[Sai D Irfan]

Dear all master aerodynamic of train
Good night, my name is irfan and i study in a small university in suburb area from south east asia, i'm beginner in Ansys fluent, i have a big problem about parallel wind aerodynamic on a train . i already try anything what i know to fix but its not solved. The simulation about eksternal flow of parallel wind around a train. There are the configuration of my simulation :
1. My enclosure is, 15m X+, 20m Y+, 10m Z+, 15m X-, 0.3m Y-, 20m Z-. i substract the train body and the enclosure using boolean
2. The mesh is Hex dominant, the details in the picture below
3. In setup, i use
model viscous-k omega-SST
methods coupled
convergent criteria, continuity 10^-5, x velocity 10^-6, y velocity 10^-6, z velocity 10^-6, k 10^-6, omega 10^-6

then i have finished iterating up to 1000 times, but it never reach the convergent. i also changed the explicit relaxation and under relaxation for any combination even to stupid number, but it doesn't work.
the best residuals about at the number 646 iteration. the details is in the picture below.


please someone help me, how to solve this problem.
I really appreciate help even if it's the smallest. Thanks a lot for all of your kindness

sincerely


Irfan

[RaiderDoctor]

Hi Irfan,


So I see a couple of issues with your setup, initially. While you constructed your fluid domain okay, you really don't need to simulate the whole thing. Cut your domain in half, through the mid-section of your train, and set the cross-sectional plane to "Symmetry". Now, you can run a bit more efficiently.


Next, we need a better pic of your mesh. While most issues can be attributed to a bad mesh, it helps to see what the mesh actually looks like close to areas where there is flow disturbance. Just showing the outside of the fluid domain is usually not enough. Also, if all of your cells look like what you showed, you might have a overall poor quality mesh. You can use metrics such as skewness and orthogonal quality to "check" if you have a good mesh, but ultimately you'll need to do a sensitivity analysis to be sure. That doesn't fix your current problem, just looking ahead for you.


Now, as to why your residuals aren't "converging". The main thing you need to understand is that "convergence" does not simply mean your residuals fall below a specific value. A solution is converged when it no longer shows dramatic fluctuation in either residuals, or quantities of interest from one iteration to the next. Check out this link for more info (https://www.engineering.com/DesignSo...nvergence.aspx). Without seeing a pic of your residuals, or more info on your setup, I'm not sure anyone would be able to give you much help.

[Sai D Irfan]

Originally Posted by RaiderDoctor

Hi Irfan,


So I see a couple of issues with your setup, initially. While you constructed your fluid domain okay, you really don't need to simulate the whole thing. Cut your domain in half, through the mid-section of your train, and set the cross-sectional plane to "Symmetry". Now, you can run a bit more efficiently.


Next, we need a better pic of your mesh. While most issues can be attributed to a bad mesh, it helps to see what the mesh actually looks like close to areas where there is flow disturbance. Just showing the outside of the fluid domain is usually not enough. Also, if all of your cells look like what you showed, you might have a overall poor quality mesh. You can use metrics such as skewness and orthogonal quality to "check" if you have a good mesh, but ultimately you'll need to do a sensitivity analysis to be sure. That doesn't fix your current problem, just looking ahead for you.


Now, as to why your residuals aren't "converging". The main thing you need to understand is that "convergence" does not simply mean your residuals fall below a specific value. A solution is converged when it no longer shows dramatic fluctuation in either residuals, or quantities of interest from one iteration to the next. Check out this link for more info (https://www.engineering.com/DesignSo...nvergence.aspx). Without seeing a pic of your residuals, or more info on your setup, I'm not sure anyone would be able to give you much help.

Dear RaiderDoctor, Thanks so much for your reply , but i have something in my mind ...

Its okay i will try to cut my domain in half,
there are the picture of my mesh in the below, from the graphic of mesh metric i think my mesh is quite good, but please help if there is something wrong,
i get the point for the convergence... thanks,
there is the picture of residuals in below, its OK if i try to find any combination value of explicit relaxation and under relaxation to get the best one and then i use that value later for new simulation start from beginning ?

Thanks a lot for your reply

[RaiderDoctor]

Sorry about the late response, life happens sometimes.

First off, your mesh is not that great. While it is good to have finer quality cells next to wall regions and areas of possible flow disturbance, the transition between large to small are much too large. Furthermore, on the exterior of your train, you have regions of small cells and large cells that will create a lot of problems for you. Try to have uniform cells that transition smoothly between themselves.

Next, you've specified 0.001 as your under relaxation factors. Holy cow, dude! These values are way too low. Consult the manual, but I believe that the lowest value you can have for pressure and momentum is about 0.5. Under relaxation factors essentially work by slowing down the particular variable they are assigned to. A small factor means the solution will be stable, but it will take a long time to converge. A large factor will mean quick convergence, but inherent instability.

[MIDHUN@CFD]

pls give inflation and try to get wall y plus reduced below 30 and give k e turbulance model with enhanced wall treatment .. pls achieve skewness below .9 initially and aspect ratio below 650 and try