Fire Safety Journal 137 (2023) 103752

Contents lists available at ScienceDirect

Fire Safety Journal

journal homepage: www.elsevier.com/locate/firesaf

Method of measuring the temperature of wood exposed to fire with type

K thermocouples
Manh Hung Nguyen *, Salah-Eddine Ouldboukhitine, Sébastien Durif, Véronique Saulnier,
Abdelhamid Bouchair
Université Clermont Auvergne, Clermont Auvergne INP, CNRS, Institut Pascal, F-63000, Clermont-Ferrand, France

A R T I C L E I N F O A B S T R A C T

Keywords: The study of wood behaviour in the event of fire is of increasing interest in the civil engineering field. Wood is
Charring rate one of the most prevalent combustible materials used in buildings. However, this material is protected by the
Timber in fire char layer during fire exposure. It can slow down the charring rate of timber in fire and protect the internal wood
Temperature measurement
structure. In fire tests, it is very important to use the most effective method of measuring the temperature of
Thermocouple K type
wood. In this paper, experimental tests are performed in furnaces and the charring rate is determined by
measuring the depth of the charred part of timber at the end of the time. This approach is compared to direct
measurements given by thermocouples positioned at different depths in the wood sample in different directions
(radial, longitudinal, and tangential). All test samples were tested in a furnace where three surfaces of the sample
were exposed to ISO-834 fire standards. The comparison between the two methods of measuring the charring
rate allowed us to evaluate the accuracy of the thermocouple measurements according to the implementation
adopted in the sample. Besides, the factors affecting the temperature measurement results are also mentioned in
this study such as the position of the thermocouple in the wood, the size of the measuring sample and the size of
the thermocouple.

1. Introduction The charring rate of wood at different depths is influenced by many

factors. It has been observed that the free water in the wood begins to
Wood materials are being used popularly as structural materials in evaporate and a part of this water migrates inside the section and plays a
buildings today because of their low carbon emissions and as a highly significant role in the thermal protection of the steel inner parts. Some of
renewable natural building material. In addition, this material has good the water vapour will migrate deeper into the sample (away from the
mechanical properties and is easy to transport and install for construc­ heat source) in the fire test of Mikkola, 1991 [4], Gronli & Melaaen,
tion buildings. Beyond its good mechanical properties, wood has inter­ 2000 [5] and Arne, 2014 [6]. When the char layer reaches its maximum
esting thermal characteristics. Despite it being a combustible material, thickness at a furnace temperature of around 1000 ◦ C the char layer will
its thermal properties imply it has good behaviour under elevated have a constant thickness as the char front advances into the wood due
temperatures. Indeed, in the case of fire, the formation of a thermally to the consumption of carbon at the surface [7]. One way to evaluate the
protective char layer allows having an important temperature gradient thermal behaviour of wood under elevated temperatures is to measure
in the section. Thus, it makes wood a good material for thermal insu­ the temperature of wood at different depths with thermocouples.
lation of other structural elements such as steel elements [1]. Various There are different types of thermocouples for different specific ap­
studies, especially concerning the behaviour of steel-timber connections plications. An exposed thermocouple will work best when high response
under fire exist [2,3]. Those studies show that timber can provide effi­ times are required, but an ungrounded thermocouple is better in cor­
cient thermal protection for steel parts. Generally, the wood at the rosive environments. The junction of the ungrounded thermocouple is
surface has a moisture content smaller than in the inner part, thus it electrically insulated from the sheath, which prevents electrical noise
implies a charring rate more important in the first centimetres but as the from interfering with the signal. This yields much greater temperature
char layer forms on the surface, the charring rate will become constant. measurement accuracy, especially with very low-level signals [8]. In

* Corresponding author.
E-mail address: [email protected] (M.H. Nguyen).

https://doi.org/10.1016/j.firesaf.2023.103752
Received 12 August 2022; Received in revised form 18 January 2023; Accepted 29 January 2023
Available online 3 February 2023
0379-7112/© 2023 Elsevier Ltd. All rights reserved.
M.H. Nguyen et al. Fire Safety Journal 137 (2023) 103752

Fig. 1. Instrumentation of solid timber panel (left) and CLT (right) with conductor direction and ply orientation [16].

Fig. 2. Fire resistance test setup: Radial direction (exposed side) [17,18].

Fig. 3. Thermocouple placement from unexposed side CLT (a) [20] and (b) [21], GLT (c) [22] and (d) [23].

addition, the influence of thermocouple size is also a notable factor. materials, many authors recommend installing thermocouples (TCs)
Until now, the focus has been on a standard thermocouple size of 1/16th parallel to isotherms, such as William, 2002 [13], Totaro, 2018 [14] and
in (about 1.5 mm) diameter. It is recognized that as the thermocouple Beck, 1962 [15]. Convection was seen to not affect (respectively not to
diameter is decreased, the convection coefficient increases [9]. In fire happen) in usual setups with drilling depths significantly larger than the
testing and research, thermocouples of type K are by far the most open volume between the borehole surface and the TC. For timber,
common because of their wide temperature range, which supports an Fahrni et al., 2018 [16] recommend only inlaying thermocouples par­
oxidizing atmosphere. However, at temperatures above 800 ◦ C, oxida­ allel to the temperature isotherms, to avoid underestimation of charring
tion may occur leading to substantial measuring errors. The thermo­ rates (Fig. 1). This thermocouple layout is done before making the
couples may also age when used for longer times at temperatures above specimen to ensure convenience and accuracy.
500 ◦ C and should therefore in such cases be calibrated about every 20 h Most of the arrangements of thermocouples in the experiments
of use [10]. According to the international standards for fire resistance depend a lot on the type of experiment. In experiments on wood beams
furnace tests, ISO-834 and EN-1363-1 thermocouples may not be used and columns, thermocouples are usually arranged on the exposed sur­
for more than 50 tests [11]. Schmid et al., 2018 [12] show that for the face of the sample to measure the charring rate of wood as in the study of
same depth, temperature changes vary greatly depending on the type of Owusu, 2019 [17], Boadi, 2015 [18] (Fig. 2) and Akotuah et al., 2015
thermocouple used and the installation method. [19], where thermocouples are inserted at different depths of the beam
To limit temperature measurement errors in low conductive section. These studies show a high difference in charring rates obtained

