Processes 1670234 Supplementary

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 21

Numerical simulations of a postulated methanol pool fire scenario in

a ventilated enclosure using a coupled FVM-FEM approach

“Supplementary Information”

Shashank S. Tiwari*1†,2, Shivkumar Bale3, Diptendu Das*4, Arpit Tripathi5, Ankit


Tripathi5, Pawan Kumar Mishra5, Adam Ekielski6 and Sundaramurthy Suresh*1

1
Department of Chemical Engineering, Maulana Azad National Institute of Technology, Bhopal 462003,
India
2
Department of Chemical Engineering, Institute of Chemical Technology, Mumbai 400019, India
3
Department of Chemical Engineering, University of Pittsburgh at Johnstown, Johnstown, PA, USA
4
Atomic Energy Regulatory Board, Mumbai 400094, India
5
Faculty of Business and Economics, Mendel University in Brno, Brno 61300, Czechia

6
Department of Production Engineering, Warsaw University Of Life Sciences, Warsaw, Poland

*
Corresponding Author: SST, DD, and SS will equally handle the correspondence at all stages
Address: Department of Chemical Engineering, Maulana Azad National Institute of Technology,
Bhopal 462003, India; Atomic Energy Regulatory Board, Mumbai 400094, India

This work has been jointly carried out in Maulana Azad National Institute of Technology, Bhopal
and Atomic Energy Regulatory Board, Mumbai.
Email Ids: che15ss.tiwari@pg.ictmumbai.edu.in (SST); diptendudas@aerb.gov.in (DD);
sureshpecchem@gmail.com / sureshs@manit.ac.in (SS)
S1. Assessment methodology for fire (FDS) and heat transfer (COMSOL) simulations

Figure S1. Assessment methodology for fire (FDS) and heat transfer (COMSOL) simulations

S2. Radiant heat flux empirical models

Heat transfer occurs by conduction, convection, and radiation simultaneously for any fire. Any
object near the fire will receive the heat from the fire by radiation. It is significant to determine the
heat flux of the object due to the heat that is incident on it from the fire to study the damages that
may occur.

S2.1. Point Source Model (Drysdale [1])

The point source model for predicting the incident radiant heat flux is based on the point source
assumption for the fire. Thus, a point source of thermal energy is located at the center of the flame,
and the energy radiated is assumed to be a specified fraction of energy released during combustion.
The incident heat flux incident on a given target from the point source will depend on the total
radiative energy output of the fire Q r and the view factor of the flame from the given target, which

varies inversely as the square of R ( R is the distance of the target from the center of the point

2
source). Figure S2 illustrates the point source model for predicting incident radiant heat flux. The
radiative energy output of fire is given as shown in the equation (2).

In Figure S2, θ is the angle between the target and line of sight from the target to the point source.
R is the distance from the point source to the leading edge of the target, L is the length from the
center of the flame to the leading edge of the target. H is the flame height.

If the pool has an L/W ratio approximately equal to unity, an equivalent diameter can be found for
non-circular pool fires using the equation(1).

4× A
D= (1)
π

Q r cos θ
q " = (2)
4π R 2

Q r = total radiative energy output of fire.

Q r = Q χ r (3)

χ r is the radiative heat fraction. The value of χ r depends on the burning of fuel. Generally,
hydrocarbon fires produce a sooty flame. Hence, the radiative fraction for hydrocarbon fires ranges
between 0.3-0.4, while alcohol fires burn cleanly. Thus, for methanol pool fires, the value of
radiative heat fraction is approximately 0.2. The values of radiative heat fraction have also been
related to the diameter of the pool, which can be given using the equation (4)

χ r = 0.21 − 0.0034 D (4)

The distance from the point source location to the edge of the target is found using the Pythagoras
theorem:

R = L2 + ( H / 2 )
2
(5)

The Point source model is a simplistic model for predicting incident heat flux from a pool. It gives
a reasonable estimation of heat flux at more considerable distances. The prediction of heat fluxes
using the point source model is within five percent of the actual incident heat flux.

