Appearance of species in cells when using F cycle

madfan123 · March 29, 2016, 10:04

Hello all,

I consider the gas flow in a pipe which in one place is connected with small channel system. Since I am considering only gas dynamics, all regions in the model are fluid and all walls are adiabatic. At some time in the inlet pipe the oxygen begins to be supplied, which propagates through the pipe and after some time begins to flow into the channels. During calculation process, when oxygen has not yet been entered into the channels in some areas of it (in the corners, where the computational grid is reduced) some elements themselves begin to emit the oxygen and become his source. At once after oxygen appears at pipe inlet it appears at many zones in model but with very low concentration (1e-20…1e-10) although it hasn’t reach these regions physically. If I turn off F cycle, concentration in these zones is 0 like it has to be.

Mesh is polyhedral (minimal orthogonal quality is 0.25 and maximum skewness is 0.75). Everywhere there are prismatic layers to resolve the boundary layer. Solver is coupled. Time step is 1e-3...1e-4 [s].

Does somebody know why this numerical problem appears and how to avoid it do not turn off F cycle?

Thanks in advance

LuckyTran · March 29, 2016, 10:40

Isn't that normal? This isn't an issue related to the F cycle either.

The problem with transport equations (advection-diffusion equations and all diffusion equations) is that they do permit solutions with long-range diffusion that you wouldn't find physically. Transport equations are inconsistent with special relativity after all.

The other thing is numerical diffusion, which significantly enhances the physical diffusion. This is where you need high quality grids and high order discretization schemes to limit the influence of numerical diffusion.

Diffusion rate is proportional to gradients, so if you suddenly add oxygen at the inlet when there was no oxygen there before then you've applied an infinite gradient.

madfan123 · March 30, 2016, 03:42

Thanks for good explanation. But anyway is it possible to avoid this problem, because it has significant influence on the results?

March 29, 2016, 10:04	Appearance of species in cells when using F cycle	#1
madfan123 New Member Jack Join Date: Sep 2012 Posts: 10 Rep Power: 14	Hello all, I consider the gas flow in a pipe which in one place is connected with small channel system. Since I am considering only gas dynamics, all regions in the model are fluid and all walls are adiabatic. At some time in the inlet pipe the oxygen begins to be supplied, which propagates through the pipe and after some time begins to flow into the channels. During calculation process, when oxygen has not yet been entered into the channels in some areas of it (in the corners, where the computational grid is reduced) some elements themselves begin to emit the oxygen and become his source. At once after oxygen appears at pipe inlet it appears at many zones in model but with very low concentration (1e-20…1e-10) although it hasn’t reach these regions physically. If I turn off F cycle, concentration in these zones is 0 like it has to be. Mesh is polyhedral (minimal orthogonal quality is 0.25 and maximum skewness is 0.75). Everywhere there are prismatic layers to resolve the boundary layer. Solver is coupled. Time step is 1e-3...1e-4 [s]. Does somebody know why this numerical problem appears and how to avoid it do not turn off F cycle? Thanks in advance

March 29, 2016, 10:40		#2
LuckyTran Senior Member Lucky Join Date: Apr 2011 Location: Orlando, FL USA Posts: 5,751 Rep Power: 66	Isn't that normal? This isn't an issue related to the F cycle either. The problem with transport equations (advection-diffusion equations and all diffusion equations) is that they do permit solutions with long-range diffusion that you wouldn't find physically. Transport equations are inconsistent with special relativity after all. The other thing is numerical diffusion, which significantly enhances the physical diffusion. This is where you need high quality grids and high order discretization schemes to limit the influence of numerical diffusion. Diffusion rate is proportional to gradients, so if you suddenly add oxygen at the inlet when there was no oxygen there before then you've applied an infinite gradient. CeesH likes this.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
[snappyHexMesh] Error snappyhexmesh - Multiple outside loops	avinashjagdale	OpenFOAM Meshing & Mesh Conversion	53	March 8, 2019 10:42
decomposePar -allRegions	stru	OpenFOAM Pre-Processing	2	August 25, 2015 04:58
[snappyHexMesh] snappyHexMesh matches wrong cells to CellZone	Siegunn	OpenFOAM Meshing & Mesh Conversion	4	July 31, 2015 06:10
reaction in particular cells	wlt_1985	FLUENT	0	April 11, 2011 02:41
[snappyHexMesh] snappyHexMesh won't work - zeros everywhere!	sc298	OpenFOAM Meshing & Mesh Conversion	2	March 27, 2011 22:11

March 30, 2016, 03:42		#3
madfan123 New Member Jack Join Date: Sep 2012 Posts: 10 Rep Power: 14	Thanks for good explanation. But anyway is it possible to avoid this problem, because it has significant influence on the results?