Failure of tutorial PremBunsen2D
mameehan5 opened this issue · 8 comments
Hi all,
It seems like the PremBunsen2D tutorial case fails to run with the given inputs. On the first time step, an massive streamwise velocity is estimated. Attached is part of the output. Any idea what is going on here? Thanks!
Mike
SUNDIALS initialized.
MPI initialized with 112 MPI processes
MPI initialized with thread support level 0
AMReX (22.05-27-g5d88558d2aef) initialized
Successfully read inputs file ...
================= Build infos =================
PeleLMeX git hash: v21.10-118-g23397a2-dirty
AMReX git hash: 22.05-27-g5d88558d2
PelePhysics git hash: v0.1-1007-gb0d3f35b
AMReX-Hydro git hash: 343e20e
===============================================
==============================================================================
State components
==============================================================================
Velocity X: 0, Velocity Y: 1
Density: 2
First species: 3
Enthalpy: 24
Temperature: 25
thermo. pressure: 26
=> Total number of state variables: 27
==============================================================================
Initializing data for 2 levels
Initialization of Transport ...
Initialization of chemical reactor ...
Creating ReactorBase instance: ReactorCvode
Initializing CVODE:
Using atomic reductions
Mixture fraction definition lacks fuelID: consider using peleLM.fuel_name keyword
25 variables found in PMF file
580 data lines found in PMF file
Making new level 0 from scratch
Making new level 1 from scratch
Resetting fine-covered cells mask
==============================================================================
Typical values:
Velocity: 0.1 3.923852814e+179
Density: 0.6650455194
Temp: 1260.890134
H: -218228.8977
Y_H2: 0.0003458573634
Y_H: 3.14226256e-05
Y_O: 0.000241012551
Y_O2: 0.2074301803
Y_OH: 0.0006121461777
Y_H2O: 0.01451381058
Y_HO2: 5.371213803e-05
Y_CH2: 6.546128626e-06
Y_CH2(S): 4.950200534e-07
Y_CH3: 0.0002058628013
Y_CH4: 0.04688699785
Y_CO: 0.006938410704
Y_CO2: 0.01541563049
Y_HCO: 9.539561314e-06
Y_CH2O: 0.0002204527497
Y_CH3O: 1.360707325e-05
Y_C2H4: 0.0002558423542
Y_C2H5: 9.233090566e-06
Y_C2H6: 0.0003360756718
Y_N2: 0.762207865
Y_AR: -2.030418186e-127
==============================================================================
Est. time step - Conv: 1.556568412e-184, divu: 1e+20
Initial dt: 1.556568412e-187
Initial velocity projection: U: 0.1 V: 3.923852814e+179
>> After initial velocity projection: U: 0 V: 0
Est. time step - Conv: 3e+199, divu: 1.831467954e-05
Initial dt: 1.831467954e-08
Initial velocity projection: U: 0 V: 0
>> After initial velocity projection: U: 0 V: 0
Doing initial pressure iteration(s)
================ INITIAL ITERATION [0] ================
Est. time step - Conv: 3e+199, divu: 7.105853017e-173
STEP [-1] - Time: 0, dt 7.105853017e-176
SDC iter [1]
- oneSDC()::MACProjection() --> Time: 0.009342319332
- oneSDC()::ScalarAdvection() --> Time: 0.002244584262
- oneSDC()::ScalarDiffusion() --> Time: 0.004823582247
From CVODE: At t = 0 and h = 6.01889e-195, the corrector convergence test failed repeatedly or with |h| = hmin.```
All other processors fail as well.
Something is definitely up. I'll have a look at it, it should be pretty easy to fix.
Sounds good, thanks!
The fix in https://github.com/AMReX-Combustion/PeleLMeX/tree/fixBunsen2D should do the trick.
It should be merged in development soon, but you can also hop on that branch if you want to give it a try.
Got it! Thanks so much! I did hop on the branch, recompiled, and reran. Unfortunately I still received an error as follows
==================== NEW TIME STEP ====================
Est. time step - Conv: 1.500837686e-05, divu: 2.070807566e-05
STEP [89] - Time: 3.371614555e-05, dt 3.372312753e-06
SDC iter [1]
- oneSDC()::MACProjection() --> Time: 0.01127888262
- oneSDC()::ScalarAdvection() --> Time: 0.001656992361
- oneSDC()::ScalarDiffusion() --> Time: 0.03207364306
From CVODE: At t = 9.69709e-07 and h = 3.47613e-29, the corrector convergence test failed repeatedly or with |h| = hmin.
amrex::Abort::2::Aborting from CVODE !!!
SIGABRT
See Backtrace.2 file for details
Abort(6) on node 2 (rank 2 in comm 0): application called MPI_Abort(MPI_COMM_WORLD, 6) - process 2
TACC: MPI job exited with code: 6
This is using the exact same inputs, compilation flags, etc. from the tutorial case. I did try both intel and gnu compilation, both erroring at the same time step.
I'll admit I haven't run this case in a while and it's not part of regression tests we perform periodically. I remember putting together an initial solution that was a bit odd. I'll re-run to completion and ensure that it is possible to reach a stable premixed bunsen flame.
The error you get is most likely the result of the initial solution/increase step size and I think a couple of options we introduced recently should fix that.
Okay sounds good! Just let me know either way. It's not critical for me that the case runs to completion, I was able to drop the CFL condition lower to run the simulation further. This was more a test for me to ensure that I installed everything correctly, and since it broke, I wanted to bring it to you attention.
I've updated the case in the latest version of development. It does go through with ~2e-5 s step size, but this is definitely very coarse and there are significant overshoots of intermediate species.