ARTICLE IN PRESS

Fire Safety Journal 40 (2005) 466–476

www.elsevier.com/locate/firesaf

Test method for spalling of fire exposed concrete

K.D. Hertz, L.S. Sørensen
Department of Civil Engineering, Building 118, Technical University of Denmark,
DK-2800 Kgs. Lyngby, Denmark
Received 19 January 2004; received in revised form 25 March 2004; accepted 27 April 2005
Available online 22 June 2005

Abstract

A new material test method is presented for determining whether or not an actual concrete
may suffer from explosive spalling at a specified moisture level. The method takes into account
the effect of stresses from hindered thermal expansion at the fire-exposed surface. Cylinders
are used, which in many countries serve as standard specimens for testing the compressive
strength. Consequently, the method is quick, cheap and easy to use in comparison to the
alternative of testing full-scale or semi full-scale structures with correct humidity, load and
boundary conditions. A number of concretes have been studied using this method, and it is
concluded that sufficient quantities of polypropylene fibres of suitable characteristics may
prevent spalling of a concrete even when thermal expansion is restrained.
r 2005 Elsevier Ltd. All rights reserved.

Keywords: Spalling; Concrete; Fire; Test; Polypropylene fibres

1. Introduction

Explosive spalling may reduce the load bearing capacity of a structure

substantially and must be considered when designing concrete structures for fire [1].
The introduction of superplasticizing additives in the 1970s made it possible to
create dense concretes by mixing ultra fine particles smaller than the cement grains,
such as silica fume, with the cement. Hertz [2–5] discovered that these new materials

Corresponding author. Tel.: +45 4525 1700; fax: +45 4593 4430.
E-mail address: [email protected] (K.D. Hertz).

0379-7112/$ - see front matter r 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.firesaf.2005.04.001
 ARTICLE IN PRESS
K.D. Hertz, L.S. Sørensen / Fire Safety Journal 40 (2005) 466–476 467

suffered from explosive spalling to an extent never seen before. For example, it was
observed that spalling occurred at a heating rate of only 1 1C/min and at
temperatures of no more than 350 1C for a super dense concrete with 14% micro
silica.
The present knowledge about spalling is discussed in the paper by Hertz [1] in
which it is concluded that spalling is not a problem for traditional (not ‘‘densified’’)
concretes with 3 wt% moisture. This means that most indoor structures of
traditional concrete are not susceptible to spalling.
Hertz [1] concluded for the first time that the risk of spalling of a dense concrete is
significantly increased if thermal expansion is restrained, and that the correct
boundary conditions are decisive for spalling tests.
Hertz [1] recommended further research in order to investigate the limits of safe
application of dense or moist concrete. This could be divided into 4 tasks:

(A) Establish reliable procedures for testing the spalling behaviour of specific
concrete materials taking the effect of thermal stresses into account.
(B) Extend the known limits of spalling by means of tests and further observations
on the nature of spalling.
(C) Understand the nature of the problem fully and develop a coherent theory.
(D) Establish a theoretical method to predict whether a concrete structure is
susceptible to spalling or not.

This paper presents a test method as a result of task (A) and some observations
made by means of the test method extending the known limits as a result of task (B).

2. Test method

As mentioned above a test method is required because the mechanisms of spalling

are yet not fully understood.
In order to ensure that all relevant parameters are taken into account the
influences of the parameters involved must be analysed. However, since the
mechanisms are still uncertain this analysis must be made by means of black-box
considerations based on experience from tests and practise. This discussion has been
made in detail in Hertz [1].
From the discussion, it is seen that the significant parameters are the moisture
content, the heating rate, and stresses from static and (mainly) thermal loading.
Parameters specific to the concrete are also significant, such as the compressive
strength, the density of the microstructure and of the aggregates and, if present, the
nature, quality and quantity of fibre reinforcement.
All properties of the concrete are specified when a test is carried out on a sample of
the actual material.
According to Hertz [1] moisture is the main reason for explosive spalling. For
the most dense concretes the water of crystallisation can give rise to spalling of
moisture-free specimens. The moisture content, the chemically bound water and the
 ARTICLE IN PRESS
468 K.D. Hertz, L.S. Sørensen / Fire Safety Journal 40 (2005) 466–476