3
Figure S2. The point source model
S2.2. Solid Flame Model

The solid flame model assumes the flame as a cylindrical and homogeneous blackbody radiator
with an average emissive power E. There are two different approaches for estimating the incident
radiant heat flux on a target depending on the position of the target. The first approach is one where
the target to which heat from the fire is incident is at the ground level, while the second approach
is for the case when the target is at a specific vertical distance above the ground level.
This model gives the following relation for finding the incident radiant heat flux from the fire to a
given target (Beyler [2]).
S2.2.1. Target at ground level

Figure S3 shows a typical flame considered in the form of a cylinder to determine the incident heat
flux at radial distances at ground level from the fire.
Here, E is the emissive power of the pool fire, which can be estimated by the equation (6).
E = 58 (10−0.00823 D ) (6)

F1→2 is the view factor between the target and the flame. The view factor is a function of the target
location, flame height, and diameter of the pool fire and lies between zero and one. When a target
is very close to a flame, the view factor approaches one since everything viewed by the target is
the flame. The flame is idealized as a cylinder with a diameter equal to the pool diameter, D, and
a height equal to the flame height, L. If the pool has a length-to-width ratio near one, an equivalent-

4
diameter relation for non-circular pools can be used to determine the flame height. Equation (7)
and Equation (8) give the formula for calculating the horizontal and vertical target orientation view
factor of cylindrical radiation sources, which are then used for finding the maximum view factor
given by the Equation (9).

Figure S3. Cylindrical flame-shape configuration factor geometry for vertical and horizontal
targets at ground level

 1  1
B−  ( B + 1 S − 1  A− 
)( ) −  S  tan −1 ( A + 1)( S − 1)
= 
S
F1→ 2, H tan −1 (7)
π B −1
2
( B − 1)( S + 1) π A2 − 1 ( A − 1)( S + 1)

F1→2,V =
1  h  h
tan −1  tan −1
( S − 1)
−
πS  s −1  π S
2
( S + 1)
(8)
Ah ( A + 1)( S − 1)
+ tan −1
π S A −1
2 ( A − 1)( S + 1)

h2 + S 2 + 1 1+ S 2 2L 2H
Where, A = , B= , S= , h=
2S 2S D D
2 2
F1→2,max = F1→ 2, H + F1→ 2,V (9)

5
S2.2.1. Target above ground level

The view factor, in this case, is calculated by dividing the cylindrical flame into two parts, the first
part represents the flame below the height of the target, while the second part represents the flame
above the height of the target.

Figure S4. Two-cylinder flame-shape configuration factor geometry above ground level
The two view factors explained above are determined using the Equation (10) and (11):

F1→2,V1 =
1  h1  h1
tan −1  tan −1
( S − 1)
−
πS  S −1  π S
2
( S + 1)
(10)
A1h1 ( A1 + 1)( S − 1)
+ tan −1
π S A −1
1
2 ( A1 − 1)( S + 1)

2L 2 H1 h2 + S 2 + 1
Where, S = , h1 = , A1 = 1
D D 2S

F1→2,V2 =
1  h2  h2
tan −1  tan −1
( S − 1)
−
πS 2
 S −1  π S ( S + 1)
(11)
A2 h2 ( A2 + 1)( S − 1)
+ tan −1
π S A2 − 1
2 ( A2 − 1)( S + 1)

2L 2H 2 h2 + S 2 + 1
Where, S = , h2 = , A2 = 2
D D 2S

The total view factor, in this case, is given by the sum of two configuration factors in Equation
(12).

6
F1→2,V = F1→2,V1 + F1→2,V2 (12)

S3. Ventilation modeling

FDS has a dedicated module for modeling Heating, Ventilation, and Air-conditioning (HVAC)
systems connected to the gas space of the fire simulation. The ventilation network is described as
a series of ducts and nodes. The nodes are placed at points where ducts intersect each other or the
CFD computational domain. The ducts are uninterrupted domains of fluid flow, which can
encompass elbows, expansion/contraction fittings, and various other fittings. The losses due to
friction and various duct fittings are assigned dimensionless loss numbers to the ducts. The node
losses are attached to the ducts as loss terms only appear in the duct equations. The module does
not presently store any mass. Therefore, mass flux into a duct is equal to the mass flux out of the
duct. The nodal conservation equations for mass, energy, and momentum are as shown in
Equations (13) and (14):

