Estimating viscosity of non-Newtonian fluids using support vector regression method: Rheological parameters of drilling fluids using data fusion

摘要

In contrast to flow studies involving water in hydraulics and hydrodynamics falling in the realm of Newtonian fluids, the flow behavior of food, oil and polymers fall into the category of non-Newtonian fluids, where the rheological parameters viscosity, density, flow velocity, Reynolds number are all interdependent. Because of these interdependencies, in addition, the viscosity is dependent on the dimension and geometry of the conduit used for transporting the medium. Many models exist for different conditions relating the shear rate to shear stress for the estimation of viscosity. For engineering applications involving drilling fluids as encountered in oil & gas industries or geothermal applications, a knowledge of viscosity is important. This paper presents the estimation of viscosity using Support Vector Regression (SVR) method. In an earlier study, without resorting to analytical techniques involving a plethora of equations, the viscosity was estimated, using Artificial Neural Networks (ANN). In this paper, measurements performed in an open Venturi channel with the non-Newtonian fluid flow are used to estimate the viscosity using ANN and SVR techniques.