You are on page 1of 19


Effect of fire on concrete and concrete structures

Gabriel Alexander Khoury Imperial College, London, UK

The behaviour of concrete in fire depends on its mix proportions and constituents and is determined by complex physicochemical transformations during heating. Normal-strength concretes and highperformance concretes microstructurally follow similar trends when heated, but ultra-highperformance concrete behaves differently. A key property unique to concrete amongst structural materials is transient creep. Any structural analysis of heated concrete that ignores transient creep will yield erroneous results, particularly for columns exposed to fire. Failure of structural concrete in fire varies according to the nature of the fire; the loading system and the type of structure. Failure could occur from loss of bending or tensile strength; loss of bond strength; loss of shear or torsional strength; loss of compressive strength; and spalling of the concrete. The structural element should, therefore, be designed to fulfil its separating and/or load-bearing function without failure for the required period of time in a given fire scenario. Design for fire resistance aims to ensure overall dimensions of the section of an element sufficient to keep the heat transfer through this element within acceptable limits, and an average concrete cover to the reinforcement sufficient to keep the temperature of the reinforcement below critical values long enough for the required fire resistance period to be attained. The prediction of spalling ` hitherto an imprecise empirical exercise ` is now becoming possible with the development of thermohydromechanical nonlinear finite element models capable of predicting pore pressures. The risk of explosive spalling in fire increases with decrease in concrete permeability and could be eliminated by the appropriate inclusion of polypropylene fibres in the mix and/or by protecting the exposed concrete surface with a thermal barrier. There are three methods of assessment of fire resistance: (a) fire testing; (b) prescriptive methods, which are rigid; and (c) performancebased methods, which are flexible. Performancebased methods can be classified into three categories of increasing sophistication and complexity: (a) simplified calculations based on limit state analysis; (b) thermomechanical finite element analysis; and (c) comprehensive thermohydromechanical finite element analysis. It is only now that performance-based methods are being accepted in an increasing number of countries.

Prog. Struct. Engng Mater. 2000; 2: 429d447

This paper presents a general brief outline of the effect of fire on both concrete material and concrete structures with emphasis being placed upon the subject areas receiving most attention in the past few years, namely: (a) deterioration in mechanical properties of concrete and especially of high-performance and ultra-high-performance concretes in fire; (b) explosive spalling and the use of polypropylene fibres; (c) the development of finite element structural analysis models capable of predicting pore pressures and spalling; and (d) fires in tunnels. The basic principles of fire engineering are also presented on the
Copyright ^ 2000 John Wiley & Sons, Ltd.

assumption that the reader is not a specialist in this field.

Research into the effect of fire on concrete and concrete structures has been conducted since at least 1922[1], primarily in relation to buildings. The main areas of interest were: (a) the understanding of the complex behaviour of the material itself; and (b) the structural safety and integrity of the building during, and after, fire. The introduction of the advanced gas-cooled (AGR) nuclear reactor system in the UK in the 1960s and 1970s provided added impetus and funding for
Prog. Struct. Engng Mater. 2000; 2:429}447

research into the effect of high temperatures on concrete, albeit in relation to service and accident nuclear reactor conditions[2], but nevertheless with useful technology transfer to fire applications. Much of the output from this research has been published in a series of biannual SMiRT conferences since 1971, and in the journal Nuclear Engineering and Design. Up to about a decade ago, fire research was focused upon the behaviour of normal-strength concrete at high temperatures, and engineers mainly employed prescriptive methods of design to ensure structural stability in fire. Since then, there have been two major developments in this field, namely: (a) the increasing use of high-performance concrete (HPC) in buildings, tunnels and bridges; and (b) the growing acceptance of the use of performance-based structural analysis and design against fire. The employment of HPC in tunnel linings, with its greater tendency to explosively spall in fire, has resulted in a series of high-profile and costly incidents of explosive spalling of concrete in tunnel fires, which further fuelled urgent research aimed at a better understanding and solution of this serious problem. The impetus for the development of performance-based structural analysis came from the limitations inherent in the traditional prescriptive methods of design. Several countries have already developed performance-based codes (e.g. UK, Sweden, Norway, New Zealand and Australia) and many more countries are in the process of doing so. While the UK has led in this field with the publication in 1992 of the Building Regulations Approved Document B3[3], performance-based structural design and analysis methods are still not as widely used by engineers in the UK, as they are in Scandinavia. This topic is currently the focus of much discussion, research and development worldwide. This has led to new ideas for improving fire safety, thus encouraging the engineer to develop new creative solutions[4].


Fig. 1. Standard fire scenarios for buildings (ISO 834 or BS 476), offshore and petrochemical industries (hydrocarbon), and tunnels (RWS, RABT); idealized fire curves established from experience in real fires which vary considerably from fire to fire

decay stage. By comparison, real fires can have a slower or longer growth phase, and once they are established, temperatures can be higher than the furnace temperatures, though they are rarely sustained because they are subject to pronounced fluctuations. The standard temperaturetime curve, therefore, corresponds to a severe fire, but not the severest possible fire.



In the 1970s, the oil company Mobil investigated hydrocarbon fuel fires and developed a temperaturetime profile with a rapid temperature rise in the first 5 min of the fire up to 900 3C (i.e. 176 3C min!1) and a peak of 1100 3C. This research laid the foundation for test procedures to assess fire- protecting materials for the offshore and petrochemical industries.


More recently, a spate of major tunnel fires have indicated that an even more severe fire scenario needs to be considered. In the Netherlands, the Ministry of Public Works, the Rijswaterstaat (RWS), and the TNO Centre for Fire Research have established a fire curve for the evaluation of passive protecting materials in tunnels[10]. This RWS Dutch fire curve models a most severe hydrocarbon fire, rapidly exceeding 1200 3C and peaking at 1350 3C (melting temperature of concrete) after 60 min and then falling gradually to 1200 3C at 120 min, the end of the curve. RWS is intended to simulate tankers carrying petrol in tunnels with a fire load of 300 MW causing a fire for 2 h, and was established on the basis of Dutch experience in tunnel fires. However, the maximum temperatures attained in recent major fires did not reach RWS levels, e.g. Channel (1100 3C), Great Belt (800 3C), Mont Blanc (1000 3C), Tauern (1000 3C). The RWS fire curve, therefore, represents the severest form of tunnel fire in terms of initial heating rates and maximum temperatures. The RABT German fire curve, with a descending branch, represents a less severe fire
Prog. Struct. Engng Mater. 2000; 2:429}447

Fire scenarios
The temperature}time curves in Standard fires (Fig. 1) used in testing, analysis and design were established from experience in real fires and fall into three main categories, depending upon the application (i.e. buildings, offshore/petrochemical, tunnels).


476 [5] OR


834 [6]

The standard furnace curve represents a typical building fire based upon a cellulosic fire in which the fuel source is wood, paper, fabric, etc. In BS 476, the temperature increases from 20 to 842 3C after the first 30 min (i.e. equivalent to an average heating rate of 27.4 3C min!1). This fire profile has a slow temperature rise up to 1000 3C over a period of 120 min. This curve represents only one possible exposure condition at the growth and the fully developed fire stages, and does not include a final
Copyright ^ 2000 John Wiley & Sons, Ltd.


scenario in tunnels than the RWS curve, reaching a maximum temperature of 1200 3C (melting point of some aggregates) sustained up to 1 h before decaying to ambient. Many standard (building) insulation materials decompose and lose their function above 1200 3C. It may be that insulation materials behave well in BS 476, ISO 834, Eurocode hydrocarbon fires and even the RABT fire, but may not withstand a fire under RWS conditions. For this reason, the manufacturer of thermal insulation coatings for tunnel concrete linings, Lightcem, have recently developed two different thermal types of insulation materials: one for temperatures up to 1200 3C based on Portland cements, and another for temperatures up to 1350 3C based on aluminous cements. It should, however, be noted that structural aluminous cement concretes have been banned because of the risk of conversion and weakening that can take place under certain temperature/humidity conditions.


Fire and concrete material

Fire resistance is a concept applicable to elements of the building structure and not a material, but the properties of a material affect the performance of the element of which it forms a part. Therefore, the effect of fire on concrete material will be discussed first in this section and measures for improving material performance will be proposed. The influence of fire on concrete structures will then be discussed in the following section. The advantages of concrete in a fire are two-fold. It is: incombustible (e.g. when compared with wood); and a good insulating material possessing a low thermal diffusivity (e.g. when compared with steel). However, there are two problems of concrete in fire. These are: deterioration in mechanical properties as temperature rises, caused by physicochemical changes in the material during heating (Fig. 2); and explosive spalling, which results in loss of material, reduction in section size and exposure of the reinforcing steel to excessive temperatures. Consequently, both the separating/insulating and load-bearing functions of the concrete member could be compromised.
Fig. 2. Physicochemical processes in Portland cement concrete during heating: (a) temperatures are for the concrete material and not the fire, except for explosive spalling where the temperature range is that of the concrete surface and depends upon the heating rate and type of concrete; (b) unsealed cement paste (i.e. allowed to dry) behaves very differently from moist sealed cement paste d above 100 3C, the former is dominated by dehydration processes, while the latter is dominated by hydrothermal chemical transformations and reactions

E applied loading; E external sealing influencing moisture loss from the surface. These factors are described below in relation to compressive and tensile strength.