ρ u A
j
j j j =0 (13)

ρ u A h
j
j j j j =0 (14)

du
ρ j Lj = ( Pi − Pk ) + ( ρ g Δz ) j + ΔPj + 0.5K j ρ j u j u j (15)
dt

Where, L is the duct length, ρ is the density, u is the duct velocity, A is the cross-sectional area of
the duct, h is enthalpy of fluid in the duct, P is the pressure and K is the dimensionless loss
coefficient of the duct, subscript i and k represent nodes, and j a duct segment.

Note that only the node equations are solved for an HVAC system. In reality, the HVAC model in
FDS only simulates the movement of the flow, accounting for the height difference, but without
any consideration on the heat transfer and transient movement of smoke flows inside the ducts.
Thus, the pressure variation inside the enclosure is considered as nodal momentum equation as
shown in the equation (15).

S4. Heat transfer modeling in FDS

7
S4.1. Fluid-solid interface

FDS assumes that solid surfaces consist of multiple layers, with each layer composed of multiple
material components that can undergo multiple thermal degradation reactions. Heat conduction is
assumed only in the direction normal to the surface. Each reaction can produce multiple gas and
solid species. The details of the heat conduction equation for solid materials, plus the various
coefficients, source terms, and boundary conditions, including the computation of the convective
heat flux qc′′ at solid boundaries are given below.

The one-dimensional heat conduction Equation (16) for the solid phase temperature Ts[x,t] is
applied in the direction x pointing into the solid (the point x = 0 represents the surface)

∂Ts ∂  ∂Ts 
r s cs =  ks  + qs′′′ (16)
∂t ∂x  ∂x 

The source term, qs′′′ , consists of chemical reactions and radiative absorption in Equation (17):

qs′′′ = qs′′′,c + qs′′′,r (17)

q s′′′,c , is essentially the convective heat flux to the solid surface given by the pyrolysis models for
different types of solid and liquid fuels. q s′′′,r is the net radiative heat flux

The boundary condition on the front surface of a solid obstruction is shown in Equation (18):
∂Ts
−ks ( 0, t ) = qc′′ + qr′′ (18)
∂x

where qc′′ is the convective and qr′′ the radiative flux. If the radiation is assumed to penetrate in
depth, the surface radiation term, qr′′ , is set to 0.

S4.2. Convective heat transfer in FDS

The convective heat transfer coefficient, h, is based on a combination of natural and forced
convection correlations (Equation (19) and (20)):

8
qc′′ = h (Tg − Tw ) (19)

 1
k 
h = max C Tg − Tw 3 , Nu  (20)
 L 

where C is a empirical coefficient for natural convection. L is a characteristic length related to the
size of the physical obstruction, and k is the thermal conductivity of the gas. The Nusselt number
(Nu) depends on the geometric and flow characteristics. For many flow regimes, it has the form in
Equation (21):

ru L
Nu = C1 + C2 Ren Pr m ; Re = , Pr ≈ 0.7 (21)
m

For planar and cylindrical surfaces, the default values are C1 = 0, C2 = 0.037, n = 0.8, m = 0.33,
and L = 1 m. For spherical surfaces, the default values are C1 = 2, C2 = 0.6, n = 0.5, m = 0.33, and
L = D, the diameter of the sphere. Note that for a sphere, the coefficient for natural convection, C,
is assumed to be zero. It is possible to change these values for a particular application, but it is not
possible to find a set of parameters that is appropriate for the wide variety of scenarios considered.

S4.3. Surface emissivity for solid materials

The following are the solid materials used in the simulations carried out in the current investigation
along with the surface emissivity prescribed from them (as per Hodgman [3], DMIC [4], and
Gubareff et al. [5]): Stainless steel: 0.75, Mild steel: 0.66, Polyethylene: 0.95, Building bricks
(enclosure walls): 0.45.

