Sei sulla pagina 1di 9

Proceedings of the Institution of Civil Engineers Structures & Buildings 160 December 2007 Issue SB6 Pages 317325

doi: 10.1680/stbu.2007.160.6.317 Paper 700049 Received 19/10/2006 Accepted 10/09/2007 Keywords: design methods & aids/ failures/safety & hazards Uwe Starossek Professor of Structural Engineering, Structural Analysis and Steel Structures Institute, Hamburg University of Technology, Germany

Disproportionate collapse: a pragmatic approach

U. Starossek
Dr.-Ing., P.E. corresponds to the alternative-paths method. The possibility of isolating an incipient collapse through segmentationa direct design approach used for the Confederation Bridge, Canada 14,15 is not included in current guidelines. The current paper attempts to make a contribution to nomenclature and procedures. First, it seems useful to distinguish the terms robustness and collapse resistance. Corresponding denitions are proposed. Next, the basis of current reliability-based design codes for general structures is reviewed with particular regard to their suitability to prevent disproportionate collapse. Three deciencies are identied. One of these results from practical limitations in considering the effect of local failure on the remaining structure. An outline is given of how this problem could be solved within a probabilistic framework on the assumption of improved analysis capabilities in the future. At present, the problems raised here seem difcult to remedy within a purely reliabilitybased approach. A framework for a pragmatic design approach, with emphasis on performance-based methods and being applicable to any kind of structure, is therefore suggested. A set of corresponding design criteria is presented. These include requirements, design objectives, design strategies and verication procedures. It is discussed how these design criteria can possibly be specied. Design strategies are based on preventing or presuming local failure. When local failure is presumed, the alternative-paths method and the segmentation method can be considered. These design strategies and methods are compared. The factors inuencing their respective applicability are discussed. 2. TWO DEFINITIONS The term robustness regularly appears in publications and discussions on progressive or disproportionate collapse. Even so, it is used differently and there is no common agreement to date on its exact meaning.4 Two denitions are given in the next two sections. They prove useful for the discussion of design criteria, which is presented later in the current paper. 2.1. Robustness It is suggested to dene the term robustness as insensitivity to local failure, where insensitivity and local failure are quantied by the design objectives, which are part of the design criteria (see section 6 below). According to this denition, robustness is a property of the structure alone and independent of the cause and probability of initial local failure. Starossek 317

Probability-based design methods face certain problems in making structures sufciently collapse resistant and thus preventing disproportionate collapse. These problems are addressed and a discussion is given to outline how they might be overcome both within and outside a probabilistic framework. Denitions for the terms robustness and collapse resistance are proposed, the former being a property of the structure alone, the latter including possible causes of initial local failure. A pragmatic approach for designing against disproportionate collapse is suggested and a set of corresponding design criteria is presented. Design strategies based on preventing or presuming local failure are compared. In addition to the better-known design methods providing specic local resistance or alternative paths, an approach based on isolation by segmentation is examined. It is found that the applicability of the various design methods depends on the design objectives and on the type of structure. 1. INTRODUCTION Research on progressive and disproportionate collapse was stimulated by the Ronan Point failure in 1968. It has intensied since the 1995 bombing of the Alfred P. Murrah Federal Building in Oklahoma City and even more so since the events of 11 September 2001, as documented by a large number of conferences 14 and papers.58 These efforts have not yet led, however, to a consolidated and generally agreed set of nomenclature and procedures. Guidance for the practising design engineer is gradually evolving but is still limited to particular types of structures. The rst code requirements for avoiding the disproportionate collapse of buildings were introduced in the UKshortly after the Ronan Point incident. The current state of British codication is represented by Approved Document A 9 and EC 117. 10 In the US, detailed design rules for buildings 11,12 have been issued, although not for the public sector, and are in use. These guidelines include both indirect (i.e. prescriptive) and direct (i.e. performance-based) design approaches, the latter comprising the specic-local-resistance method and the alternative-paths method. For the design of cable-stayed bridges, the Post-Tensioning Institute (PTI) recommends that the sudden rupture of one cable shall not lead to structural instability and species a corresponding load case loss of cable. 13 Application of such a load case in the design Structures & Buildings 160 Issue SB6

Disproportionate collapse: a pragmatic approach

This is in contrast to a broader denition of robustnessas is given, for instance, in EC 117 10 which does include possible causes of initial failure. Such a broader denition is close to the term collapse resistance, as dened in the next section. It is believed that clarity is served by distinguishing these two properties (which could be named differently if no consensus on a re-denition of the term robustness is reached).

