CFD Online Discussion Forums

CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   Main CFD Forum (https://www.cfd-online.com/Forums/main/)
-   -   Mesh independence, internal fluid-dynamic (https://www.cfd-online.com/Forums/main/214388-mesh-independence-internal-fluid-dynamic.html)

Francesco_93 January 29, 2019 16:25

Mesh independence, internal fluid-dynamic
 
Hello everyone! I'm new to the forum and I hope this is the right section for my question. Basically I deal with a problem of internal fluid dynamics, in particular I have a hydraulic conduit with water working fluid, which ends with a valve. The study requires assessing the risk of cavitation within the valve. For the assigned data I have a turbulent flow with Re of 10 ^ 5. I have an input speed of 9.5 m / s and standard conditions. The outlet pressure is 118750 Pa. The length of the initial conduit section and the terminal part are free parameters to be imposed according to the boundary conditions.
The problem must be solved with a 2D axisymmetric approach and a structured block grid. I created the geometry in gambit and I created blocks in order to respect a low equisize skewness (about 0.35). At this point I chose to solve the problem with an enhanced wall treatment approach given the low number of reynolds. I have generated two grids and I have therefore considered a y + of 1.5 for the first grid and of 1 for the second, in order to have y + <5 on the whole domain. Basically I chose interval count and then in the inflate I have increased by multiplying the interval count by 1.5 and approaching the centroid on the wall (as already said y + = 1).
In order to have adjacent elements with a low volume variation it was necessary to use a double side refinement even in areas that are not wall-mounted and therefore where it was not needed.
By exporting the mesh to fluent I chose a pressure based solver, axismetric, steady, implicit, green gauss. I chose to work in terms of relative pressure (operating preassure = 118750, pgauge = 0).
As boundary conditions I chose velocity inlet and I assumed to have a fully developed bike as the valve is part of a conduit. I then made the hypothesis that upstream there is a sufficiently long and straight duct. So I assigned as parameters, of the k-epsilon model, the input hydraulic diameter and the turbulence intensity = 3.76 calculated on the basis of the reynolds.
As a backflow condition I put the viscosity ratio = 350 and a turbulence intensity = 1%.
Reference values ​​I take those of the inlet.
In the simple algorithm I chose a 2nd order upwind.
My problem is this: the problem goes to convergence and also respects the constraints of the two layer, but when I go to thicken and evaluate the mesh independence in the most important areas, I get values ​​of 150% of the RMSEP.
Also the trends of the two grids are different for some sizes. I thought it would need to further deepen the grid but my PC can not do it because you would get millions and millions of items.
So I thought the mistake was elsewhere. It is not a problem to thicken the grid but there is some other problem. Please help me. Thanks in advance.


All times are GMT -4. The time now is 07:14.