Estimating the Dynamic Viscosity of Vegetable Oils Using Artificial Neural Networks

Abstract

In this study, dynamic viscosities of poppy and hazelnut oils were measured at 30–100°C interval with 1°C increments by using a Brookfield Engineering Co. DV-II+ Pro rotary viscometer. Viscosity measurements were realized at 100 rpm constant speed and by using a RVII spindle. Three equations were used to estimate the viscosities of poppy and hazelnut oils. Regression analyses were conducted in MATLAB® program and R2 (coefficient of determination), correlation constants, and root mean squared error were determined. The best coefficients of determination obtained by using μ = e(A+B/T–C)) relation were 0.99983 and 0.99989 for poppy and hazelnut oils, respectively. In analyses conducted by using artificial neural networks, the coefficients of determination were obtained to be 0.999937 and 0.999960 for poppy and hazelnut oils, respectively. Although coefficients of determination obtained in two methods were close to each other, the root mean squared error obtained by using artificial neural networks was smaller for both two oils.