warning:Dij_inv is too small

luckyboy · October 30, 2021, 02:52

When I run a case of hydrogen injection, the warning occured: Dij_inv is too small, something wrong with the table calculation at the time hydrogen injection.

What's the mean of Dij_inv? How to deal with the warning?

Thanks and I'm looking forward to your reply.

dprobst · November 1, 2021, 10:22

Quote:

Originally Posted by luckyboy

When I run a case of hydrogen injection, the warning occured: Dij_inv is too small, something wrong with the table calculation at the time hydrogen injection.

What's the mean of Dij_inv? How to deal with the warning?

Thanks and I'm looking forward to your reply.

That indicates there is a problem calculating the diffusion coefficients. Are you using species_diffusion_model=1 with a transport.dat? There most likely is an input error.
Please email CONVERGE support directly so we can help debug the setup.

luckyboy · November 1, 2021, 22:10

Quote:

Originally Posted by dprobst

That indicates there is a problem calculating the diffusion coefficients. Are you using species_diffusion_model=1 with a transport.dat? There most likely is an input error.
Please email CONVERGE support directly so we can help debug the setup.

Thank you very much. The warning has been solved when I use species_diffusion_model=0. However, another problem occurs as shown in attachment. what causes the problem and how to deal with it?

dprobst · November 2, 2021, 09:49

That happens when the grid is under-resolved and load balance runs into a problem with cell pairing.

Possible solutions include refining the mesh (everywhere if it is coarse or with wall embedding)
Increase parallel scale (depending on what it is currently set to)

Finally, you could run with v3.0, which has improved load balancing

luckyboy · November 2, 2021, 10:08

Quote:

Originally Posted by dprobst

That happens when the grid is under-resolved and load balance runs into a problem with cell pairing.

Possible solutions include refining the mesh (everywhere if it is coarse or with wall embedding)
Increase parallel scale (depending on what it is currently set to)

Finally, you could run with v3.0, which has improved load balancing

Thank you very much. I will consider your suggestion.

November 2, 2021, 09:49		#4
dprobst Member Dan Probst Join Date: Jan 2016 Location: Convergent Science, Madison WI Posts: 38 Rep Power: 10	That happens when the grid is under-resolved and load balance runs into a problem with cell pairing. Possible solutions include refining the mesh (everywhere if it is coarse or with wall embedding) Increase parallel scale (depending on what it is currently set to) Finally, you could run with v3.0, which has improved load balancing __________________ Dan Probst Principal Engineer CONVERGECFD

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
[ICEM] Meshing of a cylinder containing a small hole on one face	Giova82	ANSYS Meshing & Geometry	1	August 12, 2015 06:49
patch a small field in flow	wu yu	FLUENT	5	April 9, 2014 16:45
SMALL value	C.C	FLUENT	2	November 1, 2013 05:07
small numerical fluctuations in the solution	nandiganavishal	OpenFOAM Running, Solving & CFD	4	June 18, 2013 03:01
why did Time step is too small in stiffbs?	satum	FLUENT	1	December 10, 2012 09:15