Asif Usmani & Michael Rotter

Failure of Structures Under Fire

To say that the definition of failure of structures in fire is ‘complex’ is a huge understatement. It is much more than that, beginning right from the most elementary issue of ‘what kind of failure’ is it that one is trying to define. As the fire safety of a structure is of interest not only to the architect and structural engineer but also to fire safety engineers and regulators/building control officers (not to mention the owner/client). It is the highly varying perspectives of this group of people that confuses the issue and makes it very difficult for a consensus on the definition of failure to emerge, if it were indeed possible. For a fire safety engineer, failure may be said to have occurred if a breach of compartmentation and uncontrolled spread of fire occurs (with or without undesirable consequences) irrespective of whether there had been ‘structural failure’. For the owner, failure may be defined in terms of the economic losses accrued from the fire, in terms of the cost of repair, loss of business and loss of productivity etc. without regard to structural failure. Structural engineers however are concerned with structural failure, which is the purpose of this document. In fire situations even this is far from straightforward and hence the need for discussion.

In general, to structural engineers, failure invariably means the inability of a structure to continue to sustain a given loading. This may occur due to:

• The magnitude of the load being in excess of the capacity of the structure.
• Degradation of the load carrying capacity of the structure (caused by age or other factors) so that it is unable to sustain the loads normally imposed
• Design faults, where one or more failure mechanisms were not foreseen
• Construction faults, where poor materials or quality of construction caused the structure to be have a reduced capacity than that assumed in design

When design loads are applied, the structure normally responds by changing its geometry (displacement) and reverts to its original geometry after the load is removed and its capacity to carry loads is undiminished (through the property of elasticity). However, if the magnitude of the load is excessive the change in geometry required is greater than the structure can sustain elastically. The large deformations lead to permanent changes in the capacity of the structure to withstand loads. In the limit a very large load can cause the structure to ‘collapse’, which is certainly ‘failure’. In practice however, a structure that may not have collapsed may still be said to have failed, if it is unfit for any further use. On the other hand, as for instance in the field of earthquake engineering, a structure can be designed to sustain very large changes of geometry without collapse, and in this case even if the structure is unusable after an earthquake it cannot be said to have failed. Therefore ‘failure’ may be more accurately defined as the inability of the structure to ‘perform’ as designed, which leads to a situation where no absolute criteria for failure can be defined in isolation from the requirements of a ‘performance based design’. There are however large classes of structures similar enough in construction and design to have a large overlap in the performance requirements, for which it may be worthwhile to define some general ‘failure’ criteria, if only for the purpose of opening an informed debate on this complex question.

The traditional criteria of structural failure at ambient temperature relied on the following ideas (Fig. 1):

• attainment of a peak strength under increasing load, followed by a reducing load carrying capacity;
• in structures whose strength does not reach a peak, the attainment of a deflection deemed to be so large that it is unacceptable even for an ultimate limit state.

The second criterion works well when the load-displacement curve is relatively flat at large displacements, but it is more difficult to apply when continually increasing strength is seen.

Fig. 1 Possible forms of load-deflection curve at ambient temperature

When these criteria are transferred to a fire scenario, the load is no longer changing, but generally remains fixed during the fire. Thus, the attainment of a peak strength is not so easily identified. A rapid increase in deflections with a small rise in temperature, commonly called ‘runaway’ has therefore been used as an analogy for the second criterion defined above. The main problem with such a description is that it has been based upon testing simply-supported beams in furnaces subjected to ‘standard’ fires. It is widely accepted that a beam in a furnace is has little or no relevance to a large frame structure in fire. However in the absence of clear and simple alternatives it continues to be the only picture that most professionals involved in fire safety engineering associate with failure in fire. Figure 2 illustrates the huge difference in the performance of two beams (one simply-supported and one with end translations restrained) subjected to fire. Runway is delayed considerably in the beam with ends restrained against translation. Figure 2 shows clearly that for temperatures below 300 °C, the deflections for the restrained beam are much larger than that for the simply supported beam, however they have nothing to do with ‘runaway’. These deflections are caused entirely by the increased length of the beam through thermal expansion and are not a sign of loss of ‘strength’ or ‘stiffness’ in the beam until much later. In fact approximately 90% of the defelection at 500°C and 75% at 600°C is explained by thermal expansion alone. Most of the rest is explained by increased strains due to reduced modulus of elasticity. However the behaviour remains stable until about 700°C when the first signs of runaway begin to appear.

