International Centre for Mechanical Sciences (CISM)

Course on
Effect of Heat on Concrete
9-13 June 2003, Udine, Italy

Creep & Shrinkage Khoury 6

Gabriel Alexander Khoury

Imperial College, London University, UK

ABSTRACT............................................................................................................................................ 1
1. NATURE OF THE CREEP OF HEATED CONCRETE ............................................................. 2
2. UPPER AND LOWER BOUNDS OF CREEP - Importance of load-heat sequence ................... 2
3. CREEP COMPONENTS ............................................................................................................ 4
3.1 Upper bound of creep (LITS).................................................................................................. 4
3.2 Lower bound of creep (basic creep)........................................................................................ 6
3.3 Intermediate creep................................................................................................................... 8
4. ISOLATION OF SHRINKAGE AND CREEP COMPONENTS ................................................. 9
4.1 During First Heating .............................................................................................................. 9
4.2 At Constant Temperature ...................................................................................................... 9
5. Prediction of Residual Strains by Superposition.......................................................................... 10
5. CONCLUSIONS AND PRACTICAL APPLICATIONS........................................................ 11
5. REFERENCES ......................................................................................................................... 12

ABSTRACT

The lower bound of creep “basic creep” occurs when the concrete is first heated then loaded at
constant temperature after all the physico-chemical reactions have been completed and the concrete is
stabilised. Stability is achieved after a few hours or several weeks depending upon the temperature
level. The results suggest that the structural usefulness of Portland cement based concrete is limited to
temperatures below 550-600oC as evidenced by a marked surge in basic creep above this temperature.
The upper bound of creep “transient creep” occurs when the concrete is first loaded then heated under
load thus experiencing the maximum internal physico-chemical instability, as occurs in practice
during fire exposure. Strictly it should be termed the Load Induced Thermal Strain (LITS) because,
unlike basic creep, it is a function of temperature change and in real applications (such as fire) it can
be regarded as a function of temperature only and not of time. While the aggregate has a major
influence upon the thermal strains, its influence upon LITS is significantly less or negligible – for a
given aggregate content by volume - because creep is essentially “seated” in the cement paste. While
basic creep is largely time-dependent, LITS at temperatures above 100oC is largely time independent.
The trends of LITS of high-performance-concrete are similar to those for normal concrete. LITS
strains are so large that any structural analysis of concrete heated under load would be inappropriate
without them.

The shrinkage and creep components of concrete can be isolated separately for the period during
heating-up and for the period at constant temperature. Predictions of the residual strains for the loaded
specimens can be made by simple superposition of creep and shrinkage components up to a certain
critical temperature. Above the critical temperature, it is necessary to add a ‘cracking component’.
CISM - Effect of Heat on Concrete Creep & Shrinkage

1. NATURE OF THE CREEP OF HEATED CONCRETE

The creep of concrete increases markedly at high temperatures and should be included in any
structural analysis of heated concrete - regardless of whether the heat exposure is of short duration
(e.g. fire) or long duration (e.g. nuclear reactor). Fundamentally, creep develops as a time-dependent
strain caused by physico-chemical structural changes in the material under load. Heating of concrete
enhances and accelerates the physico-chemical processes (e.g. drying & hydration-dehydration) and
introduces new changes (e.g. chemical transformations related to specific elevated temperatures)
along the whole temperature route from ambient up to melting of the concrete. Normally, creep is
“seated” in the cement paste and is restrained by the presence of aggregate, although some aggregates
may themselves experience creep. The influence of the aggregate, therefore, cannot be ignored even if
the creep of the aggregate itself is negligible when compared with that of the paste.
The scientist/engineer has essentially two tasks: (a) to understand this phenomenon and (b) to quantify
the strains for input into the mathematical structural analysis model. This paper presents a only brief
outline of this wide and complex subject.

2. UPPER AND LOWER BOUNDS OF CREEP - Importance of load-heat sequence

When concrete in a stable state is loaded at constant load and temperature levels, it will experience an
instantaneous elastic strain followed by time-dependent “basic creep” strain which comprises two
components: a recoverable delayed elastic component and an irrecoverable “flow” component. The
delayed elastic component relates to the stress transfer between the paste and the aggregate, while the
flow component relates to structural re-adjustment within the paste under load in order to achieve a
lower overall energy level. The flow component being far more sensitive to temperature than the
delayed elastic component.

