Causes of Structural Failures with Steel Structures

Göran AlpstenStålbyggnadskontroll AB, Stocksund, Sweden (formerly Adjunct Professor in Steel Construction, The Royal Institute of Technology KTH, Stockholm)

Contact: goran.alpsten@stbk.se

Abstract

This paper is based on the experience from investigating over 400 structural collapses, incidents and seriousstructural damage cases with steel structures which have occurred over the past four centuries. The causeof the failures is most often a gross human error rather than a combination of “normal” variations inparameters affecting the load-carrying capacity, as considered in normal design procedures and structuralreliability analyses. Human errors in execution are more prevalent as cause for the failures than errors inthe design process, and the construction phase appears particularly prone to human errors. For normalsteel structures with quasi-static (non-fatigue) loading, various structural instability phenomena have beenobserved to be the main collapse mode. An important observation is that welds are not as critical a cause ofstructural steel failures for statically loaded steel structures as implicitly understood in current regulationsand rules for design and execution criteria.

Keywords: Steel structures; static loading; fatigue; structural failures; collapses; incidents; structuralinstability; gross human error; execution; construction phase; welds; regulations.

1 Introduction

This paper is based on the experience frominvestigating over 400 failures, incidents andstructural damage cases with steel structures inbuildings, bridges, chimneys and other civilengineering structures. These accidents haveoccurred in Sweden and elsewhere over the pastfour centuries.

The experience from being a consultant andinvestigator of accidents with steel structures willbe summarized with a few examples beingdiscussed more in detail. Because of the sensitivenature of many such accidents the reference formost cases will be made here in such a way thatidentification should not be possible.

The purpose of the discussion is to identify whatseems to be the major problems with regard tostructural safety of steel structures, in order toreduce the risk for similar events to happen again. It should be emphasized that the background forthe paper is not a research program but rather the

result of the writer being commissioned forinvestigating a large number of structuralcollapses, incidents and structural damages whichhave occurred in Sweden and elsewhere. It couldbe argued whether the observations discussed inthis paper are representative and whether theconclusions drawn in the paper are of a generalnature. However, considering the large number ofcases, constituting a fair portion of the accidentswhich have occurred with steel structures in theregion, it is believed that the findings should havesome general significance.

2 Causes of structural failures

Most of the failures with steel structuresinvestigated are with (quasi-)static (non-fatigue)loading. This is a reflection of the fact that mostcivil engineering structures are indeed actingunder service conditions where variations inloading is not great enough to cause fatigueproblems.

1

The discussion of causes of structural failures withsteel structures will focus here on three differentconditions: failures for (quasi-)statically loadedsteel structures in the erection/constructionphase, such failures occurring in the service phaseand finally fatigue damage.

2.1 Static loading, construction phase

Of the failures occurring for normal steelstructures with static loading, a large proportionoccur during the construction phase. The mostcommon cause is incorrect erection procedureswhich do not consider the various instabilityproblems during all the stages of erection andinstallation.

2.1.1 Floor deck in commercial building

A well-publicized collapse during construction of athree-story commercial center building occurredin Stockholm in 2008, see Figures 1 through 4 [1]. This case is of interest here because it is a lucidexample of gross human error. This failure ishowever not representative for most casesinvestigated, in that the error was in thestructural design, not the execution andworkmanship.

Figures 1 and 2 show a view of the floor deckbefore and after the failure. The structuralmember causing the failure is the welded 1400mm high I-girder visible in the center of Figure 2,introduced in the structural frame in the planningof the building because the column to the right ofthe I-girder had to be relocated from the normalsystem grid, the I-girder providing intermediatesupports for deck girders in the perpendiculardirection.

The I-girder was designed using a computerprogram. When transferring the dimensionsobtained from the program to the drawings, thethroat thickness of the fillet welds was confusedfor the web thickness. Thus the web thickness asspecified on the drawings of the 1 400 mm high I-girder was 7 mm, instead of at least 12 mmrequired from a correct design. However, afurther gross human error was introduced in thatno web stiffeners were provided over the right-end column support.

