In recent years, some valuable cultural heritage sites, such as the
Notre Dame de Paris cathedral and Shuri Castle in Okinawa, have
been destroyed by fire (both in 2019). These tragic accidents and
the sense of loss they provoked are still fresh in our memories. We
have a mission to protect these symbols of history, tradition, and
culture, and to pass them on to future generations by preventing such
accidents from happening again.
In East Asia, including Japan, many historical buildings are made of
wood and so there is an impelling need to prevent such fires. As a
result, many facilities today are equipped with fixed and mobile water
guns and drenchers that do not affect the surrounding landscape,
while research and development of firefighting drones to extinguish
fires in high-rise buildings continues.
The effectiveness of such firefighting equipment is generally verified
by means of installation standards and on-site water discharge tests.
However, conducting field tests that include all realistic fire conditions
are impractical due to the cost and the physical damage to valuable
buildings. It is, therefore, beneficial to evaluate the firefighting
equipment’s effectiveness first by using numerical simulation. The
simulation considers the total amount of water applied to the building
from the water discharge, the trajectory of the water discharge, and
the direction of flow. Particle method computational fluid dynamics
(CFD) is well-suited to effectively handle the large number of droplets
and the free surface of the water as it flows over the building’s exterior
in these simulations.
This article presents simulation examples for three types of firefighting
equipment using Particleworks, a CFD software based on the Moving
Particle Simulation (MPS) method: a water discharge gun, a drencher,
and a firefighting drone. Water guns and drenchers are designed to
extinguish and prevent fires by spraying water, while firefighting
drones use powdered extinguishing agents, which requires the
powder model to be defined. Particleworks includes Granuleworks
software which uses the discrete element method (DEM). By either
coupling the functionality of MPS and DEM for liquid and powder
interactions, or by using them individually depending on the target
materials, the operation of firefighting equipment can be evaluated
realistically.
Modeling airflow in the MPS method
MPS uses a formula that reproduces the incompressible flow
of water and other substances and this formula is applied to
the representation of water being released. During an actual
fire, strong winds may be blowing which would affect the
flow of the water discharged from the gun, so that the water
doesn’t flow in the direction of the target. Moreover, even
in the absence of wind, water droplets and powder flying
through the air are subject to air resistance, which gradually
reduces their momentum and shortens their flight distance.
To simulate a water discharge gun while taking air into
account, we performed calculations using an air resistance
model. One way to analyze two-phase gas-liquid flow is
to use the finite volume method (FVM), also available in
Particleworks, and perform the simulation using FVM-MPS
coupling. However, in this case, we chose to use the air
resistance model to reduce the amount of computation.
Simulation of a water discharge gun in
high wind conditions
We chose a wooden structure called a Yosemunezukuri,
as the target building for the water gun simulation. A
Yosemunezukuri has a roof that slopes in four directions and
is one of the representative architectural styles of historical
buildings in Japan where there were used in many national treasures
such as the Great Buddha Hall of Todaiji Temple. The main shape
and dimensions of the simulation model are shown in Fig. 1. Four
water discharge guns are placed diagonally across the building, and
the discharge angle is set at 60 degrees from horizontal so that the
water can reach the top of the roof. Note that Particleworks uses CAD
data in STL format to define the shape of the structure, so meshing
is not required.
For the analysis conditions, the flow rate from each water discharge
gun was set to 0.15m3/s (150l/s), and the physical properties of water
and air were defined as general values. We set the particle diameter,
which is an index of resolution when fluid is modeled as particles, to
20mm in consideration of the calculation time, and used the pressure
explicit method to calculate the pressure explicitly.
Fig. 2 shows the result of a simulation of water discharge in windless
conditions. Even in such conditions, the flying water droplets are
subject to air resistance and the distance varies. Fig. 3 is a contour
plot showing the total amount of water on the roof. The wet area can
be seen, and the water discharge gun’s angle and position can be
evaluated in relation to the extent of the area covered by water.
Next, we simulated the water discharge gun with a strong, 25m/s wind
blowing around the building. The airflow velocity field that combines
spatial coordinates and airflow vectors is imported into Particleworks
in CSV format. The air resistance is given by a 3D interpolation of the
air velocity at each particle position.
The result is shown in Fig. 4. Since the wind is blowing to the right
in the figure, we can see that the water flow is pushed away and only
a part of the roof becomes covered with water. As such, it is difficult
to change the direction of the water discharge with a fixed water
discharge gun according to changes in wind speed and direction. It
is necessary to change the direction manually at the fire site, but this
is not safe.
The next step was to simulate whether the firefighting capability
could be improved by using a movable water discharge gun,
using a remote control to adjust the direction of the water
gun according to the situation. Fig. 5 shows the area of water
coverage when the gun is rotated by 60 degrees around the
vertical axis in the same 25m/s wind. We can see that the
area covered by water is wider with the movable type of water
discharge gun than with the fixed type. The final number of
particles was approximately 350,000 and the calculation time
was 30 hours using NVIDIA’s GeForce GTX TITAN X GPU. If
a faster GPU is dedicated to the numerical calculation, the
calculation time is estimated to be several hours.
