All entries for April 2007
April 30, 2007
Spent the weekend freeing up disk space by burning all non-essential data onto disks. Have also synchronised all data between the office machine and my personal PC.
I am now downloading all from Solar so that I can free up space there and get the simulations finalised. The following name changes are to be made:
rs01-02-01 > rs01-02-A01: No approx
rs01-02-04 > rs01-02-A02: WF approx
rs01-02-05 > rs01-02-A03: IBM approx
rs01-02-07 > rs01-02-A04: Both WF & IBM approx
rs01-02-10 > rs01-02-B01: No approx (run from stratch)
rs01-02-11 > rs01-02-B02: No approx (use rs01-02-A01 as I.C.)
mt03-01-25a > mt03-01-A01: 2D RANS boundary layer (No TIR)
mt03-01-26 > mt03-01-A02: 3D RANS boundary layer (No TIR)
mt03-01-27 > mt03-01-A03: 3D DES boundary layer (No TIR)
mt03-01-28 > mt03-01-A04: 3D DES boundary layer with rescaling
Will now get to work…
Have looked at rs01-02-07a and it seems similar (although not identical) to the previous set. I think that whilst the wall functions approach does correct for having nodes too far away from the wall, I think most of the grid is ok so it does not have to do too much. With this in mind I think that the 2D results can be provisionally published and written about
I have also checked rs01-02-10 and it appears settled down. Will wait until it and rs01-02-11 have finished and then finalise them (Change the name etc.)
Finally, I have also set running mt03-01-26 in preparation to using the rescaling.
April 27, 2007
Have set running rs01-02-07 again using the new formulation (I.e. change the sense of the source term addition/subtraction if the friction velocity is negative)
Have also set running rs01-02-04 using the opposite formulation to check that I have picked the signs in the write sense.
Have set running the mt03-01-25a simulation running again as it seems ok. Will probably need one more run and it will be converged. I can then switch to 3D and start with the TIR work.
As I am waiting for both rs01 and mt03 simulations to finish running (can’t really start writing anything without them) I am going to use the time to get my hard-drives sorted. This is for a number of reasons.
a) When I delete things using, say, WinSCP I’m not sure that the hard-drive is being cleared of it but simply emptying the recycle bin doesn’t work.
b) I have hardly any space left and it is all in chunks. It is probably sufficient to burn down non-essential simulations (I.e. sims I’d rather keep but would not be seriously screwed if I lost them) and clear up a ton of room. To do all this, I will take the office machine home and network it to my personal PC in order to synchronise all data and burn it.
c) I could do with clearing up the C:\ drives in order to get them running a little quicker and prevent them having a stress every ten minutes
April 25, 2007
Have spent yesterday and today producing the plots necessary from the bump results. This includes creating the boundary layer integral thicknesses, velocity profiles and streamtrace plots.
I’ve also noticed that rs01-02-04 is different to the others and have finally tracked this down to the fact that even if the friction velocity is negative, the wall treatment is being applied in the positive direction. This has the effect of suppressing separation but is entirely fictional. Am re-running now.
April 23, 2007
Checked mt03-01-25 this morning and it started to crash just before 2000 timesteps. Not sure quite what it causing this so will run 1000 timesteps then another 1000 and see if the same thing happens.
Also, I will probably take some early mt03-01 simulation and run this as mt03-01-25a to check that I haven’t done something wrong within the code since then.
Actually, have decided to switch off the turbulence modelling first to test.
Checked and seems ok so have set running to 5000 timesteps using a shell.
April 22, 2007
Ok, I have two sets of simulations which are progressing alright but I want to make the results stronger. I will now assess the stage at which both of these are at and where to go from here.
-MT03-01: Flat plate boundary layer using TIR
This isn’t bad but I’m not really happy that I have a 3D simulation with a fully settled down mean profile. What may be the best bet is to run a 2D RANS simulation and use this as the mean profile to start with. I can then turn it 3D and add in the random fluctuations. If we choose the time averaging interval long enough that the mean is protected then I think we can go from there.
So, the overall plan of action is:
- Create a new simulation, mt03-01-25, which will be the new full 3D boundary layer in preparation for TIR.
- Run this as 2D RANS without fluct. for the moment and until the averages have settled down.
- Once this is done, switch to 3D DES with fluct. and run until the RMSs have settled down too (although these will probably not be correct, they need to have converged)
- Only after all this is settled, can we get on with applying the inflow rescaling realistically.