Compressive and tensile strength Before explaining the changes in strength of concrete as a function of increasing temperature, two positive aspects of heated concrete need pointing out. Influence of transient creep Up to the 1970s, scientists were puzzled as to why concrete does not break up when heated above 100 3C, because the differential strain resulting from the expansion of the aggregate and shrinkage of the cement paste is far too large to be accommodated by elastic strains. The discovery of the phenomenon of transient creep provided the answer. Transient creep (strictly, it should be called load-induced thermal strain, LITS[11]) develops during first heating (and not during cooling) under load. LITS is unique to concrete amongst structural materials. Above 100 3C it is essentially a function of temperature and not of time and is, therefore, relatively easy to model
Prog. Struct. Engng Mater. 2000; 2:429}447


Deterioration in mechanical properties of concrete upon heating may be attributed to three material factors: E physicochemical changes in the cement paste; E physicochemical changes in the aggregate; E thermal incompatibility between the aggregate and the cement paste; and are influenced by environmental factors, such as: E temperature level; E heating rate;
Copyright ^ 2000 John Wiley & Sons, Ltd.

mathematically in a short-term heating scenario such as fire. A further simplification, discovered by the author[11], is the existence of a master LITS for Portland-cement-based concrete in general for temperatures up to 450 3C, irrespective of the type of aggregate or cement blend used (Fig. 3). LITS is much larger than the elastic strain (compare Figs 3 and 4 for 20% stress level), and contributes to a significant relaxation and redistribution of thermal stresses in heated concrete structures. Any structural analysis of heated concrete that ignores LITS will, therefore, be wholly inappropriate and will yield erroneous results, particularly for columns exposed to fire. This phenomenon is still not fully appreciated by structural engineers and should be incorporated more fully into standards and design codes.

cooling, and is attributed to the weakening of the physical van der Waals forces as the expanding water molecules push the CSH layers further apart. This minimum, therefore, does not appear when the residual strength is measured after cooling and plotted against previous temperature. Frequently, the hot strength is shown to increase to a maximum at about 200300 3C greater than the initial strength prior to heating. Most concretes experience a strength reduction above 300 3C, but this depends upon the type of aggregate and cement blend used in the mix. Between 300 and 600 3C there is room for improving the performance of concrete by judicious mix design.

Influence of loading during heating Another positive aspect of concrete at high temperature is the beneficial influence of loading which places the material into compression, compacts the concrete during heating and inhibits the development of cracks. Again, this influence is not fully appreciated by the practising engineer and is not adequately covered in the codes and standards. An example of this phenomenon is given in Fig. 4, comparing the stressstrain relation at levels of high temperature 20700 3C for concrete heated without load and under 20% load. The influence of temperature can be markedly diminished. Both the compressive strength and elastic modulus reduce far less with increase in temperature for the concrete heated under load. However, for concrete heated without load, tests consistently show the modulus of elasticity to reduce with increase in temperature by a larger proportion of the initial value than the compressive strength[12]. Influence of temperature Compressive strength, reduces from ambient to reach a minimum at 80 3C for unsealed concrete. This is an apparent strength loss, largely reversible upon
Fig. 3. Load-induced thermal strains (or transient creep) of concrete during heating at 1 3C min!1 to 600 3C under compressive loads up to 30% of the unheated strength; heating was slow to reduce structural effects; LITS was found to be more a function of temperature than of time

Fig. 4. Effect of temperature and load level during heating-up upon the residual stressdstrain relation in uniaxial compression of unsealed C70 HITECO concrete cylindrical specimens; tests conducted at constant stress rate, descending branch not shown
Copyright ^ 2000 John Wiley & Sons, Ltd.

Prog. Struct. Engng Mater. 2000; 2:429}447


However, the author discovered a marked increase in the basic creep of Portland cement paste and concrete[13] at about 550600 3C, indicating this temperature to be critical, above which concrete is not structurally useful. Fortunately, in fire, only the first few centimetres from the concrete surface experience temperatures greater than 300 3C owing to the low thermal diffusivity of concrete in general. The normal practice after fire is to remove and replace this overheated layer. As long as the concrete does not spall during fire, this layer continues to provide thermal protection to the steel and inner concrete, and would effectively act as a thermal barrier, although its structural role becomes greatly diminished. Another important phenomenon, also not fully appreciated, is the major difference in behaviour between unsealed and moist sealed concrete at temperatures above about 100 3C. The dominant process for unsealed concrete relates to the loss of the various forms of water (free, adsorbed and chemically bound), while the dominant process in sealed concrete relates to hydrothermal chemical reactions which could result in much weaker or stronger gel, depending upon the CaO/SiO2 ratio (C/S ratio). The C/S ratio is influenced by the use of cement replacements such as slag, pulverized fly ash (PFA), or silica fume in the mix. Given the large number of material and environmental factors that influence the compressive strength of heated concrete, it is no surprise that measurements of compressive strength at 150 3C can yield results ranging from as low as 30% to as high as 120% of the original cold strength[14]. It is no wonder that data for concrete strengths at elevated temperatures taken from different sources differ substantially, and in several cases even seem contradictory. Therefore, representing typical strength behaviour of all normal-weight concrete at high temperatures with a single average curve, as given in BS 8110[15], can be misleading unless the specific mix and environmental conditions (e.g. heating rate, load during heating, moisture condition, cooling rate, etc.) are specified. Recent research[12,16] has extended the testing of compressive strength from normal-strength concrete to high-performance concrete (HPC; strengths 60100 MPa) and even ultra-high-performance concrete (UHPC containing steel fibres; strengths 100300 MPa). The results shown in Fig. 5 show that the hot strengths of the HPC concretes decline less with increase in temperature than those of the UHPC concrete, but that the reverse is true for the residual strength measured after cooling (possibly because of the steel fibres). The relatively poorer hot performance of the UHPC could be due to its very dense structure, from which moisture escapes less readily. This has two effects: a physical effect due to reduced van der Waals forces as water expands upon heating, and a chemical effect whereby detrimental
Copyright ^ 2000 John Wiley & Sons, Ltd.


Fig. 5. Effect of temperature upon the residual compressive strength of two high-performance concretes (C60-SF, C70) and two ultra-high-performance concretes (CRC, RPC-AF) after heat cycling without load at 2 3C min!1: (a) actual values; (b) percentage of initial strength at 20 3C

transformations can take place under hydrothermal conditions. This suggests that the UHPC hot strength loss could have been more pronounced had the specimens been kept for a longer time at appropriate temperatures. Normal-strength concretes and highperformance concretes microstructurally follow similar trends when heated, but ultra-highperformance concrete behaves differently, as evidenced by rate of moisture loss readings taken at constant high temperatures. Until very recently, all measurements of tensile strength at high temperatures were performed by indirect means. However, there is sufficient evidence to show that the tensile constitutive behaviour of concrete at high temperature cannot be adequately determined from indirect tension tests (i.e. splitting and bending). For this reason, stressstrain tests in direct tension were performed at high temperatures for the first time last year at Imperial College in collaboration with Milan Polytechnic. The direct tensile hot strength declines with increase in temperature. Also as expected, the results obtained from the hot tests were closer to those obtained for the residual tests for the HPC than for the UHPC concretes, owing to the presence of steel.
Prog. Struct. Engng Mater. 2000; 2:429}447

Mix design against strength loss Deterioration of mechanical properties can be reduced for the temperature range from ambient up to 600 3C by judicious design of the concrete mix (Fig. 6). This would take into consideration the behaviours of; (a) the aggregate; (b) the cement paste; and (c) the interaction between them. Aggregate The choice of aggregate is probably the more important since some aggregates, such as flint or Thames gravel, break up at relatively low temperatures (below 350 3C) while other aggregates, such as granite, exhibit thermal stability, up to 600 3C. The thermal stability of different aggregates increases in the following order: flint, Thames gravel, limestone, basalt, granite, gabbro. Other desirable features in aggregates are: (a) low thermal expansion, which improves thermal compatibility with the cement paste; (b) rough angular surface, which improves the physical bond with the paste; and (c) the presence of reactive silica, which improves the chemical bond with the paste. Cement blend As for the cement blend, an important feature is the C/S ratio. A low C/S ratio results in a low calcium hydroxide (Ca(OH)2) content in the original mix and ensures a more beneficial hydrothermal reaction. Calcium hydroxide is not desirable because it dissociates at about 400 3C into CaO and CO . The  CaO rehydrates expansively and detrimentally upon cooling and exposure to moisture. The C/S ratio is reduced in practice by the use of slag, PFA or silica fume. Tests by the author show that the use of slag produces the best results at high temperatures, followed by PFA and then silica fume. The relatively low performance of the silica fume cement paste (contrary to its high durability performance at room temperature) may be attributed to the dense, low permeability structure of the paste which does not readily allow moisture to escape from the heated concrete, thus resulting in high pore pressures and the development of microcracks.


