Skip to main content

Recent Lessons Learned in Structural Fire Engineering for Composite Steel Structures


The knowledge in the field of structural fire engineering has been greatly advanced through assessment of a number of real fires (WTC, Torre Windsor, Broadgate, etc.) and, especially, by the Cardington series of full scale structural fire tests. This knowledge has been used to validate and verify the use of computational finite element models that have expanded the range of structures that can be investigated under severe fire exposure. This paper presents a selection of key lessons learned by the authors through the assessment of structures in fire for real commercial building projects. The key areas of sensitivity that have been encountered are described and a discussion of each point presented. The paper is aimed at describing potential weaknesses that have been observed in the commercial work of the authors, often driven by the requirements for efficient ambient structural design. The paper concludes with some suggested advice for structural engineers aimed at increasing the general robustness of building structures. This is based on designing out as far as possible in the ambient design of a structure the potential weaknesses identified in past project work.


Recent decades have significantly increased the fund of knowledge available in the field of structural fire engineering and a number of buildings have been designed to withstand credible design fires based on an understanding of the performance of the structure in fire.

Starting with the Broadgate Phase 8 fire in 1990 [1] and through the extensive testing completed on the Cardington test frame [2] it has become apparent that composite steel framed buildings generally perform well under severe fire loading. However as demonstrated in the collapse of the World Trade Center (WTC) buildings, in particular buildings 1, 2 [3] and 7 [4], the full range of building response to severe fires are not yet known.

This paper presents a number of lessons that have been learned through the assessment of a variety of steel framed structures with steel–concrete composite floors under fire conditions over the past 10 years. All these lessons have been learned on real building designs through the commercial application of analysis and structural fire engineering in the UK.

The lessons relate to structural arrangements/details that are commonly observed in modern building design. The structural fire modelling that has been conducted on these commercial projects has provided important insights into the response of these common details in a fire event, i.e. whether they are weak, strong, ductile or inductile. The main issue that has been highlighted relates to structural layout. This includes the issues of expansion effects on the slab and floor framing of beams, columns offset from primary gridlines and primary members with secondary beams framing in on both sides. Additional areas of interest have been highlighted with regards to local response of cellular beams, sources of restraint and issues around the effects of fire on the response of steel to steel connections.

All of these issues are relevant to the structural designs of the WTC buildings and many other steel frame buildings around the world. They provide an insight into possible future “best practice” guidance for the design of structures under severe fire loading.

The authors have investigated the structural performance of a number of building designs in fire as part of the structural design process. Detailed finite element models have been produced (published examples in [5, 6]) using contemporary research and include the most important input parameters, such as material models, connection and shear stud properties and design fires. Most importantly the models were designed to allow the investigation of the effects of restrained thermal expansion in the context of real, whole building design. These were commercial studies considering the design of steel framed buildings with steel–concrete composite floors in the UK with the aim of justifying engineered structural fire protection strategies and checking the stability of the structure to credible fires. They also provided great insight into the response of the range of typical structural layouts to severe heating.

The research conducted following the Cardington tests [7, 8] demonstrated that the response of a structure to severe heating was related to:

  1. 1.

    Axial thermal expansion—members lengthening as they heat

  2. 2.

    Thermal bowing—downward deflections caused by differential heating, and therefore differential expansion, through the depth of slabs and beams and between beams and slabs when connected together compositely

  3. 3.

    Restraint—the inherent stiffness of parts or the whole of a structure that resist the horizontal forces generated through thermal expansion. Higher restraint typically creating additional vertical deflections in beams while less stiff structures typically exhibit greater horizontal displacement of columns

The conclusions of this paper are based on extensive numerical modelling using the ABAQUS finite element modelling software. This software has been validated extensively against large scale fire tests, including the Cardington large frame tests. However it should be noted that, as with all numerical modelling work, certain assumptions and simplifications have been made to allow the models to function efficiently.

The models referred to in this paper are considered to provide an effective representation of global structural behaviour and many local responses will be accurately captured, for instance yielding and plastic hinge formation in beams and columns. However some models will not accurately depict responses that are based on very localised mechanisms, i.e. rebar rupture or bolt and plate failures in connections. It is the responsibility of the investigating engineer using the models to consider such issues using engineering judgement based on detailed research to determine if the structure under investigation is sensitive to such detailed issues. For instance, additional sub-models may be needed to check connection response if high forces or large deformations are predicted.

Lessons Learned

Over the course of these commercial design projects a number of responses have been discovered that are common between buildings where similar structural arrangements are used. The following subsections describe the issues encountered and present a discussion on how they impact on building design.

Lessons in Structural Layout

A key aim of many modern structural designs is to optimise the design to obtain the minimum steel weight for a given building geometry. This can result in unique beam layouts that are “unbalanced”, in that adjacent bays have beams running in different directions leading to primary beams supporting secondary beams on one side only. Additionally such optimised structures often have beams that are offset from column gridlines. These issues are especially evident when the building is highly irregular in its architectural shape as indicated in Figure 1.