In the simply-supported beam, by contrast, the rapid increase in deflections results from a great loss of tangent stiffness, which corresponds exactly to the case of structures at ambient temperature subjected to increasing loads and degrading stiffness. If a small increment in load produced a great increase in deflection the intended meaning of the second failure criterion would be met. This criteria makes sense in an ambient temperature situation as the large deflections correspond to a large loss of load carrying capacity. The recognition of this relationship between an ambient temperature failure criterion and a fire criterion is valuable in assisting the development of an understanding of the basis of failure criteria for fire.

In highly redundant structures, very large deflections can develop at moderate temperatures (for instance the restrained beam in Figure 2). It might therefore be thought that a state corresponding to the second criterion of failure has been reached. However, the sensitivity to a small increase in loading is very small, so the second criterion has not been reached even when the deflections themselves are very large.

Moreover, it has now become evident that large deflections in a highly redundant structure at high temperature actually assist the structure to survive the fire with minimal damage, so a criterion which limits deflection itself would encourage the designer towards a poorer design (against an ultimate limit state of collapse). The source of the additional robustness imparted by large thermally induced deflections derives from the fact that a displaced configuration amenable to the alternative load carrying mechanism of tensile membrane action is achieved without large and damaging mechanical strains in the structure (as would be the case at ambient temperature). Figure 3 below shows the restrained beam of Figure 2 loaded with factors of a udl w that will cause a plastic hinge to occur at the beam midspan at ambient. It is interesting to see that up to approximately 600 C the load has very little practical effect on the deflection and nearly all the deflection increase (after loading) from ambient to 600 C is thermally induced. Therefore it would be simplistic to assume that the structure is in distress (or approaching failure) just by looking at the total deflection.

(Figs 2 and 3 are not available)

For these reasons, it is necessary that new criteria of failure should be developed for real structures in fire.

3. Discussion

Since the deflections of the structure due to thermal effects can be very large at even moderate temperatures and when the structure is still quite undamaged, it would be meaningless to use a criterion based solely on the deflection divided by the span, or some similar measure to indicate structural failure in fire.

Furthermore, to mobilise the alternative load carrying mechanism of tensile membrane action, large deflections are a requirement. This action has been shown to be much more reliably available at high temperatures than at ambient temperatures. This is because large deflections, and the corresponding change of geometry of the structure, can be achieved predominantly from thermal strains. As a result, large mechanical strains, which are detrimental to the structure, are not required. Moreover, the restraint to thermal expansion often causes beneficial mechanical strains (compression), which allow much greater bending strengths to be exploited.

Large deflections are used as indicators of failure at ambient temperatures chiefly because deflections are closely related to mechanical strains, and large mechanical strains are associated with severe damage to the materials of construction. Thus, large deflections at ambient temperature are a sign of structural distress. They signal that the structure has reached an ‘undesirable state’ where further loading may cause excessive material degradation and the loss of capacity of the structure to support the applied loads. This is termed an ‘ultimate limit state’.

Failure under fire can be approached using the same philosophy of identifying signs of ‘distress’ in the structure. As absolute deflections are no longer a good measure of damage to the material of construction, the concept of failure must be developed by considering what the signs of distress must now be. The behaviour of a structure under fire is considerably more complex than its behaviour under normal loading at ambient temperature, so it may be appropriate to use more than one criterion to signal a potentially catastrophic or ‘undesirable state’.

4. Proposed criteria

A computer analysis may provide clues about the levels of degradation of material properties experienced by various components of the structure. Coupling this with recently developed knowledge of structural responses that have the greatest bearing upon precipitating a collapse, it is possible to identify some of the most critical signs of distress:

1. The first criterion can be based upon the fact that deflections caused by non-thermal effects (leading to mechanical strains) are the ones that cause damage to the structure. So if the magnitude of the deflections is predominantly (say over 90%) accounted for by considering only the thermal strains (thermal expansion and thermal curvature) imposed upon the structure (taking into consideration the conditions of compatibility and boundary restraints etc.) then it can be deemed to be "safe". This can be further refined to include the deflection caused by reduction in stiffness (modulus) at higher temperatures.

By contrast, if a significant portion of the deflection (say over 10%) consists of non-recoverable strains (non-thermal and non-elastic), then the structure may be deemed to have reached the "undesirable state of irrecoverable deflection".

2. The tensile membrane capacity of the slab is terminated by rupture of the reinforcement, which in turn depends on the mechanical strain demand placed on the reinforcement. Thus the mechanical strain (not total strain) in the composite deck slab reinforcement is an important indicator of potential failure at large deflections. This criterion is especially important when cold-formed meshes are used, because the steel may have a low strain capacity before rupture. Therefore if the tensile mechanical strains (after excluding the thermal strains) in the concrete slab reinforcement exceed a specified maximum value an "undesirable state of slab tensile mechanical strain" may be deemed to exist.