For the calculation of the gray and mean absorption coefficients κ (κn), a narrow-band model,
RadCal (Grosshandler [6]), is combined with FDS. At the beginning of a simulation, the
absorption coefficients of gases are tabulated as a function of mixture fraction and temperature.
During the simulation, the local absorption coefficient is found from a pre-computed table. An
important consideration in computing the entries in the table is the fact that the radiation spectrum
is dependent on a path length due to the widening and overlapping of the individual lines. Thus
the “effective” absorption coefficient will be a function of the distance over which the line-of-sight
form of the RTE is integrated. In FDS, a fraction of the characteristic length of the computational

9
domain is chosen as the path length used by RadCal in computing effective gray gas absorption
coefficients.

S4.4. Equation of state for pressure predictions

The conservation equations are supplemented by an equation of state relating the thermodynamic
quantities density, pressure, and enthalpy; ρ, P and h. The pressure is decomposed into three
components as shown in Equation (22).

p = p0 − ρ∞ gz + p (22)

The first term on the right-hand side is the “background” pressure, the second is the hydrostatic
contribution, and the third is the flow-induced perturbation pressure. For tightly sealed enclosures,
the term p0 is allowed to increase (or decrease) with time as the pressure within the enclosure rises
due to thermal expansion or falls due to forced ventilation. The purpose of decomposing the
pressure is that for low-Mach number flows, it can be assumed that the temperature and density
are inversely proportional, and thus the equation of state can be approximated as Equation (23):

p0 = ρTR  (Yi M i ) = ρTR M (23)

S4.5. Conduction in FDS

FDS uses a one-dimensional equation to solve conduction. The equation (16) is the equation for
the x-direction where, q s′′′ is the combination of heat fluxes from chemical reactions and radiative

absorption.

S4.6. Conduction in COMSOL Multiphysics

COMSOL Multiphysics is a finite element method (FEM) based CFD solver. Like most other CFD
solvers, COMSOL also provides provisions for solving different physics problems; in addition to
this, COMSOL also allows manually entering the coupled partial differential equations. Being a
finite element method, COMSOL can numerically solve the heat transfer for complex geometries,
which are difficult to solve in FVM and FDM methods. The earlier version of COMSOL was
known as FEMLAB. COMSOL can be used when a moving mesh boundary condition is used; this
facility is not available in other commonly used FVM solvers. There is an excellent and seamless
interface between different physics, like heat transfer, fluid flow, etc. COMSOL also be coupled
10
with MATLAB readily for solving, adding new equations, etc., as there is a default option in
COMSOL for using MATLAB. There have been many successful validation and verification
studies for COMSOL, which confirm that it is one of the most promising CFD-based software for
solving heat conduction problems in particular. The Equation (24) solves conduction in
COMSOL:

∂T
ρCP − ∇⋅ ( k∇T ) = Q (24)
∂t

In steady-state simulations, the temperature does not change with time, and thus, the first term
becomes zero. In the equation (24), the density, ρ, the heat capacity, CP, and the thermal
conductivity, k, are the physical properties of the materials considered. At the same time, Q is the
heat source/ sink for which one or more equations can be added separately.

S5. Validation Studies

The velocities at distances closer to the centerline were found to have a greater value than the
center. This was because the buoyancy at the center is more significant than that near the vessel
walls, resulting in higher temperatures and velocities at the center. When observed in terms of
height, it was found that greater velocities exist for a height of 30 cm above the fuel surface,
followed by the velocities at 20, 12, and 6 cm, respectively. This can be accounted for by the fact
that at lower heights, the flow was dominated by chemical reactions. The eddies are generated due
to the reaction entrain at higher distances from the base of the flame. They are driven by the
combined effect of buoyancy and molecular mixing, resulting in greater velocities at these heights.
The centerline velocities at heights 6cm, 12cm, 20cm, and 30cm have been shown in Figure S5.

11
Figure S5. Comparison of centerline velocity at height (a) 6 cm (b) 12 cm (c) 20 cm and (d) 30
cm above the fuel layer for different meshes of present simulations against experimental results of
Weckman and Strong [7]
S6. Nomenclature for the postulated scenarios (Simulation I and II)

12
Figure S6. Nomenclature for the four sides of the polyethylene shield: Simulation I.

Figure S7. Nomenclature for the four sides of the mild steel layer: Simulation II.