Figure 1. Commercial center building underconstruction, view half an hour before structuralcollapse

Figure 2. Same view as in Figure1, after collapse

Figure 3. Welded 1 400 mm high I-girder whichinitiated the structural collapse shown in Figure 2

2

Figure 4. Local buckling of the web of the I-girderover the column support, where web stiffeners arelacking

Figure 3 shows the deformed I-girder after thecollapse, the fallen perpendicular deck girders andprefab deck elements fallen to the ground.

The typical local buckling behavior of theoverloaded web of the I-girder over the columnsupport where web stiffeners are lacking is visiblein Figure 4.

When simulating the collapse behavior in theinvestigation after the failure all details includingthe action of the web of the critical I-girder wereaccurately predicted.

Although the consequences of this collapse wereindeed very serious, with one person killed and atleast two persons injured, a cynical person might argue that it was good luck that two gross humanerrors were introduced in this case. Had only oneof them been at hand it might be that the collapsehad occurred on the occasion when thecommercial center building was opening with 50 000 persons occupying the complete building.

2.1.2 Roof trusses in sports arena

The collapse during construction of the steelstructure shown in Figures 5 and 6 isrepresentative of a class of failures where humanerror has been made in failing to provideconditions for full stability of the structure duringthe complete construction phase. The building inthis case is a track and field sports arena with rooftrusses spanning 50 m.

The subcontractor for the steel structure haderected the steel structure including temporarybracing elements of the upper chord of the rooftrusses as required to provide lateral support forthe trusses in the construction phase until finallateral support is provided by the roof deck.

However, the next subcontractor, engaged for theerection of the roof structure consisting oflaminated wood purlins and a plate roof, hadproblems due to the temporary bracing elementshindering movements of the crane used to lift theroof material. Negligent of the important purposeof the temporary bracing elements, thesubcontractor loosened them one by one, untilthe two trusses now without intermediate lateralsupports collapsed due to lateral-torsionalbuckling. The action to remove temporary bracingwithout reflection should be considered a gross

Figure 5. Roof trusses collapsed duringconstruction when roof structure is being erected

Figure 6. Joint in collapsed roof trusses fromFigure 5, with some fractured welds

3

human error. Such behavior could be eliminatedby proper instructions and supervision.

In some joints of the trusses the welds werefractured after the collapse, see Figure 6. It is incases like this easy to jump to the conclusion thatdefective welds are the cause of the collapse. Infact, in the case of Figures 5 and 6 an experiencedweld inspector had made up an initial report thatdefective welds in the joints of the trusses werethe probable cause of the collapse. However, it iseasy to verify that these 50 m trusses when noside support is provided will fail by lateral-torsional buckling when subjected to their ownweight only. The fractured welds are aconsequence of the trusses being subjected in thefall to unforseen weak axis bending, possibly alsoby dynamic effects when hitting the ground. Thenormal collapse mechanism in this class ofcollapses is instability in some form, such ascolumn buckling, lateral-torsional buckling or localbuckling.

2.2 Static loading, service phase

2.2.1 Conveyor belt gallery

The most serious of the collapse casesinvestigated by the writer in terms of theconsequences is that of a conveyor belt gallery,Figure 7. In this case two persons were killed andgreat economical losses were made due to a longinterruption in the production. Part of the steelstructure fell down some 30 m to the ground, withinstability effects to some members and fracturesin others, Figure 8.

The main reason for the collapse was corrosiondamage internally in a bolted joint carrying a steelcross-beam in which the conveyor belt gallery washanging at its upper end in a process building, seeFigure 9. Only a small portion of the cross-sectional area of the bolts were remaining at thetime of the collapse. This collapse could beattributed mainly to the gross human error inneglecting to implement a rational inspectionprogram for regular checks of the status of thisstructure in a very corrosive environment. Theinappropriate design with the cross-beam hangingfrom supports instead of resting on the supportsmay also be considered a human error.

