Time scale factor and physical timescale

saisanthoshm88 · August 8, 2012, 13:49

Could some one please let me know when to use the time scale factor of the auto timescale and when to change to physical timescale when the intention is to accelerate convergence by changing the time scale in CFX.

ghorrocks · August 8, 2012, 18:48

Starting a simulation with auto time scale is usually OK. If you have convergence problems then make it smaller. Once the simulation is converging nicely then you can start increasing the time scale size to accelerate convergence.

bratzinger · August 9, 2012, 03:29

or if u want to calculate it by yourself and u know the massflow and the typical cross-section area:
v(_char)=m_dot/(rho*A) -> t_char=length/v(_char)
u should get roughly the same (depending of the complexity of your geometrie the same value or at least ~ the same power of tchar)

with heat flux i often use e.g. 100s for phys. timescale - a time where the casing in my project should easily got the right temperatures ... in my steady state simulation (hope i understood that right that u can avoid transient influence through initial conditions in that domain if u set it suficiantly high?)

for the energy equation i often use large timescale factors(100, 500 or more) to accelerate convergence

saisanthoshm88 · August 22, 2012, 04:00

Hi Glenn,

So if there is an auto time scale of 9 [s] and if I wanted it to be 0.9 [s] then does it make the same sense with the two approaches viz..,:

1. Setting the Time Scale Factor to 0.1 in the auto time scale.

2. Directly switching over to physical time scale and setting it to 0.9 [s]

ghorrocks · August 22, 2012, 06:46

Both approaches are equivalent, they both set the solver to 0.9s. Do which ever is easiest for you.

Bdew8556 · August 11, 2017, 09:35

Is it fair to say that having a small timescale will result in the solution taking longer to converge but do so in a more stable way? Ive had some simulations where the max mach number is reached quickly and ive been recommended to apply a smaller timescale.

ghorrocks · August 11, 2017, 19:41

For a steady state simulation the advantage of larger time steps is you can converge faster. But the disadvantage is that they will go unstable. So the ideal time step size is the biggest time step size which is not unstable. If you are smaller than that it will converge but just take longer.

There is a lower bound to this - if the time step size is too small then round off errors dominate and it diverges again.

August 8, 2012, 13:49	Time scale factor and physical timescale	#1
saisanthoshm88 Senior Member --------- Join Date: Oct 2010 Posts: 303 Rep Power: 17	Could some one please let me know when to use the time scale factor of the auto timescale and when to change to physical timescale when the intention is to accelerate convergence by changing the time scale in CFX. __________________ Best regards, Santhosh.

August 9, 2012, 03:29		#3
bratzinger New Member Join Date: Jul 2012 Location: Germany Posts: 23 Rep Power: 13	or if u want to calculate it by yourself and u know the massflow and the typical cross-section area: v(_char)=m_dot/(rhoA) -> t_char=length/v(_char) u should get roughly the same (depending of the complexity of your geometrie the same value or at least ~ the same power of tchar) with heat flux i often use e.g. 100s for phys. timescale - a time where the casing in my project should easily got the right temperatures ... in my steady state simulation (hope i understood that right that u can avoid transient influence through initial conditions in that domain if u set it suficiantly high?) for the energy equation i often use large timescale factors(100, 500 or more) to accelerate convergence Last edited by bratzinger; August 9, 2012 at 03:53.*

August 22, 2012, 04:00		#4
saisanthoshm88 Senior Member --------- Join Date: Oct 2010 Posts: 303 Rep Power: 17	Hi Glenn, So if there is an auto time scale of 9 [s] and if I wanted it to be 0.9 [s] then does it make the same sense with the two approaches viz..,: 1. Setting the Time Scale Factor to 0.1 in the auto time scale. 2. Directly switching over to physical time scale and setting it to 0.9 [s] wc34071209 and Goenitz like this. __________________ Best regards, Santhosh.

August 22, 2012, 06:46		#5
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,700 Rep Power: 143	Both approaches are equivalent, they both set the solver to 0.9s. Do which ever is easiest for you. wc34071209 and roi247 like this.

August 11, 2017, 19:41		#7
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,700 Rep Power: 143	For a steady state simulation the advantage of larger time steps is you can converge faster. But the disadvantage is that they will go unstable. So the ideal time step size is the biggest time step size which is not unstable. If you are smaller than that it will converge but just take longer. There is a lower bound to this - if the time step size is too small then round off errors dominate and it diverges again. pocketprotector, Goenitz, chtabz11 and 1 others like this.

August 8, 2012, 18:48		#2
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,700 Rep Power: 143	Starting a simulation with auto time scale is usually OK. If you have convergence problems then make it smaller. Once the simulation is converging nicely then you can start increasing the time scale size to accelerate convergence.

August 11, 2017, 09:35		#6
Bdew8556 Senior Member Brett Join Date: May 2013 Posts: 212 Rep Power: 13	Is it fair to say that having a small timescale will result in the solution taking longer to converge but do so in a more stable way? Ive had some simulations where the max mach number is reached quickly and ive been recommended to apply a smaller timescale.