Professional Documents
Culture Documents
Numerical Simulation of Fire Integrity Resistance of Full-Scale Gypsum-Faced
Numerical Simulation of Fire Integrity Resistance of Full-Scale Gypsum-Faced
A R T I C LE I N FO A B S T R A C T
Keywords: Fire design requires three fire resistance criteria, namely integrity, insulation and load-bearing. In particular,
Gypsum board repartition wall systems used in residential and commercial buildings are required to provide certification of
Integrity failure insulation and integrity fire resistance and have to be analysed in the context of the characteristic time scales of a
Fire safety fire. This paper presents a numerical methodology to simulate the fire integrity resistance of full-scale timber-
CLT
based wall systems exposed to fire. The main contribution of this paper, with regard to the existing literature, is
FEM
Thermal behaviour
the presence of cross-laminated timber (CLT) panel in the central part of the wall system, where the pyrolysis of
timber was modelled explicitly in the energy balance equation system. For this end, a user-subroutine, called
UMATHT, is developed and successfully implemented in the Abaqus finite element code for thermal analysis.
The falling off of the gypsum boards under fire was considered implicitly in the FE model based on experimental
observations. The obtained results show clearly the need to take into account explicitly the cracks and falling off
of the gypsum boards for an appropriate prediction of the integrity fire resistance in order to reduce expensive
experimental procedures.
∗
Corresponding author.
E-mail address: mourad.khelifa@univ-lorraine.fr (M. Khelifa).
https://doi.org/10.1016/j.ijthermalsci.2018.06.003
Received 15 December 2016; Received in revised form 13 March 2018; Accepted 1 June 2018
Available online 14 June 2018
1290-0729/ © 2018 Elsevier Masson SAS. All rights reserved.
V.D. Thi et al. International Journal of Thermal Sciences 132 (2018) 96–103
the behaviour of full-scale structural (or non-structural) components material relative to all the characteristic times for egress.
under fire. Timber structures have traditionally been built using heavy timber
Timber has an explicit role within the context of fire growth. Timber frames, with the walls being constructed of various materials such as
is a flammable material that when exposed to a fire will ignite con- interwoven branches and split logs in the very early versions of these
tributing to the growth of the fire. Thus, it will affect the rate of growth types of structures, and later using plastered panels and bricks [4]. The
of the fire and thus alter the times to reach untenable conditions. If coming of the industrial revolution marked the appearance of in-
burning or pyrolyzing it will generate combustion products, thus will dustrially built planks, boards and nails which spawned the appearance
introduce toxic and irritant species. Timber, if part of the structure of new and modern construction techniques, especially in the Europe
(load bearing or non-load bearing) will have a significant impact on and USA, where new framing methods reduced labour costs, increased
structural integrity but also will help define heat transfer through the the flexibility in construction and allowed for prefabrication [4]. These
infrastructure boundaries, thus will affect compartment temperatures. light frame construction methods are nowadays the predominant form
It is therefore clear that the presence of timber has the potential of of construction in residential and low-rise buildings. During the last
dramatically affect fire safety. Nevertheless, the real question that decade, prefabricated cross-laminated timber panels (CLT) for load
needs to be addressed is how significant is this effect. The significance bearing wall and floor assemblies have become increasingly popular not
of timber in fire is intimately related to the characteristic times involved only for residential but also for office, retail and industrial buildings
in the degradation, burning and loss of mechanical properties of the [5,6] in particular in the actual context of multi-storey timber
Fig. 1. Geometrical illustration of the studied wall systems and location of thermocouples:
(a) steel-stud gypsum-faced partition wall and (b) gypsum-faced CLT wall.