Figure 7. Collapsed conveyor belt gallery

Figure 8. Remains from collapsed conveyor beltgallery seen from a 30 m level in process building

Figure 9. Detail of cross beam carrying theconveyor belt gallery in the process building, withbolts loaded in tension and damaged by corrosion

4

2.2.2 Floor deck in symphony hall

Another example of a failure of a steel structureduring service concerns a floor deck in the hall of abuilding designed for performances and rehearsalsof a symphony orchestra, Figure 10. At the time ofthe collapse the premises were used for a rockconcert with 700 youths jumping on the floor inpace with the rock music, instead of the limitednumber of 250 persons sitting listening to asymphony as assumed in the structural design. When a 4 x 2 m floor section collapsed due to loadaction on the floor being 2 to 3 times greater thanforeseen for the intended service, about 50 youthsfell to the floor below, of which 29 had to betreated at the hospital for broken limbs and otherinjuries.

The special floor deck in the hall consists ofwooden slabs resting on steel trusses, which in

Figure 10. Collapsed floor deck in concert hall

Figure 11. Collapsed steel truss beams in thefloor deck of Figure 10

turn may be placed at various heights on steelcolumns below. This is to make possible various arrangements on the floor, such as preparing anorchestra pit. In the collapse the overloadedtrusses fractured in some welds and somemembers in the trusses collapsed due toinstability, Figure 11. Again, the main cause of thecollapse is a gross human error, in this case byaccepting the premises to be used for activitiescausing much heavier loads than specified for thedesign. Too small welds in the trusses may alsohave contributed to the failure.

2.2.3 Roof over walkway

A further failure during service which is ofparticular interest with respect to the normalbehavior of welds is shown in Figure 12. Anumber of cantilevered IPE beams carry a roofover a walkway along the facade of a building. The beams are attached by welds to steel platesanchored in the concrete wall of the building.

The walls of the building were prepared withinsulation and bricks, and openings were left forattaching the cantilever beams as indicated inFigure 12. A welder was then asked to performthe weld attachment of the cantilever beamsinside these openings. The welder made a grosshuman error in that he did not refuse to carry outthe welding because of too bad accessability,especially for the important weld to the upperflange of the beams subjected to tensile forcesfrom vertical loads on the roof. Instead he tried to

Figure 12. Detail of fractured weldedattachment of cantilever beams for a roof overwalkway

5

make the best possible workmanship under thepresent circumstances. This resulted in a weld atthe upper flange of the beams to be completelyfilled with slag inclusions.

The roof over the walkway performed withoutproblems until the following winter. Then piles ofsnow from the higher roof of the building felldown on the lower roof over the walkway causingthe complete roof with several cantilever beamsto collapse. It could be argued whether thestructural designer in this case had also made agross human error since the actual snow load wasmuch greater than he had anticipated. The rooflocated at a lower level than the roof of thebuilding was designed for normal snow loadincreased by 20 percent (“snow pocket”). Whilethis assumption may be much on the unsafe sidefor the actual conditions with today’s knowledgeand rules for loads in snow pockets, this wasaccording to common practice at the time.

Of particular interest in the behavior of thebreakdown shown in Figure 12 is that the fracture started as expected in the weld filled with slaginclusions at the upper flanges with tensile forces.However, when the crack propagated into theweb of the beams the fracture chose to enter thebase material in the beams away from the weld. This was the case even though the weld along theweb was far from perfect (it has not even beende-slagged as may be observed in Figure 12).

2.3 Fatigue damage

Fatigue damage may occur in steel structures, butis not the most common cause for failures in civilengineering construction. This is related to thefact that most civil engineering structures are notheavily affected by varying loads causing fatigue,except those loaded by moving vehicles such astrains, trucks and cranes, and dynamic windloading. Of the total number of structural failures,incidents and structural damage casesinvestigated, roughly 10 percent relate to fatiguedamage. In many of these cases the causes aregross human errors, for instance, the steelstructure has been designed without consideringfatigue action at all, or the structures are notproperly detailed for the actual fatigue conditions.

From a safety standpoint it is relevant that anyfatigue cracks developing in a steel structure mayoften be observed and repaired before the cracklength will become critical and cause a completecollapse or breakdown of the structure.

