Prediction of Cold Flow Properties of Biodiesel Fuel Using Artificial Neural Network

Abstract

Artificial neural network (ANN) can be utilized as a tool for modeling the properties of biodiesel fuel those are related to the fatty acid (FA) composition of a feedstock. The cold flow properties (CFP) define the operability for diesel fuel which are strongly influenced by the FA composition of feedstock. Cloud point (CP), pour point (PP) and cold filter plugging point (CFPP) are used commonly to characterize CFP. Prediction of CFP based on the FA composition of feedstock can reduce the experimental effort to produce a biodiesel suitable for a regional climate. In an attempt for this, 9-6-3 back-propagation ANN architecture was implemented to estimate CP, PP and CFPP of biodiesel samples using nine FA components as input data of 103 biodiesel study collected from literature. To check the accuracy of the model developed, refined canola oil (RCO) and waste frying oil (WFO) were converted to biodiesel then, their CP, PP and CFPP temperatures were determined following the EN and ASTM standards. The CFP estimated by the ANN model were in close agreement with the experimental values. When compared with the experimental data, ANN model predicted the CP, PP and CFPP temperatures within 98%, 94% and 96% accuracy, respectively. The model developed has revealed that CFP of biodiesel were influenced primarily by saturation or unsaturation of FA components with a few exceptions. Since the ANN model can be trained from iterations, it predicted CFP with high accuracy inspite of the presence of nonlinearities, i.e. the average of the mean R2 value for the three CFP was found as 0.96.