LOAD INDUCED THERMAL STRAIN: IMPLICATIONS FOR STRUCTURAL BEHAVIOUR

ANGUS LAW1 and MARTIN GILLIE2

ABSTRACT Load induced thermal strains (LITS) have been studied for many years from a material science perspective. This paper examines the implications of LITS for structural behaviour. The basic principles are presented and the often confusing terminologies used to describe the main LITS components are defined and explained in terms relevant to structural engineering. Given the complexity of LITS, it is tempting to build complicated phenomenological models. However, for structural modelling purposes, it is found that it is not necessary to include all of these phenomena. In fact, for short term transient heating conditions such as exposure of a structure to fire it is only necessary to distinguish between the elastic and plastic components of strain. An idealised test on a pinned column demonstrates the influence that effective representation of LITS has on the response of a structure during the cooling phase of a fire. Two material models are applied, and the importance of defining an elastic modulus which is distinct from the apparent modulus of the material is demonstrated.

1. INTRODUCTION Load induced thermal strain (LITS) is an umbrella term used to refer to a number of different strain components1 in heated concrete. These components alter the strain behaviour of concrete if the material is heated under applied stress. In fact, concrete heated under high stresses can be observed to shrink rather than expand due to thermal expansion. The most significant components of LITS (but not all) are non-recoverable; hence, the effects of LITS are only evident on first heating. There has been a great deal of work concerning LITS over the past four decades2-5. Modern development of LITS started in the 1960s with the realisation and confirmation that
PhD. Student, The University of Edinburgh, School of Civil and Environmental Engineering, email: A.Law@ed.ac.uk 2 Lecturer, The University of Edinburgh, School of Civil and Environmental Engineering, email: M.Gillie@ed.ac.uk
1

SiF 2008 Singapore

the main LITS component existed5. The following decade saw a great deal of work which identified numerous LITS components and isolated them demonstrating that the main component was distinct and had its own properties. During the 1980s a large amount of work was carried out particularly at Imperial College6 which suggested the existence of a master LITS curve whose properties were independent from other concrete properties such as strength, age, moisture content, and aggregate type. Khoury et al. also showed that LITS was physically independent from free thermal expansion. Despite the volume of work on the material property aspects of LITS, the implications for structural behaviour have yet to be fully understood. The presence of LITS provides opportunities for the development of additional strains that are not identified. These could cause structures to respond differently than previously expected. Since the strains involved are largely irrecoverable, LITS has an active influence during heating, but is not active during cooling. As the development of larger, holistic, structural models becomes more commonplace, the understanding of how a structural member behaves and interacts with the rest of a building under heating and cooling becomes more important; the influence of preloading and the development of locked-in strains becomes more pertinent. Though there are a large number of components involved in LITS it is not necessary to build phenomenological models which include every strain component and its governing factors. On this basis, a number of researchers have developed mathematical models which attempt to describe LITS. Recently, several studies have reviewed the available material models7-9. This paper will focus on the models by Terro10 and Anderberg11 with reference to Schneiders model12 and Eurocode 213. Structural modellers have also begun to include these LITS models in their simulations14, 15. However, though comparisons of the different material models have been conducted7, one model has not been found to represent structural behaviour more satisfactorily than another.

2. EXPLANATION OF LITS LITS is defined as the difference between the free thermal expansion of concrete when it is heated, and the net thermal expansion when the same concrete is heated under a level of pre-stress. Figure 1. shows the difference between an unstressed sample and a pre-stressed sample. This figure is similar to that presented by Khoury16. It can be observed that under higher levels of pre-stress the material shrinks.
0.007 0.006 0.005

Total strain

Free thermal expansion Net expansion under pre-stress

0.004 0.003 0.002 0.001 0.000 100 200 300 400

}
o

LITS

Temperature ( C)

500

600

Fig. 1 Difference between strain when heated with and without applied stress

SiF 2008 Singapore

The existence of a large number of strain components contributing to LITS has led to a number of different terms being used to describe them. This has caused some confusion regarding what different terms mean. The more commonly used terms are disambiguated below. Transitional thermal creep (ttc) develops irrecoverably during, and for a few days following, first-time heating of sealed concrete under load5. Ttc is by far the largest component of LITS10. It is thought that the origins of transitional thermal creep are in the cement paste and that the addition of aggregate restrains it. Drying creep is the shrinkage experienced by the material due to the evaporation of water. This can only occur where the specimen is unsealed and, therefore able to dry out. Transient creep and transient strain, describe the same phenomena and refer to the sum of ttc and drying creep. Though both these terms are technically time dependant, it is believed that for practical purposes they can be regarded as quasi-instantaneous. Since drying creep is regularly omitted from the transient strain calculation, and because ttc is such a large component of LITS, the three terms; ttc, transient strain, and LITS have been incorrectly regarded as synonymous. Henceforth, for clarity, transient creep/strain will be referred to as transient strain. Basic creep is used to describe the strain of thermally stabilised concrete when loaded at constant temperature. In conditions where the temperature is not constant, by definition, the term basic creep cannot be used. Instead, Khoury1 adopts the term of a time dependent strain which develops under transient heating conditions. Changes in elastic strain are used to describe the elastic components of LITS which are caused by the change in elastic modulus as temperature increases. This is illustrated in Figure 2.