For this afternoon then:
- Get a 2D RANS simulation running as mt03-01-25 and get it fully settled down
-RS01-02: Lille Bump without flow control
The 3D bump simulations without any methods are still running (good!). However, I am not 100% happy with rs01-02-07 (IBM + WF). I think that the criterion for selecting which cells to apply the WF approach is screwed up by the presence of the ghost cells. If I have a rethink and fix the criterion we may see the recirculation region reduce (as it did for the wall function approach alone).
The plan of action for this then is:
- Fix rs01-02-07 so that WF nodes are being forced along the entire domain without clashing with the IBM nodes.
Right, to work!
MT03-01-25 set running…
Am just waiting for two simulations now:
- MT03-01-25: 2D RANS flat plate simulation. Need to wait until the average has settled down
- RS01-02-07a: Bump with IBM + WF. Need to check it’s stable
April 20, 2007
Have checked mt03-01-21 which is now fully stable but it is laminarising. When considering the mean profile, it is obvious that the boundary layer has not been averaged enough so I have set running for longer and will continue until the mean velocity profile is steady.
Have set running the 3D simulations again to carry on getting results. I need to keep running them on.
On looking closer at rs01-02-04a, it was clear that the correct cells were not being wall forced. I think in getting the IBM simulation to work, the criterion (nblu) did not allow all nodes to be wall forced so meaning the run up and run out were not wall forced. Have set a different criterion for the moment to test this.
April 18, 2007
Managed to get TECPLOT work yesterday but came up with the problem that rs01-02-07 is more similar to the IBM results than to the WF results. Not sure quite why this is but will run rs01-02-07 again to double check then think about the implementation a little more.
With regard to mt03-01-21, it crashed again and so I’m not why. Will mash around the subroutine again and see if I can get it working.
On thinking about it, there are a number of things I can do.
- Have a look at the time averaging scheme as per Lund et al.
- Output the inlet profile at each major output stage
- Output the rescaling parameters at eash timestep to allow an evaluation of the stability wrt time
Have just noticed that the vectors at inlet just off the wall are huge which is probably the main cause of instability. After completing the above, I need to sort that out (it is probably an artifact of the rescaling procedure which is messed up for the first off wall nodes)
Have corrected the mt03-01-21 simulation to get rid of the troublesome first off wall node (set the velocity to half the second off wall node).
Also, have set running rs01-02-04 and rs01-02-07 for 1000 timesteps to try and assess the effects of wall functions by outputting the friction velocity at each file.
Just checking rs01-02-04….are the wall functions being applied along the whole domain and if not, why not?
April 17, 2007
Right, today’s task is to beat TECPLOT into submission. I want to be able to produce all of the plots detailed yesterday within 2 minutes. This will be done using layouts and macros to literally open the file, create the plot and export it. Let’s get to it…
April 16, 2007
- MT03-01: Flat Plate Boundary Layer
Have set running mt03-01-21 again but with the number of samples increased by a factor of 10. This should make the time average more stable to change and so allow the boundary layer to settle down correctly before it has a chance to become unstable.
Also, mt03-01-24 has finished running now so if 21 becomes stable, I can get 23 running which will be the same as 24 but with rescaling.
- RS01: Lille Bump
The fresh 3D results (10) and the pre-run 3D results (11) are currently running ok. However, I will download some of them to assess how well they have run.
Have had a look at mt03-01-24 and it doesn’t appear to be great. The eddy viscosity is lifted up towards the end of the domain. Also, I ran on this sim based on previous ones so I will now try it using pure RANS and starting afresh.
Now that I have checked on all simulations, the next task is to get on with producing those 2D results for Chapter 7. I may as well do this at home as all the data is on my hard-drive there.
Had a hassle getting TECPLOT working but I know that tomorrow’s task is now to a layout file and set of macros working PROPERLY…
I need to produce the following plots:
Overall u contours
Overall mean u contours
Overall S-A viscosity contours
Velocity profiles (y)
S-A profiles (y)
Wall shear stress (x)
Integral lengths (x)
April 13, 2007
Have gone into the office and downloaded my previous work ready for preparing for Chapter 7. I’ve also managed to get working mt03-01-24 with a DES formulation.
The next main task is to get the 2D Lille bump results prepared for the thesis but there is little point starting this now. I will get on with tidying up and copying files across along with an early lunch then get on with the results this afternoon.
After having finished planning everything else (time-wise) I am now more confident about things generally. I will spend the rest of the day sorting out a task-list for the rest of the month.
- Write-up 2D results for the bump without flow control. This will include pressure coefficient, wall shear stress, recirculation sizes, and integral thicknesses.
- Complete 3D results for the bump without flow control
- Write-up 3D results for the bump without flow control
- Get rescaling subroutine stable and working
- Finish the ‘introductory’ sections and most of the results of the Lille bump without flow control for Chapter 7