learning from a structural failure

8/14/2019 Learning From a Structural Failure

1/3

Learning from a Structural Failure

At the South African Instituteof Steel Construction (SAISC),we sometimes get to reviewtechnical reports on variousauthorities opinions as to

what caused the failure of a structure. Weare not in the business of judging peopleor what happened, but rather we look atthese situations to see if there are any les-sons that should be passed on to the steelindustry. That way, at least some goodarises from an unfortunate situation.

What follows is my opinion as towhat happened in a recent failure of astructure. Even if I am wrong as to theexact cause of the collapse, the lessonsdescribed are still applicable to futureprojects.

The major component that failed wasa two-story high truss girder over 36 m(118) long, designed to carry a com-

posite steel and concrete floor about2 m (6-6) above the bottom chord aswell as a light-weight trussed roof atthe higher level. The loads from thecomposite floor were transferred to themain girder through secondary gird-ers and tertiary composite floor beams.The concrete was cast into permanentsteel decking. See Figure 1.

The main girder was analyzed ina very simple stress diagram and themain axial forces determined. The forc-es shown in Figure 2 are my estimates

and they are tweaked to highlightthe lesson.

Top and bottom chord members andall web members were chosen to besquare hollow sections. All the profileswere formed by boxing 200 mm 200 mm angles (L88) of various thick-nesses to form hollow squares. In theinstances where the boxed sectionswould be overstressed, they were to bereinforced by welding channels to theoutside or by welding flats to the insideof the profiles (before boxing).

October 2005 Modern Steel Construction

An inside look at the contributing factors to a

South African structural collapse due to weld failure.

By Spencer Erling

secondary girders

8 @ 4.5 m = 36 m 9 m

newconcrete

columns

existing concrete wall

9 m

tertiary beams

truss

21m

A

B

214.1 kN 214.1 kN230.6 kN

698 kN 685 kN685 kN 648 kN648 kN 582 kN582 kN

ROOF LOADS

SLAB LOADS

+324

2kN

2593 kN2593 kN

Figure 1.Plan view.

Figure 2.Elevation view of main truss, showing the approximated floor and roof loads. The truss

is 6 m (20) high.

What went wrong?In the opinion of the writer, the weld

connecting the first diagonal member tothe top chord of the truss (node A in Fig-ure 2) failed, causing collapse of the wholesystem. Fortunately, no deaths occurred.

It appears as though there were manycontributing causes. To try to understandthe lessons, lets look first from the de-sign point of view and then follow up

with the fabrication and constructionpoint of view.

Design ProblemsIn analyzing the axial forces, eccentric-

ities appear to have been ignored. If welook at the detail of node B (see Figure 3)and consider only axial forces, the theorywould be that the forces acting throughthe centers of gravity of the members

International

Perspective


2/3

of the concrete columndefinitely not atthe face of the column or in the middle ofthe column for this configuration.

Once again, the contractors interpre-tation of the geometry was different fromthe strict axial force assumption of theengineers, as shown in Figure 8.

Whatever geometric approach is tak-en, the eccentricity and resulting bendingmoment needs to be considered in thedesign of the members.

In the case of node A, there is a simplesolution that would have not only elimi-nated the eccentricity but also clearlyensured that the force path of reactionwould be clear to the engineer, eliminat-

ing the need for a judgment call. Thiscould be achieved by lowering the top ofthe column as suggested in Figure 9.

After detailed drawings were submit-ted for the engineers approval, it wasagreed that partial penetration weldswould be used to connect the ends ofthe web members to the top and bot-tom chords (penetration of about 75% ofthe wall thickness of the hollow squaremembers selected). What was not recog-nized by the contracting team is the factthat these partial penetration welds can

never be executed in practice, not even

e = 230 mm

Figure 5. As-built geometry of node B. Note

that other nodes had even worse eccentrici-

ties.

Figure 6.Photograph of node B after the col-

lapse.

Node B

would all act through a common work-ing point.

The engineers designed for the condi-tion in Figure 3. However, the contractorsinterpretation of the design drawings ap-pears to have been similar to Figure 4.

