FDS6/FDS5

[Francis]

Hello.

Why the calculation in FDS 660 ends for "Numerical Instability ..." and in FDS 553 the calculation runs normály.:::?

Francis

[fde]

Model is highly complex. I cannot say what might be the problem. Simplifying it might lead to a solution. I would suggest starting with sprinklers. 

[Kevin]

A numerical instability can occur for many reasons. Many times, a simulation comes close to having an instability, but then recovers. It is possible that the calculation was on the borderline. There are many differences between FDS 5 and 6. I suggest you focus on 6 and determine where the instability occurs in your domain. Is it near a boundary? Near a sprinkler? Near a mesh interface? Once you determine that, there might be some suggestions to fix the problem.

[dr_jfloyd]

I ran the case with 6.6.0 and it appears to be a droplet issue. 6.6.0 introduced a modified evaporation routine and there was an issue with it that was fixed earlier this year and your case has some similar attributes. I am running now with the current source to see if it gets past the point of failure. If it does, a possible work around might be to use FDS 6.5.3.

[Francis]

well, I will try 6.5.3 (NS will open the first open sprinkler)

Dne středa 13. června 2018 11:43:43 UTC+2 Francis napsal(a):

[Francis]

Hello,

in FDS 653 the program runs without NS, but the sprinkler does not have a fire extinguishing effect.

I have been using the fire simulation software FDS more than 11 years – I have to mention that the version FDS 6 – compared to the version FDS 5 – seems to be very instable for real building objects (obviously the unwritten programmer’s rule applies here: “every even version is bad”).

Francis

 

Dne středa 13. června 2018 11:43:43 UTC+2 Francis napsal(a):

Hello.

Why the calculation in FDS 660 ends for "Numerical Instability ..." and in FDS 553 the calculation runs normály.:::?

Francis

[fde]

Do you set the E_COEFFICIENT in the calculation?

[dr_jfloyd]

I strongly disagree with your statement. FDS 6.6 was tested against over 2,000 verification and validation cases. About an order of magnitude more cases than FDS 5.5.3 was tested against. For a number of those cases FDS 5.5.3 would not pass current testing, since the cases were created as a result of fixing some issue with the software. We do not see as many reports about instabilities and anomalous behavior as we did with FDS 5.5.3.

In this particular case the instability appears to be resulting from the roz_benzin liquid pool, there is some error in the code in handling the heat transfer between the drops and the liquid pool. However, it should be noted that we don't currently have the proper physics for droplet interactions with a liquid pool to begin. Even if there were no instability, your results may not be accurate. The droplet model treats the liquid pool VENT as a solid surface when in reality drops could penetrate into the pool and have the burning liquid float on top. 

As a work around until this is fixed, you could run your pool by itself at the same grid resolution to see what you get for a burning rate and then just prescribe that to the pool using HRRPUA plus a RAMP to give you the correct burnout time.based on the pool thickness (a 2 mm thick pool isn't going to burn for much more than 30 s and the sprinklers are operating near 15 s)

[Francis]

Yet the task does not convert to the right solution (the sprinkler does not have a fire extinguishing effect)

Dne středa 13. června 2018 11:43:43 UTC+2 Francis napsal(a):

Hello.

Why the calculation in FDS 660 ends for "Numerical Instability ..." and in FDS 553 the calculation runs normály.:::?

Francis

[dr_jfloyd]

How do you know what the "right' solution is?  Are you modeling an experiment?

Have you run your case without sprinklers to see what the HRR is?  Is there no effect or just not as large an effect as you are expecting?

How have you determined the material properties for the burning materials?  Are these measured for the exact same materials or just pulled from some reference? Have you tested these properties using FDS 6 with a virtual cone test to ensure they give the proper burning behavior?

[dr_jfloyd]

The issue with the instability in FDS 6.6 is with the pool burning out.

The motivation in changing the evaporation routine between 6.5 and 6.6 was users having instability issues with very small droplets or droplets cooling hot, low density surfaces (these issues also occurred in 5.5.3). The evaporation routine in 6.5.3 and before was only partially implicit. Drop temperature and mass were implicit but the surrounding air or wall temperature was explicit. Under some conditions this resulted in oscillating overshoots or undershoots in gas or wall temperature that grew over time and result in an instability. The change in FDS 6.6 was to also make the wall and gas temperature implicit in the evaporation routine. The new routine accounts for the enthalpy in the first node of a wall cell to prevent a cold drop on a hot surface from removing too much heat.

In your case you have a liquid pool with no solid behind it. When the pool burns out, there will no longer be any mass in the wall and the evaporation algorithm fails with an instability. The FDS 6.6 changes were tested against wall cells that didn't burn away and wall cells that were part of obstructions that burned away. In the first case since the wall isn't burning away, there is always mass in the wall cell. In the second case, if the obstruction burns away, then the wall cell went away. In your case, since it is at the exterior boundary, the wall cell doesn't go away but it now has no mass. This was a use case that wasn't tested. This will be fixed in 6.7. 

This means there is a second, perhaps better work around for this issue. You could make roz_benzin two layers. The first layer is the pool, and the second layer is the floor (looks like that is the betzon_zel material). When the pool burns away there will still be a floor beneath it.

[dr_jfloyd]

I've run your FDS 6.6 case using the current FDS source (with the pool issue fixed) with and without sprinklers.  Your sprinklers are not at the ceiling; therefore, they will only operate if the head winds up immersed in the fire plume. In both cases the pool fire ignites the bottom of the fuel packages about the time the pool burns out. In the case with sprinklers, when the sprinkler operates the water spray pushes the fire plume away from the other sprinkler heads. The initial head is the only head that activates (at least for the first couple of minutes). The sprinkler is not centered over the initially burning package, so one side remains somewhat in the lee of the spay pattern and the bottom can still burn. 

[Francis]

And thus the task converges into a non-preferred version of FDS 553

Dne středa 13. června 2018 11:43:43 UTC+2 Francis napsal(a):

Hello.

Why the calculation in FDS 660 ends for "Numerical Instability ..." and in FDS 553 the calculation runs normály.:::?

Francis

[dr_jfloyd]

How do you know that FDS 5.5.3 is the giving right answer?

Neither FDS 5.5.3 nor FDS 6.6 have all the physics required to predict sprinkler suppression; neither version accounts for potential thermoplastic behavior of fuels polystyrene, polyethylene, and PVC; your fuel properties are all constant temperature and have nice round numbers (suggesting they may not be actual measurements), and your sprinkler spray pattern is a perfectly uniform distribution which is likely not the case unless the head is just a nozzle with no frame and deflector.