2
M.H. Nguyen et al. Fire Safety Journal 137 (2023) 103752

by thermocouples and obtained by char depth. Boadi showed that the

average value of the charring rate obtained by the thermocouples is
2.07 mm/min and 1.15 mm/min by the measured charring depths.
For tests on CLT (Cross Laminated Timber) and GLT (Glue Laminated
Timber), the installation of thermocouples is done before glueing the
wood components. So, they are placed inside the wood on the unexposed
surface as in the experimental studies on fire-resistant of CLT and GLT.
Fig. 3 shows 2 tests with CLT wall panel (Zöllig et al., 2016 [20],
Osborne & Dagenais, 2012 [21]), those for the GLT column (Andersen,
2017 [22]), and the tests on GLT beams (Kinjo, 2016 [23]). In most
studies on CLT and GLT, the thermocouples are usually arranged be­
tween the layers of wood before glueing the test sample.
In Andersen’s experiment, the temperature in the furnace was
controlled to follow the ISO-834 standard fire for 90 min. The charring
rate of 0.59 mm/min was observed. Cachim and Franssen have shown
that the charring rate of wood is dependent on numerous factors, such as
wood species (density, permeability or composition), moisture or di­
rection of burning (along or across the grain) [24]. For this reason, it is
important to study the impact of the thermocouple positions, on the
exposed and unexposed sides, considering the directions of the wood
fibres at different depths in the wood. Fig. 4. Test furnace.
The precision of the temperature measurement using thermocouples
is very difficult to guarantee as the sources of errors are important as
exposed by Schmid et al., 2018 [12]. Indeed, various parameters can
affect the precision of the measure as the type of thermocouple
(sheathed or wire), the size, the exact position, the orientation con­
cerning isotherms, the heat dissipation along the thermocouple, and the
imperfect contact with the sample.
Through these studies, we have noticed that a possible way to assess
the reliability of a temperature measurement evaluating the charring
rate is to compare the obtained measurement with a visual inspection of
the residual cross-section of timber after a certain time of exposure to
fire. The results are compared to those obtained by a finite element
model used to evaluate the influence of some parameters. The experi­
mental and numerical results make it possible to give indications of the
most relevant methods for measuring the temperature in wood.
The literature presents many research dealing with methods of
temperature measurements under fire conditions and the thermocouples
implementation at the wood level to insure correct measurements.
However, there are not so much comparative measurements which help
to get a quantitative idea of the temperature differences. On the other
hand, some studies still use temperature measurements which do not
respect the recommended implementation. For all these reasons it has
been decided to present those results.
Fig. 5. Section sketch of furnace setup.
We first focus on the implementation of thermocouples through an
experimental study carried out on composite steel-timber samples. The
method consists in setting the thermocouples between the wood and the
steel on the unexposed side. Then, the charring rate is determined by
measuring the depth of the charred part of the timber at the end of the
test. The obtained results confirm and complete the existing research.
Finally, in some cases, it is much more difficult to provide a tem­
perature measurement parallel to isotherms, such as for cross-laminated
timber panels. That is why some authors are obliged to come back to TC
disposed perpendicularly to isotherms. The author believes that sharing
those datas, and especially this type of comparisons between the
different orientations of thermocouples, it can help researchers to
develop a method to calibrate measurements which could have been
perpendicular to isotherm in order to get a more “realistic” value.

2. Materials and methods

2.1. Fire test setup

The tests are carried out in an experimental furnace made on-site at

IUT Clermont Auvergne. Those tests are made following the general
requirements of EN-1363-1 which describes the basic specifications Fig. 6. Temperature in furnace compared with ISO-834 curve.

3
M.H. Nguyen et al. Fire Safety Journal 137 (2023) 103752

Fig. 7. Thermocouple placement, size of K thermocouples and drill bits.

Fig. 8. TCs on the unexposed face.

applicable to all tests. 2.2. Methodology

The furnace is approximately 1 m3 in external dimensions. The
temperature in the furnace is measured by plate thermometers fixed at The first test was performed with an empty furnace without any
the 4 corners of the furnace as shown in (Fig. 4). The walls are covered samples. The performed test aimed at evaluating the capacity of the
with a thick insulating layer having good thermal stability Superwool furnace to reproduce the ISO-834 fire curve [26]. The average furnace
BLANKET about 150 mm in thickness and with refractory bricks at the temperatures measured with plate thermometers during the test are
level of the chimney and the hearth (Fig. 5). compared to the standard curve ISO-834.
They are made of a folded steel plate base (nickel alloy strip (0.7 ± After validating the ISO-834 in the furnace, experiments were con­
0.1) mm thick), which is fixed in a sheathed type K, 1 mm diameter ducted with a timber beam to measure the temperature in wood at
thermocouple and covered by an insulating material according to different depths with a type K thermocouple.
standard EN-1363 [25]. For the same depth, the temperature changes vary greatly depending
The furnace is equipped with 2 fuel burners each having a power of on the type of thermocouple used [14] and the installation method.
29 kW. The samples are solid timber beams and temperature measure­ In-depth temperatures within timber were recorded by 1.5 mm diameter
ments are made with type K thermocouples. The furnace temperature Inconel sheathed type K thermocouples, placed in the wood parts at
was measured with 4 plate thermometers (Fig. 6). The temperature different depths and following the different directions of the wood fibres
measured in the furnace was uniform; the thermocouples give results (Fig. 7). All the thermocouples used in all the experiments are semi-rigid
quite consistent with the standard ISO-834 temperature curve. metal sheath thermocouples with the hot solder isolated (no ground
In this paper, first of all, the furnace will be tested to ensure ISO-834 loops). It is possible to deform and bend the metal sheath to adapt it to
standard temperature conditions. Then, three experiments (Test n ◦ 1, 2 the situation without affecting the performance of the sensor.
and 3) were analyzed specifically to compare the effect of thermocouple Regarding the arrangement of thermocouples (TCs) on the unex­
position. Then, 10 similar experiments on single timber beams are pre­ posed side, thermocouples were inserted into the back of the specimen
sented with the results in Section 3. and parallel to the wood grain (Fig. 8).
The beams are placed above the burners and parallel to the flame.
There are two types of beam: two hybrid steel-timber beams of dimen­
sion 74 × 220 × 500 mm (Test n ◦ 2 and n◦ 3) and a timber beam of

4
M.H. Nguyen et al. Fire Safety Journal 137 (2023) 103752

the fire experiment is following the ISO-834 standard curve.

