HRR curve and sprinkler's activation for usage in FDS

[blackarrow90]

Hello,

I am trying to design a HRR curve which represents a fire in one of the shops in a shopping mall, and use it for FDS simulation. Estimation of sprinkler's activation time is done using Alpert's correlation in the SFPE Handbook, and the fire is set to be fast growing. The maximum HRR value is determined from the literature as well. One thing that I did not get is what happens after the activation of sprinklers? "Capping" the HRR e.g. keeping the HRR constant after sprinkler's activation seems unreasonable for me, since the sprinklers extinguish the fire and thus lowering HRR. It does not become fully extinguished in seconds, it would take some time, but is there any way to estimate the last part of the HRR curve?

Thanks.

[dr_jfloyd]

Do you know exactly what will be burning in the shop (material, configuration,etc.)?

Do you know the degree to which the burning object(s) are shielded from water spray due to display racks, signs hanging from the ceiling, etc.?

If this simulation is for design purposes, the answer to these questions will be no, meaning there will be no reliable way to estimate the decay (or even state that the fire is fully extinguished).  

[blackarrow90]

I do not know exactly what will burn, even if I knew, that could change over time (it is a shopping mall).

In the paper "Swedish Best Practice Guideline for CFD" I found recommendation: once the sprinklers are activated, keep it constant for another minute, and in the next minute reduce it to 1/3 HRR and then keep it constant again till the end of simulation. Is it reasonable and are there other recommendations I could read? 

Thanks.

[Riccardo Smania]

Hi I have your same problem. 

Where do you found the paper " Swedish Best Practice"? because i need some information and validation in order to model a sprinkler activation on tunnel fire.

However I would share some my doubt with this group:

I found some information about E_COEFFICIENT but I'm little bit confused:

FDS_User Guide used only value higher than 1 on its user guide example.

 "Numerical simulations on the performance of waterbased fire suppressions systems" by VTT that seem they suggest value k = ( E_coefficient? ) = 0.4

NISTIR 4833 "A Sprinkler Fire Suppression Algorithm for the GSA Engineering Fire Assessment System" has proposed a k = 0.023 in their algorithm.

Which is true? 

Thank you

[fde]

k is dependant on many factors e.g. geometry of fuel and sprinkler type. It cannot be universal and needs to be defined a posteriori. 

[Riccardo Smania]

Thanks for reply.

Yes I understand that only through experimental test it's possible to have a correct K and there is too many factors involved in this issue.

However what I don't understand is which value I have to put in FDS MODEL:

For example I decide to consider the assumption made by D.Madrzykowski in NISTIR 4833 and consider algorithm HRRreduction factor(t)= e^(-kDT) with k = 0.023

in FDS model I have to write E_COEFFICIENT = 0.023?

I'm not very sure because on user guide is wrote that  k(t) = masswater(t)*E_COEFFICIENT.

I know that is a very big assumption but I don't have any possibility to do a fire,sprinkler tests.

Thanks again for the quickly reply.

Smania R. 

[Kevin]

E_COEFFICIENT in FDS has units of m2/(kg s) and k in Madrzykowski's report has units of 1/s. Like many empirical coefficients in engineering, the basic theory is the same, i.e. exponential decrease in HRR, but the implementation is different. The E_COEFFICIENT in FDS was inspired by Dan's work, and work being done at FM in the 1980s and 1990s on rack storage fires. However, the idea of an "extinction coefficient" to describe suppression of fire due to water spray has not really developed much in the past few decades. If you want to use it for a particular commodity, you need to run full-scale experiments because there is no other way to account for the wide variety of materials and geometric configurations. Even cardboard boxes or wood pallets are not as easy to describe as you would think.

[Riccardo Smania]

Yes, unfortunately I understand.

I was hoping that maybe I had missed some passages and maybe there was a solution.

In conclusion it's possible to confirm that sprinkler analysis on FDS nowadays is useful only for research analysis and it doesn't fit very well for design fire fighting system due to abscense of empirical data and full-scale experiments.

However I'm really glad of yours response and help that you gave to me.

Have a nice day.

Smania Riccardo

[Kevin]

Correct. Most fire protection engineers who use FDS tell me that they it use to predict the first sprinkler activation, and then apply an empirical assumption to the HRR. For example, when the first sprinkler activates, the HRR is held steady. This is based on the assumption that the given sprinkler system is working properly and has been classified appropriately for the given commodity. The model is not "predicting" anything except when the first sprinkler activates.

There is a set of experiments in the FDS Validation Guide called UL_NFPRF

https://github.com/firemodels/fds/tree/master/Validation/UL_NFPRF/FDS_Input_Files

in which we use the E_COEFFICIENT. The cases are called UL_NFPRF_P-N.fds (N=1-5). These are rack storage experiments, and the E_COEFFICIENT is based on extensive full-scale testing of the commodity. The idea of an E_COEFFICIENT makes sense in this case because the commodity is a "standard" commodity used in sprinkler listing tests. But, in general, if you are considering a houseful of "stuff", it's nearly impossible to develop a truly predictive E_COEFF for all possible fire scenarios.

[Riccardo Smania]

Thanks a lot.

Have a nice day