3. A third criterion that may be considered important is the horizontal displacement of exterior columns, caused by thermal expansion of internal structural members or the floor system. This may be seen as an undesirable due to unacceptably high additional eccentricities of load may occur in columns in the lower storeys. Therefore a limit may be required on the outward displacement of exterior columns. When this limit is exceeded the "undesirable state of exterior column displacement" may deemed to have been reached.

4. Finally, large deflections in beams at or near the compartment boundaries may lead to partition damage, leading to a breach of the compartmentation. Therefore, should the deflections of beams at the compartment boundaries exceed a specified value, the "undesirable state of compartment breach" may deemed to have been reached.

The above criteria must be developed in a quantitative manner, but the amount of information on which to determine limits for each is currently rather small. The following provide some suggestions for quantitative failure indicators:

1. Over 20% of the total deflection of a member derives from irrecoverable (or plastic) strains along its length (therefore excluding high local strains through rotation at supports) (???)

2. The rate of deflection increase at the point of highest deflection is greater than that explained by the rate of compartment temperature increase and creep effects (???)

3. The tensile strain in the mesh reinforcement is over 50% of the rupture strain at the point of largest deflection (???)

4. The tensile strain in the mesh reinforcement is over 75% of the rupture strain at the points of support (???)

5. Exterior column displacements are under 30 mm at any storey (???)

(Top)

Comments on the above by Kees Both (TNO)

In the article I miss the description of the origin of fire failure criteria and the actual description of them. To understand the current fire failure criteria one must realise how traditionally fire tests were performed. The tradition is that tests are performed on components, which in Europe are un-restrained (i.e. no restraint against thermal expansion or thermal curvatures). Even statically indeterminate composite structures mostly show typical run-away failure (i.e. if bending is the failure mode, and not e.g. shear). The statement you make that deflections in those cases correspond to high mechanical strains is in fact the most important one. Obviously mechanical strains (actual level) and mechanical strain rates are in fact the pure failure indicators in bending. The development of thermal strains due to (partial restraint to) thermal elongation "mystifies" the engineers feeling that large deflections must always correspond to high unfavourable mechanical strains.

Two other remarks concern the strains in the reinforcement meshes. It is in fact not only the effect of cold-worked reinforcement which may cause rapid increase of strains (due to its different stress-strain relationship), but also the reinforcement ratio. Low ratios cause (earlier and more) localisation of deformations and strains and therefore "undesirable" states may be reached sooner than expected. Furthermore, the (mechanical) tensile strain in the mesh should also and to my opinion more importantly by checked in regions were the composite slab is in hogging (this need not be the region with largest deflections). This because the (statically indeterminate fire exposed) composite steel-concrete slab will reach hogging plastic moment capacity sooner than sagging (and thus solicitates hogging moment capacity sooner and longer; for which also the steel decking does not play any role as opposed to sagging were the steel decking in realistic fires significantly contributes to the load bearing capacity).

(Top)

Roger Plank (1)

The following notes set out my personal thoughts in relation to how ‘failure’ may be defined. In doing this I have tried to view the problem from the standpoint of establishing criteria which may be used in computer modelling of the structural behaviour. I have not addressed in detail issues such as integrity (which is a separate performance requirement).

Arbitrary deflection limits alone are not appropriate (except for tests where furnace etc. must be protected)

Rate of increase of deflection could be used as a measure - e.g. when deflection rate = 10 x average rate of increase between 20° C and q ’ - where q ’ is the temperature at which the deflection is equal to span/30, say?

However this may not adequately allow for ‘localised’ failure - i.e. rapid deflections in one member which ‘fails’ but is subsequently supported by alternative actions, OR an initially lightly loaded element which subsequently attracts load from a ‘failing’ member.

The total deflection history could examined retrospectively, but this is a lot of work, and more importantly is subjective.

There is no evidence that horizontal deformations of columns are critical. Structural failure in this context is not concerned with limiting behaviour to prevent irreversible changes (although insurers may wish to do so).

At ambient temperature, collapse of steel structures (in the absence of buckling) is equivalent to the formation of a mechanism. The same conditions can be considered for steel structures in fire (for design) – using a modified plastic analysis. This is essentially the moment capacity method when applied to isolated beams. However it is not directly applicable to the analysis of complete structures - unless collapse mechanisms can be defined. Most computer simulations represent stress-strain-temperature functions in such a way that complete ‘plastic hinges’ do not develop, and more importantly the interaction between slab and frame invalidates a simple definition of the ‘plastic’ collapse mechanism.

