the potential of cfd simulations for the determination of

Universiteit Gent - UGent

Vakgroep Mechanica van Stroming, Warmte en Verbranding

www.FloHeaCom.UGent.be

The Potential of CFD Simulations for the

Determination of Smoke Motion in case of Fire

Prof. Bart Merci

Ghent University, Belgium

Faculty of Engineering and Architecture

Department of Flow, Heat and Combustion Mechanics




Overview

– Setting

– ACR - RCR

– Fire safety systems design

– Example: Car park

– Fire Safety Science Research

– Fire forecasting

– Concluding remarks




Setting

– This talk is NOT intended to be a review of the

SOTA.

– Detailed assessment of quality of simulations

itself is NOT the central issue of this talk.

– Important challenges, such as pyrolysis modeling

and interaction of water with smoke/fire, are NOT

addressed.

– So, what IS addressed?




Setting

– Use of computer (gas phase) simulations in the

context of smoke motion:

• Fire safety systems design

• Fire safety science development

• Fire forecasting




ACR - RCR

– ACR: Available (or ‘Applied’) Computing Resources

– RCR: Required Computing Resources

– What determines RCR?

– Is ACR > RCR?

– If not, what are the consequences?




ACR - RCR




Fire Safety Systems Design

– RCR: determine what is necessary to tackle the

problem at hand.

– Subsequently, set ACR equal to, or higher than,

RCR.

– Example: user-defined fire with soot yield and

radiative loss fraction CFD calculations for

smoke dynamics are feasible (as long as species

are not considered), unless the geometry is very

complex CFD, with SOTA models, has the

potential of providing accurate results for smoke

dynamics CFD can ‘help’ developing or

assessing a fire safety system design.





– State-of-the-art:

• CFD (note: zone modeling not to be ruled out!)

• User-defined fire (in design calculations)

• Providing source terms (valid assumption?)

and BCs: smoke concentrations can be

calculated in reliable manner.




Example: Car park fire safety

– Fire Safety Journal, Vol. 57 (2013).

– Full-scale experiments (> 100!), model-scale

experiments and CFD (>500!) simulations.

Fire source

Jet fans

Support beams (and columns)

Thermocouple lines

Extraction fans





– Flow patterns at z = 1.2m (low) and z = 2.65m

(near ceiling), 500kW fire.

OOOOO XXOXX OXXXO XXXXO

1.2m





– Flow patterns at z = 2.65m (near ceiling), 500kW

and 4MW fire.

OOOOO XXOXX OXXXO XXXXO

4MW

500kW





– But: the investment in the computational mesh

and/or subgrid scale modeling is essential.

– With classical LES: one or two orders of

magnitude higher than what we use, even in

academic research.

– Why? No technical reason… Reason: time and/or

money pressure.

– In other words: ACR < RCR is possible.





– Important questions:

• How reliable are the results if ACR < RCR?

• Do we still capture trends correctly?

• What safety margins do we need?

– Note: the proof of the pudding is often not in the

eating.

– Research needed.

– Bench-marking studies can provide important

insight.




Fire Safety Science Research

– Idea: bench-marking studies on cases relevant for

fire.

– Idea is not new. Much effort made and ongoing:

• FDS Verification and Validation (with

repository):

– Open (44 sets of experiments, 5000+ data points)

– Key target: make FDS code more ‘fail-safe’ clear

benefit

– Yet, response is reported1 to be ‘disappointing’.

1https://groups.google.com/forum/#!topic/fds-smv/jmIyOy2wzFA

https://groups.google.com/forum/#!topic/fds-smv/jmIyOy2wzFA




Fire Safety Science Development

– Idea is not new. Much effort made and ongoing:

• FM Global Open Source CFD Fire Modeling

Workshop :

– Open to all of us

– ‘Come together and share data’ – ‘collaboration’

– Key target: ‘to develop the capability of predicting

fire growth behavior and fire protection requirements

within the framework of physics-based models’2

2https://sites.google.com/site/firemodelingworkshop/

https://sites.google.com/site/firemodelingworkshop/





• Initiative: ‘MaCFP’ (International Workshop on

Measurement and Computation of Fire

Phenomena) series.