It is important to emphasise the word “stable” in relation to basic creep. Any chemical (e.g. hydration)
or physical (e.g. drying) instability within the material will induce an additional creep component
such as hydration creep or drying creep. Therefore, the experiment must ensure that the concrete is
fully stable before it is loaded (e.g. no temperature change, no carbonation, no hydration, no moisture
movement). Basic creep, defined in this way, can apply for any temperature - ambient or high. The
key criterion is that the specimen once heated to, say 300oC, should be stabilized for a sufficiently
long period of time before being loaded.

If the specimen is not stabilised then there will be additional creep components relating to the
physico-chemical transformation taking place (e.g. carbonation creep). The development with
temperature and time of the chemical reactions could follow the Arrhenius relation so that the rate of
reaction would be low at the starting reacting temperature (e.g. about 390oC for the dissociation of
Ca(OH)2) to reach a peak at a specific higher temperature. The creep relating to a specific reaction
could then develop at a rate commensurate with the underlying rate of reaction. However, the
Arrhenius relation cannot be applied to simulate the temperature function of basic creep since, by
definition, the underlying reactions would have been completed before loading.

The foregoing discussion suggests that the lower bound for the creep is the basic creep of a given
concrete loaded at constant levels of load and temperature after a period at temperature long enough to
ensure stability (point E, Figure 1). The length of this period (Figure 2) varies with temperature
because the rates and natures of the physico-chemical reactions also vary with temperature (Moisture
loss is slow at 80oC but much higher at 110oC and so on). The stability criteria used here is that of
dimensional stability of unsealed cement paste cylinders (6cm dia x 18cm long) heated to the test
temperature until the daily change in strain becomes less than 50microstrains. Although a more
stringent criteria could have been used, stability at 400oC could not be achieved even after 30 days at

Khoury, G. 2
CISM - Effect of Heat on Concrete Creep & Shrinkage

temperature. This demonstrates how important it is to ensure stability of the material before loading in
basic creep tests.

Temperature

Figure 1. Loading-heating regime for creep tests at high temperatures.

The upper bound for creep lies at the opposite end of the spectrum when the concrete is loaded and
then heated under load (point A, Figure 1). This produces the largest internal chemical and physical
instability in the material and hence the largest “creep” strains - normally called “transient creep”
because it occurs during the heating up (transient temperature) stage. Loading the concrete at
intermediate points (e.g. points B, C, D in Figure 1) produces creep strains that lie in between the
upper and lower bound values.

The principles presented in this section apply to both cement paste and concrete (the latter is
influenced by the aggregate) and to young and mature concrete, but in young concrete the role of
hydration would be prominent and the microstructure not fully developed.

Khoury, G. 3
CISM - Effect of Heat on Concrete Creep & Shrinkage

Figure 2. Dimensional stability of unsealed cement paste at constant temperatures.

3. CREEP COMPONENTS

Testing over decades has, fortunately, shown that the strain components of concrete are simply
additive and that the superposition of strain principle may be applied [1]. The creep, and other strain
components of concrete, therefore, add up to produce the “total” strain ( ε σt , ,tot
T, m
). Concrete that is first
loaded and then heated will experience elastic strain ( ε σ0, ,T, σ ,T, m
el ) and plastic strain ( ε 0 , plas ), thermal
m

expansion strain ( ε 00,,T,thm ), shrinkage ( ε 00,,T,shm ), creep strain ( ε σt, cr,T, m ), and a strain related to cracking in
the concrete ( ε 0σ, ,crack
T, m
):

ε σt , ,tot
T, m
= ε σ0, ,T,el m + ε 0σ, ,T,plasm + ε 00,,T,thm + ε 00,,T,shm + ε σt, cr,T, m + ε 0σ, ,crack
T, m
(1)

All these strain components would be functions of temperature and moisture (T & m), but not all will
be functions of stress (σ) or time (t). A detailed experimental and theoretical analysis of this subject
will be published shortly by the first author. The creep component in equation (1) would contain
“transitional thermal creep” related only to temperature increase, “drying creep” related to moisture
loss, “hydration creep” related to increased hydration etc.

3.1 Upper bound of creep (LITS)

The upper bound of creep occurs during first heating-up under load and is determined experimentally
from two test: (a) a test conducted without load to produce the thermal strain which comprises the
thermal expansion and shrinkage components and (b) the same test conducted on the same material
but under a constant load. The difference in strain between the two tests produces the Load Induced
Thermal Strain (LITS) (Figure 3) which contains “transitional thermal creep” (ttc), drying creep,
changes in the elastic strain and other components. The combination of transitional thermal creep and
drying creep is called “transient creep” which is the expression often loosely used to denote LITS.

