Prediction of Kinematic Viscosities of Biodiesels Derived from Edible and Non-edible Vegetable Oils by Using Artificial Neural Networks

In the present study, the seeds named as wild mustard (Sinapis arvensis L.) and safflower (Carthamus tinctorius L.) were used as feedstocks for production of biodiesels. In order to obtain wild mustard seed oil (WMO) and safflower seed oil (SO), screw press apparatus was used. wild mustard seed oil biodiesel (WMOB) and safflower seed oil biodiesel (SOB) were produced using methanol and NaOH by transesterification process. Various properties of these biodiesels such as density (883.62–886.35 kgm-3), specific gravity (0.88442–0.88709), kinematic viscosity (5.75–4.11 mm2s-1), calorific value (40.63–38.97 MJkg-1), flash point (171– 175∘C), water content (328.19–412.15 mgkg-1), color (2.0–1.8), cloud point [5.8–(-4.7)∘C], pour point [(–3.1)–(–13.1)∘C), cold filter plugging point [(−2.0)–(-9.0)∘C)], copper strip corrosion (1a–1a) and pH (7.831–7.037) were determined. Furthermore, kinematic viscosities of biodiesels and euro-diesel (ED) were measured at 298.15–373.15 K intervals with 1 K increments. Four different equations were used to predict the viscosities of fuels. Regression analyses were done in MATLAB program, and R2, correlation constants and root-mean-square error were determined. 1–7–7–3 artificial neural network (ANN) model with a back propagation learning algorithm was developed to predict the viscosities of fuels. The performance of neural network-based model was compared with the performance of viscosity prediction models using same observed data. It was found that ANN model consistently gave better predictions (0.9999 R2 values for all fuels) compared to these models. ANN model was showed 0.34 % maximum errors. Based on the results of this study, ANNs appear to be a promising technique for predicting viscosities of biodiesels.