Yield line analysis of the whole floor slab (ignoring the frame), and collapse analysis of the frame (ignoring the membrane action of slab) could be performed independently, and collapse be conservatively based on whichever gave the higher failure temperature.

A more sophisticated approach is to consider the combined behaviour of slab and frame.

It may be possible to establish limits on the behaviour of beam and slab cross-sections and substitute real hinges/yield lines when these limits are reached - at present most material properties allow infinite deformation. However there remains the problem of defining limits – for example over what length should ‘failure strain’ be considered? Such an approach is suitable for elastic behaviour (where the yield point or some proof limit can be used), but how applicable is it when considering plastic behaviour and rupture?

I am conscious that in outlining these thoughts I have posed more questions than suggested answers, but hopefully they will provoke some thought and discussion.

(Top)

Roger Plank (2)

Behaviour of Whole Structures

Background

Traditional approaches to the design and analysis of structures in fire are based on the behaviour of isolated elements – beams, slabs and columns with idealised support conditions. This is convenient for 'proof testing' and is consistent with the approaches traditionally used for normal design at ambient temperature. However, when exposed to fire, whole structures can behave quite differently from the way individual members might respond to the same conditions. This may be for a number of reasons, but especially

• structural 'continuity' (through connections considered as nominally 'simply supported' or 'pinned')
• 'free' thermal expansion which can cause deformations and additional (second order) stresses
• restraint to free expansion, which can lead to induced forces (axial and bending)

The effects of continuity are generally beneficial, whilst those of expansion (free and restrained) are generally detrimental. They can be very significant and it may be necessary to consider whole building behaviour in relation to not only structural performance but also the integrity of other building components (eg non loadbearing walls).

Steel framed structures

In steel framed buildings the principal sources of continuity are the connections, and, in the case of composite buildings, the interaction between the slab and frame. For non-composite buildings and in-situ concrete floors, the slab may still provide some significant continuity which may be beneficial, but this has not been adequately studied.

Evidence supporting the difference between complete steel-framed buildings and individual steel members comes from:

• Tests: Cardington, King William St, Melbourne
• Analysis: Cardington, parametric studies
• Case history of real fires: Broadgate

The potential influence of the connections is to reduce the effects of fire on beams (reduced deflections and bending stresses), extending survival times.

Connections

In steel framed structures the connections between beams and columns will almost always provide some continuity. The effect of connection rigidity is to induce bending moments at the beam ends, relieving those at midspan, and also reducing vertical deflections. A significant research effort has been devoted to establishing the moment-rotation characteristics of different connections, and methods have been developed allowing designers to account for the ‘semi-rigidity’ of the connections. However these are not widely used, and it is usual for ambient temperature design to be based on the assumption of simply supported beams.

In fire conditions the connections may have a greater influence than at ambient temperature because

Unfortunately there is little data available concerning the rotational stiffness of connections in fire, represented by moment-rotation-temperature relationships, but simplified calculation methods have been proposed ( ).

The effect of connection rigidity and strength on the performance of frames in fire has been demonstrated by experiment (Dave Latham), but this provided insufficient data to establish general conclusions. However, parametric analysis has shown that, even with only modest rotational stiffness, the beam behaviour is closer to that of a fixed ended beams. (SCI design approach?)

In composite construction the continuity of the slab at column supports increases the rigidity of the connection, although cracking of the concrete can make this unreliable.

In practice, the continuity of the slab is a much greater influence than connection rigidity for such structures, and the difference between the behaviour of the frame assuming rigid joints is very similar to the case of pinned joints.

Whilst the connection rigidity reduces bending in the beam at mid-span, the effect on the supporting columns can be to induce additional bending, which could be detrimental to its performance. However, analysis has shown that this is not significant.

Slab continuity

In composite construction the slab is generally continuous over the supports even if it is normally designed as a series of isolated spans. Evidence of real fires, tests and analysis (refs) indicates that, as the steelwork softens and deforms, so the influence of the slab increases. This reduces the beam deflections compared with the equivalent bare steel skeleton, and can increase survival temperatures and times significantly.

The precise mechanisms are not fully understood at present, but appear to involve tensile membrane action of the slab at large deflections. The implications of this have yet to be fully explored but a simplified design guide has been prepared (Cardington Design Guide). Research is continuing to establish how much further advantage can be taken of this, and what must be done to the supporting structure to enable the membrane action to develop.