Figure 1

Example structures where offset columns and unbalanced structural bays (clouded) are observed. Offset beams are circled. (a) WTC7 structure (Image from NIST Report [4]). (b) Recent commercial structural design assessment

Key Influences on Structural Form

The key load cases considered in structural design are the Ultimate Limit State (ULS) and the Serviceability Limit State (SLS) and are primarily interested in gravity and wind loading. There are a number of “accidental” load cases that are also assessed for ambient design that may include cases such as earthquake and snow loading. The Fire Limit State (FLS) is the equivalent accidental load case associated with the investigation of structural response when exposed to severe compartment fires, however as yet there is no requirement to specifically consider the FLS during design.

Standard structural optimisation software is currently designed for covering the cold/ambient load cases, such as gravity, wind, snow, etc. As the FLS is not yet accepted as a standard load case, extreme optimisation for ambient load cases can lead to a structure that has inherent weaknesses under severe fire exposure, even if the structure is fully fire protected with fire protection meeting the recommendations of modern design codes (i.e. “code compliant” protection). One of the reasons that the FLS has not become a standard assessment for structural engineers is that historically the fire protection of a structure has been the responsibility of the Project Architect rather than the Structural Engineer. This has led to the prevalent view that structural fire protection is something that only needs to be considered at the end of the structural design as a separate exercise. The authors suggest that the assessment of the fire protection needs for a building should happen early in the design process and should be led by the Fire Engineer and the Structural Engineer with appropriate input from fire protection manufacturers.

Two examples of this issue are shown in Figure 1. The first is from the WTC7 structure [4], while the second is from a recent structural analysis conducted as part of a commercial project. Figure 1 demonstrates both unbalanced structure and primary beams offset from columns in the example structural layouts.

The WTC7 building structure is a key example of an “unbalanced” layout. The bays on the whole of the east side of the building have beams running at 90° to the bays directly adjacent to the west. The interface between different beam directions is highlighted by the cloud in Figure 1a. The example structure in Figure 1b also demonstrates similarly unbalanced structure in the bays to the east and south of the structure.

Figure 2 shows an example of a well balanced structure.

Figure 2

Example balanced structure with some offset beams (examples circled)

Offset beams are evident in both examples in Figures 1 and 2 where secondary beams do not frame directly into columns.

Offset Beams and Columns

When a restrained steel member expands it rapidly reaches very large forces. Lamont [7] and the University of Edinburgh [8] demonstrated that a temperature rise of only 100°C to 200°C is enough to generate forces within a beam sufficient to cause yield and local buckling toward its connections. These structural temperatures will quickly be reached in a severe compartment fire, even in protected steel (beams are protected to prevent temperature rise exceeding 620°C rather than to maintain ambient temperatures in the structure). The forces generated are typically in the Meganewton range and therefore will lead to significant deformations in the structure. This is particularly true if the floor structure is unbalanced (c.f. Sect. 2.1).

Offset beams and unbalanced bays lead to 3 key issues when considering a structure subject to a severe compartment fire and associated thermal expansion/thermal bowing. The first is high lateral forces in connections between primary beams and columns. This is caused by the thermal expansion of the secondary beams. The short length of primary beam between the secondary beam and the column connection is relatively stiff and therefore the forces (in the order of magnitude of Meganewtons) generated by the expansion of the secondary beam is transferred directly into the column connection as lateral shear as indicated in Figure 3a. This type of shear may not have been considered in the ambient design of the member.

Figure 3

Effects of expansion in offset structure (plan view). (a) Direction of expansion forces. (b) Geometrical effects of expansion

The second key issue is high minor axis shear and bending moments at the end of the beam. As well as transferring lateral shear into the column connection the high forces generated in the secondary beam will affect the load carrying capacity of the primary beam near the column. The deformation likely to result from this lateral loading is indicated in Figure 3b.

Finally torsional moments can be transferred into the column that may not have been accounted for as part of the ambient design. Lateral movements in the primary beam combined with the offset of the column connection can generate large rotational moments around the column axis as indicated in Figure 3b.

Each issue associated with offset beams and unbalanced bays reduces the overall robustness of a structure. It is therefore recommended that structures be reviewed during design to minimise as far as possible such features. If it is impractical to remove offsets or unbalanced bays then the structural design should account for their inclusion by addressing the robustness of the beams, columns and connections locally. This could be in the form of beams with higher lateral shear capacities or greater ductility in the connections. The design of these elements should be informed by non-linear analysis of the structure in fire.

Effects on Beams and Columns

The likely deformations, indicated in Figure 3b, will lead to the following secondary effects. Firstly P-Delta moments in the primary beam can be increased as it expands against the induced out of plane deformation. This moment is induced when an axial force is moved out of the main structural frame line as indicated in Figure 4. The name comes from the typical designation of axial loads as P and the lever arm being acted over is the displacement, typically denoted with the Greek symbol δ (delta). P-δ moments are common in columns but can occur in any members.

