, additives, and hardware. The B20GO50 + TBC configuration achieved a BTE of 36.22%, an absolute gain of 6.65% points over neat diesel at full load, yielding an estimated fuel cost saving of INR 6000-7800 per 1000 operating hours at current diesel prices. Concurrently, CO was reduced by 40% and HC by 25.6%, demonstrating regulatory compliance potential under BS-VI emission norms without engine hardware modification.