Comparative assessment of artificial neural network and response surface methodology for evaluation of the predictive capability on bio-oil yield ofTithonia diversifoliapyrolysis

The present study aims to explore the thermochemical conversion feasibility of an unexplored weed biomass,Tithonia diversifolia. In this investigation, the predictive ability of response surface model (RSM) and artificial neural network (ANN) models for bio-oil yield from pyrolysis ofTithonia Diversifoliawas carried out in a fixed bed reactor. The influence of process parameters which included temperature (375-675 degrees C), heating rate (10-50 degrees C/min), nitrogen flow rate (50-250 mL/min), and particle size (< 0.25- > 1 mm) on bio-oil yield was investigated by central composite design (CCD) of RSM and feedforward backpropagation neural network with different topology. Based on statistical value for coefficient of determination (R-2), mean average error (MAE), root mean square error (RMSE), standard error of prediction (SEP), and absolute average deviation (AAD), it was found that the ANN was more efficient than the RSM model for predicting bio-oil yield. The bio-oil obtained was analyzed for different physical and chemical properties which revealed that the bio-oil can be used as renewable fuel and chemicals.