Test N◦ 1 studies the effect of the arrangement of 6 thermocouples on
the exposed and unexposed sides of the timber beam. On the exposed
and unexposed radial faces, two thermocouples are arranged at the same
depth (Table 2a).
Test N◦ 2 investigated the differences in the temperature evolution of
arrangements of thermocouples on wood in three basic directions on
wood. There are four groups of thermocouples as follows: radial group
on the exposed (3 thermocouples) and unexposed side (3 thermocou­
ples), tangential group (2 thermocouples) and longitudinal group (2
thermocouples) (Table 2b).
Test N◦ 3 studies the effect of the arrangement of thermocouples on
the exposed (3 thermocouples) and unexposed sides (6 thermocouples)
and compares the two groups of thermocouples perpendicular to the
isotherm (radial and tangential) on the unexposed side (Table 2c).
Fig. 10 shows the three basic directions in wood: radial, R (perpen­
Fig. 9. Position of the sample - Top view of the furnace.
dicular to the grain in the radial direction), longitudinal, L (parallel to
the fibre grain) and tangential, T (perpendicular to the grain but tangent
Table 1 to the growth rings). Longitudinal thermocouples were performed only
List of test samples. in Test n◦ 2.
The following experiments are introduced for the general purpose of
Test n Timber Density (kg/ Moisture content Fire
◦
m3) (%) exposure studying the effect of thermocouple position. The temperature of the
furnace during the fire experiment is following the ISO-834 standard
1 Douglas 567 12 2 sides
2 Douglas + 567 12 2 sides curve for 60 min. Tests n◦ 1 and 2 were performed on the same Douglas
Steel beam of the same size. The difference here is that test n◦ 2 has the
3 Silver fir + 435 11 1 side
Steel

dimension 74 × 220 × 500 mm (Test n ◦ 1). The sample is well protected

on both ends (Fig. 9). The timber beams are exposed on 1, 2 or 3 sides in
the fire tests. All thermocouples are installed in the middle of the beam
length (position L/2 = 250 mm).
The moisture content and density of each wood sample are evaluated
according to moisture content testing standards NF-EN-13183-1 [27].

3. Experimental results

3.1. Samples configurations

To study the effect of thermocouple positions in wood along iso­

therms under different fire exposure cases. Thermocouples were ar­
ranged parallel or perpendicular to the isotherm and the thermocouple
arrangement was on the exposed or unexposed side. Four tests are
conducted on timber specimens to study the influence of the thermo­
couple position in the wood and the number of faces exposed to fire. The
specimens are set horizontally in the furnace and are subjected to 1 and Fig. 10. Thermocouples along the radial (R), tangential (T) and longitudinal
2 sides fire exposure (Table 1). The temperature of the furnace during (L) directions of wood.

Table 2
Different configurations.
Test N◦ 1 Test N◦ 2 Test N◦ 3

a) Douglas b) Douglas with steel c) Silver fir with steel

5
M.H. Nguyen et al. Fire Safety Journal 137 (2023) 103752

stopped at 60 min to compare the temperature measurements and the

observed charred part of the different wood elements. To stop the
charring of wood just after stopping the furnace, it is necessary to put the
sample in water. The main desired result from this test is the comparison
of the value of the charring rate obtained by the calculated char layer
thickness and by the measured temperatures as shown in Table 6. As
provided by Eurocode 5 part 1–2, the position of the charred layer is
taken as the position of the 300 ◦ C isotherms. From this assumption, a
charring rate can be deduced from the thermocouples measurements. As
soon as a thermocouple reaches a temperature of 300 ◦ C it can be
considered that the carried layer reached the distance corresponding to
the thermocouple position.

3.2. Timber beam - two faces exposure test

In this experiment, there are two thermocouples for all depths in the
Fig. 11. Test N◦ 1 – Cross-section of specimen and location of thermocouples.
exposed and unexposed radial direction (Fig. 11). In this experiment,
there are two K thermocouples located at the same depth along the
addition of a C-shaped steel material on the unexposed surface. radial direction of the exposed side (2 thermocouples R10 at 20 mm and
The tests n◦ 2 and 3 were performed on the same configuration. At the 2 thermocouples R11 at 30 mm) and the same for the unexposed side.
end of these three tests, measurements of the residual cross-section were Figs. 12 and 13 presents the comparison of the temperatures for the
performed to calculate the thickness of the char layer formed after the different thermocouples disposed of in the wood samples at 20 mm and
fire test, and from there determine the average value of the charring rate 30 mm. Again, it can be observed that all thermocouples arranged from
of the wood. The value of this charring rate is compared with the an exposed face give higher temperatures compared to those measured
charring rate of wood obtained from thermocouples. The furnace was with thermocouples from an unexposed face. The average charring rate

Fig. 12. Temperature rise of timber as a function of time at 20 mm depth.

Fig. 13. Temperature rise of timber as a function of time at 30 mm depth.

6
M.H. Nguyen et al. Fire Safety Journal 137 (2023) 103752

Fig. 14. Test N◦ 3 – Test setup.

Table 3
Instrumentation of thermocouples.