Khoury, G. 4
CISM - Effect of Heat on Concrete Creep & Shrinkage

Strictly, transient creep is part of LITS which itself contains non-creep components (e.g. changes in
the elastic strain).

Transitional thermal creep is the creep component directly related to the process temperature increase
(drying creep is only indirectly related to temperature in so far as temperature influences the drying
process). Below 100oC, ttc can take up to 28 days to develop completely, but above 100oC it develops
much more rapidly and is often regarded in practice as a function of temperature only. Ttc is
irrecoverable upon cooling or upon unloading and does not occur during second heating to the same
maximum temperature – unless a long period of “healing” is allowed at room temperature between the
two heat cycles. An extensive series of tests on 5 different types of concrete containing gravel, basalt,
limestone and lightweight aggregates showed that, while the thermal strains depended largely upon
the type of aggregate used, the LITS - for a given stress/strength ratio and aggregate content by
volume - were independent of the type of aggregate used thus indicating that LITS is seated in the
cement paste. Tests conducted on pure cement paste gave considerably larger LITS values (e.g.
4,400µε instead of 1,600µε during heating at 1oC/minute under 20% load) thus proving that the
aggregate acts to restrain creep of concrete also at high temperatures [2]. LITS is strictly not linearly
related to stress but can be assumed to be so for practical purposes. It is also independent of age for
ages in excess of 3 months. The LITS trends for normal strength and high performance concretes are
also similar.

Figure 3. Definition of the Load Induced Thermal Strain (LITS)

Khoury, G. 5
CISM - Effect of Heat on Concrete Creep & Shrinkage

LITS

Figure 4. Strain behaviour of five concretes during first heating [2].

3.2 Lower bound of creep (basic creep)

Basic creep tests were performed on normal-weight (i.e. gravel) and light-weight (i.e. Lytag)
concretes at Imperial College to examine the characteristics of the lower bound of the creep of
concrete at high temperatures. Since the cement paste is the primary creeping medium of concrete,
tests were also conducted on pure cement paste. The specimens were loaded after stabilisation at
constant temperatures up to 724°C. The results suggests that it is possible, within certain limits, to
uncouple the time (t), stress (σ) and temperature (T) functions of basic creep, which can be
represented mathematically by a simple multiple of the three functions.

ε σt , ,b.cr
T
= f(t) x f(σ) x f(T) = tn x σm x f(Τ) (2)

The time function for concrete and cement paste can be expressed by a power law with constant expo-
nent (n = 0.3-0.46 depending on type of mix) within certain temperature limits (Figure 5), thus
justifying the separation of the time and temperature functions. The exponent, however, tends to
decline at higher temperatures. The same power equation seems to apply at a range of stresses thus
also justifying the separation of the time and stress functions.

The stress function is, for practical purposes, considered to be linear within the working range and the
concept of specific creep invoked. However, any departure from linearity within the working range
results in proportionately more creep at lower stresses, suggesting a value of the exponent (m) less
than 1.

Khoury, G. 6
CISM - Effect of Heat on Concrete Creep & Shrinkage

Figure 5. Basic creep of cement paste and concrete as a function of time and temperature [3].

The temperature function is non-linear and increases considerably for gravel concrete above 350°C
and lightweight concrete and cement paste above 600°C (Figure 6). The marked increase of creep of
gravel concrete is attributed to aggregate damage above 350°C. The marked increase in creep of
lightweight concrete above 600°C originates in the cement paste and cannot be correlated with
irreversible changes in the structure of cement paste but appears to be a “viscous” or “molecular
diffusion” type phenomenon related to actual temperature at loading [3].

The large increase in creep of cement paste and lightweight concrete above 600°C suggests that a
rheological criterion limits the structural, though not necessarily the refractory, usefulness of Portland
cement based concretes in general to temperatures of 550-600°C.

Figure 6. Basic creep of cement paste and concrete one day after loading at 10% of the initial
cold strength [3].

Khoury, G. 7
CISM - Effect of Heat on Concrete Creep & Shrinkage

3.3 Intermediate creep

Having described the upper and lower bounds of the creep of concrete, there exist creep strains that
fall between those two limits.

“Intermediate” creep falls under two categories – both related to the creep of concrete at constant
temperature. The first (Case 1) is for concrete that is heated under load to a given constant
temperature and then maintained at that temperature for a period of time (i.e. following a LITS test),
and the second (Case 2) is for concrete that is heated without load to a constant temperature and then
loaded before stabilisation.