Fig. 6. Concrete mix design for improved performance at high temperature; some factors that improve performance at room temperature, e.g. low permeability from the use of SF or low water/cement ratio (w/c) increase risk of explosive spalling

spalling are violent while corner/sloughing-off spalling is non-violent. It could also be argued that surface spalling is really a subset of explosive spalling which is the most serious, and hence most researched, form of spalling. The damage caused to a concrete structure by spalling can render fire safety design calculations inaccurate and lead to significantly reduced levels of safety in fire. It is, therefore, important to: (a) better understand the fundamental mechanisms responsible for explosive spalling of concrete; (b) develop a realistic predictive model; and (c) optimise (in terms of cost and effectiveness) the methods for eliminating explosive spalling in practice.

Spalling is the violent or non-violent breaking off of layers or pieces of concrete from the surface of a structural element when it is exposed to high and rapidly rising temperatures, as experienced in fires with heating rates typically 2030 3C/min\. Spalling can be grouped into four categories: (a) aggregate spalling; (b) explosive spalling; (c) surface spalling; (d) corner/sloughing-off spalling. The first three occur during the first 2030 min into a fire and are influenced by the heating rate, while the fourth occurs after 3060 min of fire and is influenced by the maximum temperature. Surface and explosive
Copyright ^ 2000 John Wiley & Sons, Ltd.

Factors influencing explosive spalling Many material (e.g. permeability, saturation level, aggregate size and type, presence of cracking and reinforcement), geometric (e.g. section shape and size) and environmental (e.g. heating rate, heating profile, load level) factors have been identified from experiments as influencing spalling of concrete in fire (Table 1). The test results were used to produce nomograms identifying spalling and non-spalling zones (e.g. Fig. 7). A detailed account of the factors influencing spalling is beyond the scope of this paper, but the main factors are the heating rate (especially above 3 3C min!1), permeability of the material, pore saturation level (especially above 23% moisture content by weight of concrete), the presence of reinforcement and the level of external applied load. Low-permeability, high-performance concrete (HPC) is more likely to explosively spall, and to experience multiple spalling, than normal-strength concrete, despite its higher tensile strength. This is because
Prog. Struct. Engng Mater. 2000; 2:429}447


Table 1 Characteristics of the different forms of spalling Spalling Time of Nature Sound Influence Main occurrence (min) influences ..................................................................................................................... H, A, S, D, W T, A, Ft, R H, W, P, Ft H, A, S, Fs, G, L, O, P, Q, R, S, W, Z ..................................................................................................................... 7d30 30d90 7d30 7d30 Splitting Non-violent Violent Violent Popping None Cracking Loud bang Superficial Can be serious Can be serious Serious


Aggregate Corner Surface Explosive

A"aggregate thermal expansion, D"aggregate thermal diffusivity, Fs"shear strength of concrete, Ft"tensile strength of concrete, G"age of concrete, H"heating rate, L"loading/restraint, O"heating profile, P"permeability, Q"section shape, R"reinforcement, S"aggregate size, T"maximum temperature, W"moisture content, Z"section size

Fig. 7. Explosive spalling empirical envelope for normal-strength concrete, showing influences of moisture content and applied stress[24]

Fig. 8. Gradients of temperature, pore pressure and moisture in normal and high-performance concrete sections during heating from one unsealed surface (Anderberg & Khoury)

greater pore pressures build up during heating (Fig. 8), owing to the materials low permeability. Also, the peak in pore pressure occurs nearer to the surface for HPC which explains why thinner concrete sections spall repeatedly from HPC concrete in fire.

Mechanisms of explosive spalling The mechanisms proposed to explain the explosive spalling of concrete fall under three categories:
E pore pressure spalling: as favoured by Shorter & Harmathy[17], Meyer-Ottens[18], Aktarruzzaman & Sullivan[19], Khoylou[20], Ichikawa[21]; E thermal stress spalling: as favoured by Saito[22], Dougill[23]; E combined pore pressure and thermal stress spalling: as favoured by Zhukov[24], Sertmehmetoglu[25], Connolly[26].

Pore pressure spalling This has been predicted using a moisture clog model[17], vapour drag forces model[18] or an idealized spherical pore model[19]. The main factors which influence pore pressure spalling are the permeability of the concrete, the initial water saturation level, and the rate of heating. The generation of pore pressure in the heated concrete
Copyright ^ 2000 John Wiley & Sons, Ltd.

(Fig. 8) is difficult to predict reliably. The models range in complexity from the simple use of steam tables to full solution of the equations of state using finite element analysis[27,28]. The majority of such models predict pore pressure levels which are substantially less than the tensile strength of concrete. Furthermore, test measurements of pore pressures in unsealed concrete specimens also usually produced levels less than the tensile strength of concrete[25,29,30]. However, Khoylou[20], suggested that large hydraulic pressures could be generated in heated saturated sealed spherical pores to cause hydraulic pore pressure explosive spalling. This model, based on elastic theory, does not consider the influence of moisture migration between pores, nor the role of creep, and may therefore greatly overestimate pore pressures. An analytical model by Connolly[26], based on the velocities of the moisture clog and the 100 3C isotherm also consistently predicted pore pressures greater than those measured. All this serves to illustrate how difficult it has been to accurately predict pore pressures and pore pressure explosive spalling by means of analytical methods.

Thermal stress spalling At a sufficiently high heating rates, ceramics can experience explosive spalling. This is attributed to excessive thermal stresses generated by rapid heating,
Prog. Struct. Engng Mater. 2000; 2:429}447

and demonstrates that factors other than pore pressures may contribute to explosive spalling. Heating concrete generates temperature gradients (Fig. 8) that induce compressive stresses close to the heated surface (due to restrained thermal expansion) and tensile stresses in the cooler interior regions (Fig. 9). Surface compression may be augmented by load or prestress, which are superimposed upon the thermal stresses. However, very few concrete structures are loaded to levels where the necessary failure stress state is reached. This makes thermal stress spalling, by itself, a relatively rare (but not impossible) occurrence. Predicting spalling based on thermal stresses would have the merit of simplicity, as reasonable estimates may be made of their magnitude using nonlinear constitutive models. However, although Dougills theory[23] is more realistic than Saitos[22], it still does not explain the obvious role of moisture in spalling.


Combined thermal stress and pore pressures Explosive spalling generally occurs under the combined action of pore pressure, compression in the exposed surface region (induced by thermal stresses and external loading) and internal cracking (Fig. 10). Pore pressure spalling may act by itself for small unloaded specimens. For larger specimens, the pore pressure will need to be considered, together with both the thermal and load-induced stresses, before the likelihood of explosive spalling, can be assessed. Cracks develop parallel to the surface when the sum of the stresses exceeds the tensile strength of the material. This is accompanied by a sudden release of energy and a violent failure of the heated surface region. Sertmehmetoglu[25] suggested that the effect of load was simply to create planes of weakness parallel to the heated face of the concrete. He demonstrated that the action of relatively small pore pressures from within such planes could initiate spalling. Pore pressure spalling and thermal stress spalling, both influenced by external loading, act singly or on combination, depending upon the section size, the type of concrete and the moisture content. In both normal and highperformance concretes, the evidence suggests that pore pressure spalling is the more dominant of the two hence the effectiveness of the polypropylene fibres. However, unpublished test results suggest that in some ultra-high-performance concretes, containing large proportions of expansive silica, thermal stress spalling may assume a greater importance and the use of large quantities of polypropylene fibres does not seem to prevent spalling. Further research is required to settle this issue.

Fig. 9. Calculated elastic thermal stress profiles in the central cross-section of a 60 mm diameter gravel concrete cylinder heated uniformly on the curved surface at a constant rate of 13C min!1, without relaxation by creep

Fig. 10. Forces acting in heated concrete[24]

have failed[26] owing to the complex microstructure and multiphase nature of heated concrete, which does not readily lend itself to simplified analytical models. The inability to predict the occurrence of spalling has been a limiting factor in the development of robust models capable of predicting the response of concrete structures to fire. It is only now that a fully comprehensive and coupled nonlinear finite element model is being developed to predict the pore pressures and spalling in heated concrete structures (see below).