Many of the fatigue cases investigated are withtall steel chimneys, where load action from vortexshedding may provide the variation in stress(stress range) necessary to cause a fatigue crack toinitiate and propagate. Today design methodsexist to model the fatigue action from oscillationsdue to vortex shedding and a detailed fatigueanalysis considering the quality of the weld toes,but this has not always been the case.

2.3.1 Steel chimney A

Figure 13 shows an example of a 90 m high steelchimney where fatigue cracks were observed after12 years of service. The cracks were made visibleby magnetic-particle testing, Figure 14. Figures 15and 16 show an example of a detected crack atthe edge between weld passes at the top of thevertical stiffeners, and after grinding down toabout half the thickness of the shell plate,respectively. The incident of fatigue cracks in thischimney was to a great extent caused by a grosshuman error behind the regulations at that timedefining load action related to vortex shedding.

Figure 13. 90 steel chimney where fatigue cracksdue to vortex shedding were detected at the base

6

Figure 14. Magnetic-particle testing of welds atthe base of the chimney in Figure 13. The lowerpart of the chimney made of weather-resistantsteel has been blast-cleaned before testing

Figure 15. Fatigue crack detected by magnetic-particle testing at the toe between two weldpasses around a vertical stiffener at the base ofthe chimney in Figures 13 and 14

Figure 16. Fatigue crack indicated by magnetic-particle testing after grinding the detail in Figure15 down to half the thickness of the shell plate

A number of fatigue cracks in the shell and inanchor bolts could be detected in this chimney. Ascorrective actions the base of the chimney wasrepaired by adding an outside skirt with newanchor bolts to unload the cracked parts, and inaddition the chimney was fitted with helicalstrakes (Scruton strakes) at the top of thechimney. This chimney has now been in servicefor many years without further fatigue problems.

2.3.2 Steel chimney B

A detail from the base of another 90 m high steelchimney where several fatigue cracks wereobserved a few months after installation of thechimney is shown in Figure 17 [2]. Most of thecracks were located in the weld toe at the top ofthe vertical stiffeners, and in some places thecracks were visible even to the naked eye, Figure18. The position of the cracks at the lower weldtoe as evident in Figure 18 is due to the unsuitabledesign of the stiffeners, causing an even greaterstress concentration effect on this side of the weldas compared to the upper weld toe to the shell.

In this case the cause for the undue fatiguedamage after a few months of service was acombination of two gross human errors. A massdamper at the top of the chimney was incorrectlyinstalled, which did not reduce large-amplitudeoscillations at vortex shedding as expected, and critical details were badly designed, causingunnecessarily grave stress concentrations at theweld toes.

Figure 17. Base of 90 m high chimney withmalfunctioning mass damper at the top

7

Figure 18. Example of fatigue crack at a stiffenerat base of chimney shown in Figure 17, in this casevisible in the cracked paint

The structure was rectified by weld repair of thecracks, correcting the malfunctioning massdamper and finally also providing a supplementarydamping device [3].

2.3.3 Steel frame supporting hydraulic press

A final example of fatigue damage occurred in asteel structure is shown in Figure 19. A frame

Figure 19. Fatigue cracks in two parallel hot-rolled beams HEB1000, initiated at the weld toe ofthe fillet weld to the cover plate at the lowerflange and then propagating up to half the heightof the web of the beams

consisting of four rolled beams HEB 1000 carries aconcrete floor with an 800-ton hydraulic press.

The beams are connected two and two via coverplates welded to the bottom flange of the beams. Cracks have initiated from the weld toe at thecover plate and propagated into the lower flangeof the two beams in Figure 19 and then furtherinto the web, in one of the beams up to half theweb height.

Apparently the initial design of this steel structurehad not considered any fatigue effects at all. Acheck analysis of the structure for the actualfatigue loading conditions revealed that the safefatigue life of the structure had indeed beensurpassed. The only surprising thing about thisincident is that the steel frame had not collapsedcompletely. The fact that the structure was still inplace with near critical crack lengths is probablydue to arch action in the slab.

3 Conclusions

In this paper a few examples of failures withdifferent types of steel structures have beenreviewed.