The principles are not necessarily applicable to slab forms other than conventional composite decking. Floor systems such as precast planks clearly do not provide significant continuity and would be unable to generate tensile membrane action. Even systems like Slimdek may have insufficient continuity to mobilise the structural actions evident with composite decks, and traditional solid slabs acting compositely with the steel frame may spall, exposing reinforcement and weakening the membrane action.

Compartmentation

Although not strictly a structural performance requirement, it is an important design criterion that any fire is contained within its compartment of origin. This means that the enclosure, including the ‘roof’, of a compartment does not allow the transmission of flames. In this context two issues may be important:

The evidence of experiments and real fires in buildings is that composite deck floors generally maintain sufficient integrity, although some large cracks were evident in some of the full scale tests at Cardington. However, spalling of slabs with an exposed concrete soffit may compromise their performance. This is an issue even in conventional approaches to fire protection.

The effect of a severely deflecting slab on compartment walls below should be considered by the designer to ensure that lateral spread of fire is contained. Two approaches are possible:

• design the compartment walls to carry the loads transferred by the slab;
• detail the slab-wall junction to allow for the expected movement (which can be considerable) without transfer of load or breach of the compartment boundary.

Reinforced concrete structures

RC structures are rarely analysed for fire conditions, but continuity in complete buildings is similar to steel frames. Thus the sources of continuity are the connections, and the interaction between slabs ands frame.

Connections

In a reinforced concrete frame, the connections are generally close to rigid This is often accounted for in ambient temperature design so there is potentially less benefit in fire. Experience of steel structures would suggest that joints should be treated as rigid, unless detailed deliberately to minimise moment transfer.

Slabs

For in situ construction it is likely that the slab continuity will demonstrate the same characteristics as for steel framed construction, improving survival times. However, it is possible that cracking of the concrete may prevent the large deflections associated with tensile membrane action from developing, and spalling may expose reinforcement, reducing membrane strength. At present there is no data available for this.

Compartmentation

The issues of maintaining the integrity of the compartment boundary are the same as for steel framed structures. However, it is more likely that the slab soffit will be bare concrete, increasing the threat from spalled concrete. However, the mass of material associated with reinforced concrete construction generally results in a much slower rate of temperature increase of the structure. This reduces deflections, and thus lessens the risk of large deflections damaging walls below.

Masonry structures

Very little information is available in relation to the performance of masonry structures in fire. In multi-storey buildings, masonry is typically used as a non-loadbearing cladding material, supported by the main frame. In order to avoid collapse of the masonry it is necessary to limit both horizontal and vertical deflections. Some guidance is given in ref () which is largely derived from the tests undertaken on the steel framed building at Cardington. These generally indicated that the floor plate effectively restrained outward expansion of the frame. The maximum movement observed was ???, which recovered to ??? after cooling. There was no apparent damage to the external masonry wall, which comprised continuous, full height blockwork on the end elevations, and a ??? high parapet wall (with windows over) on the sides.

In addition to the (minor) effects of the expanding steelwork, masonry heated on one side is subject to thermal bowing. At Cardington, where the maximum temperature difference across the wall was about ??? the horizontal deflection at mid-storey reached about ???, but this returned to almost zero on cooling.

(Top)

On 4 October Tony O'Meagher wrote ...

There are many failure conditions, but one of the key ones for composite floors in fire would appear to be local structural failure of a slab, which also happens to be breach of compartmentation - perhaps the more important condition as far as fire safety is concerned.

Slab failure would not result in more significant structural collapse unless it completely separated an area of the floor from the cores that are providing lateral stability. Then more significant structural collapse would only occur if the column framing were unable to provide stability to the separated area of the floor.

w.r.t. a further test at Cardington: Verifying the mode of failure of a slab panel certainly has merit in terms of providing confidence in the Level 1 Design Guidance and any extension of it; also it would assist the development of the more detailed computer based analysis methods. The edge beams would need to be protected (as required by the Level 1 Guidance) and the fire compartment would extend over at least a couple of bays so that the assumptions on panels acting in isolation could be confirmed (i.e. slab acts as if it was not continuous after initial heating) and also so that it could be confirmed that there is satisfactory performance at the edges of the panels (i.e. no loss of compartmentation etc.). In terms of loading - increased applied load and/or increased fuel load would probably bring about the failure condition. Protecting the edge beams on the panels means that the test would not endanger the overall stability of the Cardington frame.

w.r.t. the comments regarding floor deformations disrupting evacuation and safe refuges: Designs that call for phased evacuation almost always evacuate the fire floor and one or two floors above on first alarm or shortly thereafter. This evacuation would occur long before any significant deformation of the floor immediately above the fire occurs. Refuges for disabled etc. are typically within core areas and hence would not be directly exposed to fire from below. As far as brigade intervention is concerned the overall stability of the frame needs to be maintained, but they have always had to contend with local failures in slabs and/or significant local deformations.

