Why I get unbounded solution with bounded scheme?

jorkolino · August 24, 2012, 23:54

I simulate a scalar field with inlet value equal to 1 and no sources/sinks in the domain (there is one more inlet with zero scalar value). Naturally, I'd expect to have solution within [0;1], however I get some negative values, the same order of magnitude as the positives, in some inlet-adjacent cells

Most curious, the solution converges to a high accuracy (1e-15), which means it should be correct

WTF?!! I use first order upwind for the scalar, nothing special to note, apart from that inlet diffusion is disabled. But even enabling it only diminished the problem, did not eliminate it.

I've noticed that the scalar diffusion coefficient plays role. Setting higher diff. coeff. leads to higher oscillation, i.e. more and higher magnitude negative values. Yet again, solution converges perfectly. I;ve tried relaxing, but as you guess, it helps convergence, which is no problem by itself, and does not change the final converged solution.

Touré · August 26, 2012, 21:53

Show the geometry. You may have recirculation.

jorkolino · August 27, 2012, 00:28

Here is the inlet colored by scalar cell values after convergence. The inlet supplies fixed flow velocity (0.0113m/s) with fixed value of scalar (=1). I don't get how circulation can change the values to negatives as no parts of the domain have sinks. Nothing can decrease value below zero, physically. The only possible twist is I've disabled inlet diffusion. So no diffusive flux from the inlet is expected, only convective. But still that should only possibly increase the scalar value in the nearby cells.

Touré · August 27, 2012, 12:40

What is your scalar function? I work with the level set method and you need to initialize the scalar level set function everywhere in the domain (volume) including all boundaries at the initial time step, but not only on the inlet boundaries. At the initial, the scalar function varies between 0 and 1, but when it evolves in transient (unsteady), it may have negative values depending of the evolution of the flow or for example if it's an outlet. The maximum value of you scalar is 0.0981 instead of 1. Is the value of the scalar of the inlet really fixed to 1?

jorkolino · August 27, 2012, 17:42

I solve steady state. The value of the boundary is 1, as is the boundary condition, but you won't see it on the pictures, becuse they show cell center values after convergence. I have tried different strategies to initialize, with 0, 1 and in between for all cells in the domain. With the same result. It simply converges to what you see. Negative values at some cells adjacent to inlet. And it converges well, reeeeally well (res 1e-15). I continue to think that the disabled inlet diffusion (which you control via the UDS panel) is the culprit, but I can not see why and how it will create negatives of that order.

August 24, 2012, 23:54	Why I get unbounded solution when I shouldn't?	#1
jorkolino Member George Pichurov Join Date: Jul 2010 Posts: 52 Rep Power: 15	I simulate a scalar field with inlet value equal to 1 and no sources/sinks in the domain (there is one more inlet with zero scalar value). Naturally, I'd expect to have solution within [0;1], however I get some negative values, the same order of magnitude as the positives, in some inlet-adjacent cells Most curious, the solution converges to a high accuracy (1e-15), which means it should be correct WTF?!! I use first order upwind for the scalar, nothing special to note, apart from that inlet diffusion is disabled. But even enabling it only diminished the problem, did not eliminate it. I've noticed that the scalar diffusion coefficient plays role. Setting higher diff. coeff. leads to higher oscillation, i.e. more and higher magnitude negative values. Yet again, solution converges perfectly. I;ve tried relaxing, but as you guess, it helps convergence, which is no problem by itself, and does not change the final converged solution. Last edited by jorkolino; August 25, 2012 at 15:53.

Thread Tools	Search this Thread
Show Printable Version Email this Page	Search this Thread: Advanced Search
Display Modes
Linear Mode Switch to Hybrid Mode Switch to Threaded Mode

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
grid dependancy	gueynard a.	Main CFD Forum	19	June 27, 2014 21:22
Bounded scheme gives unbounded solution	su_junwei	OpenFOAM Running, Solving & CFD	3	November 18, 2011 02:35
Doubt on Implicit Methods	analyse In India	Main CFD Forum	10	March 9, 2007 03:01
Could anybody help me see this error and give help	liugx212	OpenFOAM Running, Solving & CFD	3	January 4, 2006 18:07
Wall functions	Abhijit Tilak	Main CFD Forum	6	February 5, 1999 01:16

August 26, 2012, 21:53		#2
Touré Member Guiliguili Join Date: Aug 2010 Location: Montréal Posts: 97 Rep Power: 15	Show the geometry. You may have recirculation.

August 27, 2012, 12:40		#4
Touré Member Guiliguili Join Date: Aug 2010 Location: Montréal Posts: 97 Rep Power: 15	What is your scalar function? I work with the level set method and you need to initialize the scalar level set function everywhere in the domain (volume) including all boundaries at the initial time step, but not only on the inlet boundaries. At the initial, the scalar function varies between 0 and 1, but when it evolves in transient (unsteady), it may have negative values depending of the evolution of the flow or for example if it's an outlet. The maximum value of you scalar is 0.0981 instead of 1. Is the value of the scalar of the inlet really fixed to 1?

August 27, 2012, 17:42		#5
jorkolino Member George Pichurov Join Date: Jul 2010 Posts: 52 Rep Power: 15	I solve steady state. The value of the boundary is 1, as is the boundary condition, but you won't see it on the pictures, becuse they show cell center values after convergence. I have tried different strategies to initialize, with 0, 1 and in between for all cells in the domain. With the same result. It simply converges to what you see. Negative values at some cells adjacent to inlet. And it converges well, reeeeally well (res 1e-15). I continue to think that the disabled inlet diffusion (which you control via the UDS panel) is the culprit, but I can not see why and how it will create negatives of that order.