97
V.D. Thi et al. International Journal of Thermal Sciences 132 (2018) 96–103
buildings. CLT panels are not only recognized for their excellent re- materials: steel, rockwool and gypsum boards. The thermal beha-
sistance and stiffness characteristics, but also for their excellent thermal viour of those materials was modelled using gradually-modified
insulation and air tightness [7]. However, even if CLT panels are fa- thermo-physical properties gathered from the published literature
vourable in case of fire, as the risk of fire spread through void cavities is [1,20] and the corresponding standards [21];
reduced, in comparison to light timber frame constructions, they in- - Combustible materials, namely CLT panel. The thermal behaviour of
crease the fire load in the room compartment. timber exposed to fire is modelled by taking into account the dif-
Nowadays, the need for predicting the structural behaviour of ferent physicochemical reactions (Fig. 5) due to the pyrolysis, at
timber structures exposed to fire has gained increasing interest in the elevated temperature (> 300 °C), explicitly in the energy balance
context of more and more intensive use of CLT in modern constructions, equation (see Eq. (1)).
including multi-storey buildings. On the other hand, appropriate and
safe fire design of buildings in the context of multi-storey timber All the numerical simulations were run based on the standard fire
buildings [8] will requires the integrity, insulation as well as the load- ISO834-1, using convection coefficient h conv = 25 [W/(K.m2)] and
bearing criteria. In this context, the authors have developed and im- emissivityε emis = 0.8.
plemented in the Abaqus software a user-subroutine material called
UMATHT dedicated to the behaviour of timber under fire. The main
novelty of this work with regard to the existing literature [9–12] 2.1. Modelling of steel-stud gypsum-faced partition wall
dealing with the behaviour of timber structures under fire, is that the
energy sources (endothermic and exothermic) involved in the pyrolysis The main objective of this example is to calibrate the thermo-phy-
of timber exposed to fire are taken into account explicitly in the energy sical parameters of the thermal model, thus all the model parameters of
balance equation, instead of gradually-modified thermo-physical gypsum, steel-stud and rockwool have been chosen from the literature
properties which is standard in the literature. In the spirit of previous [1]. Thermal behaviour under fire of light non-load-bearing steel-stud
analytical one-dimensional pyrolysis models [13–20] dedicated to small gypsum-faced partition wall systems has been extensively studied ex-
scale wooden fuel particles or biomass gasification, we develop a 2D perimentally, by many researchers [1,5,6,21,22]. These studied include
finite element pyrolysis model to simulate the behaviour of timber measurement of temperature rating at different location depths, mid-
structures under fire. The model was successfully implemented in the high deflection/buckling of steel-stud, influence of the thickness of
ABAQUS commercial software via a user-defined subroutine called steel-stud as well as the mode of failure integrity. Numerical simulation
UMATHT. of the behaviour, during fire scenario, is presented in the published
Here, the developed pyrolysis model is applied to two examples literature by considering the standard ISO fire [2] curve provided by EN
gathered from the literature. The firt example is the simulation of fire 1991-1-2 [3] and by means of gradually-modified thermal and physical
integrity resistance of full-scale steel-stud gypsum-faced repartition properties of the different constitutive materials (gypsum, steel and
wall (Fig. 1a), experimentally and numerically studied by Nassif et al. rockwool). In particular, specific heat and thermal properties of the
[1]. The second example deals with gypsum-faced CLT timber wall gypsum boards are determined as a function of temperature by many
(Fig. 1b), studied experimentally Frangi et al. [5]. researchers [21,23–27], among others. Therefore, several data base
regarding the thermo-physical properties of gypsum board are available
2. Numerical modelling strategy in the published literature, however, the users are advised to select the
best suited data to the nature of gypsum that they are using [1], since
The studied wall systems include several materials: galvanised steel- the data are different from an author to another. For example, some
stud, gypsum boards (standard and fire proof), mineral wool and cross- authors consider one pic value of the specific heat corresponding to
laminated timber panel (Fig. 1). dehydration of the gypsum at about 120 °C, while some others consider
To model such a wall, we propose to consider two kinds of mate- two or three pic values at 120 °C and 650 °C, and at 120 °C, 200 °C and
rials: 650 °C, respectively, corresponding to two or three dehydration stages.
In the present study, the data published by Nassif et al. [1] (Fig. 2) were
- Non-combustible materials, namely insulating and finishing used in the numerical simulation of the gypsum board. While the
thermo-physical properties of steel-stud as a function of temperature
98
V.D. Thi et al. International Journal of Thermal Sciences 132 (2018) 96–103
admits symmetry, only one half of the model was meshed using 2325
and 12000 solid elements (C3D8RT element of Abaqus) for the gypsum
board and the rockwool, respectively, and 1092 shell elements (S4RT
element of Abaqus) for the steel-studs (Fig. 4).