(Top)

On 5 October Mick Green wrote ...

Continuing the debate ...

I still think it is worth considering controlling deflection for a variety of building scenarios. It is true that means of escape takes place very early on compared to when structural deflections begin for a large majority of buildings. However the following conditions should be considered further.

Disabled people don't like the current solution of refuges in cores (it is not an inclusive solution so it doesn't go down very well) and there are moves being made to create alternatives either in adjacent compartments, in suitably designed corridor areas etc.

What is said about phased evacuation is true but there is the case of progressive evacuation into an adjacent compartment. The objective is to create more time but equally there may be no requirement to evacuate the second compartment if the compartmentation performs to an adequate standard

If a floor is truly a compartment floor the performance requirement depends on how important it is to maintain the operation of the upper compartment in the event of a fire in the compartment below. Reasons could be to address item 2 or for business continuity.

In cases where some of the floor beams are protected then we need to know that the fire protection will perform at large deflections. At the moment this is limited by the capability of the current test rigs. Therefore the assumption at the moment must be that we have to use the current maximum deflection limits until we know that the fire protection will still perform when the deflections are higher

In high rise buildings it can take along time to complete the vertical evacuation and a badly affected floor may have deflected significantly during this time. As we know cores are not always on the edge of a building and there is a need to traverse a series of floor slabs to get to the final exit Major deflections in this scenario would not be very good. Similarly search and rescue can take place at an advanced time in large buildings so some consideration needs to be given. We may decide there is no problem particularly if sprinklers are provided to control this serviceability condition.

I think the main point is that as we go away from traditional prescriptive solutions to performance-based designs then we have to give greater depth of consideration to a whole range of potential failure scenarios. There will no doubt be some simple things that we can recommend once we have examined the non-standard conditions and become more confident by carrying out this wider debate.

If we keep this up we should have a good paper by the next meeting ...

(Top)

August 2001

Prior to the 13 September 2001 meeting further contributions were received ...

E-mails from :

• Barbara Lane
• reply from Ian Burgess
  

On 16 August 2001 Barbara Lane wrote ...

From a structural engineering point of view failure did not occur at Cardington. In terms of the standard fire resistance test is did as there were breaches in the slab. In terms of the life safety requirements of the Building regulations, failure did occur, only because BS 476 recommendations were not met.

Question is does it matter? Do we need to push frames to a total failure, what ever that is, in order to truly define fire resistance/real fire behavior?

The Cardington tests, amongst other real fires, have shown us current fire resistance requirements are protecting structures well in fire. But we still cannot quantify by how much.

Current models, as a result of Cardington, address global behavior only, but there may be a local failure in a compartment wall/floor and therefore in terms of current requirements, failure has occurred.

Can we predict such local effects now? Will failure tests help this to be achieved to a more accurate degree? Or are these failure tests to be more of a "global" nature?

I am worried about this new "failure" theme and how it relates to how people design now. But more importantly how it relates to fire resistance tests and application of such data.

If this failure work is carried out and failure turns out to be say 10 hours longer than the furnace test what does that mean? What happens if it¹s just after current predictions? How will all this be translated into the way things are protected now?

Local failure of slabs means the main ingredient of fire resistance has failed. And whether we like it or not every single product on the market is tested using a fire resistance test, and it is not something we can delete over night. So how do we progress from here and how can failure tests help us achieve this?

How useful is overall failure prediction in terms of life safety and/or property protection?

Can we not use something like probability of occurrence factors and more importantly consequence of failure factors to address overall concerns on the gap between current ratings and total failure.

At Cardington windows had to be forcibly broken to get a flashover in the first place. In other words what can be done to make "failure" happen ­ quadruple the fire duration? And if this works and causes a global failure of the frame, what have we achieved?

Finally if we do have a local failure of a single compartment e.g. beam/column collapse, gap in slab, and structurally it does not affect adjacent structure, is this failure?

As a fire engineer I have to say if the people get out and the fire brigade can do their job safely, no failure has occurred. As a structural engineer is this relevant?

(Top)

Reply by Ian Burgess

Personally I'm reasonably relaxed about this subject, because I think we sometimes get obsessed with semantics, and the word "f*****e" is not doing us any favours.

