Natural convection through a hole in a vertical wall

Numerical methods and mathematical models of Elmer
foadsf
Posts: 81
Joined: 20 Sep 2016, 23:56
Antispam: Yes
Contact:

Re: Natural convection through a hole in a vertical wall

Post by foadsf »

According to this post by Peter, when this error happens there are multiple possibilities:
  • increasing the Linear System Max Iterations
  • increasing the Linear System Convergence Tolerance
  • use a better preconditioner "Linear System Preconditioning"
  • Using a different iterative method, the direct umfpack is recommended for 2D scalar problems

    Code: Select all

    Linear System Solver = Direct
    Linear System Direct Method = UMFPack
    
the first two I tried and didn't help. I am not sure how to incorporate the other two options into my .sif file. 🤔
Attachments
HoleNatConv.sif
(2.75 KiB) Downloaded 200 times
kevinarden
Posts: 2237
Joined: 25 Jan 2019, 01:28
Antispam: Yes

Re: Natural convection through a hole in a vertical wall

Post by kevinarden »

I made several changes along those lines, but could only get a solution if I increased the viscosity to 0.01. This means the flow is turbulent at the original viscosity and you would need a turbulent flow model.
air.png
(216.36 KiB) Not downloaded yet
case.sif
(3.43 KiB) Downloaded 103 times
foadsf
Posts: 81
Joined: 20 Sep 2016, 23:56
Antispam: Yes
Contact:

Re: Natural convection through a hole in a vertical wall

Post by foadsf »

Looking at some other posts, I wonder if Elmer supports turbulent flow for compressible materials? 🤔
foadsf
Posts: 81
Joined: 20 Sep 2016, 23:56
Antispam: Yes
Contact:

Re: Natural convection through a hole in a vertical wall

Post by foadsf »

what should be the setting for Stabilize/Bubbles in Navier-Stokes equations solver? Which one should be Logical False and which one True, in my case?
kevinarden
Posts: 2237
Joined: 25 Jan 2019, 01:28
Antispam: Yes

Re: Natural convection through a hole in a vertical wall

Post by kevinarden »

Elmer is not strong for turbulent flows, most people move to openfoam if it is too turbulent. I stay with Stabilize on the settings. I am working on a KEepsilion example to try and capture some turbulence
kevinarden
Posts: 2237
Joined: 25 Jan 2019, 01:28
Antispam: Yes

Re: Natural convection through a hole in a vertical wall

Post by kevinarden »

The other issue with turbulence is that there is no steady state so it should really be done as a transient solution.
kevinarden
Posts: 2237
Joined: 25 Jan 2019, 01:28
Antispam: Yes

Re: Natural convection through a hole in a vertical wall

Post by kevinarden »

https://github.com/mrkearden/turbulent

The problem with flow and heat transients is that the flow requires very small time steps, but the heat takes a long time to transport.

The pictures are from a Steady State run with 10 steady state itterations
foadsf
Posts: 81
Joined: 20 Sep 2016, 23:56
Antispam: Yes
Contact:

Re: Natural convection through a hole in a vertical wall

Post by foadsf »

Posted a follow-up question here on cfd-online, trying to do the simulation in OpenFOAM.
raback
Site Admin
Posts: 4812
Joined: 22 Aug 2009, 11:57
Antispam: Yes
Location: Espoo, Finland
Contact:

Re: Natural convection through a hole in a vertical wall

Post by raback »

Hi Foad et al.

Certainly this results to turbulence. I didn't dwell too much into the details as it will probably still be somewhat disappointing. In the attached sif there is multiplier for viscosity by which you can play to see how far you can go in 2D steady state. With very large viscosity heat diffusion dominates (Peclet number is small). Decreasing the viscosity will first gives you one nice convection roll but further decrease results to divergence of the nonlinear (and possibly also linear) iteration.

The 1st remedy for divergence is to go transient as there is no stationary solution for the problem. You can probably find a regime where the convection rolls look "nice" while still not fully recovering the true features. One challenge is that also the symmetry is broken early which would require 3D simulations. As 3D simulations with fine enough details would require quite a bit of computing power it may be wiser to look at turbulence models. Unfortunately in both respects OpenFOAM probably provides you a better solution. Elmer might be the code of choice if this would be some very high viscous flow keeping the Navier-Stokes equations in the elliptic regime.

-Peter
Attachments
HoleNatConv.sif
(3.05 KiB) Downloaded 106 times
HoleNatConv.grd
(868 Bytes) Downloaded 101 times
Post Reply