• Timing: final preparation in 2015, with first workshop

in 2017 (Lund).

• Systematic evaluation of the accuracy and reliability

of CFD simulation results with state-of-the-art models

as used now and with typical numerical settings

(including typical computational mesh cell size): high

on the priority list and crystallized into target test

cases.





• Note: repetitive small-scale experiments and full-

scale experiments also remain important.

• Small-scale, intermediate-scale and full-scale set-ups

provide complementary information for the engineer.

• Full-scale experiments: analyze ‘crashes’ (recall: the

proof of the pudding is often not in the eating).




Fire Forecasting

• The need for speed:

– Forecasting simulations must be faster than real-

time a clear need for speed.

– Current levels of ACR and RCR: use of CFD for

forecasting purposes practically impossible.

– Two (or more) zone modeling approach: far more

time efficient solution.

– CFD on very coarse mesh: not useful (key

advantage of CFD (solving momentum equations)

is lost if there is no accuracy in solving the

momentum equations).




Fire Forecasting

• The need for speed:

– Thus, it is better to use two-zone model

calculations and remedy deviations by taking

advantage of sensor data.

– Note: uncertainties in boundary conditions

(including the fuel and fire heat release rate)

there is a need for assistance by (sensor) data ‘on

the fly’ during the calculations anyway.




Fire Forecasting

• Sensor-assisted simulations:

– Necessary data during the simulations, in order to

circumvent uncertainties in boundary conditions.

– User-defined fire, in combination with thermal (and

ventilation) boundary conditions, as in fire safety

design calculations.

– Key difference from design calculations: UDF is

varied on the run, taking advantage of sensor

data, monitoring the fire as it evolves.

– Most important parameter: HRR (in combination

with thermal (and ventilation) boundary

conditions).




Rabot2012




Fire Forecasting

• Example: Rabot2012 full-scale tests




Fire Forecasting

• Example: Rabot2012 full-scale tests: flame spread




Fire Forecasting

• Example: Rabot2012 full-scale tests: flame height




Fire Forecasting

• Example: Rabot2012 full-scale tests: smoke height




Fire Forecasting

• Example: Rabot2012 full-scale tests: use in 2-zone

modelling




Fire Forecasting

• Example: Rabot2012 full-scale tests: use in CFD




Fire Forecasting

• Example: Rabot2012 full-scale tests: fire forecasting




Concluding remarks

– CFD simulations are a big help for:

• Fire safety system design calculations,

• Improving our understanding in fire safety

science (theory and model development and

validation),

• Fire forecasting.

– … if performed in a proper manner.




Concluding remarks

– Focus of talk was on gas phase phenomena.

– But: much research effort is required for multi-

phase phenomena (condensed phase):

• Pyrolysis,

• Interaction of water with flames / smoke,

• Fire suppression,

• …




Concluding remarks

– ACR (Available Computing Resources) vs. RCR

(Required Computing Resources):

• Determine the required accuracy, complexity

and modeling dimensionality;

• Carefully map it to what can be achieved with

available resources;

• In other words, the investment in complexity,

accuracy and modeling dimensionality must be

done with great care.




Concluding remarks

– ACR (Available Computing Resources) vs. RCR

(Required Computing Resources):

• Making the ACR and the RCR more

quantitative and assessing the consequences

if ACR < RCR are considered major and

imminent tasks for the fire safety science

community.




Concluding remarks

• Improvements in CFD fire simulations:

– By carefully importing knowledge and progress

from other research communities.

– Focus on fire specific aspects as a community.

– Bench-marking studies at all levels and scales

should play a central role in this effort, in order to

evaluate in a systematic manner the reliability of

CFD simulations of fires, with special attention to

cell sizes in the computational meshes used.

• Sensor-assisted fire forecasting simulations very

promising (with zone modeling for the time being).




Special thanks to:

• All present and former co-workers.

• All present and former sponsors.

• You for your attention (and questions coming up…).

the potential of cfd simulations for the determination of

Documents