Figure 4

P-Delta moment definition. (a) Typical column load resistance. (b) P-Delta moment

Additionally high lateral shear is observed in both the primary beam and the connection to the column which can also lead to axial torsion in the column.

All of these actions will increase the stress in the primary load bearing structure. Depending on the level of ambient design optimisation that has been implemented this additional loading may be sufficient to cause failure in the primary structure, especially as the steel strength begins to reduce above 400°C and stiffness reduces above 200°C.

It should also be noted that relying on code guidance with respect to thicknesses of fire protection may not be sufficient. Structural fire protection does not prevent members from heating up. Even in a “code compliant” structural fire protection layout, members may reach temperatures in excess of 600°C leading to large thermally induced deformations. It is therefore considered to be particularly important to consider the FLS in a highly optimised structure, even where full “code compliant” protection is proposed by the Architect or Fire Engineer.

Another secondary effect that arises from an unbalanced structural layout is the rotation of primary beams. When a primary beam has secondary beams spanning onto it from both sides the thermally induced thrust is roughly equal on both sides preventing significant lateral movement of the primary beam bottom flange. When a primary beam supports secondaries on one side only, the thermally induced thrust in the secondary beams acts against the restraint of the composite connection with the slab to push the bottom flange of the primary beam sideways. This effect on the primary beam is represented diagrammatically in Figure 5.

Figure 5

Rotation of primary beam (Note—P is the axial force in secondary beams created by thermal expansion of steel). (a) Balanced structure—minimal rotation. (b) Unbalanced structure—larger rotations

This is similar to the response observed in edge beams but may be of more concern when applied to internal primary beams due to the greater importance of these members compared to edge beams. As the specific optimisation being performed is unique to each building it is only possible to ensure that this response is not detrimental to the stability of the structure through a detailed analysis at the Fire Limit State.

Effects on the Slab

The series of tests on the Cardington test frame [2] at the Fire Research Station in the 1990s demonstrated that traditionally designed composite steel frame buildings had significant reserves of strength in the fire limit state. Research [7] into the available load carrying mechanisms indicated that at high temperature the floor loads in a building could be carried by the slab through tensile membrane action. In particular it was demonstrated that intermediate secondary beams could be left unprotected without compromising the overall stability of the structure. The large slab deflections generated in a severe compartment fire allow the floor to change from carrying loads in bending/shear to carrying them in tensile membrane action.

When considering the ability of the floor slab to carry load in tensile membrane action the most advantageous floor structure design consists of rectangular bays with columns at each corner. Indeed the simplified design guidance available for taking into account the benefits of tensile membrane action [9, 10] requires square or rectangular bays to function. If this layout is deviated from it can lead to compatibility issues as the slab is highly constrained around the column-beam connections.

Figure 6 depicts 2 cases. The first is a “standard” arrangement of a rectangular bay commonly used for the research of the response of a slab when intermediate secondary beams are left unprotected [7, 10]. This diagram also depicts the secondary load carrying mechanism of tensile membrane action that is known to occur in the slab (through activation of the steel reinforcement) at high deflections in a fire. In this case it should also be noted that relatively large tensile strains are likely to occur at the perimeter of the bay as the slab passes over the top of the protected perimeter beams.

Figure 6

Development of tensile membrane action in slab. (a) Idealised rectangular bay with unprotected intermediate secondaries. (b) Rectangular bay with offset columns and unprotected intermediate secondaries. Note—an intermediate secondary beam is a secondary beam that does not connect into a column

In the arrangement indicated in Figure 6a the perimeter beams are supported equally at the corners of the bay and, assuming similar structure surrounding the bay, will be subject to even lateral and vertical restraint.

The second diagram (Figure 6b) is a simple depiction of an offset bay where the columns on one side do not line up with the beams that would typically form the protected perimeter of the bay. In this example it is assumed that both intermediate secondary beams are to be left unprotected.

In case (b) the perimeter protected beams are not supported equally. The offset in the column location affects the vertical support at the corners of the bay. As the slab can no longer deform smoothly over the perimeter beams, stress and strain concentrations will develop around the columns. Similarly, as the stress pattern in the slab is disrupted it may not be possible for a full compression ring to develop in the slab to support tensile membrane action.

Figure 7 shows the strain in 2 areas of the same structural fire finite element model at the same time. The mid-span deflections are comparable as the spans are similar and the temperature regime is the same in both areas of the model. Protected steel temperatures peak at approximately 620°C, while unprotected steel reaches approximately 1,000°C. The key difference is that the bay on the left has a small offset at each corner, while the bay on the right has strongly offset beams.

