|
[Sponsors] |
July 14, 2017, 19:44 |
Transiant simulation and experimentation
|
#1 |
Senior Member
dilaw meda
Join Date: Jun 2017
Location: algeria
Posts: 145
Rep Power: 8 |
Hi, when performing a transiant simulation , should i do the same length time that i needed in experimentation. For example if it took me 1 hour to mesures all velocities in different points. Should the time of the simulation be 1 hour ? And if the mesurement in exerimental was not in a constant time step ( for example 10 seconds for a mesure and 15s for the next mesure) because the nature device of mesurement (you need until the numbers are stable on the screen ) does this have an effect ?
Sent from my F1f using CFD Online Forum mobile app Last edited by medaouarwalid; July 15, 2017 at 09:23. |
|
July 16, 2017, 11:26 |
|
#2 |
Senior Member
Join Date: Aug 2014
Location: Germany
Posts: 292
Rep Power: 13 |
It took one hour to measure all different points in the apparatus. But did you check if the value was changing over time at the same spot? So, did you check after 5 mins, 30 mins, 1hr the same spot? If so, is the value constant over time? If the answer is yes, my answer would be (assuming LES):
It is ok to simulate 5 mins. and check afterwards. |
|
July 17, 2017, 16:55 |
|
#3 | |
Senior Member
dilaw meda
Join Date: Jun 2017
Location: algeria
Posts: 145
Rep Power: 8 |
Quote:
Sent from my F1f using CFD Online Forum mobile app |
||
July 18, 2017, 03:27 |
|
#4 | |
Senior Member
Join Date: Aug 2014
Location: Germany
Posts: 292
Rep Power: 13 |
Quote:
|
||
July 18, 2017, 03:59 |
|
#5 |
Senior Member
Filippo Maria Denaro
Join Date: Jul 2010
Posts: 6,768
Rep Power: 71 |
While comparing a transient simulation with experimental measurements, the main problem is in setting the same initial condition otherwise you LES will produce only a numerical transient.
|
|
July 18, 2017, 04:52 |
|
#6 |
Senior Member
Join Date: Aug 2014
Location: Germany
Posts: 292
Rep Power: 13 |
This is true. But i guess finding good initial conditions for this case startup of hairdryer, or what it exactly is i don't know, can be tricky. I guess it depends also on the flow history before the inlet into the computational domain.
Last edited by BlnPhoenix; July 19, 2017 at 07:33. |
|
July 18, 2017, 04:59 |
|
#7 |
Senior Member
Filippo Maria Denaro
Join Date: Jul 2010
Posts: 6,768
Rep Power: 71 |
Right. And if it is not possible to match the initial condition, at least close to the nozzles, the only meaningful result is when the LES solution forgets the initial condition. But that does not match the transient measurements...
