Assessing the Predicting Capability of RSM and ANN on Transesterification Process for Yielding Biodiesel from Vitis vinifera Seed Oil

Abstract

In this present investigation, the process parameters to obtain maximum fatty acid methyl ester yield from Vitis vinifera seed bio-oil by transesterification were explored using the central composite design with variable input parameters like catalyst concentration (0.5–1.5% of KOH), reaction duration (30–60 min), and molar ratio (3:1–7:1). Response surface methodology (RSM) and artificial neural network (ANN) were employed to predict the optimized biodiesel yield and model the transesterification process. The experimental outputs were simulated using a quadratic model generated by RSM. The maximum biodiesel yield parameters were determined by RSM, and it was found to be 6.4246:1 molar ratio, 66.8205 min reaction time, and 1.1719% of catalyst concentration. The transesterification process performed with this experimental combination resulted in methyl ester yield of around 97.53% which correlated well with the yield predicted by RSM. The statistical analysis was carried out to determine the model validity, accuracy, and predictive capability of both ANN and RSM models. The biodiesel obtained by this process was subjected to analysis for estimating the physiochemical properties like cetane number, calorific value, density, acid value, flash and fire point, and kinematic viscosity, and it was found to be within ASTM limits.