Figure 7

Strain in slab rebar (images use the same strain scale where blue is low strain and orange/red is high strain) (Color figure online)

The strains in the slab rebar where offset beams exist are focussed around the end of the beams attached to the columns. This is in contrast to normal bays where rebar strains tend to be larger in the centre of the span of the protected bay perimeter beams. This latter response is due to the shape taken on by a rectangular slab bay, with rigid corners limiting slab deflections and greater slab curvature occurring over the middle of the bay perimeter beams.

One method of controlling strains over protected members is to locally increase the amount of reinforcing steel in areas of high strain. In ambient design in the UK it is typical to detail additional loose bars over the top of primary beams to assist in maximising shear connection efficiency between the beams and the slab [11, 12]. This additional rebar can also be used to demonstrate additional robustness in the fire limit state. Where the strains are well distributed along the length of a beam, as in Figure 7a, such rebar is easy to detail. If there are severe strain concentrations around a column for instance, then it can be difficult to incorporate additional rebar to assist. Further research is required to determine if there are robust alternative design solutions available for this issue, or if it is better to avoid offset columns.

The full implications of the effect of offset beams in the high temperature response of composite structures have not been investigated at this point, however they appear to locally increase strains in the slab around column locations. It is therefore recommended that such offsets be removed from the design or minimized as far as possible to reduce the possibility of slab rupture. Further research is required to understand the full effects of offset beams/columns. The detailed assessment of structures with this type of arrangement is recommended to ensure that concentrations of high strain do not develop in the slab over the protected members, potentially affecting the floor to floor compartmentation or overall stability in a building.

Beams with Web Openings

In recent years long span beams with openings in the web (e.g. cellular beams, castellated beams, “smart” beams) have become increasingly popular in steel framed buildings because they reduce steel weight by removing unnecessary material. This type of beam also allows a reduction in floor to floor heights by allowing services to be positioned in the structural zone below a floor by passing through the beams. In normal structural design for buildings these beams are typically provided with shear studs to rigidly connect the beam to a concrete slab above to create a composite structure.

The response of this type of member under loading is complex even under ambient conditions. The complexity of local response increases when thermal expansion forces and displacements driven by whole frame response to severe compartment fires are also considered.

Substantial research has been done [13] with regard to the key failure mechanisms of this type of beam, including but not limited to:

  • Web post buckling

  • Vierendeel action around the holes

  • Buckling/shear failure of top and bottom Ts around holes

  • Web post shear resistance (vertical and horizontal)

  • Longitudinal shear failure of slab above the beam adjacent to studs

  • Concrete crushing failure around the studs

All of these mechanisms are affected by an increase in member temperature as both strength and stiffness will be reduced. The fabricators of proprietary cellular beams in the UK/Europe tend to include an intumescent (fire protection) paint manufacturer as part of the joint venture and therefore most design software for cellular beams includes an elevated temperature check to determine the required thickness of intumescent paint.

It should be noted, however, that such checks are based on simply reducing material properties in line with design code recommendations to determine the lowest temperature at which the first failure mechanism is observed. This method of assessment only considers a single structural member. This type of design check is informative with regard to the response of the member in question; however it does not address the whole building response or the impact of thermal expansion forces. This approach is in line with current international fire testing. Single element checks are an important part of the Structural Fire Engineer’s toolbox, but they do not provide a sufficient degree of detail to allow an engineer to fully understand the response of a larger structure.

The inclusion of substantial web penetrations in cellular beams, especially in long span cellular beams, adversely affects the strength and stiffness of the member and also introduces specific local failure mechanisms. It is important to determine if any of these issues could materially affect the robustness of a building structure. However, as the response of such members is complex, and their response within a whole building structure as it is being heated is more complex, it is often necessary to use finite element modelling to fully understand the response of the building as a whole.

The use of cellular beams introduces challenges with regard to calculating the structural response at elevated temperatures. When assessing a building under severe fire exposure it is useful to be able to characterise the response of the largest area possible to reduce the effect of numerical errors as far as possible. Assessment of a whole floor, for instance, removes the problems that can arise through the use of symmetry boundary conditions, especially where the building geometry is irregular. The most computationally efficient way to model such a large portion of structure is to use beam or line elements as a single finite element can accurately represent several hundred millimetres of beam or column under the right circumstances.

However, when cellular beams are included in the structure, beam (or line) elements should be used with caution as the simplifications used in their numerical formulation cannot accurately reflect the local response of the beam around the holes. This issue can be addressed through the use of shell elements but such a model may require up to 100 shell elements to represent the same portion of beam as a single beam element. Figure 8 depicts this difference by presenting two different representations of the same members from a recent commercial project using both global and submodels as explained below.

Figure 8

Beam element versus shell element mesh. *Note—the model in (a) consists of beam or line elements between the nodes. The real section size used in the analysis has been overlaid over the line element to provide a direct comparison to the shell element model in (b)

When considering the response of a structure, the authors recommend the use of a combination of models taking advantage of both types of element, beam and shell. In a typical project where cellular beams are included, a beam element based global model will provide the response of the overall structure. A submodel is also constructed of a smaller area (typically 1 or 2 bays) using shell elements and including as many details as possible regarding holes in beam webs. This second model is used to check that the beam elements in the global model are responding correctly and that there are no other local effects in the cellular beams that might lead to early failure.