N° Depth Exposed or Direction to

Disposition
(mm) Unexposed isotherm

Group 1

Group 2

Group 3

Group 4

on the exposed side of the tangential direction (T30, T31) is 0.60 mm/ level gradually increases to the temperature level of around 100 ◦ C. This
min. The average charring rate according to the position of thermo­ phase corresponds to the drying of the wood for approximately 10 min in
couples R20 and R21 has a value of 0.56 mm/min. However, a high this test.
charring rate of 0.86 mm/min was obtained by thermocouples on the At temperatures above 100 ◦ C, the wood enters the pyrolysis phase
exposed face following radial direction (R10, R11). and the temperature level increases rapidly. Char formation forms at
temperatures T ≥ 300 ◦ C.
Thus, an average carbonization rate of 0.6 mm/min can be found for
3.3. Steel-timber beam – two and one faces exposure test the groups of thermocouples coming from inside (unexposed face),
against a higher speed obtained for thermocouples coming from the
3.3.1. Two face exposure tests with steel outside of the sample (exposed face: R10, R11, R12) from 0.8 to 1 mm/
In this experiment, a 3 mm thick steel has been combined with wood min in this test n◦ 3.
to get an unexposed surface without deformation after the test, so that There are two tangential thermocouples (T30 and T31) located at a
we can measure the thickness of the char layer more easily. depth of 60 mm from the top of the timber specimen. As the combustion
The position of sample is always placed at the back of the furnace and proceeds, the position of the T30 and T31 sensors changes from parallel
this time with only two faces exposed to the fire: the front face and the to perpendicular, which legitimizes the fact that initially, they follow the
bottom face (Fig. 14). curves of the unexposed sensors and then that of the exposed sensor, also
A solid Douglas timber beam was tested in the furnace to study the perpendicular to the isotherms. This analysis will be explained using the
fire resistance performance of this beam according to the temperature numerical model in section 4.3.
curve of the ISO-834 standard. The initial values of the density and The char layer on the surface of the wood specimen isolated the fire
moisture content of the samples tested are respectively: ρ = 440 kg/m3 from the inside wood elements, the thickness of the char layer was
and H = 12 %. increased with exposure time, and the charring rate of the wood was not
Type K thermocouples are placed on both materials: steel and wood, constant through the fire tests.
at various depths of 20 mm, 30 mm and 40 mm. These thermocouples To compare with thermocouple measurement results, the remaining
are separated into 4 groups as presented in Table 3. unburnt wood layer has been measured to deduce the thickness of the
The results of the tests in terms of temperature as a function of the charred layer (Fig. 16). Thanks to the steel, part of the combustion can
time of exposure to fire at different depths of the thermocouples are be avoided on the unexposed side of the wood to ensure the measure­
summarized in Fig. 15. ments taken at the end of the test (Fig. 16b). Of course, the speed of
For a fire exposure time of t < 7 min, the curves show a small plateau wood carbonization is more important for the lower part of the beam
and the temperature is around 20 ◦ C. Beyond this limit, the temperature

7
M.H. Nguyen et al. Fire Safety Journal 137 (2023) 103752

Fig. 15. Evolution of the temperature of the thermocouples at a depth of 20, 30, and 40 mm.

8
M.H. Nguyen et al. Fire Safety Journal 137 (2023) 103752

exposed side, radial from the unexposed side, or tangential). For the
depth of 20 mm: it can be observed an important difference in temper­
atures in the function of the thermocouple disposition. Indeed, at a time
of 20 min of fire exposure, the temperature in timber differs by more
than 300 ◦ C between the thermocouple installed from the exposed and
the unexposed faces. Furthermore, the endothermic plateau at 100 ◦ C is
only observed for thermocouples used from an unexposed face. This
observation is common to the other depths. However, it can be observed
that the differences between exposed and unexposed face dispositions
reduce as the depth increases.
Through this experiment, the density of the wood also affects the
charring rate of the wood. Specifically, in two tests n◦ 2 and n◦ 3, the
wood in the test n◦ 2 had a greater hardness (567 kg/m3) than in test n◦ 3
(435 kg/m3). Therefore, the average charring rate of wood in test n◦ 2
was smaller than for test n◦ 3.
The 3 thermocouples of each group were exploited at the tempera­
ture of the formation of the layer of char at 300 ◦ C. The average charring
rate obtained using the thermocouples introduced from the unexposed
side in the tangential direction (group 3) is 0.65 mm/min. The average
charring rate according to the position of thermocouple group 2 has a
value of 0.60 mm/min. A high charring rate of 0.97 mm/min was ob­
tained with thermocouples placed from the exposed face (group 1). This
proves that the thermal conduction of the thermocouples arranged from
the faces exposed to fire is a factor that directly affects the measurement
results. Additionally, it can be seen that for all studied depths, the
disposition which gives the lowest temperatures in the radial disposition
comes from the unexposed face. The main conclusion of these results is
that the thermocouple disposition choice has a clear impact on the
temperature measurements.
To monitor the influence of the heat transfer of steel on the unex­
posed surface of the wood in the two tests n◦ 2 and n◦ 3, Fig. 21 shows a
comparison of the wood surface condition after the experiment. We
observed that after 60 min of the fire experiment, with test n◦ 2, the
Fig. 16. Charred layer after the test. wood surface was significantly affected, and char layers were formed. In
test n◦ 3, only the four corners of the beams were burned.
Table 5 shows the value of the charring rate calculated according to
the thickness of the char layer and according to the value obtained by
the thermocouple on the exposed surface. Based on the measurements
made by the thickness of char wood, the measurement on the unexposed
is the most accurate. The charring rate in the standard test depends on
species, density, moisture content and duration of exposure by White
and Norheim, 1992 [28] and to determine the charring rate of wood, the
three experiments above have shown the influence of the location of the
thermocouples on the exposed or unexposed surfaces. For these three
test samples on the Douglas and silver fir, thermocouples measured from
the unexposed surface become more reliable and accurate. In particular,
Fig. 17. Residual section (remaining wood). after our studies, these are thermocouples on the unexposed sides (radial
and longitudinal).
than the isolated upper part as shown in Fig. 16c. To study the effect on the appearance of steel materials, Fig. 22
The work was carried out by scraping the outer layer of char formed shows a comparison of the temperature rise rate of wood at 20 and 30
during the combustion test after 55 min with the digital caliper to obtain mm between two experiments n◦ 1 (timber beam) and n◦ 2 (steel - timber
the thickness of the sound wood inside (Fig. 17). We took 10 measure­ beam).
ments and calculated an average value of unburnt wood. Then calculate These two experiments were performed at the same ISO-834 furnace
the thickness of the char layer formed after the test. temperature. In test n◦ 2, the heat transfer on steel material made the
heat increase rate on wood happen faster than in test n◦ 1.
3.3.2. One face exposure test with steel In addition to the 3 experiments mentioned above, 10 other experi­
In this test, 3 mm thick steel has been combined with silver fir timber ments were performed on 4 different types of wood of the same
(Fig. 18). The thermocouples were well placed on both exposed and dimension: 2 tests on GLT, 1 test on the poplar, 3 tests on the silver fir
unexposed sides with type K thermocouples of 1.5 mm diameter. and 4 tests on the Douglas (Fig. 23). The positions of thermocouples are
Table 4 summarizes the disposition of each thermocouple used in this arranged in three basic directions as in previous experiments.
test. Table 6 shows the charring rate values of timber at different positions
Figs. 19 and 20 show the evolutions of temperatures at different obtained based on thermocouples and some values obtained based on
depths inside the wood sample. Each depth (20 mm, 30 mm and 40 mm) char thickness measurements.
is considered from the exposed face. The temperature in each depth is Similar to the previous 3 tests, these 10 tests were first measured for
measured through different thermocouple dispositions (radial from the moisture content and density before experimentation. Then, the wood
charring rate was measured by K thermocouples in different directions:

9
M.H. Nguyen et al. Fire Safety Journal 137 (2023) 103752

Fig. 18. Test N◦ 2 - Test sample preparation and thermocouple layout.

Table 4
Group of thermocouples.

Fig. 19. Temperature in the timber specimen during the fire tests with thermocouples in the radial direction (R: radial).

radial exposed, radial unexposed, tangential and longitudinal. The wood 4. Numerical simulation
charring rate results in Table 6 show a good agreement that the charring
rate measured on the exposed radial surface is much larger than that on In the thermal model, the heat transfer in the wood and the steel is
the other surfaces. And the most important parameter in this Table 6 is simulated by a two-dimensional (2D) finite element program on the
the charring rate of the wood based on the measurement of the char Abaqus software [29]. The validation is based on the comparison be­
layer after the experiment. There were 3 experiments (tests 1, 2 and 3) tween calculated values and measured temperature values.
that measured the thickness of the coal layer and obtained the charring From these results of numerical simulation, it is easier to determine
rate consistent with the measured results on the sides: radial unexposed, the charring rate of wood in the model and experiment for an accurate
tangential and longitudinal. choice of thermocouples positioning in the wood.

10
M.H. Nguyen et al. Fire Safety Journal 137 (2023) 103752

Fig. 20. Temperature in the timber specimen during the fire tests with thermocouples introduced from the unexposed side (R: radial, T: tangential).

Fig. 21. Unexposed side after 60 min of fire test.

Table 5
Charring depths and charring rates of timber samples.
N◦ Dimension Remaining wood Char layer 300 ◦ C Vchar by char layer Vchar by Thermocouple (unexposed radial) Vchar by Thermocouple (exposed surface)
b × h (mm× (mm) isotherm (mm) (mm/min) (mm/min) (mm/min)
mm)

Test 74 × 220 36.68 37.32 0.635 0.568 0.854

n◦ 1
Test 74 × 220 38.71 35.29 0.588 0.604 0.955
n◦ 2
Test 74 × 220 36.72 37.28 0.621 0.605 0.952
n◦ 3

4.1. Thermo-physical properties of wood In the case of wood, there is a peak in the reported values of specific
heat, most often taken between 99 and 120 ◦ C. This accounts for the
At high temperatures, the physical and thermal properties of wood vaporization of water in the wood and is presented by a large increase in
and steel materials vary with temperature; factors including specific the energy required to facilitate this phase change from liquid to gas.
heat, thermal conductivity and density have been defined based on the The thermal conductivity values of the char layer were increased at
EN 1995-1-2 standards for timber (Fig. 24) and the EN 1993-1-2 stan­ temperatures above 500 ◦ C, to account for the influence of cracks and
dards for steel [30]. shrinkage of the char layer [31].

11
M.H. Nguyen et al. Fire Safety Journal 137 (2023) 103752

( )
ḣnet,c = αc . θg − θm (3)

The net radiativeheat flux component per unit surface area is

determined by:
( )
ḣnet,r = ∅. εm .εf .σ. θr 4 − θm 4 (4)

In the heat transfer model, the required parameters of thermal

emissivity εm = 0.8 for timber and 0.7 for steel, the coefficient of con­
vection on exposed sides: αc = 25 W/(m2.◦ K), on the unexposed side: αc
= 4 W/(m2.◦ K) and the emissivity of fire: εf = 1.0, Stephan Boltzmann
coefficient: σ = 5.67 × 10− 8 W/m2K4. θm is the surface temperature of
the member, Φ is a configuration factor that is usually taken as equal to
1.0 and θr is the radiation temperature of the fire environment taken as
equal to θg in the case of fully engulfed members. The boundary con­
ditions applied to each experiment for simulation are shown in Fig. 25.
DC2D4 type elements 4-node linear heat transfer quadrilateral, are
used in these simulations and available in the ABAQUS library. Time of
exposure to fire t = 3600 s for each test.
The mesh size of 1–6 mm is used for simple heat transfer analyses in
Fig. 22. Effect of a steel barrier on the charring rate of timber.
timber, to ensure accuracy while saving on computational time [33]. To
confirm the mesh size and element type, sensitivity analysis has been
4.2. Thermal model
performed, as shown in Fig. 26. In this thermal model, we selected a
mesh size of 5 mm.
The net heat flux on surfaces exposed to fire, under thermal action, at
time t, is defined by the sum of convective and radiative flows, according
to Eqs. (2)–(4) by Eurocode 1 [32]. 4.3. Validation of the numerical model

ḣnet = ḣnet,c + ḣnet,r (2) The thermal models in this study were compared to the test results
The net convective heat flux component should be determined by: according to Konig’s tests and calculations to be validated [34]. In
Konig’s tests, the wood samples were spruce with a furnace, dry density

Fig. 23. Type of timber.

