How to couple navier-stokes to potential flow in a finite volume solver?

PhilipS · March 10, 2015, 11:03

Hello,
Does anyone know an efficient way of coupling steady NS equations (or RANS) to potential flow on non-overlapping domains, when both flows are solved by a colocated finite volume method? For transient flows, I found a solution that uses the velocity from the PF as a bc for the NS solver, and the pressure from NS is then used in the transient Bernoulli eq. to update the velocity potential on the interface for the PF solver.
For the steady case I tried a kind of pseudo time stepping in the Bernoulli eq but did not manage to get things nice & stable. Any suggestions, references...?

Thanks!

Philip

sbaffini · March 26, 2015, 18:22

The scalar potential gradient is, actually, a velocity field. Couldn't you use that for the coupling? Just guessing, never tried...

PhilipS · March 27, 2015, 02:56

Hi,

thanks for answering, and yes, you can/should use that. I am now skipping the Bernoulli eq for the coupling, and doing it in a more direct way by only matching the velocities with a deferred-correction method. Matched velocities yield both mass & momentum conservation over the interface - that seems good enough.
The steady Bernoulli eq is mainly tricky because if you plug in the pressure (from the NS side) to get an updated velocity/flow potential, the 1/2V.mag^2 term has 2 solutions and taking one of them would probably introduce locking the solution to either in- or outflowing, where you want to leave it free to move from in- to outgoing on the interface, and vv.

Ph

H0T_S0UP · March 27, 2015, 08:19

Could you solve the potential flow field first, then use the results as a boundary condition for the region where you use N-S? I would think this would work pretty well if you know where your flow field will be potential.

H0T_S0UP · March 27, 2015, 10:24

Actually, it just occurred to me there are codes already that do something similar to this. Interactive Boundary Layer Codes (XFoil) is an example), solve the potential field, then employ a boundary layer solver to resolve the flow near the surface. About a year ago I read a thesis where the Cebeci-Smith model was employed to approximate separation. Ill see if I can get the name

PhilipS · March 27, 2015, 11:05

Ok, would be nice if you could give a reference. I am just looking for a general coupling of Pot Flow with NS, in fact not with a special boundary layer code. The idea is to only use NS where it is necessary (viscous effects) and Pot Flow in the rest. As far as I know (and could find), this is often done when combining BEM with FVM or FEM for free surface flows around ships.

Philip

H0T_S0UP · March 27, 2015, 11:21

You need to look up interactive boundary methods. They are well documented. Im at work now I cant get yoy anything until later.

March 10, 2015, 11:03	How to couple navier-stokes to potential flow in a finite volume solver?	#1
PhilipS New Member Nederland Join Date: Mar 2015 Posts: 4 Rep Power: 11	Hello, Does anyone know an efficient way of coupling steady NS equations (or RANS) to potential flow on non-overlapping domains, when both flows are solved by a colocated finite volume method? For transient flows, I found a solution that uses the velocity from the PF as a bc for the NS solver, and the pressure from NS is then used in the transient Bernoulli eq. to update the velocity potential on the interface for the PF solver. For the steady case I tried a kind of pseudo time stepping in the Bernoulli eq but did not manage to get things nice & stable. Any suggestions, references...? Thanks! Philip

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Finite Volume Roe Solver	morteza08	Main CFD Forum	5	January 15, 2011 21:59
LBM Vs navier stokes equations in turbulent fluid flow modeling.	sharad_shevate	Main CFD Forum	0	August 3, 2009 02:25
transform navier-stokes eq. to euler-eq.	pxyz	Main CFD Forum	37	July 7, 2006 09:42
Free 2D Axisymmetric Potential Flow Solver for Matlab?	Jonas Larsson	Main CFD Forum	0	August 13, 2003 10:13
Eigenvalues and Eigenvectors for 3-D TVD Finite Volume Solver	Axel Rohde	Main CFD Forum	3	April 2, 2000 18:42

March 26, 2015, 18:22		#2
sbaffini Senior Member Paolo Lampitella Join Date: Mar 2009 Location: Italy Posts: 2,195 Blog Entries: 29 Rep Power: 39	The scalar potential gradient is, actually, a velocity field. Couldn't you use that for the coupling? Just guessing, never tried...

March 27, 2015, 02:56		#3
PhilipS New Member Nederland Join Date: Mar 2015 Posts: 4 Rep Power: 11	Hi, thanks for answering, and yes, you can/should use that. I am now skipping the Bernoulli eq for the coupling, and doing it in a more direct way by only matching the velocities with a deferred-correction method. Matched velocities yield both mass & momentum conservation over the interface - that seems good enough. The steady Bernoulli eq is mainly tricky because if you plug in the pressure (from the NS side) to get an updated velocity/flow potential, the 1/2V.mag^2 term has 2 solutions and taking one of them would probably introduce locking the solution to either in- or outflowing, where you want to leave it free to move from in- to outgoing on the interface, and vv. Ph

March 27, 2015, 08:19		#4
H0T_S0UP Member Alex Join Date: Jan 2014 Posts: 54 Rep Power: 12	Could you solve the potential flow field first, then use the results as a boundary condition for the region where you use N-S? I would think this would work pretty well if you know where your flow field will be potential.

March 27, 2015, 10:24		#5
H0T_S0UP Member Alex Join Date: Jan 2014 Posts: 54 Rep Power: 12	Actually, it just occurred to me there are codes already that do something similar to this. Interactive Boundary Layer Codes (XFoil) is an example), solve the potential field, then employ a boundary layer solver to resolve the flow near the surface. About a year ago I read a thesis where the Cebeci-Smith model was employed to approximate separation. Ill see if I can get the name

March 27, 2015, 11:05		#6
PhilipS New Member Nederland Join Date: Mar 2015 Posts: 4 Rep Power: 11	Ok, would be nice if you could give a reference. I am just looking for a general coupling of Pot Flow with NS, in fact not with a special boundary layer code. The idea is to only use NS where it is necessary (viscous effects) and Pot Flow in the rest. As far as I know (and could find), this is often done when combining BEM with FVM or FEM for free surface flows around ships. Philip

March 27, 2015, 11:21		#7
H0T_S0UP Member Alex Join Date: Jan 2014 Posts: 54 Rep Power: 12	You need to look up interactive boundary methods. They are well documented. Im at work now I cant get yoy anything until later.