The “creep” of the concrete during the transient thermal period is described above by the LITS. The
creep measured at constant temperature immediately following the thermal transient under load (i.e.
Case 1) depends upon the reactions that continue to take place at that temperature. Had all reactions
been completed during the thermal transient and the material becomes completely stable at the start of
the constant temperature stage, then the creep would be essentially that of “basic” creep with a
structure dissimilar to that had the material been heated without load. In practice, concrete will
experience “delayed transient creep” at the end of the thermal transient which is evident from the test
results presented in Figure 7 which show the creep of basalt concrete measured from the end of the
transient (< 0.5 day) up to 5 days (i.e. h-5) when compared with that measured from one day up to 5
days (i.e. 1-5). The “bulge” at 500oC effectively disappears in the latter case indicating that the
underlying reaction (dissociation of Ca(OH)2) has been completed at that temperature before 1 day.
For Case 2, the creep measured is greater than the basic creep by an amount which depends upon the
temperature level. Again this relates to the reactions taking place in the concrete before stabilisation.
The Case 2 creep for cement paste, loaded at 10% after preheating for 5 days at constant temperatures
of 300oC, 400oC and 600oC is about 450µε, 1,400µε and 1,000µε respectively. After about 20 days of
preheating, these values are reduced to 300µε, 700µε and 1,000µε respectively. The largest reduction
is for 400oC because the dissociation of Ca(OH)2 is very slow at this temperature and the least
reduction is at 600oC at which temperature the material stabilises in a short period of time.

Figure 7. Creep of basalt concrete at constant temperature following heating under load at
1oC/minute [6].

Khoury, G. 8
CISM - Effect of Heat on Concrete Creep & Shrinkage

4. ISOLATION OF SHRINKAGE AND CREEP COMPONENTS

When isolating the shrinkage and creep components it is initially assumed that cracking does not takes
place during heating, at constant temperature, or during cooling. The residual strains would then
contain only contractive non-recoverable contributions from shrinkage and creep. It is also assumed
that the shrinkage components that develop in the loaded specimen (Sσ) are the same as those that
develop in the unloaded specimen (S0). The non-shrinkage load induced strains are loosely designated
as ‘creep’. Strictly, changes in elastic strains that have taken place for the loaded specimens also exist,
but these were shown to be relatively small. The shrinkage and creep components that develop during
first heating and at constant temperature are then evaluated in terms of directly measured strains as
follows:

4.1 During First Heating

Shrinkage ( ε tr , shr ) is evaluated indirectly from the difference between residual shrinkage strains
measured after first cooling and the shrinkage measured at temperature for the unloaded specimen.

ε tr , shr = ε 0res - ε 0c , tr ⇒ 5 [1 specimen required S0] (3)

where subscript 'tr' = initial thermal transient, subscript 'shr' = shrinkage, and subscript 'res' = residual

Creep ( ε σtr , cr ) is evaluated indirectly from the difference between the 'total' strain measured during
first heating for the loaded and unloaded specimens.

ε σtr , cr = ε 0tr , tot + ε σtr , tot [2 specimens required S0 & Sσ] (4)

where subscript 'cr' = creep.

4.2 At Constant Temperature

Shrinkage ( ε c , shr ) is measured directly from the strain at temperature of unloaded specimen:

ε c , shr = ε 0c , tr ⇒ 5 [1 specimen required S0] (5)

where subscript 'c' = constant temperature.

Creep ( ε σc , cr ) is evaluated indirectly as the difference in strain between the loaded and unloaded
specimen measured at constant temperature:

ε σc , cr = ε σc , tr ⇒ 5 - ε 0c , tr ⇒ 5 [2 specimen required S0 & Sσ] (6)

It is important to note that none of the above four components was evaluated using the measured
residual strain ( ε σres ) of the loaded specimen (Sσ).

Khoury, G. 9
CISM - Effect of Heat on Concrete Creep & Shrinkage

Figure 8. Relative proportions of Shrinkage, LITS and creep during the first heating up transient
to temperatures of 100-400oC at 1oC/minute and subsequently at constant temperature
for up to 5 days from start of heating. Basalt concrete heated under 0% & 20% load [6].