P(C) P(CjLE)P(LjE)P(E)

2.2. Collapse resistance It is suggested that the term collapse resistance be dened as insensitivity to accidental circumstances. Again, the accidental circumstances, which comprise low-probability events and unforeseeable incidents, are quantied by the design objectives. Collapse resistance is a property that is inuenced by numerous conditions including both structural features and the possible causes of initial local failure. The structural system is of particular importance. It would intolerably limit the range of design choices, however, if only those structural systems were permitted that are clearly robust. Nor is such a limitation necessary because a non-robust structure, the system of which tends to promote collapse progression, can possibly be made collapse resistant by other measures such as a particularly safe design of key elements. Furthermore, collapse resistance may not be required for every structure.

where P(E) is the probability of occurrence of E, P(L|E) is the probability of local failure, L, given the occurrence of E and P(C|LE) is the probability of disproportionate collapse given the occurrence of L owing to E. 7 The factor P(C|LE) is not reected in current design codes. The probability of disproportionate collapse thus remains likewise disregarded. Thus, probabilitybased design does not deliver on the task it sets out to accomplish, namely to provide uniform reliability. Even if this is a well-known fact to reliability theorists, it seems to have remained unknown to many practising engineers. The second shortcoming of current design methods is that lowprobability events (i.e. events E for which P(E) is very small) and unforeseeable incidents are not taken into account. Within the scope of a probabilistic design concept, such a simplication is necessary because the supporting statistical data, derived from experience and observation, are unavailable. 17 In the case of a non-robust structure, however, this simplication becomes inadmissible. This follows from the rst reason discussed above: a structure with primarily serial load transfer, for example a high-rise building, is considered. For the sake of argument, the initial local failures shall be mutually exclusive events. The probability of collapse is then in the order of the sum of the failure probabilities of all the constitutive elements, that is of the buildings individual storeys, when it is assumed that, owing to impact forces, collapse is induced by the failure of any one storey. If the number of elements is sufciently large (simply, if the area of attack is large), even very low probabilities of local failure can result in a probability of global failure that is high enough to be taken seriously. The latter statement also holds when the initial local failures are not disjoint events. The third problem with current design methods is that the underlying probabilistic concept requires specication of an admissible probability of failure. The target failure probabilities of probabilistic design codes are usually derived from calibration with previous deterministic design codes. Hence, no new societal consensus seems necessary when probabilistic design is adopted. Considering the extreme losses that can result from disproportionate collapse, however, it might be difcult to reach an informed and true societal consensus on the numerical value of the admissible probability of such an outcomea problem that risks of the type low probability/high consequence are typically up against. 18 This problem could be evaded, but not solved, by not bringing the question to the attention of the public. 4. POSSIBLE IMPROVEMENTS OF CURRENT DESIGN METHODS The rst problem outlined in the previous section is not inherent to reliability theory. As illustrated by equation (1) and further outlined by Ellingwood and Dusenberry, 7 it is in principle possible to account for disproportionate collapse within a probabilistic framework. The difculty arises from practical limitations that appear when actual structural systems are considered. The system response to local failure needs to be considered. That response involves large deformations and displacements, separation of structural elements, falling Starossek

3. THE INADEQUACY OF CURRENT DESIGN METHODS Modern design codes and procedures of verication are based on reliability theory. Actions and resistances are statistically determined on the basis of empirical data. After choosing an allowable probability of failure, the design values for actions and resistances can be computed using probabilistic methods. Such an approach is based on a mathematically sophisticated and, seemingly, sound foundation. It is reected in the design codes by partial safety factors and a series of load combination schemes. If the application of the ensuing code rules is often cumbersome, the design engineer might take comfort in the idea that, by working on a rational mathematical basis, a uniform safety level is reached. Even so, it transpires that such an approach fails with regard to the identication and proper treatment of a potential for disproportionate collapse. There are three reasons for this failure. 14,16 The rst reason lies in the consideration of local instead of global failure. Design equations are usually dened and applied on a local level only (check of cross-sectional forces or element stability). Structural safety, therefore, is likewise accounted for on a local level only. The global safety, that is, the safety against the collapse of the entire system or a major part thereof, is a function of the safety of all the elements against local failure but also of the system response to local failure. The latter inuence is neglected. Different systems will respond differently to local failure. The underlying condition that uniform reliability of a structural system is reached by a uniform reliability of its elements, therefore, is not generally fullled. Such methods when applied to non-robust structures will produce unsafe designs. The problem can be illustrated by expressing the probability, P(C), of a disproportionate collapse, C, owing to an event, E 318 Structures & Buildings 160 Issue SB6