compactness of the microstructure are factors determining the vapour pressure

which causes explosive spalling. However, when the test is made on a specific
concrete only the moisture content is varied.
The moisture content is therefore carefully measured by weighing the specimen,
including all spalled material, before and after the test and after a period of drying at
105 1C until there is no further weight loss. This means that the test result is related
to the actual moisture content. A series of tests is required to obtain the result as a
function of the moisture content.
As explained by Hertz [1] a rapid temperature increase is more likely to
cause spalling than a slow one and is therefore preferable for testing purposes.
Although a slow heating rate of 5 1C per minute has given spalling at 600 1C after 2 h
for a dense and sufficiently wet concrete during the development of the test, this slow
heating rate would not be reliable for testing specimens of less susceptibility to
spalling.
Furthermore, the temperature at the depth of a typical spalling flake
(approximately 2 cm) should be increased during the test to above the range, where
spalling is observed to take place. This is found to be near the critical point for water
at 374 1C. On the other hand, the temperature at the surface must not exceed a level,
where the concrete melts (approximately 1150 1C). Therefore, a simple procedure is
used to establish the temperature increase by removing a cover to the oven at
1000 1C, increasing the temperature at the surface of the cylinder to 800 1C, within
20 min.
During the development of the method a test series of 6 dense and moist
specimens, which were known to be susceptible to spalling, were tested by quickly
removing the cover from the oven at a temperature of 800 1C increasing the surface
temperature to 600 1C. In these tests, only one specimen spalled while the others
showed little or no spalling.
This suggested that a surface temperature of 800 1C would be required, which
means that the oven temperature should be 1000 1C.
This thermal exposure gives a temperature increase at the surface which is
comparable to the increase from the hottest fires which do not melt the surface. The
test is therefore on the safe side for the assessment for all fires relevant for design.
Temperature profiles have been calculated and measured for cylinders, tiles and
full scale tested walls as reported by Hertz [1] and Kristiansen et al. [7]. These
temperature profiles support the conclusion that a suddenly applied surface
temperature of 800 1C is necessary for some fires and safe for others.
As further explained by Hertz [1], stresses parallel to a fire-exposed surface are of
importance for explosive spalling. In addition, it is found that stresses from
restrained thermal expansion can be considerable and of a magnitude comparable to
the compressive strength of the concrete. By establishing a total restraint for thermal
expansion these stresses will be at maximum, and at this level additional stresses
from the static load would be reduced by transient strains at the surface layer as
explained elsewhere (e.g. Hertz [8]).
It is therefore concluded that restrained thermal expansion leads to stresses, which
are most likely to cause explosive spalling compared to other stress combinations,
 ARTICLE IN PRESS
K.D. Hertz, L.S. Sørensen / Fire Safety Journal 40 (2005) 466–476 469

and that the application of this boundary condition will be conservative for the
purpose of the test method.
At first Hertz proposed a test, where a cover was removed from an oven exposing
a central part of a concrete tile. The idea was that the unexposed outer part of the tile
should counteract the thermal stresses (Hertz [1]). Gunnarsson [6] carried out this
test for 10 samples of concretes and it was proved that such stresses are decisive for
explosive spalling of several dense concrete qualities, because the spalling stopped
when the outer parts of the tiles cracked in tension, thus unloading the thermal
stresses.
However, the tiles were heavy and the test had to be stopped when they cracked,
which meant that the tile test would not show spalling if it occurred at a larger stress
level.
Hertz therefore proposed the test method presented in this paper, where a steel
mantle is counteracting the thermal stresses, and where the specimen can be a small
cylinder. The cylinder is used for measuring mechanical properties of concrete in
many countries. It is easily produced and is known as a cheap and handy specimen.
The end surface is exposed through a circular hole and the thermal stresses will be bi-
directional and uniform. A ring shaped steel mantle is well suited to counteract the
thermal compression without causing any irregular stresses.
In some countries, cubes are used instead of cylinders. However, if a cube was used
for this spalling test, it would be more complicated to establish the restraints and the
temperature and stress distribution would most likely be uneven, thus confusing the
test result.
By means of the test method, it should be possible to assess whether a particular
concrete with a known moisture content has a risk of spalling. The method should
take the effect of restrained thermal expansion into account and thus give results ‘‘on
the safe side’’ for applications where thermal stresses are not present. It should also
be on the safe side for other possible applications where the fire exposure may not be
extreme. This means that if a concrete at a certain moisture content is deemed to be
free of spalling in the test, it should be safe to apply at that moisture content. If not,
further investigations are necessary.
This test method was further refined as a part of the project ‘‘Resource Saving
Concrete Structures’’, also called ‘‘Green Concrete’’ [7].
The test specimen is a concrete cylinder of diameter 150 mm and height 300 mm,
identical to the cylinders used for testing for compressive strength.
The cylinder is placed in a steel mantle consisting of two 50 mm thick parts
connected by 12 bolts of diameter 36 mm (See Figs. 1 and 2). A pressure-distributing
layer of neoprene is placed between the concrete cylinder and the steel mantle to
compensate for irregularities of the concrete surface.
One end of the cylinder is suddenly exposed through the 100 mm diameter hole to
heat from an oven at 1000 1C (Fig. 2), resulting in a temperature of 800 1C within
20 min at the surface of the cylinder.
The dimensions of the steel mantle ensure that thermal stresses are counter-
acted as if the end of the cylinder was a part of a fire-exposed surface of a concrete
wall.
 ARTICLE IN PRESS
470 K.D. Hertz, L.S. Sørensen / Fire Safety Journal 40 (2005) 466–476