99
V.D. Thi et al. International Journal of Thermal Sciences 132 (2018) 96–103
Blasi [18]. In the first model, it is assumed that wood decomposes on Table 1
three products, namely char, water/vapor and gas by three reactions. Di Thermo-physical parameters of the pyrolysis [17,18].
Blasi's model, however, assumes moisture-free wood which decomposes Pyrolysis reaction parameters
on char, tar and gas by three primary reactions, and then a portion of
the tar decomposes on gas and char by successive secondary reactions w→c (kc) w→t (kt) w→g (kg) t→c (kc2) t→g (kg2)
[18]. The primary reactions are identical to that of the Shen's model −1 6 10 6 6
Ai (s ) 2,66 .10 1,48.10 5,16.10 1.10 4,28.106
and they are endothermic reactions, while the secondary reactions are Ei (kJ/mol) 106.5 112.7 88.6 108 108
exothermic. Δhi (kJ/kg) −418 −418 −418 42 42
It is concluded that the thermal characteristics of the wood pyrolysis
is highly depending on the endo/exothermic mechanisms [16–20]. It is, Thermo-physical properties
therefore, proposed to develop and implement a user-defined sub- H Ci [J/(kg.K)] ρ0 [kg/m3] T0 [K]
10% Cw = 2300 470 298 [K]
routine in the Abaqus finite element code to obtain close-to-reality fire
Cl = 4200
behaviour of timber. Cc = 1100
The evolution of the temperature gradient, incorporating the energy Cv = 1580
source due to the pyrolysis, in a 3D wood sample can be described by Cg = 1100
Ct = 1980
the energy conservation equation as follows:
∂ ∂ ⎛ ∂T ⎞ ∂ ⎛ ∂T ⎞
[T(ρw c w + ρc cc + ρt ct + ρl cl)] = λ sx + ⎜λ sy ⎟ Note that all the thermo-physical properties of the CLT timber as
∂t ∂x ⎝ ∂x ⎠ ∂y ⎝ ∂y ⎠
well as the parameters of the pyrolysis reactions were adopted from
∂ ⎛ ∂T ⎞ [17,18] and are listed in Table 1. Several simulations have been per-
+ λ sz + Q''r
∂z ⎝ ∂z ⎠ (1) formed by the authors on small-scale applications showing that the
adopted parameters are generally well suited for spruce timber.
It is assumed that the energy source induced by the pyrolysis de-
2D finite element mesh was used to model the gypsum board-faced
composes on endothermic and exothermic reaction sources as follows:
CLT wall involving 268 DC2D4 elements of Abaqus for the gypsum and
Q''r = Q''endot + Q''exot (2) the rockwool and 402 DC2D4 elements of Abaqus for the CLT panel.
where
3. Results and discussion
Q''endot = k c ρw [Δh0c + (cc −c w)(T−T0)]+ k g ρw [Δh0g
3.1. Results from steel-stud gypsum partition wall
+ (cg −c w)(T−T0)]+ kt ρw [Δht0 + (ct −c w)(T−T0)]+ kv ρl [Δh 0v
+ (c v −cl)(T−T0)] (3) Fig. 6 displays the comparison between the measured thermal
profiles and the numerically-predicted ones at different thermocouple
and
locations (T1 to T4). It can be observed a fairly good agreement up to
Q''exot = k c 2 ρt [Δhc02 + (cc −ct )(T−T0)]+ k g 2 ρt [Δh0g 2 + (cg −c t)(T−T0)] about 50 min time of exposure to fire. Beyond this point the numerical
results show divergence as compared to the experimental measure-
(4)
ments. This is mainly caused by the falling off of the gypsum boards
The mass conservation equations for wood, char, gas, tar and liquid/ leading to direct exposure to fire of the thermocouples, as documented
vapor are as follows: in [1]. Thermocouples T2 to T4 demonstrate the effect of the insulating
∂ρw property of the rockwool, where the temperature profiles are still al-
= − (k c + k g + kt )ρw most constant up to 30 min time of exposure to fire.