Table 6
Value of charring rate of wood.
Type of Number Moisture Density Vchar radial Vchar tangential Vchar longitudinal Vchar radial Vchar (char
timber exposure faces content (%) (kg/m3) unexposed (mm/ (mm/min) (mm/min) exposed (mm/min) layer) (mm/min)
min)

GLT 1 3 9.5 486 0.67–0.75 0.86–0.91

GLT 2 3 9.5 486 0.58–0.62 0.67–0.71
Poplar 3 11 445 0.62–0.75 0.82–1.02 1.11–1.17
Silver fir 1 3 11 435 0.50–0.55 0.71–0.76 0.81–1.11
Silver fir 2 2 11 435 0.54–0.57 0.65–0.66 0.80–1.10
Silver fir 3 1 11 435 0.57–0.61 0.61–0.71 0.82–1.11 0.62–0.64
(Test n◦ 3)
Douglas 1 3 12 567 0.55–0.56 0.58–0.65 0.67–0.95
Douglas 2 3 12 567 0.57–0.60 0.58–0.64 0.75–0.92
Douglas 3 2 12 567 0.53–0.55 0.59–0.64 0.79–0.91 0.62–0.65
(Test n◦ 1)
Douglas 4 2 12 567 0.59–0.62 0.59–0.65 0.59–0.60 0.91–1.02 0.58–0.62
(Test n◦ 2)

12
M.H. Nguyen et al. Fire Safety Journal 137 (2023) 103752

Fig. 24. Thermal conductivity (a); density of wood (b) specific heat of wood as a function of temperature (c) according to Eurocode 5 [31].

Fig. 25. Boundary conditions.

13
M.H. Nguyen et al. Fire Safety Journal 137 (2023) 103752

Fig. 26. Sensitivity analysis of mesh size. Fig. 29. The peak specific heat of wood in our model.

heating at the height of the sample. However, this peak poses certain
problems of numerical convergence or calculation time. Also, to over­
come this problem, an equivalent representation of this peak is proposed
to obtain a more stable model and representative of the results obtained
from the literature.
To avoid the problems of numerical convergence due to the infinite
slope of the specific heat curve at 99–120 ◦ C, it is necessary to decrease
the slope of this peak, as illustrated in Fig. 29. This triangle peak will be
used for all other model validations.
Fig. 30 compares the results of our simplified model with Konig’s
tests as well as the modelling by Werther et al., 2012 [33], showing that
the simulation calculations with the triangular peak model are
satisfactory.
We do observe a good correspondence between all numerical and
Fig. 27. Thermocouple in test wood by König & Walleij, 1999.
experimental curves. The two FEM models compared in Fig. 30 (Werther
et al., 2012 and the simplified model) are made by Abaqus software
considering the properties of Eurocode 5. A discrepancy can be observed
between them, possibly due to the variation of wood density with
temperature. However, the simplified model approaches well the
experimental results for the depths of 18, 30, 42, and 54 mm.

4.4. Thermal results

To understand the impact of the number of faces exposed to fire on

thermocouples, especially tangential thermocouples, the numerical
simulation will show the distribution of isotherms at different times: 10,
20, 40 and 60 min (Fig. 31). When the 20 mm deep thermocouple is
placed on top of the isotherms in the numerical result, it can be seen that
initially it is parallel and then it progresses to a perpendicular
arrangement.
This result justifies the fact that the thermocouple is parallel to the
isotherms at the beginning, this explains the bifurcation on the curve
because the part still parallel to the isotherm strongly decreases after 30
Fig. 28. Sample temperature at 90 min.
min because the carbonized part which increases (which is less than the
case in the exposed configuration 2 faces - test n◦ 2), which legitimizes all
of about 480 kg/m3 and relative humidity of 12 %. A 45 mm × 95 mm the more the interest to have protected the top surface, to make it
solid wood beam was exposed to the standard ISO-834 fire curve for 90 possible to maintain a condition parallel to the isotherm on the trans­
min. versal thermocouple.
The experimental sample was modelled with temperature points The results of the numerical models are compared with experiments.
recorded at 0, 6, 18, 30, 42 and 54 mm, with one side exposed to the ISO- In this section, the numerical model is made and compared based on the
834 fire (bottom surface) (Fig. 27). The simulations were run for a total experimental results of test n◦ 2.
of 90 min with a 5 mm mesh size. Figs. 32 and 33 compare different measurements made with ther­
According to Konig, Eurocode 5 does not give any definition of mocouples positioned either from an exposed or unexposed face, passing
charring rate, nor how appropriate design charring rates should be through or not isothermal curves. It can be seen that the temperatures
derived from fire test results. For this reason, Konig & Walleij presented measured by thermocouple R10, are especially high compared to the
their research work. The suggested thermal properties are obtained by others. And the temperatures obtained with simulation (FEM) coincide
calibration according to the experimental results. The numerical results with the results of other thermocouples. However, it can be observed
are presented in Fig. 28. that initially, despite a lack of plateau at 100 ◦ C, the SI temperatures
Fig. 28 shows the temperature gradient across the sample at 90 min remain lower than all other measurements, and after 25 min of fire
for exposure to unidirectional heat flow with a specific heat peak. exposure, starts to pass over all the curves. Excluding the results of
It can be observed that the specific heat peak allows for a delay in the thermocouple R10, the simulation gives a global shape of the evolution

14
M.H. Nguyen et al. Fire Safety Journal 137 (2023) 103752

Fig. 30. Comparison between experimental and numerical temperatures at different depths by different authors.

Fig. 31. Temperature variation with the char layer in test n◦ 2.

of temperature rather close to all other thermocouples. representation on the cross-section of the beam at 60 min (test n◦ 3), steel
Additionally, the numerical value of the charring rate was not con­ material has a higher thermal conductivity than wood as shown by the
stant because the material’s density, thermal conductivity and specific light blue area on the top and bottom flange of the beam. This numerical
heat vary with the temperature in Abaqus. During the initial period of model contributes to the analysis shown in Fig. 21 for the test n◦ 3.
10–20 min, the model results have a slower temperature rise than the
experimental results. The relative error between the experimental and 5. Discussion
numerical results narrowed with the exposure time increased from
300 ◦ C. For thermocouples installed from the side exposed to fire, there is an
From measurement results and model results in Fig. 34, it is easy to influence of the temperature in the furnace on the measurement results,
see that the thickness of the char layer is 35 mm and the unburnt wood is due to the impact of the thermal conduction on the thermocouple. To
38.8 mm, which shows that there is a good agreement between the observe the effect of heat conduction at locations closer to the mea­
simulation and experimental data with an average wood charring rate surement point of the thermocouple. Two K thermocouples with di­
from 0.59 to 0.62 mm/min. ameters of 1.5 mm were used to make measurements. A flame was
Fig. 35 shows the comparison of temperatures on wood at different attached at 3 positions on the thermocouple: at the head of the ther­
depths of 20 mm, 30 mm and 40 mm between the numerical model and mocouple, and a distance of 20 mm and 30 mm from the head. Practical
the experiment. Through this comparison, it is easy to observe that the observation shows that, even when the heat source is 20 mm and 30 mm
results from the model closely follow the experimental results measured away from the thermocouple point, the added temperature difference is
in the tangential direction. In addition, through the temperature 200 ◦ C and 100 ◦ C respectively due to the heat transfer occurring across

15
M.H. Nguyen et al. Fire Safety Journal 137 (2023) 103752

Fig. 32. Experimental (Test n◦ 2) and numerical temperature-time curve at 20 mm.