Disproportionate collapse: a pragmatic approach

elements striking other elements below and other kinds of interaction, which all require a fully non-linear dynamic analysis in the time domain. Even the precise modelling of such scenarios is difcult because system properties as, for instance, the behaviour of joints under failure loading are not well established yet. These difculties are compounded by the need to consider many initial failure scenarios and by the fact that, owing to the non-linear dependencies appearing here, small errors in the modelling assumptions can produce large deviations in the computational outcome. Even a deterministic analysis of the system response to local failure poses tremendous difculties. A probabilistic analysis of that response and the computation of global safety would add further dimensions of difculty and seems out of reach of todays analysis resources. If such analysis becomes feasible in the future, one could attempt to consider the inuence of the system response to local failure by an additional partial safety factor on the resistance side of the design equations, thereby maintaining the framework of current design methods. This factor would take a value of one for robust structures and a value smaller than one for non-robust structures. Provisions in some codes are indeed equivalent to such an approach. In that case, however, the reduction of the design value of resistance of non-robust structures is based on judgement rather than on thorough analysis. Such a reduction factor would have to be specied based on parametric analyses and reliability assessments for all the different types of structures covered by the respective design code. The robustness of these structures can be expected to vary widely. A classication of structures would thus be required to assign to each class of structure its respective reduction factor. Moreover, the reduction factor would have to be specied differently for different structural elements of one class of structure according to the respective global consequences of their failure. Such differentiation corresponds to the specic-local-resistance method discussed below. Another possibility would be to pursue a fully probabilistic analysis in a given design situation. The second and third problems outlined in the preceding section are fundamental challenges to a purely probabilistic design approach. If the low probability of an event that leads to local failure can add up to a large probability of global failure, then that quantity needs to be known. Also, if societal consensus on the admissible probability of a catastrophic event cannot be reached, another basic ingredient to a numerical probabilistic computation would be missing. 5. SUGGESTED DESIGN APPROACH It follows from the discussion above that the shortcomings of current design methods can only partly be overcome within the framework of reliability theory. The possibilities of improvement which do exist are not yet explored today and might prove insufcient in the future. Still, guidance is needed on how to design a collapse-resistant structure that is insensitive to accidental circumstances. It is therefore suggested that, for the time being, the following pragmatic approach is used. 16,19 On the one hand, the design methods as specied in current codes are applied. They are based on reliability theory, which is Structures & Buildings 160 Issue SB6

reected in the codes by partial safety factors and load combination schemes. Where necessary and possible, as for instance concerning the risk of ship impact on major bridges, this code-based design is complemented by direct probabilistic analysis and risk assessment. In view of the inconsistencies outlined above, it could be argued that the number of load combinations prescribed by some codes should be reduced because it is exaggerated when compared with the accuracy actually achieved. On the other hand, additional measures are taken with particular regard to collapse resistance. The procedure is further described in the next section. It is not necessarily based on reliability theory but rather on judgement and a decisionmaking process. Emphasis is put on performance-based methods. Structural analyses are carried out deterministically. This approach is called pragmatic because it lacks the stringency of a purely mathematical basis. Nevertheless, it enables the engineer adequately to address the disproportionate-collapse problem in the sense that safety and economy are reasonably balanced and analysis is tractable. It seems that such an approach has tacitly been used in the few cases where disproportionate collapse was considered in the design of actual structures. 14,15 Recent regulations 912 indicate that codication might implicitly be moving towards such a pragmatic approach already. The same holds for the ASCE SEI PreStandard Prospectus 20 that is inuenced by this proposal. 6. DESIGN CRITERIA In the assessment and the design of a structure with regard to its collapse resistance, the following design criteria are of importance (A) (B) (C ) (D ) requirements design objectives design strategies verication procedures.

First, the requirements, particularly the question of whether collapse resistance is necessary, should be claried. The necessity depends on the structures signicance with respect to the consequences of a collapse, including the immediate material and immaterial losses but also indirect effects such as the possible impairment of the infrastructure and of civil and national defence. Another criterion for the determination of requirements is the structures degree of exposure to accidental circumstances such as hazards of war, malicious action and natural disasters. The exposure can be considered particularly high for public buildings, major bridges, and other lifeline structures. If collapse resistance is deemed necessary, the following design objectives must be specied (a) (b) (c) (d ) (e) assumable extent of accidental circumstances assumable extent of initial local failure acceptable extent of collapse progression acceptable extent of damage to the remaining structure applicable safety factors and load combinations.

Design objectives (b), (c) and (d) can be used when testing for robustness; design objectives (a), (c) and (d) can be used when testing for collapse resistance according to the respective Starossek 319