∂t (5)
Fig. 7a depicts the thermal deflection measured at the mid-span of
∂ρc the steel-stud. It can be observed from this Figure that the numerically-
= k c ρw + k c 2ρt
∂t (6) predicted thermo-mechanical behaviour of steel-studs compares well
∂ρg with the experimental measurement as well as with the simulation
= k g ρw + k g2ρt
∂t (7)
∂ρt
= kt ρw − k c 2ρt − k g 2ρt
∂t (8)
∂ρl ∂ρv
= − kv ρl ; = kv ρl
∂t ∂t (9)
The chemical reactions for pyrolysis of wet wood are assumed to be
governed by the first-order Arrhenius law as follows:
−Ei
⎛ ⎞
ki = Ai . e⎝ RT ⎠ (10)
The thermal conductivity value of the solid wood was obtained from
those corresponding to wood and char and they are taken from [16]:
λ w = 0.166 + 0.369 × H (11)
λ c = 0.105 (12)
100
V.D. Thi et al. International Journal of Thermal Sciences 132 (2018) 96–103
Fig. 7. Thermo-mechanical behaviour of steel-studs: (a) thermal deflection curves at the mid-span of the steel-stud and (b) deformed shape.
results from [1]. However, beyond 50 min time of exposure, the final regarding the thermal profiles within the CLT panel, at various location
predicted deflection seems to be more much quick and higher than the depths, in order to appropriately validate the pyrolysis model. The
experimental one. This can be explained by the partial composite action authors, however, have run several simulations on the small size ex-
between the gypsum board and the steel-studs (screwed connections) of ample presented by Shen et al. [16] and the results seems to be in good
the non-exposed side which was neglected in the present study. On the agreement [29].
other hand, contrarily to that documented in [1], the simulations were Furthermore, Fig. 10 depicts the evolution of the thermal con-
run with and without taking into account the geometrical nonlinearity ductivity of the solid timber (see Eqs. (11)–(13)). The thermal con-
(model-GL and model-GNL) of the steel-stud and the results showed ductivity deceases following two stages: the first stage at about 100 °C
that this do not have any effect on the thermal deflection (Fig. 7a). corresponds to the drying process, while the second stage of decrease
takes place at about 300 °C and it corresponds to the conversion to
3.2. Results from gypsum-faced CLT wall charring.
Fig. 11 illustrates the numerically-predicted product fractions.
Fig. 8 displays the numerically-predicted temperature rise at ther- When temperature exceeds 300 °C, it seems that more gas and char are
mocouples T1 and T2 compared against experimental measurements. It produced and less tar is produced. Tar, however, seems to decrease
can be seen that the numerical predictions are reasonable as compared beyond 350 °C because of the secondary tar decomposition, leading to
to the experiments up to about 48 min time of exposure to fire. Beyond additional gas and char (see Fig. 5).
this time of exposure, the experimental measurement, recorded by
thermocouple T2, exhibits a quick rise in temperatures. This corre- 4. Conclusion
sponds to the time where the gypsum boards cracked and crumbled,
with it came the mineral wool, leading to the direct exposure of the This paper aimed to provide a numerical methodology for full-scale
thermocouples to fire [7]. Note that the cracks and degradation of the simulation of the fire resistance of load-bearing or non-load bearing
gypsum board exposed to fire is a complicated issue which is not ex- gypsum-faced wall systems used in modern constructions. The numer-
plicitly modelled here. ical modelling considers nonlinear thermo-mechanical behaviour of
It is worth noting that the divergence of the numerically-predicted steel-stud, falling off of the gypsum boards as well as the pyrolysis of
temperature at the thermocouple T2 (Fig. 8) is due to the insulating timber (CLT panel). The thermal behaviour of the gypsum and the steel-
property of the mineral wool. stud was considered through gradually-modified thermo-physical
Based on these results, the authors propose to carry-out the simu-
lation on two steps, to take into account, implicitly, the falling off of the
gypsum under fire: the first step corresponds to the time scale up to
48 min time of exposure, after that a second step takes place where the
insulating properties of both the gypsum and the mineral wool are
disregarded in order to heat directly the interface between the mineral
wool and the CLT panel (location of thermocouple T2). With this
methodology, it seems that the numerically-predicted thermal profile at
T2 is in line with the experimental measurement (Fig. 9). However,
during time exposure comprised between 12 and about 48 min the
predicted thermal profile at T2 exhibits a slight difference (Fig. 9). We
do not know for sure, at the time of this writing, the reason for this
difference but it could be, as documented in [1], that this is believed to
be caused by energy transportation via steam movement within the
insulating and finishing material cavities.