Prediction of spalling Until now, the prediction of spalling during heating has been largely an imprecise empirical exercise. Attempts to predict spalling by analytical methods
Copyright ^ 2000 John Wiley & Sons, Ltd.

Design against spalling While a whole raft of measures (Table 2) have in the past been proposed to combat explosive spalling, the only really effective method has so far been the use of a thermal barrier, protecting the surface of the concrete from the fire. It is only now that polypropylene (pp)
Prog. Struct. Engng Mater. 2000; 2:429}447


Table 2 Evaluation of preventive measures for the spalling of concrete


Method Effectiveness Comments ............................................................................................................................................... Polypropylene fibres Very effective, even in high-strength concrete May not prevent spalling in expansive ultra- highstrength concrete. Does not reduce temperatures d only pore pressures But can reduce strength Also reduces concrete increases fire resistance temperatures and

Air-entraining agent Thermal barrier Moisture content control Compressive stress control Choice of aggregate

Very effective Very effective Reduces vapour pressure Reduces explosive pressure It is best to use low expansion and small-size aggregate Reduces spalling damage Reduces spalling damage

Normal moisture content is usually above the no spalling limit for most buildings Not economical as section sizes increase If low moisture lightweight concrete is used, additional fire resistance is possible, but in highmoisture conditions violent spalling is promoted. Presence of reinforcement limited spread of spalling in the Channel Tunnel fire Difficult to use in small and narrow sections

Reinforcement Supplementary reinforcement

Choice of section shape Thicker sections reduce spalling damage Important for I-beams and ribbed sections ...............................................................................................................................................

fibres are being considered for use in the concrete mix for the purpose of increasing permeability during heating above 160 3C, thus reducing pore pressures and the risk of spalling. Polypropylene fibres melt at about 160 3C and provide channels in the concrete for moisture to escape. A further, recent advance has been the development by the company Fibrin of low-melt pp fibres (130 3C) which promise to be even more effective, particularly for low-permeability dense and saturated concrete. Nevertheless, there remain two key areas for further research and development, namely: (a) the exact mechanism by which pp fibres operate is not properly understood (e.g. the influence of the melting/evaporation process of the fibres upon the microstructure of the concrete and the additional role of microcracking); and (b) the use of pp fibres in practice is yet to be optimized (e.g. type and amount of fibres in relation to applied load and concrete strength). There is unpublished evidence that even large quantities of pp fibres may not prevent explosive spalling of some ultra-high-performance concretes in fire. When such a material contains high levels of expansive silica, thermal stress spalling may become more dominant than in normal concrete.

E initiation of strength loss critical temperature: this depends upon the type of concrete and for siliceous concrete it is about 300 3C, but can be lower for flint aggregate concrete and higher for granite aggregate concrete. E generic loss of load-bearing capacity critical temperature: Portland-cement-based concretes experience considerable creep at about 550600 3C, at which temperature the material would not be structurally useful in the hot state.

Fire and concrete structures

Fire impacts upon concrete structures insofar as it generates a heat flow into the concretes exposed surface, producing temperature, moisture and pore pressure gradients within the concrete mass. Thermal strains, stresses and cracking develop within the heated structure. Explosive spalling of the concrete can take place. Also, both the concrete and steel, as well as the bond between them, experience strength losses upon heating. The designer should, therefore, ensure that all these factors combined do not prejudice the structures primary separating and/or loadbearing functions for the duration of, and following, the fire.

Three critical temperatures for structural concrete can be identified (Fig. 2): E spalling critical surface temperature: experimental evidence indicates that the concrete surface temperatures when spalling occurs fall in the range of about 250420 3C[19,26,31], depending upon the heating rate and characteristics of the concrete.
Copyright ^ 2000 John Wiley & Sons, Ltd.

Knowledge of the development of temperature distribution in concrete structures is the first key step in the understanding of the structures behaviour in fire. Air temperatures in fires frequently exceed 900 3C. However, the good insulating properties of concrete mean that the temperature gradient is large
Prog. Struct. Engng Mater. 2000; 2:429}447

and only the temperature of the outside layer is markedly increased, while the temperature of the internal concrete remains comparatively low. This fact is illustrated by the example of a 160-mm, wide beam exposed on three sides to a standard ISO 834 (BS 476) fire[9]. The temperature after 1 h of fire exposure in the region away from the influence of the corner will be as follows: 600 3C at 16 mm depth and 300 3C at 42 mm depth (Fig. 11). The 500 3C isotherm reaches 10 mm at 30 min, 20 mm at 60 min, and 32 mm at 90 min. The temperature at a given distance from the exposed surface will be higher at corners of an element due to the transmission of heat from the two surfaces. Thus, the profile of equitemperatures of a cross-section is rounded at corners (Fig. 11).


The concept of fire resistance has been for decades at the heart of research, design and assessment of concrete structures exposed to fire. Fire resistance can be defined as the ability of an element (not a material) of building construction to fulfil its designed function for a period of time in the event of a fire. Fire resistance time of a structural element exposed to a standard furnace test is defined as the time elapsed before a fire limit state is violated. These fire limit states are given in BS 476: Part 20 as follows for separating (E and I) and load-bearing (R) functions[5]: E limit state for insulation (I): A fire on one side of a wall, or underside of a floor, acting as a compartment boundary, should not cause combustion of objects on the unexposed side. Limits of temperature rise of 140 3C (average) or 180 3C (peak) above ambient temperature are specified in the standard fire resistance test. E limit state for integrity (E): A wall or floor acting as a compartment boundary should not allow passage of smoke or flame from one compartment to another as a result of breaks or cracks in the wall or floor. Both insulation and integrity criteria also apply to members embedded in walls or floors. E limit state for load-carrying capacity (R): The members in a structural assembly should resist the applied loads in a fire. Hence, normally, each part of the structure will have a different function in fire, according to its type and position. This function could be to contain a fire (as with non-load-bearing walls), to support the design loads (as with beams and columns), or both (as with a floor). The fire design process for separating structures comprise only thermal analysis, whilst for load-bearing structures both thermal and structural analyses are required. Design for fire resistance also aims to ensure: E overall dimensions of the section of an element sufficient to keep the heat transfer through this element within acceptable limits;
Copyright ^ 2000 John Wiley & Sons, Ltd.

Fig. 11. Temperature contours in a quarter section of a concrete beam heated on three sides in ISO 834 (or BS 476) fire: (a) contours after 60 min; (b) 5003C isotherms after 30, 60 and 90 min[9]

E average concrete cover to the reinforcement sufficient to keep the temperature of the reinforcement below critical values (500 3C for reinforcing steel and 350 3C for prestressing steel) before the required fire resistance period is attained. Concrete cover in fire applications is defined as the distance between the nearest heated face of the concrete and the surface of the main reinforcement, or an average value determined in BS 8110: Part 2[15]. This cover is defined differently from the nominal cover of BS 8110: Part 1[15]. The cover thickness for a specified steel temperature limit depends upon the thermal diffusivity (k/ C where k"thermal conductivity, C"specific heat, and "density) of aggregate used. The cover has to provide lasting protection to the reinforcement from both fire and environmental attack. Choice of cover thickness should be on the basis of the more onerous requirement[15]. Reinforced or prestressed members generally fail in fire as a result of excessive temperature rise in the steel. This failure applies mainly to simply supported flexural members. Loss of concrete cover by spalling of the concrete, especially explosive spalling in the region of the tensile steel, endangers the carrying capacity because of the increased rate of heat transfer to the steel and reduction in overall thickness of concrete. Hence, a maximum cover to reinforcement is recommended to reduce the potential for spalling, but a minimum
Prog. Struct. Engng Mater. 2000; 2:429}447


thickness is required for thermal insulation. Therefore, the actual thickness should be between these two limits. The fire resistance of a whole concrete structure would not necessarily be that ascribed to its individual elements. Better fire behaviour could arise from such factors as robustness, adequate continuity of reinforcement, reduced level of loading, composite construction, and the availability of alternative paths for load support. The provision of continuity of reinforcement in the design allows the redistribution of forces and moments to gradually take place towards the parts of the structure not exposed to the fire as the exposed parts are weakened, thus ensuring an improved fire resistance relative to the situation of the simply supported single element. It is, therefore, necessary to pay particular attention to detailing. Factors affecting fire resistance of concrete elements according to BS 8110[15] are: E size and shape of elements; E disposition and properties of reinforcement or tendon; E the load supported; E the type of concrete and aggregate; E protective concrete cover provided to reinforcement or tendons; E conditions of end support. E the overall thickness of the section in order to keep heat transfer through the floor or wall within acceptable limits; E provision of surface insulation. Fire engineering calculations, or a full-scale test, allow interaction between these factors to be taken into account (see next section).