I think we should see ourselves as moving towards a situation where:

The performance of a building (not just structurally but functionally) should stay within acceptable limits under a range of conditions, and that these limits should be set largely on the basis of the severity of the perceived outcomes - as indeed they generally are. In fire these limits may include (perm any number as appropriate from a much longer list):

• All the life safety & escape issues ...
• Environmental risks outside the building
• Insulation
• Integrity
• Resistance to collapse
• Damage to specialised contents
• Repairability after local or widespread events
• Safety and efficiency of firefighting

A true limit state philosophy should be applied, so that partial safety factors are applied to the loadings and fire conditions (fire load etc..) on the basis of the occurrence statistics and the uncertainty of prediction.

These should be combined with a "natural fire" prediction so that we can get rid of most of the witchcraft (the ISO834 Standard Fire, time-equivalence, fire resistance times ...) that currently muddies the waters.

Structurally the main reason for continuing to do research is that we still don't have good simplified guidance to give engineers about local loss of integrity. On collapse things are getting a little better, but this isn't a major problem in reality.

(Top)

Paper tabled by Brian Kirby in September 2001

Mechanisms of Fire Spread

Conduction

The solid boundaries of a fire enclosure will have one surface exposed to fire conditions whilst the other non-exposed surface will face into the adjacent enclosure/space. An excessive flow of heat from the exposed to the non-exposed surfaces of the boundary elements may lead to transmission of fire to adjacent spaces. Traditionally, fire spread by this mechanism has been referred to as "insulation" failure of the enclosure. Heat may be transmitted from the enclosure by way of direct conduction to the non-exposed side of boundary elements or by indirect conduction through building components which are continuous to outside the enclosure, eg pipes, ducts, beams, columns. Whether the heat conducted to the non-exposed surface causes transmission of fire will depend on the effect such heat may have on adjacent spaces. The heat conducted to the non-exposed surface from the fire enclosure may precipitate fire spread as follows;

• Ignition of the non-exposed surface
• Conduction of heat from non-exposed surface to combustibles with which it has direct contact
• Convection of heat from non-exposed surface to adjacent combustibles
• Radiation of heat from non-exposed surface to adjacent combustibles

It is possible to inhibit this fire spread mechanism through prevention of the above scenarios. However, the conductive heating of the non-exposed surface might need to be considered separately in terms of its potential effect on building occupants.

Convection

The excessive flow of hot gases or flames through openings in the enclosure may cause ignition of combustible items in adjacent spaces. The flow of hot gases from the enclosure may be by way of the fixed openings from the enclosure or openings which have occurred as a result of fire. Traditionally, fire spread by this mechanism is termed integrity failure of the enclosure. In addition collapse of the boundary element, eg due to its failure to remain sufficiently load-bearing under fire conditions, may also permit transmission of fire through excessive convection. Heat flow through openings is one of the most difficult parameters to quantify, particularly in the stage between initial integrity failure and total collapse.

Radiation

The transmission of heat from openings in the enclosure may cause ignition of adjacent combustible items. Heat may be radiated from fixed openings (e.g. doors, windows) or openings which have occurred as a result of fire.

Mass transfer

It is possible that burning fuel items within the fire enclosure may be transferred from the enclosure through fixed or fire created openings. Examples include the projection of flying brands and the flowing of liquid pool fires under doors having no bund protection.

Direct pyrolysis and reaction to fire

Where boundary elements are combustible and continuous outside the fire enclosure, it is possible that pyrolysis may extend beyond the enclosure. Examples include lateral fire spread within the thickness of combustible walls or roofs. Successful fire stopping of such pyrolysis routes will be influenced by the reaction to fire characteristics of the materials present as well as the mechanical stability of the overall system. For example, continuous members extending beyond the enclosure of fire origin may permit fire spread by pyrolysis via some continuous combustible component. Fire stopping may be impaired by local collapse or deformation of the non-combustible part of the system. The collapse of enclosure boundaries may also permit fire to spread by direct pyrolysis.

Download presentation slides

Download illustrations of failure mechanisms and flowchart from BS7974 PD3

(Top)

Paper tabled by Ulf Wickström in January 2004

Comments on calculation of temperature in fire exposed bare steel structures in prEN 1993-1-2
Eurocode 3 – Design of steel structures – Part 1-2: General rules – Structural fire design

The Final Draft of prEN 1993-1-2, December 2003 for calculation of the fire resistance of fire exposed steel structures is out for Final Vote (Januari 2004). I have some specific comments which I would like to draw your attention to. It relates to a procedure which has been modified in comparison to the corresponding ENV on how to calculate temperature in fire exposed bare (uninsulated) steel sections. This procedure or formula is from a commercially as well as from a safety point of view a very important, the single most important in the above standard, as it is in many cases decisive whether a steel structure need to be fire insulated or not.