Disproportionate collapse: a pragmatic approach

denitions given above. The following design strategies and methods to prevent disproportionate collapse are mentioned in the literature and have at least partially made their way into design codes (a) prevent local failure of key elements (i) specic local resistance (SLR) (ii) non-structural protective measures (b) presume local failure (i) alternative paths (AP) (ii) isolation by segmentation (S) (c) prescriptive design rules (PDR) The segmentation method is explained in the next section; the other methods are well known. A detailed description is given elsewhere. 16,19 Instead of segmentation, the term compartmentalization was used there. The latter term has also been used by others, 6,21 but in a different sense, denoting a particular kind of residential building structure consisting of small, box-like and monolithically connected reinforced concrete compartments. The benet regarding collapse resistance of such a structure would result from enhanced structural integrity and the formation of alternative paths, not from a deliberate isolation of collapse at dened segment borders (see below). Furthermore, the term compartmentation is used in re engineering. To avoid confusion, the term segmentation is used here. It is of interest to note that the assumable extent of accidental circumstances (design objective (a)) must be specied only if the design strategy prevent local failure is used. Otherwise, the structures collapse resistance is provided through robustness, by presuming local failure, so that a specication of design objective (b) (instead of (a)) is necessary, or through indirect design (PDR), which does not explicitly require (and consider) any design objectives. The prediction of the structural response to local failure requires suitable verication procedures. It is suggested to use deterministic structural analyses, which should be as accurate as possible and consider all relevant scenarios. 16,19 The proper choice of triggering events, out of the many scenarios possibly implied by design objectives (a) and (b), and the admissibility of simplied analysis methods, if necessary, are highly dependent on the structure at hand and require genuine engineering work. 14,15 It is difcult to codify. Still, the development, validation and codication of simplied but admissible verication methods would be a worthwhile undertaking. If the design criteria listed above are considered as a framework to be specied further, note that some of these specications are already being discussed1,17 or are in use. 912,21,22 These efforts need to be continued to arrive at more detailed, generally applicable and consistent criteria. In EN 199117 10 and the General 320 Structures & Buildings 160 Issue SB6

Services Administration (GSA) guidelines, 11 for instance, indirect effects on infrastructure and civil and national defence and the degree of exposure are not yet considered; both documents are concerned solely with buildings. EN 19911 7 10 recommends limiting the extent of collapse progression to a local damage of 15% of the oor area, but not more than 100 m2 in each of two adjacent storeys. The corresponding limits set in the GSA guidelines 11 and in Approved Document A9 are smaller. The acceptable extent of damage to the remaining structurea design objective related to the necessity of immediate or early resumption of usehas not yet expressly been codied. Where applicable design criteria cannot be made available in codied form or be developed from rst principles and reliability theory, they should be agreed upon by the contracting and other affected parties or established by the building authorities. It is anticipated that the design criteria can only partly be developed from rst principles and reliability theory. Quantifying the assumable extent of accidental circumstances related to low-probability and unforeseeable events is particularly challenging, if not impossible. Establishing the acceptable extent of collapse progression can be facilitated by cost/benet considerations which become controversial, however, when human life is to be included. Advice can be taken from the insurance industry that has developed methods and parameters for such considerations. Even so, there will remain the necessity for judgement and a decision-making process when assets, the environment or human life are affected on a large scale. On the other hand, the choices to be made here tend to be fairly transparent. The design criteria for the Confederation Bridge, Canada, for instance, stipulated the acceptable extent of collapse progression to be the equivalent of about a 700 m bridge length out of a total length of 12 910 m. 14,15 The only alternatives would have been not to address the risk of disproportionate collapse at all or to abandon the project. These three choices are clear even to a lay person. A societal consensus or an administrative or political decision is therefore easier when compared with the choice of an abstract quantity such as a safety index even when that consensus or decision is that certain structures remain unbuilt. 7. ISOLATION BY SEGMENTATION For certain structures, the alternative-paths method will reach its limits. The Confederation Bridge, Canada, consists of 43 continuous main spans of 250 m each and shorter approach spans (Fig. 1). If this structure were to be made robust and

New Brunswick abutment

El. 4080 El. 000 (C.G.D.)

El. 5906

El. 4080 00%






Approximate bedrock 1320 10 990 12 910 600

Fig. 1. Confederation Bridge, Canada

Disproportionate collapse: a pragmatic approach


Prince Edward Island abutment

C Navigation L span

collapse resistant by the alternative-paths method, an initial local failure to be considered would be the failure of a bridge pier. This in turn would require designing a prestressed concrete frame with a doubled span length of 500 marguably a vain endeavour even if resorting to catenary action. The design method chosen was to limit the extent of collapse progression by isolating the collapsing sections (Fig. 2). In consultation with the supervising authority, Public Works Canada, the acceptable extent of collapse progression was determined and, based thereupon, the location of collapse boundaries was derived (Fig. 2, pier D and hinge H1). The collapse must not transgress these boundaries; the collapsing part of the structure is thus isolated from the remaining structure. The structure is segmented by the collapse boundaries, which are chosen based on the design objectives and become segment borders. Such an approach requires investigation of the remaining structure for the loads resulting from partial collapse. Special attention has to be paid to the structural elements that form the segment borders (in the considered case pier D and the region between and including hinge H1 and pier A); they isolate the collapse and become the key elements in this approach. They are different, however, from the key elements of the design strategy prevent local failure and, therefore, are referred to as relevant elements instead. One design possibility is to provide these elements with specic local resistance. Verifying that their resistance is adequate, however, may be difcult because of the high loading and because of analytical uncertainties. Both problems can be solved or mitigated by selectively eliminating continuity at or close to the segment borders. By inserting joints, break-away hinges or structural fuses, or by providing plastic hinges, the loading of the segment borders is reduced and analysis is simplied. In the case of the Confederation Bridge, it proved particularly important to interrupt the continuity of the prestressing tendons to allow for an early separation of the falling drop-in girder (located in the centre region of every main span) from the remaining structure. Otherwise, the collapse could progress into the adjacent span (Fig. 2, left of pier D), and then further on. It was attempted to design a structural fuse within the castin-place joint between cantilever and drop-in girder. No secure way of automatically cutting the continuity tendons after the onset of collapse was, however, found and the idea was abandoned. The preliminary design was therefore changed by inserting one additional hinge in every second span. Instead of using monolithic cast-in-place joints, the drop-in girder in those spans was connected to both cantilevers by hinges (Fig. 3). If hinge H2 fails, the drop-in girder, extending between the inserted hinge and hinge H2, will fall and disengage from the

