There is a great number of factors which affect the fire dynamics simulation time. Among them you can identify the following main groups of factors:
-
Simulation volume (size and characteristics of calculation area, fire spread time, during which simulation must be cunducted).
-
Physical phenomena, which must be considered during simulation (ventilation, leaks through doorways and windows, presence of wind and etc.)
-
Usage of multithread simulation tools.
Simulation volume
A simulation object is regarded as a physical volume where you must conduct fire dynamics simulation (i.e. a number of cells this object is devided into) as well as fire spread time, during which simulation must be conducted.
In both cases the dependency is almost directly proportional. That is, if you encrease a number of cells or fire spread time by 2 times, simulation time is also encreased by 2 times.
Simulation time dependecy on a number of cells and fire spread time creates possible optimization methods. However, while trying to cut down the simulation time you must try to keep a balance in order to get valid simulation results, which are applicable for further use.
Number of cells in the calculation area is defined as a relation of calculation area volume to the total cell volume. Therefore, to decrease a total amount of cells you must decrease calculation area size and increase calculation area cell size. If a scenario contains several calculation areas, then a total number of cells equals to a sum of cells in every calculation area.
Recommendations for positioning calculation areas
-
Place the calculation area only in those places of the building where the distribution of the dangerous fire factors is possible.
-
Position the calculation area so that it covers as little empty space outside the building as possible.
-
Avoid duplicating and overlapping calculation areas.
-
If the vent size or fire source size (or the size of the object on which they are located) is less than the size of the cell of the calculation area, position an additional calculation area with a smaller cell size covering only vent or fire source.
-
When the vent size or the fire source size (or the size of the object on which they are located) is less than the size of the calculation area cell, then it is impossible to run a simulation or simulation results can be incorrect. Vent or fire source cannot be taken into account in the simulation in this case.
-
Do not create multiple small calculation areas. It is better to create one large calculation area.
This recommendation in some way comes in conflict with the assertion, that the less the calculation area is, the less time the simulation takes. But you should consider the fact that when you increase a number of calcualtion areas, the application spends time for processing their results and this time can surpass the time which is spent for processing each single calculation area.
Recommendations for selection of the calculation area cell size
To select the cell size of the calculation area, you can use the following formula:
{width=20%}
where:
$D^{*}$ — the characteristic diameter of the fire source;
$Q^{*}$ — the capacity of the fire source;
$p_{∞}$ — air density;
$C_{p}$ — heat capacity of air;
T∞ — air temperature (m.u.: K);
g — the acceleration of gravity.
The ratio of the characteristic diameter of the fire source to the cell size does not have a definite value. During the validation, FDS developers used values from the range of several to one hundred units. See the Cell Resolution section of the FDS user manual. The most commonly used values belong to the range between 4 and 16. In other words, the following expression is true:
{width=15%}
where a is the cell size.
For example, with a fire capacity of 1500 kW (vehicle ignition over an area of 2 m²) with a cell size of 0.25 meters, the ratio of the characteristic diameter to the cell size is 4.5. This value belongs to the range of the most commonly used values in FDS validation.
Recommendations for selection of the fire spread time for a simulation
You must not conduct the fire dynamics simulation longer than neceasary.
To define a fire risk value, the fire spread time for a simulation must be specified with regard to the safety factor (i.e. fire spread time must exceed evacuation simulation time by 1.25 times). Do not enter a great value because it does not affect the validity of fire risk calculation results, but fire dynamics simulation takes more time.
In standard projects, when calculating fire separation distances, the most dangerous scenarios are considered, where the entire surface of the fire load is engulfed in flames from the very beginning of the simulation. However, it takes some time for a steady state situation to be established. Typically, this requires 2-3 minutes.
Phisical phenomena which are regarded during simulation
Phisical phenomena (ventilation, leaks through doorways and windows, presence of wind and etc.) which are regarded during the fire dynamics simulation increase simulation time.
The greater value you use for the above-mentioned factors, the more it affects the simulation time. For example, a high capacity smoke exhaust system increases the simulation time in a greater way than a low capaсity one.
Deploying multi-thread simulation tools
Fenix+ 3 application can conduct a fire dynamics simulation in the multi-thread mode. In this mode all processing load is devided among several processes (threads) which are run in parallel. In addition, every calculation area (in fact, a cell group in the FDS source file) is processed in a dedicated thread.
Typically, a total simulation time in the multi-thread mode is less than when all calculations are performed in a dedicated thread.
But you must consider that the more a number of threads is, the more time the applicatopn needs to align the results between them. If a number of threads is large, the deciding factor is not the time the application spends on processing calculation areas in several threads but the time it spends on aligning the results. If a number of threads is too large, then the simulation can require more time than when it is run with a fewer threads.
By default, a fire dynamics simulation is conducted with a single thread.
The number of threads must not exceed the number of logical cores of the processor.
The number of threads, which allows you to conduct the simulation with minimum time depends on the size of calculation areas, their mutual location, and the number of processor cores.
Therefore, it is impossible to create generic rules for selecting the optimal number of threads for simulation.
How to calculate RAM required for fire a dynamics simulation
Memory size (m.u.: Gb), which is required for running a fire dynamics simulation in one calculation area can be defined using the following formula:
{width=30%}
where:
width, length and height are the sizes of the calculation area;
a is the cell size.
If a scenario has several calculation areas, you must define the total memory required for all of them.
Total memory required for the simulation defined using the above formula is approximate. In fact, a real amount of required memory can differ. In practice the real amount of required memory is less than the calculated value.