The pic observed at the thermocouple T2 at about 55 min time of
exposure to fire is due to the exothermic reaction involved in the pyr-
olysis of the CLT panel.
After about 55 min time of exposure to fire, the numerically-pre-
dicted the value of the charring depth for the CLT panel (8 mm) pre-
sents a fairly good agreement with the experimentally-measured values Fig. 8. Comparison between numerically-predicted temperature rises against
at different locations which were comprised between 5 and 10 mm [7]. experimentally-recorded ones at thermocouples T1and T2 for gypsum-faced
At this stage of the study, we do not have enough measurements CLT wall.
101
V.D. Thi et al. International Journal of Thermal Sciences 132 (2018) 96–103
Acknowledgement
References
[1] A.Y. Nassif, I. Yoshitake, A. Allam, Full-scale fire testing and numerical modelling of
the transient thermo-mechanical behaviour of steel-stud gypsum board partition
walls, Construct. Build. Mater. 59 (2014) 51–61.
[2] ISO 834-1, Fire-resistance Tests. Elements of Building Construction. Part 1: General
Requirements, International Organization for Standardization, Geneva,
Switzerland,, 1999.
[3] Eurocode 5 – Design of timber structures. Part 1–2: General – Structural Fire Design.
CEN 2004 (European Committee for Standardization), EN 1995-1-2, Brussels,
Fig. 10. Evolution of the thermal conductivity of the solid timber during the Belgium.
conversion process. [4] G.C. Foliente, Developments in performance-based building codes and standards,
For. Prod. J. 50 (7/8) (2000).
[5] M. Feng, Y.C. Wang, J. Davis, Thermal performance of cold-formed thin-walled
properties of the constituent materials, while the thermal behaviour of steel panel systems in fire, Fire Saf. J. 38 (4) (2003) 365–394.
[6] M. Feng, Y.C. Wang, An experimental study of loaded full-scale cold-formed thin-
timber panel was considered by considering explicitly the reactions walled steel structural panels under fire conditions, Fire Saf. J. 40 (1) (2005) 43–63.
involved in the pyrolysis of timber at high temperature (> 300 °C). [7] Frangi A., Bochicchio G., Ceccotti A., Lauriola M.P., Natural full-scale fire test on a 3
The following main remarks can be mentioned: storey Xlam timber building. Proceedings of the 10th World Conference on Timber
Engineering (WCTE), Curran Associates, Inc., 2–5th June 2008.
[8] V.D. Thi, M. Khelif, M. Oudjene, M. El Ganaoui, Y. Rogaume, Finite element ana-
- The obtained results can be considered as satisfactory until falling lysis of heat transfer through timber elements exposed to fire, Eng. Struct. 143
off of the gypsum boards; (2017) 11–21.
[9] Eurocode 5 – Design of timber structures. Part 1–2: General – Structural Fire Design.
- The geometric nonlinearity of the steel-stud under fire did not affect CEN 2004 (European Committee for Standardization), EN 1995-1-2, Brussels,
the thermo-mechanical behaviour; Belgium.
- Even if the composite partial action between the gypsum board and [10] S. Schnabl, I. Planinc, G. Turk, S. Srpcic, Fire analysis of timber composite beams
with interlayer slip, Fire Saf. J. 44 (2009) 770–778.
the steel-stud from the exposed side to fire can be neglected, the
[11] A. Menis, Fire resistance of Laminated Veneer Lumber (LVL) and Cross Laminated
non-exposed side seems to contribute to the global stiffness of the Timber (XLAM) elements. Italy: Università degli studi di Cagliari, (2012) (PhD
entire wall system; Thesis).