Simulation of fire prevention using a drencher
A drencher is a fire extinguishing device in which pipes are
installed on the roof or exterior walls of a building, and the
sprayed water forms a curtain to prevent the spread of the
fire from the surrounding area. Water is also sprayed from
pipes embedded in the ground surrounding the building to
wrap the building in a curtain of water. Since it does not spoil the
landscape, it is often used as fire-fighting equipment for cultural
heritage properties.
The same building model was used for the drencher simulation.
Water spray nozzles with a diameter of 100mm were configured at
intervals of 5m, with a flow rate of 0.2m3/s per nozzle and a spray
angle of 6° (as shown in Fig. 6) so that the water sprayed by the
nozzles would spread radially. Particle diameter was set to 25mm.
To verify the drencher’s effectiveness in preventing the spread of the
fire, we created a situation where debris from a fire that had broken
out near the building would fly in. The fire debris was modeled using
DEM particles and the simulation was performed using MPS and DEM
coupling. A coarse-grained DEM model was used to avoid excessive
computational load, and the particle diameter was set to 50mm.
Fig. 7 shows the drencher’s process of forming the curtain of water.
The water spray reaches a height of around 27m above the ground
about two seconds after the water is discharged. It then descends
and, about five seconds later, the curtain of water is formed, and
a dynamic steady state is reached. We tracked the behavior of the
moving fire debris in this state. The simulation result showed that five
seconds after the fire debris reached the curtain of water, almost all of
it was blocked by the curtain of water and did not reach the building.
Fig. 8 shows the results of the MPS-DEM coupling analysis of the
fire debris. In the simulation, some debris from the fire reached the
vicinity of the building. But the actual fire was extinguished by the
water flow, so the drencher’s fire extinguishing effect is deemed to
be good enough. The final number of particles in this calculation was
about 2 million and it took about 230 hours using an NVIDIA GV100
GPU board. In fact, further reduction of calculation time is possible by
using the multi-GPU calculation capability in Particleworks.
Simulation of a firefighting drone
Research and development of firefighting
drones for high-rise buildings is currently
underway, some of which have reached the
practical stage. This drone firefighting activity
also considers the effects of wind, whether the
drone can be guided to its desired destination
without damaging valuable buildings, and
whether a limited amount of firefighting agent
can accurately reach the target. Here, we used
a five-story pagoda as the model of a high-rise
building, simulating the process of a drone
approaching the top floor and spraying the fire
extinguishing agent.
The five-story pagoda is about 31m high, and
the drone has a diameter of about 1.5m. An inlet
with a diameter of 60mm was defined for the
nozzle at the tip from which the extinguishing
agent was sprayed. The simulation assumed that the drone would
rise from the ground to the height of the top floor (18m) in ten
seconds, and that the fire extinguishing agent would be sprayed at
rate of 30m/s immediately after the drone rose. When a liquid fire
extinguishing agent is used, the drone reaches the top floor (18m)
in ten seconds. However, since the use of liquid fire extinguishing
agents can damage wooden structures, we used DEM to model a
powder-based fire extinguishing agent that is believed to cause less
damage to wooden structures. A coarse-grained model was used to
reduce the computational load, and the particle diameter was set to
6mm.
Fig. 9 shows the result of the
simulation and the reach of the
extinguishing agent from the
drone. It was evaluated in windless
conditions, but the extinguishing
agent’s trajectory in the presence
of wind can be tracked like that of
the water in the example described
above. Furthermore, by linking to
the motion dynamics simulation, it
is possible to configure more realistic conditions in which the drone
flies and fights fire under the influence of wind. In this simulation, the
number of DEM particles was about 90,000, and the calculation time
using GeForce TITAN X was 46 hours.
This article discussed the use of Particleworks and Granuleworks to
simulate fire extinguishing activities to protect important architectural
buildings from fire. MPS and DEM can handle not only industrial
products such as the water discharge equipment and drones
discussed here, but also those used for disaster prevention in floods,
tsunamis, and landslides, and in a wide range of environmental issues
such as clean energy fields, which are of increasing concern in the
world today. I hope that simulation technology will be used wherever
possible to help protect our heritage, traditions, and culture and to
realize a safe and comfortable future for everyone.
Fig. 8 – The drencher’s method of preventing fire debris from reaching the building (right: CG rendering)
References
[1] Koshizuka, S., Shibata, K., Kondo, M., Matsunaga, T.: “Moving particle
Semi-implicit Method”, Academic Press, ELSEVIER, 2018
[2] Cundall, P. A., Strack, O. D. L.: “A discrete numerical model for granular
assemblies”, Geotechnique 29, No.1, 1979, 47-65
[3] The conference proceedings of the Japanese Society of Computational
Engineering and Science, 2021