Figures 9 and 10 present 2 models used for the recent structural fire assessment of the Pinnacle building; a tall office building in the City of London. This project has been previously published as a case study [14]. Due to the irregularity of the structure a whole floor was assessed. Figure 9 is the global model of a whole floor. Boundary conditions are restricted to columns above and below the floor of interest and where the slab frames into the cores (grey areas in Figure 9). Figure 10 presents the sub-model used for this assessment consisting of a small section of the structure, designed to capture the local response of the cellular beams. Figure 11 indicates the expected deformations in the beams when the sub-model is heated to steel temperatures of approximately 1,000°C.

Figure 9

Global model using beam elements

Figure 10

Submodel using shell elements (slab omitted for clarity)

Figure 11

Shell element submodel demonstrating local deformations in cellbeams

Figure 12 presents a representative comparison between the global model and the submodel. Beam 27 is the protected secondary beam between columns indicated in Figure 11, while Beam 28 is the beam to the left of Beam 27, which was left unprotected in the analysis. In general the trends are similar in both models; however it is apparent that the submodel is exhibiting different behaviour in Beam 27. The details of this difference are not for this paper, suffice it to say that it was demonstrated to be related to a local buckling issue in the submodel that the global model could not accurately assess numerically in its beam elements. This led to the underestimation of floor deflections in the global model as the simpler beam elements cannot accurately capture local buckling effects in cellular beams. The use of 2 models in this way identified a potential issue in the structure and allowed the design to be modified to mitigate this weakness. In this case the fire protection layout was modified to address the issue. An alternative approach would be to modify the beam size or look at other modifications to the structure to overcome the problem (such as altering member lengths, grid layouts or size of reinforcement mesh), however these were not feasible solutions for this project.

Figure 12

Comparison of mid-span vertical deflection results between Pinnacle Global and submodel

It should be noted that the kind of lateral torsional buckling observed in the Pinnacle project, amongst others, has recently been demonstrated in a full scale fire test conducted by Nadjal et al. [15] and reported in The Structural Engineer (Figure 13). The test included long span (15 m) cellular beams with a large proportion of holes in the web. This type of failure is not necessarily detrimental to the response of the overall structure when observed in intermediate secondary beams, provided that the slab and primary beams can support the floor loading in tensile membrane action.

Figure 13

Example image from Nadjal et al. [15] [Figure 16: Internal beam near mid-span and end connection after fire (Beam 4)]


The response of a structure under severe compartment fires is sensitive to the restraint available within the structure. The horizontal restraint to individual members from the surrounding structure provides a partial support that the expanding member will push against. As previous research has proven [1, 15] deformations in a structure subject to fire loading is largely caused by restrained thermal expansion. The mechanisms providing restraint therefore become important to the detailed response of a structure.

Based on the authors’ experience on projects the following have been identified as key restraint providers:

  • Core structures

  • Internal and perimeter columns

  • Shear Studs

The following sub-sections describe the effect of each type of restraint.

Core Structures

Building cores are often designed to provide lateral and vertical stability to resist wind and earthquake loading. At the FLS this also provides areas of high rigidity for heated members to push against. This rigidity comes primarily from the high lateral strength that underpins the design, however core structures will also tend to be cool due to the use of thick concrete walls with low thermal conductivity, further increasing relative rigidity.

Due to their stiffness, stability cores will tend to become a point within a floor where thermal expansion will extend from. Figure 14 compares the type of expansion pattern expected in a building where lateral restraint is provided by rigid cores and a building that uses a perimeter vierendeel stress frame to provide lateral stability. As the latter has no points of rigidity within the floor plate the expansion pattern is smoother as all deflections are generated from the centre of the floor plate. Vertical floor deflections may be larger, however, as the external perimeter frame is more rigid. The expansion pattern presented in Figure 14a presents a more complex interaction of expansion as the floor has 3 points of rigidity to push against. This type of arrangement leads to high membrane forces and/or larger deflections in the regions directly between the cores. Additionally outward movement of perimeter columns can be larger when rigid stability cores are located. As the core cannot displace, all of the horizontal thermal expansion in a beam must be transferred into the perimeter columns, rather than sharing the displacement equally between both ends of the beam.

Figure 14

Expansion pattern (arrows) comparison between building types. (a) Building with cores. (b) Building without cores

Another response that can be observed in buildings with multiple cores is in-plane rotations being induced in the floor structure also caused by the expansion pattern. This can lead to column twisting that may affect vertical stability of the building.

Again the complexity of the floor response will depend on the design of the building and therefore it is difficult to provide a definitive general conclusion with regard to the best type of design. The best solution is that the response of the building has been reviewed at elevated temperature so that high forces and thermally induced deformations can be accommodated by ductility or by other design inclusions.

