ANSYS CFX Adaptive Timestep

aeroman · July 23, 2009, 20:59

I am interested in a time accurate solution for a multiphase (air/water) problem, whereby it is important that the courant number fall below 1. This is easy enough to set up in CFX, however I have an expression for the flow velocity which is dependent on the value of the current timestep.

Sooo...My question is: What expression (if any) can I use to call the current timestep into my equation, since it will be changing based on the adaptability criteria.

Any help help you can provide would be a huge help

aeroman · July 23, 2009, 21:58

errr..is it dtstep?

ghorrocks · July 23, 2009, 22:30

Yes, I think you are right. Have a look in the CEL expression reference section of the manual for further details.

Also CFX is an implicit solver and therefore is not restricted to Courant number 1 for most types of simulations. Why do you say you need Courant Number below 1?

Glenn horrocks

aeroman · July 23, 2009, 23:20

Glen,

Thanks for your reply. In a nutshell the problem in hand is "quite simply" a free surface penetration problem of a sphere entering water at a given velocity (thus the equation for velocity with time). Using CFX, since it doesn’t have a dynamic mesh capability, I am controlling the "ball velocity" by changing the inlet velocity as a function of the force at the wall which represents the spherical section. Understanding that CFX can solve discrete nonlinear systems at each time step (implicit), I have found that the accurate time stepping evolution of the fluid phenomenology is most important. This makes the time step key, especially in the early development of the flow. By defining the max courant number, I have been pleased with results obtained by essentially "driving" the time step in this fashion. HOWEVER....I would be most interested in your thoughts on this.

Thanks again for responding.

Sam

ghorrocks · July 23, 2009, 23:31

Hi,

The timestep size you use for a transient simulation should be determined by a sensitivity analysis. It is not uncommon for multiphase flows (especially ones with surface tension) to require very small timesteps of the order of Courant Number = 1, but don't be fooled into thinking Courant Number = 1 is some sort of hard limit. It is for explicit solvers, but not implicit ones. Implicit ones just get more accurate as the timestepping gets smaller and you just have to pick the timestep size which gives you the accuracy you require.

Does the sphere go through the surface at constant velocity or does it move in reaction to the forces on it?

Glenn Horrocks

aeroman · July 23, 2009, 23:45

Glen,

The sphere moves as Vnew=(g-(f/m))+Vold. Where f is force_x()@ball (ball is the spherical wall boundary), m is the ball mass and g is gravity. I am applying Vnew to an inlet boundary with each timestep.

ghorrocks · July 24, 2009, 02:28

Hi,

Have you considered doing it with the new immersed solid and 6DOF solver? It should work nicely for this type of problem. The 6DOF solver is a beta feature in V12.

Glenn Horrocks

aeroman · July 24, 2009, 09:58

Glen,

Yes, however I am still waiting for version 12. I won't have it untill the end of the month. Seems like an eternity.

ghorrocks · July 25, 2009, 07:37

You will need to contact the CFX distributor for access to the 6DOF solver features. While you are at it, download the V12 iso from the ANSYS Customer page website and save yourself a wait.

Glenn Horrocks

July 23, 2009, 20:59	ANSYS CFX Adaptive Timestep	#1
aeroman New Member Join Date: Jul 2009 Posts: 22 Rep Power: 17	I am interested in a time accurate solution for a multiphase (air/water) problem, whereby it is important that the courant number fall below 1. This is easy enough to set up in CFX, however I have an expression for the flow velocity which is dependent on the value of the current timestep. Sooo...My question is: What expression (if any) can I use to call the current timestep into my equation, since it will be changing based on the adaptability criteria. Any help help you can provide would be a huge help

July 23, 2009, 21:58	pesky users manual!	#2
aeroman New Member Join Date: Jul 2009 Posts: 22 Rep Power: 17	errr..is it dtstep?

July 23, 2009, 23:20	Thanks for the response	#4
aeroman New Member Join Date: Jul 2009 Posts: 22 Rep Power: 17	Glen, Thanks for your reply. In a nutshell the problem in hand is "quite simply" a free surface penetration problem of a sphere entering water at a given velocity (thus the equation for velocity with time). Using CFX, since it doesn’t have a dynamic mesh capability, I am controlling the "ball velocity" by changing the inlet velocity as a function of the force at the wall which represents the spherical section. Understanding that CFX can solve discrete nonlinear systems at each time step (implicit), I have found that the accurate time stepping evolution of the fluid phenomenology is most important. This makes the time step key, especially in the early development of the flow. By defining the max courant number, I have been pleased with results obtained by essentially "driving" the time step in this fashion. HOWEVER....I would be most interested in your thoughts on this. Thanks again for responding. Sam

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
how to map resultd from cfx to ansys?	ritesh	CFX	2	June 1, 2011 08:52
ANSYS CFX Adaptive Timestep	aeroman	Main CFD Forum	1	July 23, 2009 20:53
Exporting results from CFX to ANSYS ??	sohail ahmed	CFX	1	December 20, 2007 02:10
FSI using CFX and ANSYS	Bi Chang	CFX	2	May 10, 2005 05:47
ANSYS to acquire CFX	Fred	Siemens	0	February 18, 2003 22:03

July 23, 2009, 22:30		#3
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,850 Rep Power: 144	Yes, I think you are right. Have a look in the CEL expression reference section of the manual for further details. Also CFX is an implicit solver and therefore is not restricted to Courant number 1 for most types of simulations. Why do you say you need Courant Number below 1? Glenn horrocks

July 23, 2009, 23:31		#5
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,850 Rep Power: 144	Hi, The timestep size you use for a transient simulation should be determined by a sensitivity analysis. It is not uncommon for multiphase flows (especially ones with surface tension) to require very small timesteps of the order of Courant Number = 1, but don't be fooled into thinking Courant Number = 1 is some sort of hard limit. It is for explicit solvers, but not implicit ones. Implicit ones just get more accurate as the timestepping gets smaller and you just have to pick the timestep size which gives you the accuracy you require. Does the sphere go through the surface at constant velocity or does it move in reaction to the forces on it? Glenn Horrocks

July 23, 2009, 23:45		#6
aeroman New Member Join Date: Jul 2009 Posts: 22 Rep Power: 17	Glen, The sphere moves as Vnew=(g-(f/m))+Vold. Where f is force_x()@ball (ball is the spherical wall boundary), m is the ball mass and g is gravity. I am applying Vnew to an inlet boundary with each timestep.

July 24, 2009, 02:28		#7
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,850 Rep Power: 144	Hi, Have you considered doing it with the new immersed solid and 6DOF solver? It should work nicely for this type of problem. The 6DOF solver is a beta feature in V12. Glenn Horrocks

July 24, 2009, 09:58		#8
aeroman New Member Join Date: Jul 2009 Posts: 22 Rep Power: 17	Glen, Yes, however I am still waiting for version 12. I won't have it untill the end of the month. Seems like an eternity.

July 25, 2009, 07:37		#9
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,850 Rep Power: 144	You will need to contact the CFX distributor for access to the 6DOF solver features. While you are at it, download the V12 iso from the ANSYS Customer page website and save yourself a wait. Glenn Horrocks