Bond failure Failure of reinforced concrete members may occur in fire when heating reduces the bond strength between the steel and concrete. Bond failure is usually combined with concrete tensile failure as the tensile strength decreases rapidly with heating. Shear`torsion failure Shear or torsion failure in fire is influenced by the concrete tensile strength and is much more complicated to determine than bending failure, because of limited experimental experience. As torsion members with a defined fire resistance are rare, the solution of the problem may be shifted to individual cases, if any. There is, however, practical interest in shear beams which are usually designed with a special shear reinforcement. Compressive failure Reinforced concrete members under compression (e.g. columns) usually fail in fire due to concrete failure in the compression zone as its strength diminishes with heating. The loss in strength of the reinforcement with heating is of minor importance in cases where the steel content is small, but not at high steel contents, where both the concrete and steel influence the load-bearing behaviour.


The engineer has at his/her disposal three methods of assessment of fire resistance (Fig. 12), namely: E fire testing; E prescriptive methods; E performance-based methods. The first two have been established for several decades. It is only now that performance-based methods based on fire engineering calculation are being accepted in an increasing number of countries. For any of the three methods of design, the detailing of the structure should be such as to implement the design assumptions for the changes during a fire in the distribution of load and the characteristic strengths of the materials. In particular, the reinforcement detailing should ensure that both elements and the structure as a whole contain adequate supports, ties, bonds and anchorages for the required fire resistance[16].


The failure of structural concrete in fire varies according to the nature of the fire (e.g. rate of temperature increase and maximum temperature); the loading system; and the type of structure exposed to the fire. Failure could occur due to:

loss of bending or tensile strength; loss of bond strength; loss of shear or torsional strength; loss of compressive strength; spalling of concrete (discussed above).

Bending`tensile failure Bending failure of load-bearing bending elements generally occurs when the reinforcement fails as the tensile strength of the steel is reduced on heating. Failure due to bending in the centre of the span of length L of simply supported load-bearing members is indicated by deflections of the order of L/30.
Copyright ^ 2000 John Wiley & Sons, Ltd.

Fire testing Fire testing of an element or subassembly when exposed to a standard temperaturetime regime (e.g. according to BS 476[5]), is the most expensive of the three options, particularly for larger, more complex structures. Any form of concrete element covered by a valid fire test report may be deemed to have the fire resistance ascribed to it by such a test, provided that the element has similar details of construction, stress
Prog. Struct. Engng Mater. 2000; 2:429}447


predetermined requirements, based on generic occupancies or classes of fire risk. Prescriptive codes are rigid and restrictive, and do not allow for engineering thinking. Although the cheapest to implement, they are the least accurate of the three methods. The safety level achieved by this method can vary significantly. In many cases it is very conservative and not cost efficient, but sometimes it can also be unsafe[4]. The fire resistance requirement in the prescriptive method is expressed by target fire resistance ratings for members exposed to BS 476 or ISO 834 fire. This means that a structural member should be designed in such a way that it does not collapse within 30, 60, 90 or may be 120 min. No classes in between are available. Hence, if a test specimen collapses after 59 min, it will be categorized in the R30 fire class, irrespective of the loading level. Such deficiencies have provided the driving force for the development and wider acceptance of performance-based methods (see below). Engineers apply prescriptive methods for the fire resistance of building elements or subassemblies from tabulated data presented in the codes and standards. BS 8110: Part 2: 1985[15] gives tables specifying minimum dimensions, including cover (in mm) for fire resistances ranging from 30 min to 4 h for concrete beams, columns and floors. The standard refers, erroneously, only to density of the concrete, hence the distinction made in the tables just between dense and lightweight (i.e. (2000 kg m!3) concretes and their strength change with temperature. The foregoing discussion on materials properties shows that the behaviour of heated dense concrete is very dependent upon the type of aggregate used. Some aggregates (e.g. Thames gravel) break up at 350 3C while others (e.g. granite and gabbro) can be thermally stable even at 600 3C. Also, the behaviour of the lightweight aggregates depends upon their density. The data also distinguishes between simply supported and continuous constructions for flexural members for both reinforced and prestressed concrete. The tables are based on the assumption that the elements are supporting the full design load. The tabulated data for simply supported elements are based on the steel reinforcement retaining a proportion of its strength at high temperatures; reinforcing bars and prestressing tendons are considered to retain about 50% of their ambient strength at 550 and 450 3C respectively. If any dimension of a particular construction is less than the minimum specified in the tables and it is not possible or desirable to increase it to meet the requirements, the fire resistance may be enhanced by the application of a protective coating, system or membrane[32]. It is necessary to appreciate that these periods do not signify the duration of an actual fire. For example a 60-min fire does not imply that a construction is expected to withstand a fire of 60 min duration, but that it will withstand a fire of a longer or a shorter duration whose severity corresponds to the 60-min
Prog. Struct. Engng Mater. 2000; 2:429}447

Fig. 12. Three main options of concrete fire assessment and design process; performance-based methods are gaining acceptance in an increasing number of countries

level, and support as the test specimen. Standard tests for fire resistance are usually conducted on single building elements where it is not feasible to reproduce in the test furnace the nature and magnitude of restraint and continuity of the adjoining construction. In some cases, therefore, the fire performance of structural elements in the building could be expected to be much greater than that of the simple element when tested in the furnace. In other cases, thermal movement can reduce the fire resistance. Another advantage of fire testing over prescriptive methods is that it provides an indication of temperature distributions within, and deflections of, the element during heating as well as detailing weaknesses not discovered without tests. However, its accuracy is sensitive to the testing apparatus and method employed, hence the considerable discussions at international committee level on harmonization of laboratory testing between different countries. The preparation for testing and performing the test is lengthy, and the cost of setting it up and execution is high. Furthermore, testing of a complete construction in fire is a formidable task, as evidenced by the full-scale fire tests undertaken on a building by the Building Research Establishment in the great airship hangar in Cardington, England.

Prescriptive methods Current building fire engineering practice is largely based on the application of prescriptive codes whereby the engineer designs in accordance with
Copyright ^ 2000 John Wiley & Sons, Ltd.


furnace test. Specific provisions of test for fire resistance of elements of structure in terms of the three performance criteria mentioned above are given in the Building Regulations[3]. The regulations also set out the minimum periods of fire resistance in minutes for elements of structure at basement, ground and upper levels of various types of building. The regulations state that where one element of structure supports or carries or gives stability to another, the fire resistance of the supporting element should be no less than the minimum period of fire resistance for the other element (whether that other element is load-bearing or not). Circumstances for varying this principle are also given[3]. Software for thermal analysis fall under two categories:


Performance-based methods Performance-based methods are based on fire engineering calculations, and provide a cost-effective and flexible method of assessment superior to prescriptive methods. A given problem can be studied for different fire scenarios, geometries, material properties, loading or support conditions. This can be performed in a relatively short period of time, thus allowing a better understanding of the behaviour of the structure subjected to fire until collapse. Moreover, computer programs can even simulate structural conditions that are very difficult to study in a fire test. In the performance-based design, the structure is not allowed to collapse during the complete fire process, including the cooling phase[4]. Performance-based methods can be classified into three categories of increasing sophistication and complexity. These are (Fig. 12):
E simplified calculations based on limit state analysis; E thermomechanical finite element analysis; E comprehensive thermohydromechanical finite element analysis.

E general programs developed by professional software houses. The most well-known are: ABAQUS, ADINA, and ANSYS in the USA, and PAFEC and LUSAS in the UK. E fire-dedicated programs designed by research workers in the field of fire. The most reputed are: FIRES-T3 which was developed in 1977 at the University of California, Berkeley, in the USA by Bresler, Iding & Nizamuddin[33]; TASEF-2, developed in 1979 by Wickstrom[34] in Sweden; and TEMPCALC, developed in 1986 by Anderberg[35] in Sweden. TEMPCALC continues to be improved and used in practice on a range of fire applications, including the simulation of the loss of surfacespalled material and the thermal behaviour of intumescent thermal coatings.

Thermal analysis For separating functions, only thermal analysis is required. For the load-bearing function, thermal analysis will need to be conducted in all three performance-based categories as part of the structural analysis. For the simplified limit state method, it would be a stand-alone finite element calculation to determine, for example, the location of the 500 3C (or other) contour. Without taking into consideration moisture migration, thermal analysis would provide approximate results, particularly in the temperature range 100200 3C when moisture migration and evaporation play a significant role. Normally this can be partly accounted for by introducing a latent heat of evaporation component to the specific heat capacity, but this still does not entirely solve the problem. This component may not be significant in terms of load-bearing capacity assessment, but in other examples (e.g. spalling or in nuclear reactor concrete) it is important.
Copyright ^ 2000 John Wiley & Sons, Ltd.

