Wrong position of subroutine PARTICLE_MASS_ENERGY_TRANSFER in part.f90

[gforney]

Please complete the following lines...

Application Version: FDS actual version source part.f90
SVN Revision Number:
Compile Date:
Operating System:

Describe details of the issue below:

The subroutine is called inside the subroutine SUBROUTINE
UPDATE_PARTICLES(T,NM) after calling the subroutine MOVE_PARTICLES(T,NM)
but it should be called from inside the subroutine MOVE_PARTICLES(T,NM)
within the loop 
SUB_TIME_STEP_ITERATIONS: DO N=1,NITER.
When called from inside the subroutine MOVE_PARTICLES(T,NM)the time step DT
has to be changed to DTSP.

Justification: time step subcycling is used in the subroutine
MOVE_PARTICLES(T,NM)to make sure that no particle crosses a mesh cell
without touch-down. The only reason for this is to make sure that the
evaporated mass from the droplet is continously fed to the cells along the
droplet trajectory. This is only the case when at each position of the
subcycling mass is fed to the cell by the subroutine
PARTICLE_MASS_ENERGY_TRANSFER(T,NM). 

Additional suggestion: Since the right hand side of the evaporation
differential equation seems to be highly nonlinear, additional accuracy can
be obtained by subcycling the PARTICLE_MASS_ENERGY_TRANSFER(T,NM) inside
the DTSP time loop - maybe this can be implemented as an option.

Original issue reported on code.google.com by mur@grs.de on 2008-03-25 09:46:13

[gforney]

We do need to do more work on evaporation, but we would prefer to integrate the 
evaporation and movement equations separately. In short, the time step for movement

is based on the droplet's speed, and the time step for evaporation would be based 
(mainly) on the droplet's size. 

I will mark this as "On Hold" to keep it as a suggestion when we look at this 
routine again. In the meantime, if you can suggest a robust integration scheme for

the evaporation, we'd like to hear it. 

Original issue reported on code.google.com by mcgratta on 2008-03-25 12:07:13

• Labels added: OpSys-All, Performance

[gforney]

Let's keep in mind though, that we may have to start integrating them jointly, since
they are so tightly coupled. Has anyone checked what fraction of heat and mass
balances  do those droplets, that require movement sub-time stepping, actually represent?

Adding myself to Cc.

Original issue reported on code.google.com by shostikk on 2008-03-26 12:59:45

[gforney]

The biggest issue with moving droplets is not the sub-time steps, but rather the 
drag coefficient. Currently, we assume the droplets are completely indepedent of 
each other. When the spray is dense, the drag coefficient should be reduced because

droplets are "protected" by their neighbors. 

For evaporation, it is the very small droplets that cause the problems, because they

evaporate so rapidly, and again, each evaporates independently of neighbors.

Original issue reported on code.google.com by mcgratta on 2008-03-26 13:16:36

[gforney]

Simo -- I have not had a chance to work on this. Has the issue arisen in your work 
with water mist?

Original issue reported on code.google.com by mcgratta on 2008-11-07 18:37:05

[gforney]

Kevin, before going home this friday, I just started runs of a single mist nozzle. 
One with standard FDS, and one with drag coefficient reduction formula that was 
published in Int. Journal of Multiphase flow. The current version is an extreme - as

if all the droplets were moving in a row. That is, I'm using a formula for the 
second of the two trailing spheres. If this shows any significant effect, I must 
find a way to compute how strongly I actually reduce the drag, based on local volume

fraction of liquid (which is sometimes quite grid sensitive, though).

Original issue reported on code.google.com by shostikk on 2008-11-07 20:22:38

[gforney]

I started this discussion and I would like to make a clear statement: The problem for
a single isolated droplet is very simple: There are three equations to be solved for
each droplet: mass, impulse, energy. These equations should be solved simultaneously
using the same integration scheme. 
Because of obstacles in the way, the simple explicit EULER is a good choice to keep
things simple. To keep mathematics straight all droplets should be integrated using
the same time step. In order to obtain reasonable results the time step should be
chosen small enough, i.e. the droplet should interact with each grid cell it is
passing through. This is what has to be done if we only look at the droplet. 

But now comes the big problem: the Lagrangian droplet is interacting with the
Eulerian flow field which is modelled as turbulent Smagorinski. I have not the
slightest idea how this interaction has to be modelled adequately.
But I have modified the part.f90 such that the droplets are integrated as described
above using a subtime step of Delta L / Umax = 0.1 and I have run many tests. The
finding is that the interaction has a substantial effect on the results. Cuuting off
by FLUXMAX is not the solution.
My personal findings are:
1) the droplet spray is not modelled adequately: primary break up, secondary break
up
and reduced drag are missing
2) The droplet - flow interaction is an unresolved issue but of substantial influence
 on the results
3) Final conclusion: We need specific experiments to develop reliable simulation models.

Original issue reported on code.google.com by mur@grs.de on 2008-11-08 19:45:16

[gforney]

I am transferring control over this Issue to Simo Hostikka, who has been working 
with the droplet routine recently. Before I leave, let me just say that NIST has 
invested millions of dollars in water spray experiments over the past few decades.
I 
appreciate that there are still issues, like break-up and turbulence interaction, 
but I do not think that these are as important as fire suppression, both solid and

gas phase. I would prefer to put resources into suppression. We can model water 
sprays reasonably well, but suppression is still in its infancy. For me, the biggest

single issue for droplets is the reduction in drag of the droplets near the nozzle.

If we can solve this problem, then I would suggest focussing on suppression issues.

Original issue reported on code.google.com by mcgratta on 2008-11-10 20:22:20

[gforney]

I am closing this issue. The droplet routines have been changed significantly since
this posting. Let's start a new issue if there are still remaining problems.

Original issue reported on code.google.com by mcgratta on 2013-02-26 15:00:15