- The falling off of the gypsum boards under fire was taken into ac- [12] C. Erchinger, A. Frangi, M. Fontana, Fire design of steel-to-timber dowelled con-
nections, Eng. Struct. 32 (2010) 580–589.
count implicitly based on experimental observation. [13] B. Benkoussas, J.-L. Consalvi, B. Porterie, N. Sardoy, J.-C. Loraud, Modelling
thermal degradation of woody fuel particles, Int. J. Therm. Sci. 46 (2007) 319–327.
The developed numerical model, however, needs further and deeper [14] A. Fuentes, J.L. Consalvi, Experimental study of the burning rate of small-scale
forest fuel layers, Int. J. Therm. Sci. 74 (2013) 119–125.
improvements before it can be considered as a robust and a general tool [15] O. Séro-Guillaume, S. Ramezani, J. Margerit, D. Calogine, On large scale forest fires
for fire design. For example, the cracks and falling off of the gypsum propagation models, Int. J. Therm. Sci. 47 (2008) 680–694.
[16] D.K. Shen, M.X. Fang, Z.Y. Luo, K.F. Cen, Modeling pyrolysis of wet wood under
102
V.D. Thi et al. International Journal of Thermal Sciences 132 (2018) 96–103
external heat flux, Fire Saf. J. 42 (2007) 210–217. gypsum board/wood-stud walls exposed to fire, Fire Mater. 18 (1994) 297–305.
[17] W.C. Park, A. Atreya, H.R. Baum, Experimental and theoretical investigation of heat [24] G. Thomas, Thermal properties of gypsum plasterboard at high temperatures, Fire
and mass transfer processes during wood pyrolysis, Combust. Flame 157 (2010) Mater. 45 (2002) 37–45.
481–494. [25] D.J. Hopkin, J. El-Rimawi, V. Silberschmidt, T. Lennon, An effective thermal
[18] C.D. Blasi, Analysis of convection and secondary reaction effects within porous solid property framework for softwood in parametric design fires: comparison of the
fuels undergoing pyrolysis, Combust. Sci. Technol. 90 (1993) 315–340. Eurocode 5 parametric charring approach and advanced calculation models,
[19] R. Bilbao, J.F. Mastral, J. Ceamanous, M.E. Aldea, Modeling of the pyrolysis of wet Construct. Build. Mater. 25 (5) (2011) 2584–2595.
wood, J. Anal. Appl. Pyrol. 36 (1996) 81–97. [26] D.J. Hopkin, T. Lennon, J. El-Rimawi, V. Silberschmidt, Full-scale natural fire tests
[20] S.S. Alves, J.L. Figueiredo, A model for pyrolysis of wet wood, Chem. Eng. Sci. 44 on gypsum lined structural insulated panel (SIP) and engineered floor joist as-
(1989) 2861–2869. semblies, Fire Saf. J. 46 (8) (2011) 528–542.
[21] F. Alfawakhiri, M.A. Sultan, D.H. MacKinnon, Fire resistance of loadbearing steel [27] T. Lennon, D. Hopkin, J. El-Rimawi, V. Silberschmidt, Large scale natural fire tests
stud wall protected with gypsum board: a review, Fire Technol. 35 (4) (1999) on protected engineered timber floor systems, Fire Saf. J. 45 (3) (2010) 168–182.
308–335. [28] EN1993-1-2. Eurocode 3: Design of Steel Structures—Part 1–2: General Rules –
[22] M.A. Sultan, V.R. Kodur, Light-weight frame wall assemblies: parameters for con- Structural Fire Design.
sideration in fire resistance performance based design, Fire Technol. 36 (2) (2000) [29] V.D. Thi, M. Khelifa, M. El Ganaoui, Y. Rogaume, Finite element modelling of the
75–88. pyrolysisÐf wet wood subjected to fire, Fire Saf. J. 81 (2016) 85–96.
[23] J.R. Mehaffey, P. Cuerrier, G. Carisse, A model for predicting heat transfer through
103