Internal and Perimeter Columns

These members typically act as a spring restraint against floor lateral displacements (expansion and contraction) in both compression and tension. The level of restraint provided is variable and depends on the applied heating regime, column size, column temperature, column weak/strong axis and column spacing.

Research by Flint [16] into the effects of multi-storey fires on structures similar to the WTC Towers describes the key issues around the restraint provided to and by columns. The two key phases, i.e. during heating and then during cooling, are indicated in Figure 15. The diagrams are based around multi-storey fires, however the same mechanisms have been demonstrated in single storey fires.

Figure 15

Effects of floor heating on columns (figure recreated from Flint [16]). (a) Heating phase. (b) Tensile membrane action

During the heating phase the columns are pushed out at the level(s) of the floor affected by fire. Depending on the number of floors affected by fire this can lead to high tensions and/or compressions being generated in the floor beams due to compatibility of the column geometry to the floor expansion. The outward movement of the column can also lead to additional p-Delta moments being generated in the column. This additional loading can be sufficient to lead to column failure in columns already subject to high axial forces. The positive connection between the floor and the column during heating means that this can be a stable mechanism, provided that the connections can withstand the applied force.

During cooling, or if the floor beams move into catenary action (Figure 15b), the column is pulled inward relative to its original position. This will also generate tension in the beam-column connections and the inward deflection will also lead to additional p-Delta moments being generated in the column. This is a less stable configuration as the floor will be acting in compression to restrain the column. The large deflections generated in the floor during severe fire exposure make it less able to restrain the column, especially if the beams are acting in catenary action.

As noted above the degree to which the columns are moved out of plane will depend on the relative in-plane stiffness of the floor compared to the bending stiffness of the various columns making up internal and perimeter supports. If the columns are relatively stiff compared to the floor then this will lead to larger deflections in the floor. If the columns are relatively flexible then floor deflections will be lower and outward movement of the columns will be increased.

Additionally, as indicated in Sect. 2.3.1, the outward movement of the columns will depend on the relative stiffness of the columns at either end of the beam/floor system. The limiting bounds of this movement are indicated in Figure 16.

Figure 16

Limiting bounds of column displacement. (a) Greater outward movement of column as the core provides a rigid point for the beam to expand from. (b) Horizontal expansion of beam shared between columns

Shear Studs

Composite connection between slab and beams is a very important driver of thermal curvature through the mechanism of differential thermal expansion. This is especially true where unprotected beams are present.

Breakage of shear studs can lead to substantially different patterns of restraint and deflection in a floor structure compared to full composite connection. By breaking the link between the beam and the slab thermal curvature can only be driven by the differential temperatures in the slab and beam individually. In addition the overall strength of the floor system is severely weakened by stud breakage which makes beam failure more likely (Figure 17).

Figure 17

Differential thermal expansion. (a) Composite action maintained. (b) Composite action lost

Another key role of shear studs in the restraint of a structure is by anchoring the slab to the primary superstructure. If the studs are broken then the slab is able to move laterally over the supporting beams, potentially causing the slab to slide in toward the bays with the largest spans, deflections and/or loads.

While the superstructure helps keep the slab in place, the slab also helps keep the beams in line. If the shear studs break a toppling failure or lateral torsional failure of the beam is more likely.

Shear studs are typically considered to remain in place during a structural fire analysis for design. As with many assumptions in structural fire engineering this is based on the observations of the Cardington tests. However due to the difference in maximum stud spacing between countries and the difference between the Cardington frame and modern long-span floor construction further research is required to ensure that this assumption remains valid.


Historically most structural fire research has focussed on the relatively small elements and assemblies that can fit in a standard testing furnace and therefore proportionally more research has been done on the structural members themselves. When considering whole frame response to severe compartment fires it is important to consider the means by which the structural elements are being connected.

The response of connections under fire loading has been the subject of several recent studies [17, 18]. It has also become increasingly apparent in design that connections must be considered to ensure a robust response from the structure.

The forces generated in members during a fire can be extremely large. This is because when a restrained beam, for example, expands during heating, the force required to restrain it is equivalent to the compressive capacity of the whole section. Similarly in cooling, the whole capacity of the section will be activated if a rigid restraint is provided. The forces generated are often in the Meganewton range and therefore it is impractical to design the strength of the connections to withstand such forces.

A more robust solution is to provide connections designed for ambient loading, but with ductility sufficient to allow thermally induced movements to be accommodated. The following key deformations have been observed in connections in a fire affected building.

Large axial compression and tension forces can develop during heating and cooling. Where all beams in a fire compartment are protected they will expand and contract at the same time. Where all beams are the same length, this leads to a simple response of compression in all connections during heating and potentially tension during cooling. Where the structure has unprotected secondary beams or beams that differ significantly in length connection forces will vary during the heating phase as the faster expansion of the unprotected secondaries can induce significant tension forces in the connections of protected beams. It should be noted that the magnitude of the forces are not significantly different in fully protected structures compared to those generated in structures with partial protection, however the times at which they are experienced will differ.