Fig. 1. Steel mantle with end of cylindrical specimen to be exposed.

Fig. 2. Test rig.

 ARTICLE IN PRESS
K.D. Hertz, L.S. Sørensen / Fire Safety Journal 40 (2005) 466–476 471

Fig. 3. (a) Spalling, (b) radial cracks, (c) shear top failure.

The outer 25 mm of the radius of the cylinder is not exposed and can be considered
as a layer of concrete through which the thermal stresses are applied.
Radial cracks are formed in this layer as shown in Fig. 3b.
An end stop of steel (Fig. 2) ensures that the volume of the neoprene remains
constant and does not expand along the length of the cylinder which would reduce its
thickness.
This would leave space for radial expansion of the heat-exposed surface of the
cylinder. Until this device was introduced, a shear crack was often formed near the
exposed end of the cylinder separating the top of the cylinder from the rest as shown
in Fig. 3c.
The end of the cylinder is exposed to heat for an hour. If spalling occurs, it is
usually accompanied by a sound, the time of which is recorded.
After the exposure, the area of spalling is measured and the spalled material
collected and weighed. Then a successive drying and weighing procedure in the
following weeks determine the moisture content of the specimen at the time of the
test. The result of the test is an indication of whether the concrete tested is likely to
spall at the measured moisture content.
If a test series is carried out with variable moisture content, a critical content can
be found within which the concrete will not be susceptible to spalling.
In most cases it is recommended to chose a content of 3 wt% moisture at the time
of test, so that the method will indicate whether the actual concrete is more likely to
spall than traditional normal concretes, which can be considered not to spall if
located indoors. This means that if a concrete does not spall in the test at 3 wt%
moisture, it will be safe to use indoors. It takes only a test at one moisture level to
reach this important conclusion.
 ARTICLE IN PRESS
472 K.D. Hertz, L.S. Sørensen / Fire Safety Journal 40 (2005) 466–476

Table 1
Test results

Tem- Moist Micro- Self- Fibre Fibre Spalling No

perature weight silica com- con- diameter number spalling
(1C) (%) (%) pact tent (mm) number
(kg/m3)

Tiles, dense 800 4 10 — 0 — 10 0

Cylinders drilled out of a tile, dense 800 4 10 — 0 — 3 0
Cylinders cast of a tile material, 800 4 10 — 0 — 3 0
dense
Wet cylinders, dense slow heated 600 46 5 — 0 — (3) (0)
at 5 1C/min (not (not
valid) valid)
Moist cylinders, dense at 600 1C 600 6 5 — 0 — (1) (5)
(not (not
valid) valid)
Moist cylinders, dense 800 6 5 — 0 — 6 0
Moist cylinders, not densified 800 6 0 — 0 — 6 0
Cylinders, dense 800 4–5 5 — 0 — 12 0
Cylinders, not densified 800 4–5 0 — 0 — 1 8
Cylinders, dense for fibre effect test 800 5 10 — 0 — 4 0
Cylinders, dense for fibre effect test 800 5 10 — 6 30 1 4
Cylinders for blind test, not densified 800 4 0 — 0 — 0 1
Cylinders for blind test, not densified 800 4 0 — 2 0.18 0 1
Cylinders for blind test, densified 800 4 5 — 0 — 1 0
Cylinders for blind test, densified 800 4 5 — 2 0.18 0 1
Cylinders for blind test, densified 800 4 0 X 0 — 1 0
Cylinders for blind test, densified 800 4 0 X 2 0.18 0 1

