I am modelling a simple pipe flow of a non-newtonian fluid (Ansys CFX). I'm aware that this is an area of developing knowledge.
The viscosity model for the non-newtonian fluid is the Carreau-Yasuda model. The fluid properties are taken from literature (Escudier et al, 2005). I thought it would be interesting to plot the variation of viscosity with shear strain rate in CFX, and compare that to the original viscosity relationship.
The attached image shows this plot. Grey line: Original data (Escudier et al 2005). Yellow line: Original data plotted using CFX slightly modified Carreau-Yasuda equation (I need to look into this separately). Blue line: Laminar flow, approx Re=780. Orange line: Turbulent flow, approx Re= 45000. (N.B. picked these Reynolds numbers for a comparison, they're not arbitrary).
The question is: Why do the CFD results for shear strain rate vs. dynamic (apparent) viscosity deviate so far from the input relationship?
My thoughts are that this is a mathematical issue in the CFX solver. I'm having difficulty pinpointing an explanation though, my knowledge in this area is still developing.
For reference, I'm a Masters student in Mechanical Engineering.
Have you investigated mesh sensitivity and sensitivity to near-wall mesh resolution? What is the used turbulence model?
Hi Thomas,
I Initially checked mesh sensitivity (element size), using a velocity monitor. Later, I adjusted my boundary inflation (increased thickness slightly, more layers) as the velocity profile for fully developed flow wasn’t looking well resolved near the wall. The profiles of velocity and shear rate for fully developed flow look sensible now.
I am using BSL Reynolds Stress turbulence model.
Thank you.
Have you tried to use SST turbulence model for comparison? Cannot see a clear reason for using a Reynolds Stress Model for a simple pipe flow. More complex models are not necessarily more accurate.
I haven’t tried SST, I could give that a try. I initially ran the problem using k-epsilon, but the results were quite erratic. I read a paper from Gavrilov (2016) that provided some reasoning as to why a Reynolds stress mode is more appropriate than k-epsilon for non-Newtonian flows.
thanks for your help, I’ll try a different turbulence model and see the impact.
- Badges
- Science topic
- Similar topics
- Engineering
- Mechanical Engineering