Fig. 33. Experimental and numerical temperature-time curve at 30 mm.

Fig. 34. Comparison of charring depths - Graphic visualization of a residual cross-section.

the thermocouple (Fig. 36). This difference depends on the power of the around the thermocouple. The necessity here for good measurement
flame. results is to arrange the thermocouples in a hole with a diameter
In these tests, one of the factors affecting the results is spacemen equivalent to that of the thermocouple and to ensure that there is no

16
M.H. Nguyen et al. Fire Safety Journal 137 (2023) 103752

Fig. 35. Experimental (Test n◦ 3) and numerical temperature-time curve at 20 mm, 30 mm and 40 mm.

Fig. 36. Effect of heat conduction.

displacement during the test preparation. the exposed face give less accurate results than the unexposed face
thermocouples under ISO-834 fire conditions. This observation could
6. Conclusion depend on the depth at which the measurement was made, in test n◦ 2,
the differences in the temperature of these two sides are not the same
This paper has compared a method of measuring temperature on from 20 mm to 40 mm. These differences decrease according to the
low-conductive materials such as wood with K-type thermocouples. The depth of the thermocouple in the wood, perhaps until a certain depth,
scientific consensus is to prefer thermocouples parallel to isotherm as it this difference will be eliminated. And this is also an experiment that
minimizes the conduction flux perturbation in the studied material. should be done in the future with greater depth.
However, in some cases, it can be rather difficult to get a thermocouple In a conclusion, thermocouples placed from an exposed face give the
perfectly parallel to isotherms due to a non-symmetric exposition or due least accurate results compared to thermocouples from an unexposed
to long samples like beams or floor panels. Thus, the present study was face. A solution could be to provide additional insulation to the exposed
interested in giving a comparison of various thermocouples arrange­ thermocouples but this solution has not been tested in the present paper.
ments. Three types of arrangements have been studied, the longitudinal, Thus, the authors recommend to put thermocouples along isotherms if
the transversal and the radial. Both longitudinal and transversal are possible. The transversal direction in the case of beams is the simplest
always thermocouples, which are unexposed to fire whereas the radial way to follow isotherms. Indeed, in the case of beams more than 1 m
thermocouples could be arranged either from the exposed or from the long, longitudinal thermocouples can be very difficult to arrange.
unexposed side of the sample. Finally, as the differences between radial (from unexposed face) and
From this parametrical study, it has been observed on one hand that parallel to the isotherms remain under 10 %, this latter solution can be
transversal thermocouples from unexposed faces provide very similar employed in the case of more complex samples like timber panels or
results compared to longitudinal measurements. Indeed, even if the cross-laminated timber (CLT) panels.
transversal thermocouple passes through isotherms in the very first Other research on the specific applications for CLT should be done in
centimeter of the sample, it becomes then parallel to the isotherm. On order to evaluate the precision of the radial measurements for this
the other hand, it has been observed that thermocouples distributed on specific case. Indeed, in this case, having thermocouples parallel to the