5. Prediction of Residual Strains by Superposition

A simple superposition of strain approach is adopted to predicted the residual strain after first cooling
of the loaded specimen for the test temperature 'T' and stress level 'σ '. It is assumed to be the sum of
the shrinkage and strain components that develop during heating and at constant temperature and
which have been evaluated separately in the previous section:

ε σres (predicted) = ε σtr , cr (T) + ε tr , shr (T) + ε σc , cr (T) + ε c , shr (T) (7)
⇓ ⇓ ⇓ ⇓
{ Creep + Shrinkage } + { Creep + Shrinkage}
⇓ ⇓
During Heating At Temperature

Considering first the unloaded specimens, the shrinkage taking place during heating ( ε tr , shr ) for test
temperature 110oC was a small proportion (<15%) of the total residual shrinkage. In other words, the
bulk of drying shrinkage has taken place after the test temperature of 110oC has been reached (Figure
8). This conclusion is compatible with the moisture loss measurements which show that the basalt
concrete specimens heated to 105oC at 1oC/minute would have lost only 3% of the ultimate water
loss for that temperature [5]. The shrinkage component developing during the thermal transient
increases steeply with test temperature in the range 100-200oC to reach about 60% of the total
recorded after 5 days. This is also compatible with the moisture loss data which show the peak of
moisture loss to occur at about 200oC [5].

Khoury, G. 10
CISM - Effect of Heat on Concrete Creep & Shrinkage

Figure 9: Effect of thermal cycling on the total displacement of basalt concrete [7].

Turning attention to the loaded specimens, the predicted ε σres values for test temperatures up to 400oC
for the BI and BII mixes are very close to 100% of the actual measured residual strains (Figure 8).
The average ε σres value of the 10 predictions taken for test temperatures between 110oC and 400oC
for the three concretes is 100.3%. Even for the 600oC test temperature, the predicted values for
limestone and basalt concrete mixes are 83% and 93%-99% of the recorded value. Equation 7 does
not take into account changes in the elastic strains or the relative cracking expansion taking place
during cooling between the unloaded and loaded specimens. The good predictions made for the basalt
concretes for test temperatures up to 400oC using equation 7 indicate that no such relative cracking
has taken place during cooling within that temperature range. Likewise, the poor prediction value for
the limestone mix at test temperature of 600oC suggest that it is necessary to include the relative
expansion during cooling into equation 7.

In Figure 8 the total residual strain is divided into its four constituent components of shrinkage and
creep as specified by equation 7. For the 110oC temperature, the total contribution of the transient
stage is small at about 25%, but for the test temperature range 200-400oC it rises to about 65-70%.
The bulk of this is the Load Induced Thermal Strain (LITS or approximately ε 20 tr , cr ) developing during
first heating.

5. CONCLUSIONS AND PRACTICAL APPLICATIONS

• The majority of applications relate to concrete structures exposed to fire of short durations, for
which the “creep” component is the upper bound creep (i.e. LITS) which can be simplified to a
strain that is linearly related to stress and a function of temperature only. For nuclear reactor
applications (e.g. the British AGR type reactors), the prestressed reactor concrete vessel would be
exposed to elevated temperatures for years. In this case, LITS would be followed by “constant
temperature” creep. The former could be modelled as an “instantaneous” strain followed by a
time-dependent strain [4].

Khoury, G. 11
CISM - Effect of Heat on Concrete Creep & Shrinkage

• The shrinkage and creep components of concrete can be isolated separately for the period during
heating-up and for the period at constant temperature.

• Predictions of the residual strains for the loaded specimens can be made by simple superposition
of creep and shrinkage components (equation 7) up to a certain critical temperature. Above the
critical temperature, it is necessary to add a ‘cracking component’.

5. REFERENCES

1. Khoury, G.A. Strain components of nuclear-reactor-type concretes during first heat cycle.
Nuclear Engineering & Design, no. 156, 1995.

2. Khoury, G.A., P.J.E. Sullivan and B.N. Grainger, Strain of concrete during first heating to
600oC under load. Magazine of Concrete Research, Vol. 37, No. 133, Dec 1985.

3. Khoury, G.A., P.J.E. Sullivan and W.P.S. Dias, Deformation of concrete and cement paste
loaded at constant temperatures from 140oC to 720oC. Materials and Structures, Vol. 19, No.
110, March 1986.

4. Khoury, G.A., Majorana, C.E., Pesavento, F. and Schrefler, B.A., Modelling of heated
concrete, Magazine of Concrete Research, 54, No. 2, April 2002, pp. 77-101.

5. Khoury, G.A., B.N. Grainger, and P.J.E. Sullivan, Transient thermal strain of concrete:
Literature review, conditions within specimen and behaviour of individual constituents.
Magazine of Concrete Research, Vol. 37, No. 132, September 1985.

6. Khoury, G.A. Separation and prediction of irrecoverable strain components of concrete during
the first thermal cycle. Proc. 12th SMiRT conference, Elsevier Science Publishers, 1993.

7 . Khoury, G.A.
Transient thermal creep of nuclear reactor pressure vessel type concretes.
PhD thesis, University of London, 1983.

Khoury, G. 12