3. CONSTITUTIVE RELATION Though LITS appears to manifests itself as a change in the thermal expansion under different pre-stresses and temperatures, it can instead be thought of as being part of the stressstrain definition of the material. Figure 2. (a) shows a material whose elastic modulus degrades with increasing temperature. The influence of this change is demonstrated in Figure 2. (b) where the total strain experienced by the material does not follow free thermal expansion. Instead, the strain at each temperature is reduced by the change in elastic strain. The path followed by the horizontal line in Figure 2. (a) demonstrates how the stress remains the same while the strain increases. The degree to which the total strain is reduced is dependant upon the level of applied stress. Of course, the LITS developed in Figure 2. (b) will be fully elastic, and were the material to be cooled back to the starting temperature the strains would be fully recovered. If, however, plastic strains were introduced they would not recover on cooling; the combination of these different strains is what is meant by the multiple components of LITS. Defining LITS in this way allows it to be included in the material model for the concrete.

SiF 2008 Singapore

1.0

0.5

Total strain

Stress

20 C o 100 C o 200 C o 300 C o 400 C o 500 C

1.0
Free thermal strain Strain with pre-stress

0.5

0.0 0.0 0.5 1.0

0.0 0 100 200 300

o

400

500

Strain
(a) Elastic modulus at different temperatures

Temperature ( C)
(b) Effect of different elastic moduli on total strain

Fig. 2 Modelling LITS as part of the constitutive curve

4. MATERIAL MODELS Several different models have been developed to describe LITS. However, because of the way that the components of LITS are inseparable from the constitutive curve, a model for LITS cannot be regarded in isolation. For a model to be used in finite-element analysis, the full constitutive curve must be defined. This paper will only describe two academic models which develop a full constitutive relation. The model presented in the Eurocode will also be examined. 4.1 Anderbergs model In 1976 Anderberg11 developed a constitutive model for concrete. The present authors were unable to obtain the original presentation of this model, though it has been described extensively in a number of different publications7, 8. The instantaneous stress-strain curve is composed of two branches: an ascending branch, represented by a quadratic curve which traverses the peak stress; and a linear descending branch which joins with the ascending branch when their gradients are equal and intersects with the strain axis at a predefined value of peak strain. The model includes a formulation for creep strain. As its name suggests, this is a time dependent component. It is recognised that for short term transient problems such as fires creep strain can be regarded as negligible. When temperature is less than 550C degrees, Anderbergs model relates the transient strain term (the largest LITS component) to thermal expansion and the ratio of stress-to-peakstress. For temperatures greater than 550C transient strain is given only in terms of stress-topeak-stress ratio. For T 550C

tr = k tr
For T 550C

th u0

(1)

SiF 2008 Singapore

 tr = 0.0001 T u0

(2)

Where Anderberg gives ktr as 2.35 for calcareous concrete. The existence of the link to free thermal strain is the subject of some debate as it has been demonstrated that there is no physical evidence to link it with transient strain. In fact, transient strain has been shown to be independent of free thermal strain. This is exemplified by the fact that a lightweight concrete with zero free thermal strain can be shown to have the same transient strain curve as a normal weight concrete. Neilsen 9, proposed a modification to the Anderbergs formulation of transient strain which removes the link to free thermal strain.

tr = 0.000038

u0

(3)

Unlike Anderbergs transient strain component, this can be used for all values of temperature and stress. Nielsens formula has been demonstrated to provide a good match to experimental data. The version of Anderbergs model used in this paper adopts the modification proposed by Nielsen. 4.2 Terros model The introduction mentioned the large body of work carried out at Imperial College. Though no mathematical model has been directly proposed by the authors of that work, Terro10 used the experimental data to create a constitutive model which includes LITS components. The main ascending branch is based on a simplified version of the model proposed by Schneider12. In recent studies of constitutive models, though, there has been inconsistency in the application of the instantaneous stress-strain curve. The study by Li7 regarded the relationship to be linear elastic, whereas, the study by Youssef8 used a simplified Schneider method. This study will use the latter method since this permits a descending branch and, in the study by Youssef, it was concluded that the model achieved good accuracy. The equation for LITS is fitted to Khourys master curve6 and is:

LITS = LITS (T ,0.3) 0.032 + 3.226 u0

Where: LITS (T ,0.3) = 43.87 2.73T 6.35 10 2 T 2 + 2.19 10 4 T 3 2.77 10 7 T 4 10 6

(4)

(5)

Two versions of the latter equation are provided by Terro since Thames gravel concrete departs from the master curve above 400C. A fifth order polynomial is proposed for use in this case. A number of formulae are also presented for strain at peak stress to account for different levels of pre-stress. In summary, both of the models have key features in common: they both present a constitutive stress strain curve which varies with temperature; the curves are based on the way
SiF 2008 Singapore

that material strength and elastic modulus degrades with increased temperature; these curves are then modified by the transient strain component (or the LITS component in the Terro model). This three step method of obtaining a final constitutive curve is demonstrated in Figure 3. (a).

1.0

1.0

Stress (normalised)

Stress (normalised)

0.5
Elastic modulus Instant stress-strain Curve with LITS

0.5
Actual modulus Apparent modulus Constitutive curve

0.0 0.000

0.005

0.010

0.0 0.000

0.005

0.010

Strain
(a) Three step formulation of constitutive curve

Strain
(b) Apparent and actual elastic moduli

Fig. 3 Construction of the final constitutive curve

5. APPARENT ELASTIC MODULUS For most stress-strain curves, it is reasonable to assume that the initial tangent modulus of the constitutive relation is equal to the elastic modulus. Put another way, the first section of the stress-strain curve is within the elastic region of the material. The situation becomes less clear with the inclusion of LITS. The modification of the original stress-strain curve by the LITS components causes plastic strains to develop from very low stresses. The implication of this is that the tangent modulus of the constitutive strain curve is never equal to elastic modulus of the material even at zero strain, or where the ascending branch is straight. It is, therefore, impossible to identify the elastic modulus of the material from simple inspection of the constitutive curve. If this is attempted, a value of elastic modulus will be obtained which is the not the actual modulus, but the apparent modulus of the material. This will result in significant underestimation of the true plastic strains which would develop if the actual modulus was in use. Figure 3. (b) illustrates the difference between the apparent elastic modulus of a material and the actual modulus as defined by the constitutive equations. In finite-element applications the difference between the actual and apparent material moduli must be carefully considered. There are circumstances under which the differences between the moduli will not effect the response of a structural model. For example, under direct strain or heating the solution will simply follow the defined curve or move between curves in the appropriate way. However, if the applied strain is reduced, there are multiple heating phases, or there are more complex loading patterns to be considered, the maldefinition of the plastic strains becomes significant. This has not been a major issue for most models as the emphasis has predominantly been on the heating phase of the fire. However, as attention turns to cooling it is necessary to consider these components more carefully. The concrete constitutive curves described in Eurocode 213 have the potential to suffer from these problems. The model is defined by the specification of peak stress, peak strain, and strain-at-peak-stress not elastic modulus. Thus, the user has to define an elastic modulus to associate with the model. A reasonable assumption might be to assume that the elastic

SiF 2008 Singapore

modulus should be equal to the initial gradient of the constitutive curve. However, as illustrated above, this may not be a useful or accurate assumption.

6. SIMPLE EXAMPLE The influence of the definition of elastic modulus is demonstrated using a series of simple models. A finite-element model was created using the commercially available finiteelement package ABAQUS. The model was of a simplified pure concrete column consisting of a single member pinned at both ends. The temperature was raised uniformly to 500C and then cooled back down to ambient temperature. Five different models were applied: the Anderbreg model; the Anderberg model with use of the apparent elastic modulus; the Terro model; the Terro model with the use of the apparent elastic modulus; and the Eurocode model. The constitutive curves for ambient and 500C are shown in Figure 4. The resulting stress from the heating and cooling regimes was then plotted against the temperature (Figure 5.).
1.0
Terro Anderberg Eurocode

1.0
Terro Anderberg Eurocode

Stress

0.5

Stress
0.01 0.02 0.03

0.5

0.0 0.00

0.0 0.00

0.01

0.02

0.03

Strain
(a) Constitutive curves at ambient

Strain
(b) Constitutive curves at 500C

Fig. 4 Constitutive curves at different temperatures Thermal expansion caused compressive stresses to build up as the column was heated due to the fixed ends of the column. Though the associated material models followed the same path through heating, the stains that built up in the models with the apparent moduli were predominantly elastic, whereas, in the models with the actual moduli, most of the strain was plastic. This only became evident on cooling, where the apparent models slowly released the stresses and, were only pulled into tension when the temperature was close to returning to ambient. The models which used the actual moduli, on the other hand, were unable to recover the strains developed and rapidly progressed into tension.

