Machine learning based prediction of diesel engine emissions and performance using hemp biodiesel enriched with nano additives.

The present study was focused on the experimental and machine learning approach to evaluate the diesel engine parameters behavior with hemp biodiesel blends enriched with nano additives (Al2O3, TiO2 and MWCNT's) at different loads. The Hemp biodiesel was characterized by using FTIR and GC-MS analysis. The nano additives were examined through a SEM and XRD analysis for identifying the morphological structure and crystalline phases. The experimental results revealed that HMBD20 + MWCN100 fuel shows higher BTE (5.60%) and lower BSFC (18.58%) compared to HMBD20 fuel. The CO and HC pollutants for HMBD20 + TINP100 fuel were diminished by 22.24% and 16.12% but NOx emissions were enhanced by 13.7% than HMBD20 fuel at peak load. For predictive modelling, Decision Tree (DT), Support Vector Machine (SVM) and Artificial Neural Network (ANN) were employed using the engine load and fuel type as input features. The DT model exhibits a higher prediction accuracy of R2 0.9899 & 0.998 and lowest RMSE of 1.085 & 0.0264 for BTE and BSFC data. The greater accuracy of R2 of 0.9865, 0.9899 and 0.9975 were achieved by DT model for CO, HC and Smoke emissions but NOx emissions were predicted by 0.9997 by using ANN model. The optimal condition was recorded for HMBD20 + 100 ppm MWCNT fuel at 75% load, achieving BTE (32.56%), BSFC (0.30 kg/kWhr), CO (0.12%), HC (42 ppm), NOx (980 ppm), Smoke opacity (35.2 HSU) and CO2 (7.48%) which confirms the strong potential of nano additive hemp biodiesel for efficient and cleaner operation of a diesel engine.