Floating Point Errors using RhoCentralFoam for Shock Tube

matthewshorter · May 31, 2016, 21:00

Hello,
I am trying to simulate an axisymmetric shock tube with a 150:1 pressure ratio across the driver and driven regions. Currently, I can run the simulation with a lower 15:1 pressure ratio and get the expected results for that case but I run into floating point errors almost immediately under the actual conditions (after <1e-6 seconds). The simulation uses room temperature helium with a driven section initial pressure of 0.1atm. The driver and driven sections are of unequal radii, with the driver being about twice the diameter of the driven section. All parameters are using the physical dimensions in the relevant SI base units except for the inviscid assumption. For BCs, I am using a fixedValue noslip condition for the wall velocity (slip was tried but with no effect), zeroGradient for wall pressure, and wedge for the front and back faces. What type of approach should I take to resolving this issue? I have rhoCentralFoam set up with the same solver configuration as the shockTube tutorial.

matthewshorter · June 9, 2016, 09:07

Update: using blockMesh instead of GMSH (hex instead of tets) seems to help. I think the infinitesimally thin shock wave is messed up when the nodes are all jumbled with a tet mess, introducing errors that eventually blow up the solver. Hex mesh looks much, much better.

May 31, 2016, 21:00	Floating Point Errors using RhoCentralFoam for Shock Tube	#1
matthewshorter New Member Matthew Shorter Join Date: Feb 2016 Posts: 10 Rep Power: 10	Hello, I am trying to simulate an axisymmetric shock tube with a 150:1 pressure ratio across the driver and driven regions. Currently, I can run the simulation with a lower 15:1 pressure ratio and get the expected results for that case but I run into floating point errors almost immediately under the actual conditions (after <1e-6 seconds). The simulation uses room temperature helium with a driven section initial pressure of 0.1atm. The driver and driven sections are of unequal radii, with the driver being about twice the diameter of the driven section. All parameters are using the physical dimensions in the relevant SI base units except for the inviscid assumption. For BCs, I am using a fixedValue noslip condition for the wall velocity (slip was tried but with no effect), zeroGradient for wall pressure, and wedge for the front and back faces. What type of approach should I take to resolving this issue? I have rhoCentralFoam set up with the same solver configuration as the shockTube tutorial.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
icoFoam - Floating point exception (core dumped)	File	OpenFOAM Running, Solving & CFD	26	March 6, 2021 04:26
Floating point exception error	lpz_michele	OpenFOAM Running, Solving & CFD	53	October 19, 2015 02:50
Temperature floating point exception	ayckrom	FLUENT	1	May 13, 2015 09:02
Floating point error and divergence detected	aannjj	FLUENT	0	July 2, 2013 03:44
Convergence moving mesh	lr103476	OpenFOAM Running, Solving & CFD	30	November 19, 2007 14:09

June 9, 2016, 09:07		#2
matthewshorter New Member Matthew Shorter Join Date: Feb 2016 Posts: 10 Rep Power: 10	Update: using blockMesh instead of GMSH (hex instead of tets) seems to help. I think the infinitesimally thin shock wave is messed up when the nodes are all jumbled with a tet mess, introducing errors that eventually blow up the solver. Hex mesh looks much, much better.