17
M.H. Nguyen et al. Fire Safety Journal 137 (2023) 103752

isotherm is even more complicated, which leads many researchers to use [9] A.L. Brundage, A. Burl Donaldson, W. Gill, S.P. Kearney, V.F. Nicolette, N. Yilmaz,
Thermocouple response in fires, part 1: considerations in flame temperature
radial thermocouples on the unexposed face.
measurements by a thermocouple, J. Fire Sci. 29 (3) (2011) 195–211, https://doi.
Finally, the model results show that the average charring rate is in org/10.1177/0734904110386187.
good agreement with the recommended values proposed in Eurocode 5. [10] N. Burley, R. Powell, G. Burns, M. Scroger, The nicrosil versus nisil thermocouple:
This thermal model with Abaqus allows us to study in a precise way the properties and thermoelectric reference data, Nat. Bureau Standard Monograph.
161 (161) (1978) 1–167 [Online]. Available: http://adsabs.harvard.edu/abs/
evolution of the isotherm at 300 ◦ C and to confront itself with the 1978nbs..reptU....B.
experimental measurements. The simplified model according to Euro­ [11] U. Wickström, Temperature calculation in fire safety engineering, Book J. Fire Sci.
code 5 in this paper shows that simulation results are in good accordance 34 (6) (2016) 530–533, https://doi.org/10.1177/0734904116664535.
[12] J. Schmid, R. Fahrni, M. Klippel, A. Frangi, Correct Temperature Measurements in
with measurements made along isotherms. In addition, this is a simu­ Fire Exposed Wood,” WCTE 2018, World-Conference-on-Timber-Engineering,
lation model for wood with a moisture content of 12 %. Therefore, when 2018.
performing experiments with wood with a moisture content smaller or [13] B. William D, Effect of thermocouple wire size and configuration on internal
temperature measurements in a charring Ablator, Mechatronic. Handbook. (2002)
greater than 12 %, the model must be modified to meet the simulation 19-73–19–88, https://doi.org/10.31399/asm.tb.smnm.t52140201.
requirements on the specific heat peak values in particular. [14] A. Totaro, Temperature measurements in wood specimens, Fire Mater. (2018)
The present study is considered preparatory work and presents the 1–13.
[15] J.V. Beck, Thermocouple temperature disturbances in low conductivity materials,
basic elements necessary to carry out a future research study that con­ J. Heat Tran. (1962) 131–132.
sists in evaluating the fire resistance of the mixed steel-timber structure [16] R. Fahrni, J. Schmid, M. Klippel, A. Frangi, Investigation of Different Temperature
under a mechanical load. Measurement Designs and Installations in Timber Members as Low Conductive
Material,” SiF 2018 - the 10th International Conference on Structures in Fire
FireSERT, 2018.
Declaration of competing interest [17] A.A. Owusu, Structural Performance of Hybrid Timber Connections with Varying
Bolt Patterns at Ambient and Elevated Temperatures, Thesis - Carleton University,
The authors declare that they have no known competing financial 2019.
[18] S.B.A. Boadi, Full Scale Tests on the Performance of Hybrid Timber Connections in
interests or personal relationships that could have appeared to influence Real Fires, Thesis - Carleton University Ottawa, Ontario, 2015.
the work reported in this paper. [19] A.O. Akotuah, S.G. Ali, J. Erochko, X. Zhang, G.V. Hadjisophocleous, Study of the
Fire Performance of Hybrid Steel-Timber Connections with Full-Scale Tests and
Finite Element Modelling, Applications-of-Structural-Fire-Engineering, 2017,
Data availability pp. 15–16, https://doi.org/10.14311/asfe.2015.039. October.
[20] S. Zöllig, A. Frangi, S. Franke, M. Muster, Timber structures 3.0 - new technology
No data was used for the research described in the article. for multi-axial, slim, high performance timber structures, 1, in: WCTE 2016 -
World-Conference-on-Timber-Engineering, 27, 2016, pp. 32–40.
[21] L. Osborne, C. Dagenais, Preliminary CLT fire resistance testing report, Adv. Build.
Acknowledgements Syst. 53 (9) (2012) 1689–1699.
[22] E. Andersen, Fullstendig brannforløp i limtrekonstruksjoner Vurdering av
resultater fra branntest, 2017.
The authors acknowledge the financial support of the Carnot Insti­ [23] H. Kinjo, Experimental Study on the Decrease in Yield Strength of Structural
tute (MECD) and Clermont Auvergne University (Institut Pascal). More Laminated Beam Beams Subjected to Standard Fire Heating during the Cooling
specifically, the authors are grateful to M. Moissaing and S. Roy (IUT Process, Thesis - Chiba University, Graduate School of Engineering, 2016.
[24] P.B. Cachim, J.M. Franssen, Assessment of Eurocode 5 charring rate calculation
Clermont Auvergne) for their technical support and valuable contribu­
methods, Fire Technol. 46 (1) (2010) 169–181, https://doi.org/10.1007/s10694-
tions to the research project. 009-0092-x.
[25] Afnor, “EN 1363-1 Fire Resistance Tests –Part 1: General Requirements.”.
References [26] ISO-834-1, Fire Resistance Tests-Elements of Building Construction, Part 1: General
Requirements, International-standard, 1999.
[27] British Standards Institution, “EN 13183-1:2002 Moisture Content of a Piece of
[1] M.T. Tsai, T.D.H. Le, Determination of initial stiffness of timber-steel composite Sawn Timber Part1: Determination by Oven Dry Method, 2007.
(TSC) beams based on experiment and simulation modeling, Sustain. J. MDPI. 10 [28] R.H. White, E. V Nordheim, Charring rate of wood for ASTM E 119 exposure, Fire
(4) (2018), https://doi.org/10.3390/su10041220. Technol. 28 (1) (1992).
[2] M.H. Nguyen, S.E. Ouldboukhitine, S. Durif, V. Saulnier, A. Bouchair, Passive fire [29] Dassault Systemes, “Abaqus/CAE User’s Guide,” Abaqus User’s Guid., p. vol. 1138,
protection of steel profiles using wood, Eng. Struct. 275 (March 2022) 2023, 2013.
https://doi.org/10.1016/j.engstruct.2022.115274. [30] EN-1993-1-2, Eurocode 3, Design of Steel Structures - Part 1-2, in: General Rules -
[3] E. Katayama, Steel construction today & tomorrow_ kunimi town office building, Structural Fire Design, vol. 1, The-European-Union, 2005.
Japan Iron. Steel. Federate. 73 (7) (2001) 50. [31] EN-1995-1-2, Eurocode 5, Design of Timber Structures - Part 1-2, vol. 1, General -
[4] E. Mikkola, Charring of wood based materials, Fire Saf. Sci. 3 (1991) 547–556, Structural fire design, 2011.
https://doi.org/10.3801/iafss.fss.3-547. [32] EN-1991-1-2, Eurocode 1, Actions on Structures - Part 1-2, in: General Actions -
[5] M.G. Gronli, M.C. Melaaen, Mathematical model for wood pyrolysis-comparison of Actions on Structures Exposed to Fire Eurocode, vol. 1, 2011, 2005.
experimental measurements with model predictions, Energy Fuel. 14 (4) (2000) [33] N. Werther, et al., Parametric Study of Modelling Structural Timber in Fire with
791–800, https://doi.org/10.1021/ef990176q. Different Software Packages,” 7th-International-Conference-On-Structures-In-Fire,
[6] I. Arne, Evaluation of the burning behaviour of wood products in the context of 2012.
structural fire design, Sereal Untuk 51 (1) (2014) 51. [34] J. König, L. Walleij, One-dimensional Charring of Timber Exposed to Standard and
[7] K.L. Friquin, M. Grimsbu, P.J. Hovde, Charring rates for cross-laminated timber Parametric Fires in Initially Unprotected and Postprotection Situations, 1999.
panels exposed to Standard and Parametric fires, in: 11th World Conf. Timber Eng. Stockholm, http://ri.diva-portal.org/smash/get/diva2:1079803/FULLTEXT01.pdf
2010 vol. 4, WCTE, 2010, pp. 3293–3301, 2010, 7491. %0Ahttp://urn.kb.se/resolve?urn=urn:nbn:se:ri:diva-28629. Available:.
[8] OMEGA, How to choose a thermocouple, Comput. YouTube 56 (9) (2009) 1–11
[Online]. Available: https://www.omega.com/en-us/resources/selecting-a-the
rmocouple.

18