The other key issue in fire affected connections are rotations around the major bending axis of the beam. These deformations are caused by the fire induced deflections in the beam. Rotations for a 14 m long beam can easily reach between 6° and 12° depending on fire exposure and degree of protection. Figure 18 indicates the type and extent of deformations that might be experienced in a structure under a severe fire.

Figure 18

Numerical modelling [19] of testing conducted by the University of Sheffield [17]

As well as these deformations it is possible for lateral shear and minor axis rotations in the beam to become critical issues, especially where secondary beams are offset from columns by a short distance as indicated in Sect. 2.1.3.

Research by the University of Sheffield [17] has investigated the response of a range of common connection types under deformations that are considered typical when a structure is heated. This research program suggests the following order of robustness:

  • Double Angle Cleats—Most ductile (Figure 18)

  • End plates (Figure 19a)

    Figure 19

    Connection types. (a) End plate. (b) Fin plate

  • Fin plates—Least ductile (Figure 19b)

On-going research in Europe [20] into “reverse channel” connections, where the legs of a channel are welded to a column and the beam bolted to the web, also indicates good potential ductility.

Due to the large deformations that can be expected in a structure and its connections at elevated temperatures, even in protected members, it is recommended that critical connections, i.e. connections between protected members and columns, always be designed with ductility in mind in a similar manner to seismic design. By providing connections with ductile plate elements it is possible to allow for the expected deformations without rupturing plates and compromising the vertical carrying capacity of the connection. Only a relatively small amount of movement need be provided (10 s of mm) to significantly reduce connection forces at the FLS and significantly reduce the likelihood of connection failure based on the fundamental theory of thermal expansion in beams.

Connections between unprotected secondary beams and primary members may be less critical, as long as it can be demonstrated that the floor slab is capable of supporting the floor loading in tensile membrane action. In this case the connections can be of a lower ductility/robustness as it is less important if they fail at the FLS.


The authors have conducted several detailed studies of buildings at the FLS in the last 10 years. The studies have been conducted as part of the design of modern, contemporary tall buildings and have implemented cutting edge research. This paper presents a variety of “lessons learned” over the course of these design projects that the authors consider to be relevant to the design of all buildings. The authors will be implementing these findings as part of best practice design in future projects.

As well as being relevant to future design, it is also possible to point out examples of most of the issues discussed here in the design of either the WTC towers or WTC7. It is therefore highly probable that the collapse of these buildings was affected by these mechanisms.

As all buildings are designed differently, especially tall signature buildings, it is becoming increasingly clear that structures need to be assessed at the FLS as well as under the more conventional load cases of gravity, wind, snow and earthquake. Without a detailed assessment it is difficult to determine exactly how a specific structural form will respond to elevated temperatures and it is possible that the building design may contain an inherent weakness. This type of assessment is relevant to all buildings, as it is possible to generate damaging mechanisms at elevated temperatures, even if the building is fully protected to standard code levels.

This paper describes a series of “lessons learned” about the detailed response of structures under fire loading. The following recommendations are made to practising structural engineers when considering structural design to resist fire.

Structural Layout

Square or rectangular bays with columns at each corner have been demonstrated to be the most robust layout. Irregularly shaped and unbalanced bays and offset beams can induce complex additional forces and deformations in structural members in a fire and connections making it more likely that an unforeseen failure mechanism might occur. If the building geometry is highly irregular then automated optimisation processes may return highly irregular bays and structure. It is recommended that the structural layout be reviewed and as many irregularities as possible are removed. If it is impractical to remove offsets or unbalanced bays then the structural design should account for their inclusion by addressing the robustness of the beams, columns and connections locally. This could be in the form of beams with higher lateral shear capacities or greater ductility in the connections. Due to the complexity of the interactions in a full building structure Finite Element Modelling is required to inform on the best solutions for a particular structure.

Local Response of Cellular Beams

The use of cellular beams, or other types of fabricated steel beam with web penetrations, is increasing as they provide significant benefits in the servicing strategy for a steel framed building. When assessing such members it is important to consider the effects of the web holes in the overall structural response. Local buckling issues can occur that might not be picked up using simplified analysis techniques. These can affect deflections and strains in the floor and can affect column displacements leading to an over- or under-estimation of the structural robustness of a building in fire.

Sources of Restraint

The key driver in the response of a structure to fire loading is the degree of restraint available to each individual structural member. Thermal expansion acts against this restraint producing the specific deflections and forces for that structure. Restraint can come from various sources, such as rigid cores, columns and the shear connection between the slab and steel members. The compatibility of the structure around these sources of restraint can be complex. In order to properly understand the response of a structure the sources of restraint need to be identified and considered in detail to ensure that a structure can remain stable at the FLS.


