[fpena]

Hello there, i have a terrible problem, it results that when i run a simulation with the app buoyantBoussinesq in my personal cumputer with 16 cores in openfoam 3.1 it goes well, but when i try to run the same app in a cluster in openfoam 3.2 the courant number goes up and the first iterations en it crashes, and i do not why it is happening , is the precompiled app in both versions, and i have tryed run a diferrent solver (a custom one) in the cluster with foam 3.2 and runs well as in my personal pc in foam 3.1, could someone help me plz ty

Time = 5400.0300000000007

Courant Number mean: 26.5121 max: 1666.96 velocity magnitude: 2966.55
DILUPBiCG:  Solving for Ux, Initial residual = 0.911188, Final residual = 7.42636e-06, No Iterations 62
DILUPBiCG:  Solving for Uy, Initial residual = 0.919441, Final residual = 9.4579e-06, No Iterations 55
DILUPBiCG:  Solving for Uz, Initial residual = 0.941604, Final residual = 2.72516e+09, No Iterations 1000
DILUPBiCG:  Solving for T, Initial residual = 0.998317, Final residual = 4.1805e-06, No Iterations 60
DICPCG:  Solving for p, Initial residual = 1, Final residual = 0.0950819, No Iterations 9
time step continuity errors : sum local = 1.64311e+10, global = -662.716, cumulative = -662.716
DICPCG:  Solving for p, Initial residual = 0.675918, Final residual = 9.89838e-08, No Iterations 240
time step continuity errors : sum local = 37805.4, global = -1002.59, cumulative = -1665.3
DILUPBiCG:  Solving for epsilon, Initial residual = 1, Final residual = 9.35798e-06, No Iterations 84
DILUPBiCG:  Solving for k, Initial residual = 1, Final residual = 1.12121e-06, No Iterations 15
bounding k, min: -3.27486e+06 max: 5.86224e+17 average: 5.59259e+13
ExecutionTime = 4.21 s  ClockTime = 6 s

Time = 5400.0350000000008

Courant Number mean: 2.457e+11 max: 4.17318e+15 velocity magnitude: 2.10058e+16
DILUPBiCG:  Solving for Ux, Initial residual = 1, Final residual = 6.92186e-06, No Iterations 357
DILUPBiCG:  Solving for Uy, Initial residual = 1, Final residual = 0.00108087, No Iterations 1000
DILUPBiCG:  Solving for Uz, Initial residual = 1, Final residual = 4.36233e-06, No Iterations 397
DILUPBiCG:  Solving for T, Initial residual = 1, Final residual = 0.0356456, No Iterations 1000
DICPCG:  Solving for p, Initial residual = 1, Final residual = 0.0991199, No Iterations 137
time step continuity errors : sum local = 1.13764e+12, global = -4.05105e+06, cumulative = -4.05271e+06
DICPCG:  Solving for p, Initial residual = 0.803679, Final residual = 46.9312, No Iterations 1000
time step continuity errors : sum local = 1.96568e+15, global = -1.13509e+08, cumulative = -1.17562e+08
DILUPBiCG:  Solving for epsilon, Initial residual = 0.992903, Final residual = 1.99109e-06, No Iterations 20
bounding epsilon, min: -3.30441e+15 max: 7.96184e+26 average: 1.1842e+25
DILUPBiCG:  Solving for k, Initial residual = 0.999995, Final residual = 3.99017e-06, No Iterations 13
bounding k, min: -1.24997e+14 max: 6.30641e+22 average: 1.19204e+20

[matejfor]

Just to make sure, you have exactly the same settings and on a case where it runs your Co_max looks reasonable not like in the log you posted, right?

Then your Initial condition and your numerical setup is very fragile and such a thing like run on a different system with different decomposition could be behind the problem. So I would try to: limit gradients for velocity and k and epsilon with cellLimitation. You can switch to upwinding everything at the beginning to see any difference in behaviour.

Maybe use to power of PIMPLE algorithm and increase the outer correctors to 2?

[Светлана]

What matejfor said, or do you have an opportunity to compile openfoam 3.1 on the cluster?