Monolithic connection






Fig. 3. Insertion of hinges: (a) preliminary design; (b) nal design

remaining structure in a predictable way. The disengagement would be forced and dened by the geometry of the hinge corbel (Fig. 4). Verication of the remaining structure was performed for the load impulse acting on the cantilever tip during the fall and disengagement of the drop-in girder. The formation of plastic hinges was deemed acceptable for this load case, and the plastic reserves of the structural system were utilised. Detailed accounts of the disproportionate-collapse investigation and design of the Confederation Bridge are given elsewhere. 14,15 This section on isolation by segmentation will be concluded by some more general remarks. Limiting the consequences of a local failure without consideration of the possible cause or the probability of its occurrence corresponds to the design strategy

H2 D H2 D C B C B

H1 A H1 A

Fig. 2. Limiting the extent of collapse progression by isolating the collapse

Fig. 4. Disengagement of drop-in girder

Structures & Buildings 160 Issue SB6

Disproportionate collapse: a pragmatic approach



presume local failure that is reected in some standards and guidelines already. 9,11,23 Accomplishing this goal by structural segmentation, however, is not yet an option included in structural design codes. It bears resemblance with compartmentation for re protection by means of maximum foreseeable loss (MFL) walls, as required by the insurance industry. Adopting the segmentation method for structural design can lead to the insertion of hinges. Such a measure reduces the degree of statical indeterminacy, and thus the level of continuity. In the case of the Confederation Bridge, the redundancy of the structure (as dened in the next section) was not lessened by this measure because it did not remove any feasible alternative load path, keeping in mind that designing a 500 m span, even though a theoretical option offered by the original structural system, was impracticable. On the other hand, the robustness of the structure, that is, its insensitivity to local failure, was increased. This shows that associating continuity with redundancy and equating redundancy with robustness, even if valid for particular types of structures, is not generally justied. These terms should be carefully distinguished. 8. REDUNDANCY VERSUS SEGMENTATION The existence of alternative paths is referred to as structural redundancy, which means redundancy of the structure with regard to its ability to carry loads. Isolating an incipient collapse by segmentation may require a strengthening or a reduction of continuity at the segment borders. In other words, the segment borders must be able to sustain either large forces or large displacements. 19 (A further possibility for creating a segment border is to provide it with energy dissipation capacity, which requires the ability to sustain both large forces and large displacements. 24 ) The segmentation method, including the consequence of reducing continuity, has been used in the design of the Confederation Bridge. Its application in that case was substantiated by the infeasibility of alternative paths. It may still be preferable, though, even when alternative paths could be provided. Furthermore, the continuity required for the formation of alternative paths may, in certain circumstances, promote rather than prevent collapse progression. This view is supported by eye-witness accounts of controlled demolition experts and re ghters who have observed the collapse of buildings.1 Such observations seem plausible when considering that collapse progression requires interaction, which in turn could mean a certain degree of connectivity, between structural elements. In the light of these considerations, the failure of Ronan Point, an often-cited example of disproportionate collapse, can be interpreted differently. Triggered by an explosion in one of the upper storeys, one corner of the building collapsed over nearly its entire height. The larger part of the building, however, remained undamaged. The progressive collapse of oor slabs has been ascribed to a lack of continuity in the slab reinforcement. On the tentative premise that the overall collapse of a building must be prevented, and contemplating the segmentation approach, such a lack of continuity does not seem so bad after all. Stimulated by the Ronan Point failure, requirements for continuity have been included in building codes in the form of prescriptive design rules (in particular, rules for tying structural 322 Structures & Buildings 160 Issue SB6

elements). These provisions were intended to increase structural robustness. They appear to have been successful against the kind of event experienced in the IRA bombing campaigns. Nevertheless, they may not be appropriate for other kinds of event and other types of structure. In any case, the design objective cannot be achieved if the resulting alternative paths become overloaded. In this context, an observation made by Corley et al.5 concerning the collapse of the Alfred P. Murrah federal building is of interest. If only one main column was immediately destroyed by the bomb blast (one of the possibilities discussed 5 ), it is argued that the two adjacent main columns could have been pulled down by the connections to the falling structural components in-between. This assessment is supported by the fact that the collapse stopped at a main column shortly beyond a discontinuity in the transfer girders top reinforcement (Fig. 5). Two examples where segmentation accomplished by discontinuity has possibly prevented widespread collapse are the Pentagon building and the Charles de Gaulle airport terminal in Paris. The Pentagon Building consists of three building rings, each divided in ve parts separated by expansion joints. The aircraft impact near an expansion joint caused several columns on both sides of the joint to fail. The more affected section, the outer ring on the right of the joint, collapsed while the less affected section, the outer ring on the left of the joint, did not (Fig. 6). A connection might have