A general conclusion is that the cause of failuresmost often is a gross human error, in a few cases acombination of two gross human errors. Thus,gross human error is the main cause for collapsesof steel structures, rather than a combination ofunfortunate variations in parameters affecting theactions and response of the structures, as may beconsidered in the probabilistic and semi-probabilistic design of structures [4, 5].

Human errors in execution of steel structures aremore prevalent as cause for the failures thanerrors in the design process. Human errors appearparticularly serious in causing collapses fromstructural instability during the constructionphase.

For normal steel structures with (quasi-)static(non-fatigue) loading, various structural instabilityphenomena have been observed to be the main collapse mode.

The general experience with failures due tofatigue in modern welded steel structures is thatfatigue cracks almost invariably initiate at the toes

8

of a weld, that is, at the transition between weldand base material (Figure 18), or at any edgesbetween individual weld passes (Figure 15). Thus,the important criteria for such structures are thequality of the weld toes. Visual inspection, ifnecessary supplemented by magnetic-particletesting, is the most efficient way to ensureadequate fatigue properties in the production ofsuch structures. This is true also for inspectingexisting steel structures with fatigue loading.

Based on experience from failures, of which only afraction and then mostly with major structures arepresented in the literature, several changes wouldbe required in regulations and rules for the designand execution of steel structures. One importantmatter regards the evaluation of welds. Thecurrent emphasis on internal discontinuities inwelds should be pared with a more strictevaluation of external discontinuities, inparticular, for welded structures with fatigueloading.

Since most failures in steel structures are causedby gross human errors, in particular, in theconstruction phase, it seems imperative to extendin regulations and rules the requirements forcompetence of the individual performing checkingof design and inspecting execution of steelstructures. In the current European standard forexecution of steel structures, EN 1090-2 [6], thereis an exemplary requirement for 100 percentvisual inspection of welds. However, there are nodetailed and relevant competence requirementsfor the individual performing such tasks. Also,since other deficiencies than weld discontinuitiesare of great concern with respect to the safety ofsteel structures, then it seems important to widenthe competence of the inspectors to cover notonly the welds of the structure but the completetechnology of steel structures. An attempt in thisdirection has been made by the Swedish industryfor steel construction with a program for trainingand competence evaluation of supervisors andinspectors engaged in steel structures [7].

It is important that lessons be learned fromstructural failures in a more coordinated andguided way than hitherto, in order to makeeffective use of manpower and other resources inthe design and execution of structures, and to

improve the safety of structures by avoidingsimilar events to happen in the future. In thisrespect much can be learned from the aviationindustry, “black-box thinking” [8]. Instead oftrying to hide the information on structuralfailures we should consider every collapse orincident a learning opportunity for improving thedesign and execution of structures.

4 References

[1] Alpsten, G. Confidential technical reportreferred to in the judicial proceedings of thecollapse in question, StålbyggnadskontrollAB, 2008.

[2] Tranvik, P., and Alpsten, G. DynamicBehaviour Under Wind Loading of a 90 mSteel Chimney, Alstom and Stålbyggnads-kontroll AB, March 2002.

[3] Tranvik, P., and Alpsten, G. StructuralBehaviour Under Wind Loading of a 90 mSteel Chimney, Wind & Structures, Volume 8,No. 1, January 2005.

[4] Alpsten, G. On the Use of ReliabilityAssessment Methods for Structural Design,Euro-SiBRAM Colloquium, Prague June 2002.

[5] Alpsten, G. Uncertainties and Human Errorsin the Design and Execution of SteelStructures, IABSE Workshop on Ignorance,Uncertainty and Human Errors in StructuralEngineering, Helsinki 15-16 February 2017.

[6] Execution of steel structures and aluminiumstructures - Part 2: Technical requirementsfor steel structures, EN 1090-2:2208+A1:2011.

[7] Alpsten, G. Handbok TR-stål/N (HandbookTR-steel/N, in Swedish), Stålbyggnads-kontroll AB, December 2012.

[8] Syed, M. Black Box Thinking, John Murray(Publishers), London, 2015.

9