|
[Sponsors] |
May 11, 2020, 01:15 |
Judging convergence of External Aero
|
#1 |
New Member
Ruhinsan
Join Date: Nov 2019
Posts: 4
Rep Power: 6 |
Hi,
Currently i'm doing an aerodynamic simulation of a semi-highspeed train. The boundary conditions is as follows -Velocity inlet: 44 m/s -Pressure outlet: 0 Pa -Bottom moving wall: 44 m/s -Left, top, and right are symmetry I meshed my model in a small tetramesh filled box surrounding the train and then i used hexamesh to fill the rest of the domain. I simulate it using steady K omega sst turbulence model with y+ about 30. In fluent i converted the mesh into polyhedra mesh and checked the quality. Min orthogonal quality is 0.06 and max skew (before converting to polyhedra) is about 0.85. During the run, i encountered "turbulent viscosity limited to ... cells" warning which go up to several thousand cells and then it's reduced until the warning is completely gone. Below is the image of my residual and Cd report. I want all of my residual to drop to at least 10-4, but the continuity residual got stuck on 1.6x10-3 even though my Cd already converged. So my question is 1. Is my simulation can be judged as converged even though the residual is still high? (by the the way, those last two spikes on the residual are caused because i changed the pressure outlet to negative pressure, thinking that the residual will come down) 2. Is there any way to reduce the residuals even further? 3. From several journal that i've read, they simulate the train using DES which initialized from converged steady state k omega SST model, so is it possible that my case is transient by nature? Thanks guys! Ruh cd.PNG res.PNG |
|
May 11, 2020, 10:52 |
Setup
|
#2 |
Senior Member
|
You should check the quality after polyhedral conversion as well. Polyhedral conversion can also lead to reduction of quality. If it is severe, then do not convert to polyhedral mesh.
Convergence looks good. Don't worry about 1.6e-3 or 1e-4. Those are not of significance and do not relate to the accuracy. If the fields parameters and coefficients are stable, then you can consider the case properly converged. For accuracy, you need to compare against some data.
__________________
Regards, Vinerm PM to be used if and only if you do not want something to be shared publicly. PM is considered to be of the least priority. |
|
May 12, 2020, 10:15 |
|
#3 |
New Member
Ruhinsan
Join Date: Nov 2019
Posts: 4
Rep Power: 6 |
Thank you for your reply.
So i've been reading from the other thread and i found that the residuals are scaled from the initial guess, so if the initial guess is pretty close to the solution, the residuals wont drop much. Suppose i deliberately change the initial guess values, is it possible that i can get similar Cd value for the simulation but with lower residuals? |
|
May 12, 2020, 10:38 |
Residuals
|
#4 |
Senior Member
|
It does not work that way. Residuals are scaled with respect to initial residual not with initial flow field and their definition is based on solution of the linear system . For an exact solution, both sides are equal. But numerically, that is not possible. So, their difference is Residual. Furthermore, lower residuals do not imply correct results.
__________________
Regards, Vinerm PM to be used if and only if you do not want something to be shared publicly. PM is considered to be of the least priority. |
|
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
CFX overwhelming Fluent in mass convergence of boundary layer separation case | Pierre1 | FLUENT | 7 | March 26, 2015 21:43 |
Experiences with OpenFOAM and Automobile External Aero | kyle | OpenFOAM | 5 | October 24, 2013 06:52 |
help with convergence - external automotive | me3840 | FLUENT | 14 | July 6, 2011 07:58 |
Usefulness of Similarity theory in External Aero | Amod Kumar | Main CFD Forum | 2 | January 10, 2007 04:58 |
Defect correction and convergence | ganesh | Main CFD Forum | 4 | June 30, 2006 14:20 |