|
|
July 18, 2017, 06:25 |
|
#8 | ||
Senior Member
dilaw meda
Join Date: Jun 2017
Location: algeria
Posts: 145
Rep Power: 8 |
Quote:
Quote:
Sent from my F1f using CFD Online Forum mobile app |
|||
July 18, 2017, 06:36 |
|
#9 | |
Senior Member
dilaw meda
Join Date: Jun 2017
Location: algeria
Posts: 145
Rep Power: 8 |
Quote:
Sent from my F1f using CFD Online Forum mobile app |
||
July 18, 2017, 07:07 |
|
#10 | |
Senior Member
Filippo Maria Denaro
Join Date: Jul 2010
Posts: 6,768
Rep Power: 71 |
Quote:
I suggest to take some measurements in the room but with all nozzles turned out. See in several position of the room the velocities to understand if you have a convective motion as base flow. You could try to replicate first a simulation (without nozzles) of the room base fow and use it as initial condition. Conversely, if you measure an almost zero velocity field you could set an initial condition corresponding to the rest with superimposed a small random fluctuations. Furthermore, the boundary conditions for the temperature should consider the correct heat flux leaving the room from the walls. Otherwise, if you set all adiabatic conditions, the heat flow from the nozzle produces an increasing in the internal energy that can drive to a numerical instability. You need to model you problem at the best you can. |
||
July 18, 2017, 08:51 |
|
#11 | |
Senior Member
dilaw meda
Join Date: Jun 2017
Location: algeria
Posts: 145
Rep Power: 8 |
Quote:
Sent from my F1f using CFD Online Forum mobile app Last edited by medaouarwalid; July 18, 2017 at 12:41. |
||
July 18, 2017, 10:50 |
|
#12 | |
Senior Member
Blanco
Join Date: Mar 2009
Location: Torino, Italy
Posts: 193
Rep Power: 17 |
Quote:
|
||
July 18, 2017, 11:43 |
|
#13 | |
Senior Member
dilaw meda
Join Date: Jun 2017
Location: algeria
Posts: 145
Rep Power: 8 |
Quote:
Sent from my F1f using CFD Online Forum mobile app |
||
July 18, 2017, 12:02 |
|
#14 | |
Senior Member
Filippo Maria Denaro
Join Date: Jul 2010
Posts: 6,768
Rep Power: 71 |
Quote:
The run seems too slow for this grid size... |
||
July 18, 2017, 12:40 |
|
#15 |
Senior Member
dilaw meda
Join Date: Jun 2017
Location: algeria
Posts: 145
Rep Power: 8 |
time step is 10^-5 so i need 100 000 time step to have 1 seconde of simulation , if the time step increments every 1 seconde ( which is not the case) , it takes 27 hours to complete the simulation , so it is logical somehow
|
|
July 19, 2017, 06:41 |
|
#16 |
Senior Member
Join Date: Aug 2014
Location: Germany
Posts: 292
Rep Power: 13 |
Try the RANS approach first and compare your velocity value from the experiment with the CFD solution. It will give you some idea about accuracy. Afterwards you decide what further you can do to improve.
|
|
July 19, 2017, 07:29 |
|
#17 |
Senior Member
Blanco
Join Date: Mar 2009
Location: Torino, Italy
Posts: 193
Rep Power: 17 |
If you're going to simulate multiple jets entering into a quiescent room then I would definitely start with a RANS. It will give you something usable in the short term and you can also use the RANS results to eventually do a LES simulation later (i.e. you can initilize the flowfield with the RANS results with some perturbation, this will speed-up the "flush out" of initial conditions in the LES sim)
|
|
July 19, 2017, 07:56 |
|
#18 | |
Senior Member
dilaw meda
Join Date: Jun 2017
Location: algeria
Posts: 145
Rep Power: 8 |
Quote:
Sent from my F1f using CFD Online Forum mobile app |
||
July 20, 2017, 14:32 |
|
#19 |
Senior Member
dilaw meda
Join Date: Jun 2017
Location: algeria
Posts: 145
Rep Power: 8 |
hi , after your suggestions i did 3 seconds of simulation using unsteady realizable K-epsilone , always with that low numbers of elements mesh ( refinement only in close to the nozzles) and a time step of 0.001 seconde just to have an idea. residuals was 10^-4 for continuity ; 10^-6 for k and epsilon ; 10^-6 for x,y,z velocities and 10^-8 for energy. For the radial velocity, the results was good even with that low mesh ( figure) but for the axial velocity not so good (as you see). now what do i have to do ? should i refine the mesh and use the results as initial conditions for LES or only count on Urans if it gives me better results .
|
|
July 20, 2017, 14:42 |
|
#20 |
Senior Member
Filippo Maria Denaro
Join Date: Jul 2010
Posts: 6,768
Rep Power: 71 |
I would first refine the grid to check if the solution changes
|
|
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
Guide to set initial parameters and settings for external flow simulation | amarnath_sivasubramanian | FloEFD, FloWorks & FloTHERM | 1 | August 6, 2015 01:43 |