Simplified limit state analysis Having determined the temperature distribution for the structural element, either from the published literature on a similar element or from thermal analysis, a simplified limit state analysis is carried out, using a technique first proposed by Anderberg[36]. He suggests a very simple method of analysis, based on the hypothesis that the thickness of the damaged siliceous concrete is assumed to equal the average depth of the 500 3C isotherm in the compression zone of the cross-section. Damaged concrete (i.e. concrete with temperatures in excess of 500 3C) is not expected to contribute to the load-carrying capacity of the member, whilst the residual concrete cross-section is assumed to retain its full initial values of strength and modulus of elasticity. Anderberg suggests that this method, the 500 3C isotherm, or Effective crosssection method, is applicable to a reinforced concrete section with respect to axial load, bending moment and their combinations. For this method to apply, there should be minimum dimensions of the member, depending upon fire resistance time or fire load density[37]. One note of caution is that the choice of the temperature isotherm will depend upon the type of concrete used and its characteristic strength loss against temperature. For certain concretes, the isotherm temperature may well be below 500 3C, or even below 400 3C. Thermomechanical finite element analysis The bulk of performance-based, fire-dedicated software dating back to the 1970s falls into this category. The thermal and mechanical analysis are normally interfaced and not integrated. In other words, the thermal calculation is carried out first for the entire duration of the fire and then fed into the mechanical analysis program to produce the stresses and strains for the member/structure there being no interaction between the two analyses and moisture
Prog. Struct. Engng Mater. 2000; 2:429}447

effects are also absent. Nevertheless, validation of the results from such programs has consistently shown that the deformations of simple elements can be reasonably accurately predicted by such methods, providing transient creep (or LITS) is incorporated into the model, particularly for columns. The modelling of LITS is relatively simple in fire-dedicated programs where the duration of the analysis is in minutes or a few hours. This is because LITS is largely a function of temperature rather than time. Therefore, time-dependent creep functions can be dispensed with, without prejudicing the outcome. The absence of a moisture migration analysis means that the evaporation plateau and explosive spalling cannot be predicted. The first such fire-dedicated concrete program was initially developed in 1974 at the University of California and was called FIRES-RC by Becker & Bresler[38]. That program was improved later by Anderberg[39] who introduced a more realistic concrete behavioural model. This program was followed by CONFIRE, developed in 1982 by Forsen[40] in Norway (valid so far only for siliceous concrete); STABA-F, developed at the Technical University of Braunschweig[41] in 1985; CEFICOSS developed in 1987 by Franssen[42] of the University of Liege; STRUCT, developed by the author and his PhD student at Imperial College in 1991[43]; and the most recent FIREXPO, developed in 1999 by the contractor Bouygues, with input from the author, as part of the HITECO research programme funded by the European Commission[28].


Fig. 13. Simplified flow chart of thermohydromechanical finite element analysis of heated concrete structures including spalling; most fire-dedicated finite element software are thermomechanical and exclude the hydral component. The thermal and mechanical analyses are usually interfaced and not integrated, hence the inputs and outputs are shown separately. With a fully integrated and interactive model, the input data are not separated because they all contribute to the overall analysis

Comprehensive thermohydromechanical finite element analysis A comprehensive analysis would incorporate thermal, hydral and mechanical analyses in a fully integrated and interactive model (Fig. 13), capable of predicting explosive spalling (e.g. in tunnels) or the moisture state of the concrete containment of nuclear reactors (an important function as a biological radiation shield). The first such model was developed in the 1970s by Bazant & Thonguthai[27], albeit with limitations in the modelling. For example, the hydral component is considered to be a single-phase smeared fluid and spalling is not incorporated. A more advanced model was developed in 1999 at Padua University in co-operation with the author and ENEA in Rome as part of a multinational programme of research funded by the European Commission[28]. Called HITECOSP (high-temperature concrete spalling), it is a fully coupled nonlinear model, designed to predict the behaviour, and potential for spalling, of heated concrete structures for fire and nuclear reactor applications. Concrete is considered as a multiphase material consisting of a solid phase, two gas phases and three water phases. Refinements (such as the effect of damage on permeability and nonlinearities due to temperature)
Copyright ^ 2000 John Wiley & Sons, Ltd.

are included and chemicalphysical phase changes are incorporated, such as hydrationdehydration, evaporationcondensation and adsorptiondesorption. Phase changes in concrete are, for the first time, incorporated directly in the transport mechanism of any concrete model. Melting and evaporation of pp fibres are modelled as phase changes. The complete behaviour of concrete in the elastic, inelastic and plastic ranges up to fracture is described. Previous models were based on elastic fracture only. The physical model is described with emphasis being placed upon the real processes occurring in concrete during heating, based on tests carried out in several major laboratories around Europe as part of the wider HITECO research programme. A number of experimental and modelling advances are presented in this work. The stressstrain behaviour of concrete in direct tension (determined experimentally for the first time) is input into the model. The hitherto unknown microstructural, hygral and mechanical behaviour of HPC/UHPC were determined experimentally and the information is also built into the model. It is also the first time that such a complex model has been developed with the aim of predicting the behaviour of concrete at high temperatures in general and explosive spalling in particular. The fluid phase is considered to consist of
Prog. Struct. Engng Mater. 2000; 2:429}447


water, dry air and vapour. The water phase consists of free capillary and physically bound water. Chemically bound water is considered part of the solid skeleton until it is released on heating. The solid is fully deformable and able to experience elastic, damage, thermal, creep, shrinkage, plastic, cracking strain processes. Both basic and, importantly, transient creep phenomena are included. Outputs are presented in the form of three-dimensional colour diagrams of temperature, vapour pressure, water saturation and damage distributions. An example output for a tunnel lining segment subjected to fire is shown in Fig. 14.


BRE. The review presented above is derived mainly from fire concrete research related to buildings and has, therefore, been adequately covered within the scope of this paper.

There have been at least nine major fires in European tunnels since 1990[4652]. Two road-tunnel fires in 1999 (Mont Blanc, France/Italy and Tauern, Austria) claimed 51 lives, and in November 2000 a funicularrailway-tunnel fire in Kaprun, near Salzburg, Austria claimed 159 lives. Two earlier fires in the Great Belt Tunnel (Denmark, 1994) and the Channel Tunnel (UK/France, 1996) did not claim any lives, but resulted in significant damage and spalling to the tunnel segments, made of high-performance concrete (Table 3). The maximum temperatures reached in these fires ranged from 800 3C (Great Belt) to 1100 3C (Channel), although the Dutch report even higher temperatures of 1350 3C in their tunnel fires[10]. The duration of the fires (9 h for the Channel Tunnel) were significantly longer than those encountered in buildings (normally up to 2 h). This is at least partly due to the difficulties the fire services encounter in


Engineers and scientists have devoted attention to fire in buildings for decades, and there is an abundance of books[44,45] and other literature on the subject. The website of the Building Research Establishment ( currently has 198 publications related primarily to fire in buildings, most of them available from The Stationery Office or directly from

Fig. 14. Profiles of temperature (shown by depth of shading) and gas pressure (shown by undulations) in an HPC concrete beam 20 min into a fire[28]

Table 3 Concrete damage in recent tunnel fires Tunnel Maximum Fire Length affected Segment depth temp. duration (h) affected (3C) ............................................................................................................................................... 76 MPa, 28 day 800 7 16 segment rings (1.65 m long) damaged in crown 500 m with 50 m severely affected by spalling Up to 68% spalled in layers along 10 segments Up to 100% of segment thickness spalled showing grout Concrete strength

Great Belt (1994) [50] Channel (1996) [47,48]

110 MPa, mature


Mont Blanc Not reported 1000 50 900 m; tunnel crown Not reported (1999) [49] most affected ............................................................................................................................................................................................................................................................
Copyright ^ 2000 John Wiley & Sons, Ltd.