In order to properly consider whole frame behaviour the response of the structural connections must also be considered. At the FLS high forces can be generated by the expansion and contraction of the structure. Therefore it is impractical to design connections for strength alone, especially when material strength is being affected by elevated temperature. It is suggested that a more robust approach would be to consider the ductility and overall robustness of the connection in a similar way to seismic design. By providing connections with ductile plate elements it is possible to allow for the expected deformations without rupturing plates and compromising the vertical carrying capacity of the connection. Only a relatively small amount of movement need be provided (around 10 mm to 50 mm) to significantly reduce connection forces at the FLS and significantly reduce the likelihood of connection failure.


  1. 1.

    Steel Construction Institute (1991) Structural fire engineering: investigation of Broadgate Phase 8 fire, Ascot.

  2. 2.

    British Steel plc (1998) The behaviour of multi-storey steel framed buildings in fire. Technical report, Swinden Technology Centre.

  3. 3.

    NIST (2005) Federal building and fire safety investigation of the World Trade Center disaster: final report on the collapse of the World Trade Center Towers, NIST NCSTAR 1, September 2005.

  4. 4.

    NIST (2008) Federal building and fire safety investigation of the World Trade Center disaster: structural fire response and probable collapse sequence of World Trade Center Building 7, NIST NCSTAR 1–9, November 2008.

  5. 5.

    Heise A, Flint G, Lane B (2007) Effect of fire on tall buildings: case study. In: 3rd International conference on steel and composite structures 2007, ICSCS07, pp 705–712.

  6. 6.

    Pozzi C, Lane B, Lamont S, Heise A (2006) The role of large strains and consequent mitigation measures for buildings in fire. In: Proceedings of the 4th international workshop on structures in fire, SiF 2006, vol 1, pp 487–498.

  7. 7.

    Lamont S (2001) The behaviour of multi-storey composite steel framed structures in response to compartment fires. PhD Thesis, University of Edinburgh.

  8. 8.

    Partners in Technology Report (2001) Behaviour of steel framed structures under fire conditions. Main report, University of Edinburgh.

  9. 9.

    Allam AM, Burgess IW, Plank RJ () Simple investigations of tensile membrane action in composite slabs fire, paper 03.02. In: Proceedings of the international conference on steel structures of the 2000s, Istanbul, (2000) pp 327–332.

  10. 10.

    Bailey C (2003) New fire design method for steel frames with composite floor slabs, FBE Report 5.

  11. 11.

    BSI (2004) Eurocode 2: design of concrete structures. General rules and rules for buildings, BS EN 1992-1-1.

  12. 12.

    BSI (2004) Eurocode 4: design of composite steel and concrete structures. General rules and rules for buildings, BS EN 1994-1-1.

  13. 13.

    SCI, Design of composite beams with large web openings: in accordance with Eurocodes and the UK National Annexes, Publication 355, 2011.

  14. 14.

    Jowsey A, Lim L, Lane B (2009) Structural fire analysis for a perimeter braced-frame structure. Application of structural fire engineering. Prague.

  15. 15.

    Nadjal A et al. (2011) Full-scale fire test on a composite floor slab incorporating long span cellular beams. Struct Eng 89(21):18–25.

    Google Scholar 

  16. 16.

    Flint G (2005) Fire induced collapse of tall buildings. PhD Thesis, University of Edinburgh.

  17. 17.

    Al-Jabri KS, Davison JB, Burgess IW (2008) Performance of beam-to-column joints in fire—a review. Fire Saf J 43(1):50–62.

    Article  Google Scholar 

  18. 18.

    Dai XH, Wang YC, Bailey CG (2009) Effects of partial fire protection on temperature developments in steel joints protected by intumescent coating. Fire Saf J 44:376–386.

    Article  Google Scholar 

  19. 19.

    Flint G, Lim L, Jowsey A, Sarrazin H, Lamont S, Lane B (2010) Intermediate detail modelling of structural connections in fire using shell elements. In: Proceedings of the 12th international conference—Interflam 2010, vol 1, pp 45–55.

  20. 20.

    European Research Fund for Coal & Steel (2009–2012) Design of composite joints for improved fire robustness—COMPFIRE, Grant Agreement Number RFSR-CT-2009-00021.

Download references


The authors would like to acknowledge the assistance of Dr Allan Jowsey in several of the projects upon which this paper is based. Similarly the work in this paper has been greatly assisted by on-going collaboration with the University of Edinburgh.

Author information



Corresponding author

Correspondence to G. Flint.

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Flint, G., Lamont, S., Lane, B. et al. Recent Lessons Learned in Structural Fire Engineering for Composite Steel Structures. Fire Technol 49, 767–792 (2013).

Download citation


  • Structural fire engineering
  • Fire
  • Structures
  • Finite element modelling
  • Composite steel frame
  • Connections
  • Restraint
  • Thermal expansion