3. Test results

A test series of large square flat concrete tiles exposed to high temperatures
at the centre of one side is recorded in Hertz [1] and Gunnarsson [6]. The tiles
spalled until they cracked indicating that thermal stresses have a significant influence
on the spalling mechanism for concretes containing about 10% micro silica
(Table 1).
Concretes like this have been used for several tunnel projects where spalling has
occurred.
Cylinders have been cut out of the same tiles and subjected to the new test method.
These cylinders and cylinders cast of the same material showed the same tendency of
spalling, and thereby confirmed that the mantle test can substitute the larger and
more expensive tile test (Figs. 4 and 5).
As mentioned, the spalling stopped in the tile test when the tiles broke. This might
be a problem for the tile test if the spalling of a particular sample did not take place
before cracking occurred. In these cases, the tile test will not show that the concrete
 ARTICLE IN PRESS
K.D. Hertz, L.S. Sørensen / Fire Safety Journal 40 (2005) 466–476 473

Fig. 4. Cylinders cut out of tiles and spalled in mantle test.

Fig. 5. Previous tested tile from which test cylinders have been cut.

has a spalling problem. Since the mantle test prevents cracking (which will stop the
spalling process), it appears to be more reliable.
The test method has proved to be suitable for determining the spalling risk of a
number of new proposed concretes from the larger project ‘‘Resource Saving
Concrete Structures’’ [7,9].
All four concrete samples designed for an aggressive environment of which two
contained 5% micro silica spalled, but all of them had moisture contents of 6 wt%,
so the influence of the micro silica could not be concluded.
 ARTICLE IN PRESS
474 K.D. Hertz, L.S. Sørensen / Fire Safety Journal 40 (2005) 466–476

Four out of seven concretes designed for a passive environment showed clear sign
of spalling. All four concretes contained between 4.2 and 5.1% micro silica, which
must be the reason for spalling. One of the concretes, which did not spall, was a
reference concrete used for 25 years at the university. For this concrete, we have
experienced that it should not spall, unless it is wet.
These results are in accordance with the general statement of Hertz [1] that
traditional not densified concrete has no problem for an indoor climate, and it is
satisfactory that the new test indicates that only the densified or moistures concretes
spalled.
In a M.Sc. project by Jensen [10], a concrete with 10% micro silica was tested with
and without addition of polypropylene fibres. The concrete had a water/cement ratio
of 0.37, a compressive strength of 58 MPa and the moisture content was 5.2%. The
strength was increased to 69 MPa when 6 kg polypropylene fibres were added per
cubic meter. The monofilament fibres were 13 mm long, 30 mm in diameter and had a
softening point of 150 1C.
Four cylinders spalled out of four without fibres, and only minor spalling occurred
at one cylinder out of five with polypropylene fibres.
This small test series showed for the first time, that a sufficient quantity of
polypropylene fibres could reduce the risk of spalling, even when thermal expansion
is restrained at the fire exposed surface.
The large amount of fibres was added in order to be sure that the effect
should be seen, if they had any effect at all when thermal expansion is
restrained.
As mentioned by Hertz [1], an effect of polypropylene fibres has previously been
observed for slabs. A theory was proposed that one reason for the positive effect
could be that the melting of the fibres can make it easier for the concrete to crack in a
depth of about 50 mm from the fire exposed surface which may unload the thermal
stresses necessary for spalling at the surface. This theory is still applicable and it
explains why polypropylene fibres in some cases do not help for columns, but now as
one out of more explanations, because additional theories are needed to explain the
positive effect observed by the tests reported above. One such theory is stated by
Bertil Persson (personal contact and [11]), who has made a thorough research project
on fire exposed self-compacting concrete densified by a limestone filler [12]. This
theory suggest that melted fibres may inhibit the movement of moisture into the cold
depth of a cross section and thereby hinders the built up of large steam pressures in
the material.
A blind test series of six different concretes supplied by a client has been carried
out. The compositions were not known to the laboratory before the results were
reported. Two of these concretes spalled.
All concretes were designed for an aggressive environment with a water cement
ratio of 0.38 and moisture content of 4–5%. One concrete had 2 kg polypropylene
fibre per cubic meter with a diameter of 0.18 mm. One was densified with about 5%
micro silica, one had micro silica and fibres, one was densified by means of a
limestone filler, one had filler and polypropylene fibres, and one had none of these
special components.
 ARTICLE IN PRESS
K.D. Hertz, L.S. Sørensen / Fire Safety Journal 40 (2005) 466–476 475