Collapse stopped at rebar discontinuity in transfer girder

Fig. 5. Partial collapse of the Alfred P. Murrah federal building (Oklahoma City, 1995)

Fig. 6. Partial collapse of the Pentagon building (2001)

Disproportionate collapse: a pragmatic approach


promoted a collapse progression, since the left section was heavily damaged as well and might not have been able to carry additional loads. The isolation of collapse on the other side of the collapsed section was achieved by strong structural elements which resisted the collapse loads and thus likewise formed a collapse boundary and segment border. The partial collapse of the Charles de Gaulle airport terminal was initiated by the failure of a portion of the roof owing to poor workmanship and deciencies in design. The collapse came to a halt at structural discontinuities on both sides of the collapsing section (Fig. 7). If continuity had been provided, it seems unlikely that the forces which occurred during collapse would have been sustained by the adjacent sections since these sections suffered from construction deciencies as well. The potential value of continuity will not be called into question. It should be kept in mind, however, that continuity can be harmful when the resulting alternative paths are not provided with the strength required to withstand the forces transmitted by continuity. This remark also applies to prescriptive design rules which are based on the idea of providing alternative paths by increasing continuity. If it is impossible, or overly expensive, to provide alternative paths with sufcient strength, the segmentation method (implemented, if necessary, by selectively eliminating continuity) has the advantage. This is also the case if alternative paths (or segment borders) are strong enough, but the corresponding verication proves difcult or unconvincing. The alternative-paths method (and continuity), on the other hand, can be required if the fall of components or debris must be prevented by any means. This applies particularly to cases in which falling parts could strike key elements of the remaining structure because the impact loading produced by such an event is difcult to design for. Such conditions are found in structures with primarily vertical alignment, such as high-rise buildings; they are less typical for horizontally aligned structures, such as bridges or large low-rise buildings. The suitability of the two design methods compared here thus depends on the type of structure and its alignment in space. The alternative-paths method requires an increase of either or both continuity and strength. Segmentation, on the other hand,

can be accomplished by less continuity or more strength (or both). Other differences between these two methods concern the spatial distribution of design measures, the dependency of their efciency on the size of initial failure, and the minimum extent of collapse. The alternative-paths method leads to changes that are distributed throughout the structure; its efciency decreases with an increase in initial failure size; it is therefore preferable for small initial failure size; the extent of collapse decreases with initial failure size. The segmentation method requires changes at discrete locations; its efciency tends to be insensitive to the initial failure size; it is preferable for large initial failure size; the extent of collapse is xed and comparatively large. Both methods can be combined. When the alternative-paths method is used within individual segments, structural robustness is increased for both small and large initial failure sizes (and, thus, for accidental circumstances of both small and large extent). 9. LOCAL FAILURE: PREVENT OR PRESUME? Direct design can be based either on trying to prevent an initial local failure of key elements or on designing for such a case. From each of these two strategies, two design methods are derived, as listed above. If collapse resistance is to be achieved by preventing local failure, the design methods specic local resistance and non-structural protective measures can be considered in order to provide high safety against local failure. These methods do not aim at enhancing structural robustness. On the other hand, if local failure is presumed, the alternativepaths method or the segmentation method (or a combination of both) can be pursued to make the structure robust and to limit an incipient collapse to an acceptable extent. Again, the safe performance of certain relevant elements is crucial and must be veried. In contrast to the key elements of the design strategy prevent local failure, these relevant elements are under better control of the engineer: they are selected by choosing the alternative paths or the locations of segment borders, a design freedom whose magnitude depends on the design objectives. Thus, the number of relevant elements can be comparatively small, particularly when the segmentation method is used where the segment borders are the relevant elements. A further advantage of methods based on presuming local failure, and particularly of the segmentation method, is a more favourable reliabilitycost ratio. 16,19 When it is assumed, for instance, that the relevant elements of the S structure (designed according to the segmentation method) are provided with the same enhanced resistance as the key elements of the SLR structure (designed according to the specic-local-resistance method), the probability of disproportionate collapse will be smaller for the S structure. This is a direct result of the underlying design strategies. Because the failure of any one key element of the SLR structure can induce progressive and disproportionate collapse, the probabilities of failure of these key elements combine to an increased probability of disproportionate collapse of the SLR structure. On the other hand, the failure of a relevant element of the S structure, that is the failure of a segment border, is more likely not to lead to a collapse that transgresses the next relevant element and segment border because local failure has explicitly been considered as a load case for designing that element. Therefore, the probability of disproportionate collapse of the S structure Starossek 323

