Loading...

Non-Newtonian fluid flow dynamics in rotating annular media: Physics-based and data-driven modeling

Ershadnia, R ; Sharif University of Technology | 2020

986 Viewed
  1. Type of Document: Article
  2. DOI: 10.1016/j.petrol.2019.106641
  3. Publisher: Elsevier B.V , 2020
  4. Abstract:
  5. A thorough understanding and accurate prediction of non-Newtonian fluid flow dynamics in rotating annular media are of paramount importance to numerous engineering applications. This is in particular relevant to oil and gas industry where this type of flow could occur during, e.g., drilling, well completion, and enhanced oil recovery scenarios. Here, mathematically we report on physical-based (numerical) and data-driven (intelligent) modeling of three-dimensional laminar flow of non-Newtonian fluids driven by axial pressure gradient in annular media that consist of a coaxially rotating inner cylinder. We focus on the dynamics of pressure loss ratio (PLR)—the ratio of total pressure loss in presence of rotation to that in stationary condition. We develop a novel and computationally efficient machine-learning based predictive model based on a Least Square Support Vector Machine (LSSVM) optimized with a Coupled Simulated Annealing (CSA) algorithm, which is trained and tested by 730 experimental data collected for this study. We then perform numerous CFD simulations governed by Navier-Stokes set of equations, which correspond to each of the experimental cases. Comparing the results with those obtained from a widely used empirical correlation for PLR, we find that the both of our numerical and machine-learning based models are more accurate and present a larger applicability domain. The predictive LSSVM model enjoys a supervised learning paradigm and leads to the highest accuracy, while the physically originated CFD approach captures more consistently the nonlinear behavior of PLR versus key system features. Following the validation of our models, we study the less understood effects of fluid viscosity, axial velocity and rotational velocity on PLR dynamics. We employ a dimensionless parameter analogous to Strouhal number and identify specific regimes where PLR shows a distinct behavior. Our study elucidates the role of shear-thinning effect and its interplay with inertial forces on the dynamical behavior of pressure loss (ratio) in rotating yield-power-law fluids. © 2019 Elsevier B.V
  6. Keywords:
  7. CFD ; LSSVM ; Non-Newtonian fluid ; Pressure loss ratio ; Rotating Couette flow ; Computational fluid dynamics ; Enhanced recovery ; Flow measurement ; Gas industry ; Laminar flow ; Machine learning ; Natural gas well completion ; Navier Stokes equations ; Newtonian flow ; Non Newtonian liquids ; Oil well completion ; Oil well drilling ; Rheology ; Shear flow ; Shear thinning ; Simulated annealing ; Support vector machines ; Viscous flow ; Computationally efficient ; Couette flows ; Engineering applications ; Least square support vector machines ; Non-Newtonian fluids ; Three-dimensional laminar flows ; Non Newtonian flow ; Couette flow ; Flow modeling ; Pressure gradient ; Rotating flow
  8. Source: Journal of Petroleum Science and Engineering ; Volume 185 , 2020
  9. URL: https://www.sciencedirect.com/science/article/abs/pii/S0920410519310629