The formula for calculating temperature in bare steel structures is given in Chapter 4.2.5 in the above standard. It is a simple and well established formula as written in the ENV version of the standard,. However, in the new standard the formula has been changed in two ways. First the theoretical concept of “shadow effects” has been introduced without any references or proofs of test results, and secondly an unnamed factor of 0.9 has been introduced which has no physical explanation what so ever.

In comparison with the preliminary ENV standard this means that if all other parameters remain the same, the required steel section factor may be increased by up to 40% and still calculated steel temperatures would be the same. The higher value refers to common open I-sections. (An increase of the section factor corresponds to proportionally the same decrease in the steel thickness.) Thus the formula in the new standard yields a considerably lower safety level for bare steel structures.

As the calculation procedure most likely predicts considerably lower temperature and thereby longer fire resistance times than standard furnace tests, the classification system of these type of structures becomes evidently inconsistent.

The favourable heat transfer theory mentioned above is only introduced in Eurocode 3. If correct, it should of course have been introduced in Eurocode 1 so that it could have been available for other construction materials as well.

In summary:

In the new draft Eurocode 3 standard for calculating fire resistance of steel structures a very favourable heat transfer formula has been introduced for bare steel sections. Thus

1. A new concept ‘shadow effects’ has been introduced which yields considerably lower calculated steel temperatures. The introduced formula is not tried out and verified in practical tests.
2. In addition a new unnamed and unexplained factor equal 0.9 has been introduced which further reduces calculated steel temperatures. 06 January 2004
3. Required steel thicknesses based on calculations may be reduced by up to 40 % for open steel sections.
4. The formula yields lower theoretical steel temperatures than would be obtained in standard fire tests and thereby longer fire resistance times.
5. The formula is exclusive to steel structures.

(Top)

Reply by Jean-Marc Franssen

Liège, le 20 janvier 2004

This document is a tentative answer to the document "Comments on calculation of temperature in fire exposed bare steel structures in prEN 1993-1-2" dated 06 January 2004 made by Ulf Wickström and contained in the file "UWickstrom_2004.pdf".

Although the author of this answer was one of the member of the draft team of prEN 1993-1-2, this answer by no means represent the opinion of the draft team; it is the sole opinion of the author at the time.

What we think U. Wickström failed to notice is that the boundary conditions for bare steel sections have changed in the prEN's compared to the preliminary ENV's. The emissivity, either called relative emissivity or surface emissivity, has been changed

• from 0.50, see 4.2.5.1 (2) in ENV,
• to 0.70, see 2.2 (2) in prEN.

The energy introduced in a bare steel section is proportional to :

with ksh = 1,0 in the ENV

For an I-section heated of 4 sides, the radiative part of the energy, which is dominant compared to the convective part, is thus proportional to

for the ENV

for the prEN

The following table gives the value of this factor, noted k in the table, for several I-sections of different sizes and shapes. This table shows that the difference in term of radiative flux is by far smaller than 40% and, also, that the modification is not a systematic decrease. The change from ENV to prEN, for the considered sections, is from -7% to +12%. If one considers that the modification in calculated temperature is by an order of magnitude smaller than the modification in the flux (because the re-emitted flux depends on the temperature of the section and this dampens the effect), it may be seen that the difference will finally be marginal. Hence, because we agree with U. Wickström that the formula was "… simple and well established … in the ENV version…", we may assume that the safety level is still appropriate in the prEN version.

This example is only one case in the domain of temperature calculations when simple, well established, engineering solutions that had proved to work for years in the ENV versions have had to be modified, usually in the sense of complexity, in order to finally reach very similar results. In this case, it has been argued that the emissivity of steel cannot physically be equal to 0.50 as was supposed in the ENV, and another equally "physical " factor has had to be taken into account in order not to change the final results, namely the fact that the amount of radiative energy that crosses the box contour of a section cannot be increased when it reaches the perimeter contour of the section.

We have had personally the opportunity to give our opinion on other similar cases for other materials where we also disagreed with the draft team but believe that, at that stage, i.e. so close to the final vote, it is better to stick to the agreement/solution/compromise that has been found on that matter.

Evaluation of the radiative flux on unprotected steel sections This flux is proportional to the factor k calculated hereafter:

(Top)