13
S7. Physicochemical Properties of Methanol Fuel

The properties of methanol fuel burning and the rate at which burning occurs are the boundary
conditions required to be set up. The properties of methanol used are listed below in Table S1.
Table S1. Properties of Methanol

Methanol [8]
Property Values Used
Density 791.0 kg/m3
Specific Heat 2.535 kJ/kg.K
Thermal Conductivity 0.2007 kW/m.K
Boiling Temperature 65.0 ºC
Heat of Combustion 20000 kJ/kg
Heat of Formation 1100 kJ/kg

S7. Thermal properties of the materials used for postulated scenario

Table S2. Thermal properties of the materials used for postulated scenario

Material Emissivity Density Conductivity Specific heat


(kg/m3) (W/m.K) (kJ/kg.K)
Stainless Steel 0.8 7850 45.8 0.46
Mild steel 0.8 7880 54 0.49
Polyethylene 0.8 950 0.45 0.44
Concrete 0.85 2100 2 0.95

S8. Burning duration

The duration for which the fire burns is estimated using the Equation (25) given below.

V 4V
tb = = (25)
Aν π D 2ν

Here tb is the burning duration to be found, V is the volume of the liquid in m3, A is the area of

the pool in m2, ν is the regression rate, and can be calculated in Equation (26) :

m ′′
ν= (26)
ρ

14
 ′′ is the mass burning rate per unit area (kg/m2.sec), ρ is the density in kg/m3. Now, the
Here, m
known quantities for this particular case are as follows:

 ′′ = 0.022kg / m2 .sec , V = 200l = 0.2m3 , ρ = 792kg / m3 , A = 2 m 2


m
(calculations as per Peatross and Beyler [9])

The regression rate is calculated by substituting the values of mass burning rates in the equation
(26), which is then substituted in the equation (25) to get the burning duration. In the present
postulated case, the total burning duration of the fire comes out to be 3600 seconds. Depending on
this value, the simulation was run for 3600 seconds.

S9. The heat release rate for the postulated scenario

The heat release rate was calculated by FDS using its combustion model. As shown in Figure S8
below, the HRR shoots up to around 1000 kW within few seconds after ignition.

Figure S8. The temporally varying heat release rate for the 2m2 postulated methanol pool fire as
obtained from FDS simulations.
The HRR attains such high values within such as methanol is a flammable liquid, and hence the
entire pool catches fire instantly as soon as it comes in contact with the ignition source. Once the
ignition begins and the entire upper surface of the methanol pool starts burning, the heat release

15
rate reaches a peak value, and then a steady-state value is attained, which is found to be around
850-900 kW. This value calculated by the combustion model of FDS is analogous to the value of
880 kW that has been determined by the formula for heat release rate described by calculation for
which is below.

Q = m " ΔH c,eff A
= 0.022 ( kg / m2 sec ) × 20000(kJ / kg ) × 2(m2 )
= 880kW

Thus, it is concluded that the heat release rate values obtained by hand calculations and those
predicted by FDS are in the same range. This also means that the simulation of the methanol pool
fire has been set up appropriately, and predictions of other parameters would also yield reliable
results.

S10. Selection of the thickness of mild steel shield

250

200
Temperature (ºC)

150

100

50

0
10 50 100 150 180 200
Thickness (mm)

Figure S9. Maximum temperature of the polyethylene shield attained for mild steel shield ranging
from a thickness of 10mm to 200mm

16
As shown in Figure 11 of the present study (main manuscript file), the polyethylene shield cannot
withstand the temperature rise caused by the fire during a methanol pool fire. Thus, a mild steel
shield being fire resistant is used to prevent the polyethylene from getting damaged. Even when a
mild steel shield is applied over the polyethylene, the mild steel shield receives heat from the fire
(by radiation). The heat penetrates into the mild steel shield through the conduction mode of heat
transfer. Thus, the thickness of the mold steel shield should be such that the heat it receives from
the fire does not penetrate deep enough, preventing the polyethylene from attaining high
temperatures. As shown in Figure S9, a mild steel layer thinner than 180mm yields a temperature
of around 100ºC, which is close to the melting point of polyethylene and thus not safe. The
thickness of 180mm, although sufficient, it was decided to go for 200mm thickness to be on an
even safer side.