Prog. Struct. Engng Mater. 2000; 2:429}447

reaching and extinguishing tunnel fires. Also, the closeness of the tunnel roof above the fire means that the thermal plume has less height in which it can entrain cooler air, thus resulting in higher temperatures than in buildings. The form of the tunnel also affects its ability to resist fire. Rectangular tunnels of the cut-and-cover type, formed of reinforced concrete, can be categorized as structures whose principal loading condition is controlled by bending considerations. Thus, a fire in such a structure only has to cause spalling in the soffit and loss of intrados rebar to seriously undermine the stability of the structure. Circular tunnels behave differently from rectangular ones as their principal loading condition is hoop compression. In fire, the hoop loads increase, owing to restrained expansion near the heated surface. This loading could contribute to increased risk of concrete spalling. The damage to the concrete extended for considerable lengths (500 m in the Channel Tunnel and 900 m in the Mont Blanc tunnel) owing to the fact that the fire spread from one goods vehicle to another. The maximum depth of concrete damage to the segment linings reached 68% of the thickness (i.e. 270 mm out of 400 mm) in the Great Belt tunnel to 100% of the thickness in the Channel Tunnel, caused by multiple spalling of the high-performance concrete, particularly in the lightly reinforced regions[46]. Two factors limited the spread of multiple spalling: the presence of reinforcement, and evaporation of moisture from the heated concrete. Had the Channel Tunnel not been bored (deliberately) in the impermeable chalk layer, the loss of the concrete lining would have resulted in flooding of the tunnel, with a much more serious outcome. Firemen reported that, on entering the fire zone in the Channel Tunnel, they were showered with small, hot pieces of concrete in an environment that resembled hell. The cost of repair after the Channel Tunnel fire was @45 million[47]. Damage to the concrete segments would have been largely avoided had it been protected by a layer of thermal insulation. Nowadays, designers have a second weapon in their armoury to protect concrete from spalling, and that it the use of pp fibres in the mix. Despite the problems outlined above, even the latest versions of design guides for tunnels do not adequately cover this issue. For example, there is no mention of the need for adequate fire protection against explosive spalling in the 1999 version of the Design Manual[53]. The manual emphasizes when protection is not to be used, rather than recommends when protection should be used. The two prime recommendations of concrete cover and the use of additional mesh reinforcement are not fully adequate measures to protect concrete in the event of a medium large fire. The statement in the manual that concrete is inherently fire resisting is negated by the phenomenon of explosive spalling.
Copyright ^ 2000 John Wiley & Sons, Ltd.

The optimum passive fire protection design for tunnels could involve the use of both pp fibres and a thermal barrier. Without the thermal barrier, only the problem of explosive spalling would be addressed by the fibres, and temperature levels within the concrete would not be reduced. Thermal barriers are effective in reducing: (a) the risk of explosive spalling; (b) deterioration of concrete mechanical properties; and (c) steel temperatures. In Switzerland, a concrete surface temperature limit is set at 250 3C by the use of protective panels[46]. However, thermal barriers are dearer than pp fibres, and a cost-effective design for a new tunnel could, therefore, optimize on the use of the thermal barrier to the areas that need most protection (identified from knowledge of the fire scenario, and from thermal calculations) while the concrete would incorporate the fibres. The use of fibres is, however, not possible in existing tunnels, and thermal barriers are the only effective option for the provision of passive fire protection, if deemed necessary. A cost, benefit and risk analysis would identify the optimum passive fire protection design. Some tunnels are so lightly loaded, and employ such relatively low-strength concrete, that designers may feel that an increased concrete cover combined with the use of the fibres would be an adequate protection against fire.

Assessment after fire

An immediate and thorough appraisal is normally required after a fire. Such an appraisal should begin as soon as the building can be entered, and generally before the removal of debris. After a fire, an estimate is made of the severity of temperature exposure in terms of an equivalent standard test. A visual examination and classification of damage for each structural member is carried out. The maximum concrete temperature profile during a fire can be estimated from results of previous tests, from computer simulations, and from post-fire assessment of the concrete (e.g. through its colour change or by a thermoluminescent technique)[54]. Key diagrams and schedules are then prepared. Following this, a general assessment of the likely repairs required may be drawn up. Normally, concrete exposed to temperatures above 300 3C is replaced if possible. Otherwise the dimensions are increased (e.g. reinforced columns), depending upon the design load. The fire resistance of a concrete structure is frequently well above minimum requirements. Because of the structural continuity present in most buildings, there are reserves of strength which may enable the structure to survive fires and be reinstated. Reinstatement by repair will usually be economically preferable to demolition and rebuilding in terms of capital expenditure and earlier reoccupation.
Prog. Struct. Engng Mater. 2000; 2:429}447

The behaviour of concrete material in fire depends very much on the specific concrete mix proportions and constituents used, and is determined by complex physicochemical transformations during heating. However, all Portland-cement-based concretes lose their load-bearing capacity at temperatures above 550600 3C. At lower temperatures (i.e. that of the bulk of the concrete member during fire) the deterioration in mechanical properties during heating can be reduced by judicious concrete mix design whereby thermally stable aggregates of low thermal expansion are employed, and cement blends are selected that produce a low CaO/SiO2 ratio, but not too low a permeability (also important to reduce the risk of spalling). A key property unique to concrete amongst structural materials is the load-induced thermal strain (LITS), also called transient creep. Any structural analysis of heated concrete that ignores LITS will be wholly inappropriate and will yield erroneous results, particularly for columns exposed to fire. Normalstrength concretes and high-performance concretes microstructurally follow similar trends when heated, but ultra-high-performance concrete behaves differently.


Failure of structural concrete in fire varies according to the nature of the fire, the loading system and the type of structure exposed to the fire. Failure could occur due to loss of bending or tensile strength; loss of bond strength; loss of shear or torsional strength; loss of compressive strength; and spalling of the concrete. The structural element should, therefore, be designed to fulfil its separating and/or load-bearing function without failure for the required period of time in a given fire scenario. Design for fire resistance aims to ensure overall dimensions of the section of an element sufficient to keep the heat transfer through this element within acceptable limits, and an average concrete cover to the reinforcement sufficient to keep the temperature of the reinforcement below critical values long enough for the required fire resistance period to be attained.


The engineer has at his/her disposal three methods of assessment of fire resistance: (a) fire testing; (b) prescriptive methods; and (c) performance-based methods. The first two have been established for several decades. Prescriptive methods are rigid and restrictive, and do not allow for engineering thinking, nor can they be applied to whole structures. Although the cheapest to implement of the three methods, they are the least accurate. The safety level achieved can vary significantly. Such deficiencies have provided the driving force for the development and wider acceptance of performance-based methods. Performance-based methods employ fire engineering calculations and provide a cost-effective and flexible method of assessment of fire resistance that is superior to prescriptive methods. Performance-based methods can be classified into three categories of increasing sophistication and complexity: (a) simplified calculations based on limit state analysis; (b) thermomechanical finite element analysis; and (c) comprehensive thermohydromechanical finite element analysis, also capable of predicting moisture migration, pore pressures and explosive spalling. It is only now that performance-based methods, first introduced in the UK, are being accepted in an increasing number of countries.

The risk of explosive spalling in fire is significantly reduced in high-permeability concrete. In lowpermeability concrete, spalling could be eliminated by the appropriate inclusion of polypropylene fibres in the mix (this requiring further research) and/or by protecting the exposed concrete surface with a thermal barrier. Until now, the prediction of spalling during heating has been largely an imprecise empirical exercise. Attempts to predict spalling by analytical methods have failed, owing to the complex microstructure and multiphase nature of heated concrete. The inability to predict the occurrence of spalling has been a limiting factor in the development of robust models capable of predicting the response of concrete structures to fire. This prediction is now becoming possible with the development of thermohydromechanical, nonlinear finite element models capable of predicting pore pressures and hence spalling in heated concrete structures. Experts differ as to the mechanisms responsible for explosive spalling. The balance of evidence suggests that it is the combination of pore pressure spalling and thermal stress spalling, with pore pressure spalling being the dominant mechanism in both normal-strength and high-performance-concretes. However, thermal stress spalling may assume greater importance in ultra-highperformance concretes containing a high proportion of expansive silica. Further research is required to settle this issue.
Copyright ^ 2000 John Wiley & Sons, Ltd.

A spate of major tunnel fires in the past few years has resulted in significant spalling of the concrete lining, up to 100% of its thickness. Based on experience in the Netherlands, the Dutch authorities proposed a fire scenario, RWS (more severe than the normal hydrocarbon fire) with temperatures rapidly rising and peaking at 1350 3C (melting temperature of concrete) to simulate fire in tankers carrying petrol. However, the maximum temperatures attained in recent major fires in the UK, France, Denmark and
Prog. Struct. Engng Mater. 2000; 2:429}447

Austrian tunnels did not reach RWS levels, but the durations of those fires far exceeded anything in the standards. Guidelines for tunnel design urgently require updating to incorporate the latest developments in fire protection engineering, particularly against spalling.