Fig. 7. Partial collapse of the Charles de Gaulle airport terminal (Paris, 2004) (source: AP Photo; Jerome Delay)

Structures & Buildings 160 Issue SB6

Disproportionate collapse: a pragmatic approach

will arguably not exceed the probability of failure of one relevant element. One might thus choose not to provide enhanced resistance to the relevant elements of the S structure. This advantage is shared, to a lesser extent, by the alternativepaths method. A more detailed comparison concerning reliability and cost is given elsewhere. 24 For these reasons, design methods based on presuming local failure are generally preferable for structures of high signicance or exposure. They allow high safety against disproportionate collapse at relatively low additional costas long as such a design is possible. Moreover, they are more satisfying from a reliability standpoint because the failure probability of their relevant elements is easier to determine than the failure probability of key elements and because their efciency is less dependent on that probability. Uncertainties related to accidental circumstances are altogether irrelevant. Methods based on preventing local failure and, in particular, the design method specic local resistance are of interest for structures that are not particularly suitable for other methods. They can be preferable, for instance, for providing collapse resistance to the primary load-bearing system of slender highrise buildings, even such buildings of high signicance and exposure. 24 Furthermore, such methods can generally be preferable for smaller structures, for structures with few and well-identiable key elements and in cases where the assumable extent of accidental circumstances is small. Again, the various methods can be favourably combined. While the primary load-bearing system of a slender high-rise building is possibly made collapse resistant through SLR, the design of its secondary load-bearing system can be based on providing APs or by segmentation (S). 24 When all three methods are combined in the design of the same system or subsystem, disproportionate failure can be averted for small (SLR), medium (AP) and large (S) extent of accidental circumstances. 10. CONCLUSIONS Clearer and more practical denitions are arrived at when the term robustness is distinguished from the term collapse resistance, the former being a property of the structure alone, the latter including possible causes of initial local failure. In regard to progressive and disproportionate collapse, non-robust structures are of particular concern and require specic consideration. The necessity of such consideration follows from an inspection of current design methods, which are based on reliability theory. Because of fundamental difculties and owing to the number and complexity of inuencing factors which appear after failure initiation, a purely probability-based design of such structures where disproportionate collapse comes into play seems impracticable for the time being. Possibilities of future improvement within a probabilistic framework have been outlined. To give more immediate guidance, a pragmatic design approach is proposed in which probability-based design procedures, as described in the codes or in direct application of reliability theory, are complemented by additional assessment and design measures with particular regard to disproportionate collapse. In that assessment, structural analyses are carried out deterministically. Emphasis is put on performance-based measures. The suggested approach is applicable to any kind of structure. 324 Structures & Buildings 160 Issue SB6

Even if a purely probability-based design turns out to be impracticable, reliability theory has an important role to play in a more detailed comparison of the design methods discussed here and in determining design criteria such as exposure, assumable extent of accidental circumstances or initial local failure, and safety factors. Nevertheless, these and other design criteria might be difcult to quantify and in the end be left to judgement. The dependency on the accuracy of such judgement, and the importance of reliability theory, is relatively high when using the design strategy prevent local failure, less so for the alternative-paths method and minimum for the segmentation method. Yet other design criteria need to be stipulated in a decision-making process. The choices to be made in that process, for instance, concerning the acceptable extent of collapse progression, are relatively transparent so that an informed societal consensus should be possible. The alternative-paths method, and the prescriptive design rules based on that idea, should be applied with discretion. Forces should be determined based on the overstrength of elements introduced for continuity and the force transfer should be checked down to the foundation. An alternative way of designing for a presumed initial local failure is provided by the segmentation method. Segmentation is accomplished by segment borders that are able to sustain either large forces or large displacements (or both), which may require a strengthening or a reduction of continuity at the segment borders. For certain structures, segmentation is the more suitable approach to prevent disproportionate collapsea fact that has found little attention in the structural engineering community so far. If this option has nearly been overlooked, one reason might be that the terms continuity, redundancy and robustness are intuitively equateda tacit assumption that is justied at best only for particular types of structures. The adequacy of a particular design method depends on the design objectives, and on the type of structure and its alignment in space. Impact loading produced by falling components is more of a concern for vertically aligned structures such as high-rise buildings, not so much for horizontally aligned structures such as low-rise buildings and bridges. The need to provide continuity to prevent such impact loading differs correspondingly.

