I am attempting to write data to a monitor file at a rate of 1.0E-5 s. The above settings produces two CSV files: one file containing what I am assuming is instantaneous data, the other, with the suffix _tavg, produces time averages using the 0.01s averaging interval. Nevertheless, it seems that the instantaneous data is only written at a rate of 0.01 s (the fluid timestep, whereas the solid timestep is ~7E-7). Is the write interval constrained by the fluid timestep? Or am I not understanding the monitor setup?
The subroutine WRITE_OUTPUTS is responsible for checking time intervals and writing outputs (monitors and other files) if appropriate.
Examining the code, this subroutine is called in time_step.f (the fluid time step) and also in des/des_time_march.f, but in the latter case it is only called for DEM only (no fluid phase) cases:
! Initialize time stepping variables for coupled gas/solids simulations.
IF(DES_CONTINUUM_COUPLED) THEN
IF(DT_TMP.GE.DTSOLID) THEN
FACTOR = CEILING(real(DT_TMP/DTSOLID))
ELSE
FACTOR = 1
DTSOLID = DT_TMP
ENDIF
! Initialize time stepping variable for pure granular simulations.
ELSE
FACTOR = CEILING(real((TSTOP-TIME)/DTSOLID))
DT = DTSOLID
CALL WRITE_OUTPUTS(.FALSE., .FALSE.)
ENDIF ! end if/else (des_continuum_coupled)
...
! The following section targets data writes for DEM only cases:
IF(.NOT.DES_CONTINUUM_COUPLED) THEN
! Keep track of TIME and number of steps for DEM simulations
TIME = S_TIME
NSTEP = NSTEP + 1
! Call the output manager to write RES and SPx data.
DLB = .TRUE.
CALL WRITE_OUTPUTS(.FALSE., .FALSE.)
ENDIF ! end if (.not.des_continuum_coupled)
I think the assumption is that for coupled cases, the WRITE_OUTPUTS in the fluid time step will be sufficient.
Ok maybe perfectly is the wrong word. this significantly alters what is written to *LOG. I can no longer see the progress of the simulation, i.e., fluid and dem number of iterations, solver wall times, etc. Currently looking into how to recover what I would consider ‘normal’ behavior.
regardless, the fix above does what I was looking for in the original post.
I see the first iterations (t = 0.0). After, I only get indications when VTPs are written. See attached for examples of what stdout (left) and the LOG (right) files look like.
! The following section targets data writes for DEM only cases:
!IF(.NOT.DES_CONTINUUM_COUPLED) THEN
IF(.TRUE.) THEN ! PZ 09 16 2025
! Keep track of TIME and number of steps for DEM simulations
TIME = S_TIME
NSTEP = NSTEP + 1
! Call the output manager to write RES and SPx data.
DLB = .TRUE.
!CALL WRITE_OUTPUTS(.FALSE., .FALSE.)
CALL WRITE_OUTPUTS(.FALSE., .TRUE.) !PZ 09 16 2025
ENDIF ! end if (.not.des_continuum_coupled)
IF(CALL_USR) CALL USR2_DES
IF(ADJUST_PARTITION) THEN
EXIT_LOOP = .TRUE.
RETURN
ENDIF
this at least outputs the elapsed time and estimated eta (very much like pure DEM simulations, which makes sense considering the logic).
I’m a bit late in following up on this. But I would like to note that I spoke way too soon: the fix Charles suggested ended up disabling the fluid solver after the initial timestep.
If you really need to save monitor data at such high frequency, here is something you can try:
Open model/output_manager.f, go around line 204: IF(any(MONITOR_DEFINED)) THEN
and copy that block.
Then open model/des/des_time_march.f go around line 451 (below the block that writes vtk files). Paste the block that writes monitors here (you may need to adjust it). This should force the monitor to be written within a DEM loop.
