[charryzzz]

hi Foamers:


I am running a self-developed solver with combustion using seulex as OED solver.


The case is running with a time step of [deltaT 5e-07];


Then, I will get such warning:


FOAM Warning :
From function seulex::solve(scalar& x, scalarField& y, stepState&
in file ODESolvers/seulex/seulex.C at line 255
step size underflow :8.470329473e-22



So does anyone know why this warning appears? Will this warning affect the calculation result?


Here is the link for seulex.c

https://github.com/OpenFOAM/OpenFOAM...eulex/seulex.C

[HPE]

It looks like the step size is very small to the extent that it cannot be represented in floating-point computation. Is varying time-step size setting on, so that your input for the time-step (i.e. deltaT 5e-07) has been changed during the simulation?

[charryzzz]

Originally Posted by HPE

It looks like the step size is very small to the extent that it cannot be represented in floating-point computation. Is varying time-step size setting on, so that your input for the time-step (i.e. deltaT 5e-07) has been changed during the simulation?

Yes, it seems to be the case, and there is no such warning when I reduce the time step from 5E-6 to 5E-7, but the case will cost a much longer time.

[HPE]

Use fixed time step, and fix the time-step 1e-7 or lower instead of using variable time step.

[bany]

Originally Posted by HPE

Use fixed time step, and fix the time-step 1e-7 or lower instead of using variable time step.

Sorry for recalling this Thread.

When i use chemFoam to calculate the ignition delay time for PRF80 with seulex_KLU, the infinite warnings "step size underflow :0" occur, as shown in the attachment. The mechanism creates a Jacobian with a sparsity of 98%. The calculation is performed by chemFoam with fixed time step 1E-7. In fact, small time steps like 1E-8, 1E-9 are also attempted but with the same warnings.

seulex_KLU is the solver seulex with KLU sparse matrix library, provided in the article https://doi.org/10.1016/j.combustflame.2016.09.018.

[bany]

Originally Posted by bany

Sorry for recalling this Thread.

When i use chemFoam to calculate the ignition delay time for PRF80 with seulex_KLU, the infinite warnings "step size underflow :0" occur, as shown in the attachment. The mechanism creates a Jacobian with a sparsity of 98%. The calculation is performed by chemFoam with fixed time step 1E-7. In fact, small time steps like 1E-8, 1E-9 are also attempted but with the same warnings.

seulex_KLU is the solver seulex with KLU sparse matrix library, provided in the article https://doi.org/10.1016/j.combustflame.2016.09.018.

The problem has been solved. We changed another computer and cluster and the warning disappeared.