[21] Ichikawa Y. Prediction of pore pressures, heat and moisture transfer leading to spalling of concrete during fire. PhD Thesis. University of London. 2000. [22] Saito H. Explosive spalling of prestressed concrete in fire. Occasional Report No. 22. Japan: Building Research Institute. 1965. [23] Dougill JW. Modes of failure of concrete panels exposed to high temperatures. Magazine of Concrete Research 1972; 24(79): 71d76. [24] Zhukov VV. Explosive failure of concrete during a fire. Translation No. DT 2124. Borehamwood, England: 1975. Joint Fire Research Organisation. [25] Sertmehmetoglu Y. On a mechanism of spalling of concrete under fire conditions. PhD Thesis. Kings College, London. 1977. [26] Connolly RJ. The spalling of concrete in fires. PhD Thesis. Aston University. 1995. [27] Bazant Z & Thonguthai W. Pore pressure in heated concrete walls d theoretical prediction. Magazine of Concrete Research 1979: 31(107): 67d75. [28] Khoury GA, Majorana C, Kalifa P, Aarup B, Giannuzzi M & Corsi F. The HITECO BRITE project on the effect of heat on high performance concrete. Proceedings of the International Workshop on Durable Concrete Structures, Weimar, Germany, October 1998. [29] Bremer F. Multi-layer (double-wall) prestressed concrete pressure vessel. Nuclear Engineering and Design 1967: 5: 183d190. [30] Thelanderson S. Mechanical behaviour of heated concrete under torsional loading at transient high temperature conditions. Bulletin No. 46, Institute of Technology. Lund, Sweden: 1974. [31] Jumpannen U-M. Effect of strength on fire behaviour of concrete. Nordic Concrete Research Publication No. 8. 1989. [32] Morris WA, Read REH & Cooke GME. Guidelines for the construction of fire-resisting structural elements. Building Research Establishment Report BR 128. 1988. [33] Iding R, Bresler B & Nizamuddin Z. FIRES-T3, A computer program for the fire response of structures-thermal. Report No. UCB FRG 77-15 Berkeley: University of California. 1977. [34] Wickstrom U. TASEF-2, A computer program for temperature analysis of structures exposed to Fire. Lund, Sweden: Institute of Technology. 1979. [35] Anderberg Y. TEMPCALC User Manual. Lund, Sweden: Institute of Fire Safety Design (IFSD). 1986. [36] Anderberg Y. Analytical fire engineering design of reinforced concrete structures based on real fire characteristics. Proceedings of the FIP Congress, London 1978. Part 1: 112d121. [37] Anderberg Y. Simplified cross-section calculation methods to be used for bending moments and axial forces. Annex B CEN/TC92/SC2/PT1-2 Doc 58 Lund, Sweden: Institute of Technology. 2000. [38] Becker J & Bresler B. FIRES-RC, A computer program for the fire response of structures-reinforced concrete frames. Report No. UCB FRG 74-3 Berkeley: University of California. 1974. [39] Anderberg Y. Fire-exposed hyperstatic concrete structures. An experimental and theoretical study. Bulletin 55 Lund, Sweden: Institute of Technology. 1976. [40] Forsen NE. A theoretical study on the fire resistance of concrete structures. SINTEF Report No: STF65 A82062, Cement and Concrete Research Institute, Norwegian Institute of Technology. 1982. [41] Rudolph K, Richter E, Hass R & Quast U. Principles for calculation of load-bearing and deformation behaviour of composite structural elements under fire action. Proceedings of 1st International Symposium on Fire Safety Science, 1985. [42] Franssen JM. A study of the behaviour of composite steel-concrete structures in fire. PhD. Thesis, Liege University, Belgium. 1987. (in French). [43] Terro M. Computer modelling of the effect of fire on structures. PhD Thesis, University London. 1991. [44] Anchor RD, Malhotra HL & Purkiss JA (eds). Design of structures against fire. UK: Elsevier. 1986, p. 251. [45] Shields TJ & Silcock GW. Buildings and Fire. London: Longman. 1987. [46] Lance GA. Effect of fire on tunnel lining stability. Tunnels and Tunnelling International 1998: October, pp. 39}41. [47] Coates J. Chunnel: They have been told. Construction News No. 6494, 5 December 1996: 1. [48] Arches brace Chunnel for six month repair schedule. New Civil Engineer, 28 Nov 1996: 3d4. [49] French Minister of the Interior, Equipment, Transportation and Housing. Task force for technical investigation of the 24 March 1999 fire in the Mont Blanc vehicular tunnel, Report of 30 June 1999. [50] Tait JC & Hoj NP. Storebaelt eastern railway tunnel: Dania tunnel boring machine fire d analysis and recovery. Proceedings of the Institution of Civil Engineers 1996: Supplement, 114 Special Issue 1: pp. 40d48.

The author wishes to express his appreciation to Dr Yngve Anderberg, of Fire Safety Design, for kindly reading the manuscript, and for his useful comments.

References and recommended reading

[1] Lea F & Stradling R. The resistance to fire of concrete and reinforced concrete. Engineering. 1922: 114(2959). [2] ACI Seminar. Concrete for nuclear reactors, Berlin 1970. Special Publication SP34, Farmington Hills, MI: American Concrete Institute. [3] Building Regulations. Approved Document B 3: Internal fire spread (structure). The Building Regulations Approved Document B, Fire Safety, 1992. [4] Anderberg Y. Fire engineering design of structures based on design guides. 2nd International Conference on Performance-Based Codes and Fire Safety Design Methods, Maui, USA, 7d9 May, 1998. [5] British Standards Institution. BS 476, Fire tests on building materials and structures. London: BSI. 1987. [6] ISO 834. Fire resistance test d elements of building construction. International Standard 834, 1975-11-01. [7] Eurocode 1: Basis of design and actions on structures, Part 2.2: Actions on structures exposed to fire. EN 1991. Brussels, Belgium: European Committee for standardization. [8] Eurocode 2: Design of concrete structures, Part 1: General rules d Structural fire design. EN 1992. This is currently being revised, see [9]. [9] Anderberg Y. Temperature Profiles. Annex A CEN/TC92/SC2/PT1-2 Doc 58, Lund Sweden: Institute of Technology. 2000. [10] Varley N & Both C. Fire protection of concrete linings in tunnels. Concrete May 1999: 27d30. [11] Khoury GA, Sullivan PJE & Grainger BN. Strain of concrete during first heating to 600 3C under load. Magazine of Concrete Research 1985: 37(133): 195d215. [12] Khoury GA & Algar S. Advanced mechanical characterisation of HPC and UHPC concretes at high temperatures. Proceedings of the International Workshop on Durable Concrete Structures, Weimar, October 1998. [13] Khoury GA, Sullivan PJE & Dias WPS. Deformation of concrete and cement paste loaded at constant temperatures from 140 3C to 720 3C. Materials and Structures 1986: 19(110): 97d104. [14] Khoury GA. Compressive strength of concrete at high temperatures: a reassessment. Magazine of Concrete Research 1992: 44(161): 291d309. [15] British Standards Institution. BS 8110: Part 1:1985, Structural use of concrete. Code of practice for design and construction. Part 2:1985, Structural use of concrete. Code of practice for special circumstances. Section four, Fire resistance. London: BSI. 1985. [16] Khoury GA, Algar S, Felicetti R & Gambarova P. Mechanical behaviour of HPC and UHPC concretes at high temperatures in compression and tension. ACI International Conference on State-of-the-art in High Performance Concrete, Chicago, March 1999. [17] Shorter GW & Harmathy TZ. Moisture clog spalling. Proceedings of lnstitution of Civil Engineers 1965: 20: 75d90. [18] Meyer-Ottens C. The question of spalling of concrete structural elements of standard concrete under fire loading. PhD Thesis. Technical University of Braunschweig, Germany. 1972. [19] Akhtarruzaman AA & Sullivan PJ. Explosive spalling of concrete exposed to high temperature. Concrete Structures and Technology Research Report. London: Imperial College. 1970. [20] Khoylou N. Modelling of moisture migration and spalling behaviour in nonuniformly heated concrete. PhD Thesis. University of London. 1997.

Copyright ^ 2000 John Wiley & Sons, Ltd.

Prog. Struct. Engng Mater. 2000; 2:429}447


[51] PIARC Committee on Road Tunnels (C5). Fire and smoke control in road tunnels. PIARC World Road Association, France, 1999. (ISBN 2-84060-064-1). [52] Proceedings of the International conference on fires in tunnels, Boras, Sweden, October 1994, Swedish National Testing and Research Institute Fire Technology, SP Report 1994: 54.


[53] The Highways Agency (and other UK organizations). Design Manual for Roads and Bridges: Design of Road Tunnels. Document BD 78/99, London: The Stationery Office. 1999. [54] Placido F. Thermoluminescence test for fire-damaged concrete. Magazine of Concrete Research 1980: 32(11): 112d116.

Gabriel Alexander Khoury Bsc MSc PhD DIC Eurlng CEng MIStructE MINucE MIFE MRAeS Civil Engineering Department, Imperial College, London SW7 2BU, UK E-mail:

Copyright ^ 2000 John Wiley & Sons, Ltd.

Prog. Struct. Engng Mater. 2000; 2:429}447