SiF 2008 Singapore

20 0 0 -2 -4 Stress (MPa) -6 -8 -10 -12 -14

0 20

500 Terro Apparent Terro Actual Anderberg Apparent Anderberg Actual Eurocode

20 2

Heating

Cooling
2 20
o

500 Temperature ( C)

Fig. 5 Stress response to heating and cooling regime Also presented in Figure 5. is the stress path that the Eurocode 2 model takes under the same heating regime. The first notable difference between this model and the two academic models is that the stresses generated were significantly higher. This was due to the steeper gradient of the constitutive curve in the ascending branch. Though this could be considered as an important difference between the models, it is of no specific relevance to this study which is primarily concerned with the development of elastic and plastic strains. Analysis of the cooling stress path demonstrated that though the model does not exhibit either strain type mostly plastic or mostly elastic in the extreme, it is much more similar to the apparent models. This is expected since the elastic modulus was derived from the gradient of the constitutive curve. The plastic strains are more significant than the other apparent models because the curvature of the ascending branch in the Eurocode is greater than that of the other models. Though this is a very simplified model, it demonstrates that it is necessary to consider the difference between the apparent and the actual elastic modulus of a model. It shows that a difference of definition can have significant impact on the structural response of during cooling or unloading.

7. CONCLUSIONS There are several conclusions which can be drawn from this study. These are summarised below: LITS contains a number of different components which influence the behaviour of concrete when it is heated under applied stress. The most significant component of LITS is transitional thermal creep which is irreversible. However, this is frequently considered in combination with drying creep and termed transient strain. The terms transient strain and transient creep refer to the same phenomena.

SiF 2008 Singapore

Several models exist with which to model LITS. One model has not been found to be significantly superior to another. There are significant differences between a constitutive curve which includes LITS, and a model which accurately represents LITS components. Consideration of the apparent and actual elastic moduli is necessary to accurately represent LITS in structural models.

REFERENCES 1. Khoury, G.A., "Strain of heated concrete during two thermal cycles. Part 3: isolation of strain components and strain model development", Magazine of Concrete Research, 2006. 58(7): p. 421-432. Thelandersson, S., "Modeling of Combined Thermal and Mechanical Action in Concrete", Journal of Engineering Mechanics, 1987. 113(2): p. 893-906. Schneider, U., "Concrete at High Temperatures - A General Review", Fire Safety Journal, 1988. 13: p. 55-68. Bazant, Z.P. and M.F. Kaplan, "Concrete at High Temperatures: Material Properties and Mathematical Models". 1996, Harlow: Longman Group Limited. 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, 1985. 37: p. 131-144. Khoury, G.A., B.N. Grainger, and P.J.E. Sullivan, "Strain of concrete during first heating to 600C under load", Magazine of Concrete Research, 1985. 37(133): p. 195215. Li, L.-y. and J. Prukiss, "Stress-strain constitutive equations of concrete material at elevated temperatures", Fire Safety Journal, 2005. 40: p. 669-686. Youssef, M.A. and M. Moftah, "General stress-strain relationship for concrete at elevated temperatures", Engineering Structures, 2007. 29: p. 2618-2634. Nielsen, C.V., C.J. Pearce, and N. Bicanic, "Theoretical model of high temperature effects on uniaxial concrete member under elastic restraint", Magazine of Concrete Research, 2002. 54(4): p. 239-249. Terro, M.J., "Numerical Modeling of the Behaviour of Concrete Structures in Fire", ACI Structural Journal, 1998. 95(2): p. 183-193. Anderberg, Y. and S. Thelandersson, "Stress and deformation characteristics of concrete, 2 - experimental investigation and material behaviour model". Bulletin 54. 1976, Sweden: University of Lund. Schneider, U. "Modelling of concrete behaviour at high temperatures". in Design of Structures against Fire. 1986. Aston, Birmingham: Elsevier. Eurocode2, "Design of Concrete Structures". 1992. Bratina, S., M. Saje, and I. Planinc, "The effects of different strain contributions on the response of RC beams in fire", Engineering Strucutres, 2007. 29: p. 418-430. Nechnech, W., F. Meftah, and J.M. Reynouard, "An elasto-plastic damage model for plain concrete subjected to high temperatures", Engineering Structures, 2002. 24: p. 597-611. Khoury, G.A., "Strain of heated concrete during two thermal cycles. Part 1: strain over two cycles, during first heating and at subsequent constant temperature", Magazine of Concrete Research, 2006. 58(6): p. 369-385.

2. 3. 4. 5.

6.

7. 8. 9.

10. 11.

12. 13. 14. 15.

16.

SiF 2008 Singapore