Forcasting of an ANN model for predicting behaviour of diesel engine energised by a combination of two low viscous biofuels.

This study is focused on artificial neural network (ANN) modelling of non-modified diesel engine keyed up by the combination of two low viscous biofuels to forecast the parameters of emission and performance. The diesel engine is energised with five different test fuels of the combination of citronella and Cymbopogon flexuous biofuel (C50CF50) with diesel at precise blends of B20, B30, B40, B50 and B100 in which these numbers represent the contents of combination of biofuel and the investigation is carried out from zero to full load condition. The experimental result was found that the B20 blend had improved BTE at all load states compared with the remaining biofuel blends. At 100% load state, BTE (31.5%) and fuel consumption (13.01 g/kW-h) for the B20 blend was closer to diesel. However, the B50 blend had minimal HC (0.04 to 0.157 g/kW-h), CO (0.89 to 2.025 g/kW-h) and smoke (7.8 to 60.09%) emission than other test fuels at low and high load states. The CO emission was the penalty for complete combustion. The NO emission was higher for all the biodiesel blends than diesel by 6.12%, 8%, 11.53%, 14.81% and 3.15% for B20, B30, B40, B50 and B100 respectively at 100% load condition. The reference parameters are identified as blend concentration percentage and brake power values. The trained ANN models exhibit a magnificent value of 97% coefficient of determination and the high R values ranging between 0.9076 and 0.9965 and the low MAPE values ranging between 0.98 and 4.26%. The analytical results also provide supportive evidence for the B20 blend which in turn concludes B20 as an effective alternative fuel for diesel.