REFERENCES 1. MULTIHAZARD MITIGATION COUNCIL OF THE NATIONAL INSTITUTE OF BUILDING SCIENCES. Prevention of Progressive Collapse. Workshop, Chicago, 1012 July 2002, MMC Report, Washington, D.C., 2003. 2. AMERICAN INSTITUTE OF STEEL CONSTRUCTION. Steel Building Symposium: Blast and Progressive Collapse Resistance. New York, 45 December 2003; Proceedings AISC, Chicago, 2003. 3. PRECAST/PRESTRESSED CONCRETE INSTITUTE. Blast and Progressive Collapse Resistance of Precast and Prestressed Concrete Structures. Workshop, Chicago, 21 April 2004, Proceedings PCI, Chicago, 2004. 4. JOINT COMMITTEE ON STRUCTURAL SAFETY. Robustness of Structures. Workshop, British Research Establishment, Watford, 2829 November 2005, JCSS, International Starossek

Disproportionate collapse: a pragmatic approach










Association for Bridge and Structural Engineering. Available online at CORLEY W. G., SOZEN M. A., THORNTON C. H. and MLAKAR P. F. The Oklahoma City Bombing: Improving Building Performance Through Multi-Hazard Mitigation. Federal Emergency Management Agency Mitigation Directorate, Washington, DC, FEMA Report 277, 1996. CORLEY W. G., MLAKAR P. F, SOZEN M. A. and THORNTON C. H. The Oklahoma City bombing: summary and recommendations for multihazard mitigation. Journal of Performance of Constructed Facilities, 1998, 12, No. 3, 100112. ELLINGWOOD B. R. and DUSENBERRY D. O. Building design for abnormal loads and progressive collapse. Computer-Aided Civil and Infrastructure Engineering, 2005, 20, No. 3, 194205. HAYES JR J. R., WOODSON S. C., PEKELNICKY R. G., POLAND C. D., CORLEY W. G. and SOZEN M. Can strengthening for earthquake improve blast and progressive collapse resistance? Journal of Structural Engineering, 2005, 131, No. 8, 11571177. THE STATIONERY OFFICE. The Building Regulations 2000: Structure: Approved Document A (2004 edition). Ofce of the Deputy Prime Minister, London, 2006. COMITE EUROPEEN DE NORMALISATION. Eurocode 1, Part 17. BS EN 199117 General ActionsAccidental Actions. CEN, Brussels, September 2006. US GENERAL SERVICES ADMINISTRATION. Progressive Collapse Analysis and Design Guidelines for New Federal Ofce Buildings and Major Modernization Projects. GSA, Washington, DC, 2003. US DEPARTMENT OF DEFENSE. United Facilities Criteria Design of Buildings to Resist Progressive Collapse. January 2005, UFC 402303. POST-TENSIONING INSTITUTE. Recommendations for Stay Cable Design, Testing and Installation, 4th edn. Cable-Stayed Bridges Committee, PTI, Phoenix, 2001.

14. STAROSSEK U. Zum progressiven Kollaps mehrfeldriger Bruckentragwerke. Bautechnik, 1997, 74, No. 7, 443453. 15. STAROSSEK U. Progressive collapse study of a multi-span bridge. Structural Engineering International, 1999, 9, No. 2, 121125. 16. STAROSSEK U. Progressiver Kollaps von Bauwerken. Betonund Stahlbetonbau, 2005, 100, No. 4, 305317. Available online at Index.htm. 17. ALEXANDER S. Application of standard events in design against disproportionate collapse. IABSE Symposium Structures and Extreme Events, Lisbon, Portugal, 1417 September 2005. 18. VAN BREUGEL K. Storage system criteria for hazardous products. Structural Engineering International, 1997, 7, No. 1, 5355. 19. STAROSSEK U. and WOLFF M. Design of collapse-resistant structures. Robustness of Structures, Workshop, British Research Establishment, Watford, 2829 November 2005, Joint Committee on Structural Safety, IABSE. Available online at 20. STEERING GROUP OF THE SEI PROGRESSIVE COLLAPSE STANDARDS AND GUIDANCE COMMITTEE. A PreStandard Prospectus: Robustness and Collapse Resistance for Buildings. ASCE, Structural Engineering Institute, 2006. Available online at progressivecollapse06.pdf. 21. AMERICAN SOCIETY OF CIVIL ENGINEERS. Minimum Design Loads for Buildings and other Structures. ASCE, Reston, 2002, ASCE 7-02. 22. AMERICAN SOCIETY OF CIVIL ENGINEERS. Structural Design for Physical Security. ASCE, Reston, 1999. 23. COMMISSION OF THE EUROPEAN COMMUNITIES. EN 1990:2002 Eurocode 0. Basis of Structural Design. CEN, Brussels, 2002. 24. STAROSSEK U. Progressiver Kollaps von Bauwerken. Betonkalender 2008, Ernst & Sohn, Berlin, Germany, 2008.

What do you think? To comment on this paper, please email up to 500 words to the editor at Proceedings journals rely entirely on contributions sent in by civil engineers and related professionals, academics and students. Papers should be 20005000 words long, with adequate illustrations and references. Please visit for author guidelines and further details.

Structures & Buildings 160 Issue SB6

Disproportionate collapse: a pragmatic approach