S11. Maximum Temperature values, used as input for COMSOL simulations

Table S3. The maximum value of temperature for each of the sides at different time durations

Time Side A Side C Side B Side D Unshielded polyethylene


0 25 25 25 25 25
100 26 25 25 25 40
200 27 25 25 25 50
300 28 27 27 27 58
400 30 28 28 28 70
500 35 31 33 33 90
600 40 35 36 36 110
700 44 40 41 41 120
800 47 42 43 44 137
900 50 44 47 47 138
1000 53 50 52 53 140
1100 55 51 53 54 142
1200 58 53 55 55 144
1300 64 55 59 59 146
1400 68 58 63 64 148
1500 72 65 67 66 150
1600 75 68 69 69 151
1700 78 70 72 72 153
1800 81 73 75 75 155
1900 84 76 78 78 157
2000 87 79 80 82 159
2100 90 81 83 83 161

17
2200 93 84 85 85 163
2300 96 87 89 89 165
2400 98 90 91 91 167
2500 100 92 93 93 169
2600 102 94 95 95 171
2700 104 96 97 97 173
2800 106 98 99 99 175
2900 108 100 101 101 177
3000 110 102 103 103 179
3100 112 104 105 105 181
3200 114 106 107 107 182
3300 116 108 109 109 183
3400 118 110 111 111 184
3500 120 112 113 113 185
3600 122 114 115 115 185

18
Figure S10. Temperature profile on (a) side A (b) side C (c) side D (d) side B and (e) top side
respectively after 800 seconds

19
Figure S11. Temperature profile on (a) side A (b) side C (c) side D (d) side B and (e) top side
respectively after 1600 seconds
References

[1] D. Drysdale, An Introduction to Fire Dynamics, John Wiley & Sons, 2011.
[2] C.L. Beyler, Fire Hazard Calculations for Large, Open Hydrocarbon Fires, in: M.J. Hurley,
D. Gottuk, J.R. Hall, K. Harada, E. Kuligowski, M. Puchovsky, J. Torero, J.M. Watts, C.
Wieczorek (Eds.), SFPE Handbook of Fire Protection Engineering, Springer, New York,
NY, 2016: pp. 2591–2663. https://doi.org/10.1007/978-1-4939-2565-0_66.
[3] C.D. Hodgman, CRC handbook of chemistry and physics., CRC Press, Cleveland, Ohio,
1978.

20
[4] W. Wood, DMIC Report 177: Thermal Radiative Properties of Selected Materials, Battelle
Memorial Institute, Defense Metals Information Center, 1962.
[5] G.G. Gubareff, J.E. Janssen, R.H. Torborg, Thermal Radiation Properties Survey: A
Review of the Literature, Honeywell Research Center ; Minneapolis-Honeywell Regulator
Company, Minneapolis, MN, 1960.
[6] W.L. Grooshandler, Radcal -- A Narrow-Band Model for Radiation Calculations in a
Combustion Environment, (1993).
https://www.govinfo.gov/app/details/https%3A%2F%2Fwww.govinfo.gov%2Fapp%2Fdeta
ils%2FGOVPUB-C13-d0301b96642f0167a45468dd42912cd7 (accessed April 17, 2022).
[7] E.J. Weckman, A.B. Strong, Experimental investigation of the turbulence structure of
medium-scale methanol pool fires, Combustion and Flame. 105 (1996) 245–266.
https://doi.org/10.1016/0010-2180(95)00103-4.
[8] M.J. Hurley, D.T. Gottuk, J.R.H. Jr, K. Harada, E.D. Kuligowski, M. Puchovsky, J.L.
Torero, J.M.W. Jr, C.J. Wieczorek, eds., SFPE Handbook of Fire Protection Engineering,
5th ed., Springer-Verlag, New York, 2016. https://doi.org/10.1007/978-1-4939-2565-0.
[9] M.J. Peatross, C.L. Beyler, Ventilation Effects On Compartment Fire Characterization, Fire
Safety Science. 5 (1997) 403–414.

21

You might also like