The two spalling concretes were the one with micro silica without fibres and the
one with limestone filler without fibres.
This is a clear result, which accord well with the previous experience as described
above (see also [1]), and confirms that the test method performs well.
In total 54 cylinders, representing 20 concrete recipes, have been tested by the final
method. All test results are shown in Table 1. The table also includes additional
results of the 10 tile tests and of nine cylinders of three concrete recipes which were
tested during the development of the test procedure at too low a surface temperature.

4. Conclusions

A test method has been established by means of which the spalling behaviour of a
specific concrete with a specific moisture content can be predicted. The effect of
restrained thermal expansion of the surface layer is taken into account.
It is concluded that the test method leads to results which accord well with the
experience available.
Since the specimen is a cylinder, which is standardised in many countries, the test
is cheap and easy to include in the process of specifying concretes for structural use.
It is shown by means of the test method that polypropylene fibres may hinder
explosive spalling of a dense concrete, also if the thermal expansion is restrained.
From the few results available, there is a tendency that thin fibres of 0.18 mm more
effectively hinder spalling than thick fibres of 30 mm.

References

[1] Hertz KD. Limits of spalling of fire exposed concrete. Fire Safety J 2003;38(2):103–16.
[2] Hertz KD. Explosion of silica-fume concrete. Fire Safety J 1984/85;8:77.
[3] Hertz KD. Eksplosion og reststyrke af varmepåvirket silikabeton. (Explosion and residual strength of
fire exposed silica fume concrete.) Report 162. Institute of Building Design (now Department of Civil
Engineering), Technical University of Denmark; 1982. 12p. [in Danish].
[4] Hertz KD. Heat induced explosion of dense concretes. Report 166. CIB W14/84/33(DK). Institute of
Building Design (now Department of Civil Engineering), Technical University of Denmark; 1984.
20p.
[5] Hertz KD. Danish investigations on silica fume concretes at elevated temperatures. Keynote speech at
technical session fire resistance of materials and/or members made with high-strength concrete. ACI
Spring Convention; 1991. ACI Mater J 1992;89(4):345–7.
[6] Gunnarsson JG. Eksplosiv afskalning af beton. (Explosive spalling of concrete.) M.Sc. thesis.
Department of Buildings and Energy (now Department of Civil Engineering). 1998 [in Danish].
[7] Kristiansen FH, Hertz KD, Sørensen LS. Eksplosiv afskalning af beton (Explosive spalling
of concrete.) Report BYG.DTU R-050. Department of Civil Engineering; December 2003. 76p
[in Danish]. http://www.byg.dtu.dk.
[8] Hertz KD. Betonkonstruktioners brandtekniske egenskaber. Part 1 and 2 of Ph.D. thesis
(Fire properties of concrete constructions), Report 140. Institute of Building Design (now
Department of Civil Engineering), Technical University of Denmark; 1980. 210p.
[9] Sørensen LS, Hertz KD, Kristiansen FH. Explosive spalling of fire exposed resource saving
concrete structures. Nordic Mini-Seminar on Concrete and Fire, Danish Technological Institute;
May 2003. 3p.
 ARTICLE IN PRESS
476 K.D. Hertz, L.S. Sørensen / Fire Safety Journal 40 (2005) 466–476

[10] Jensen L. Spalling of concrete exposed to fire—the mitigating effect of polypropylene fibres.
M.Sc. Project. Department of Civil Engineering, Technical University of Denmark;
April 2003. 84p.
[11] Persson B. Fire resistance of self-compacting concrete. SCC, Materials and Structures; 2002. 20pp,
accepted for publication.
[12] Persson B. Self-compacting concrete at fire temperatures. Report TVBM-3110. Lund Institute of
Technology, Division of Building Materials, Lund; September 2003. 200p.