Drawn simply to scale, using the con-tractors interpretation on his detail draw-ings, an eccentricity of about 150 mm (6)occurs which results in a bending mo-ment of about 390 kN-m (290 ft-kips) tobe shared between the members framinginto the node (obtained using a first ordermoment distribution analysis). Whateverthe moment the first diagonal carries, ifthe member is already highly stressed

in axial force alone, the addition of themoment puts the member in trouble: itwould be severely overstressed. The situ-ation was exacerbated by poor workman-ship as shown in Figure 6, which addedmore eccentricity.

Similar eccentricity issues occurredat node A, but the determination of theeccentricities calls for some engineeringjudgment. This is a result of not beingpositive of just where the reaction occurson the rather wide column, as shown inFigure 7. Most engineers seem to agree

that it is likely to be at the quarter point

200 mm, typ.

Figure 3. Strict interpretation of line diagram

at node B.

Figure 4. Contractors interpretation of line

diagram at node B.

e = 150 mm

2593 kN

2593 kN

under good workshop conditions.Consultation at this stage with a

welding expert would have stopped thejob and forced a redesign of the trussgirder, taking into account the transfer ofthe large design forces through the joints.The tubular solution would have beenabandoned.

Construction ProblemsFor practical reasons such as crane

access, lifting capacity in the shop, andtransportation concerns, it was decidedto weld the truss girder together on-site.This single decision had a very seriousimpact on the contractors ability to fab-

ricateespecially to weld the steelworkto acceptable standardsfor two mainreasons:1. Half of the welds to internal members

ends now had to be done in the over-head position.

By its very nature, gravity pulls themolten metal away from the moltenpool being formed, making it almostimpossible for the average welder toexecute these welds competently. Inaddition, the welder has to deal withdrops of molten metal falling on his

helmet, shoulders, etc.



3/3

600 mm

reaction at

center line?

2593 kN

e =170 mm

probable

reaction

location 2593 kN

2593 kN vertical component

share of bending moment

Figure 7.Strict interpretation of the line dia-

gram at node A.

Figure 8.Contractors interpretation of the line diagram at node A.

assumed

reaction

point

Figure 9. One possible solution that would

eliminate eccentricity at node A.

It is for this reason that the SAISCstrongly recommends that overheadposition welds should not be used forstructural welds. The detail should

have been changed to a design withan alternative welding position.

2. An important feature of fabricatingstructural steel under workshop con-ditions is the ability to ensure thatthe work is done to a suitable qualitystandard. By transferring this activ-ity to a site without implementingadequate controls, quality standardswill drop.Lack of quality standards is, in fact,

what happened, with many unaccept-ably poor workmanship errors contribut-ing to the demise of the structure:1. Poor cutting of web members to

length, and poor end bevels2. No preparation at the ends of the web

members to ensure the partial pene-tration welds shown on the drawingscould be achieved.

3.No one ever told the welder that par-tial penetration welds were required.Lacking weld preparations to guidehim in the right direction, he proceed-ed to try to put down 6 mm (1/4) fil-let welds everywhere.

4. Very poor set up of joints. The best

time to inspect penetration welds iswhen the joint has been set up.

5. Large gaps in the joints were reducedby introducing reinforcing bars of un-known quality into the joint to reducethe amount of welding required. Thispractice should never be condoned inany steel structure!

6. Eccentricities were exacerbated bypoor workmanship, as we have seen.

7. No quality plans, weld procedures,

proof of welder qualifications, in-spection, non-destructive testing, orany other methods of ensuring goodquality work were requested by the

professional team nor offered by thecontractor. These items are required interms of the American Welding Soci-ety D1.1 welding code (referenced inSouth African design codes). Theseitems exist to ensure that good qualitywelds can be made.

8. Nobody on the project team recog-nized that special qualifications arerequired for welding hollow squaresections to each other. This was a ma-jor contributor to the collapse. Thewelder who actually did the workwas a reasonably competent workerwhen it came to welding hot-rolledprofiles, but he had not been trainedin the additional implications of weld-ing hollow shapes together.

9. Field-welded splices in the mainchords were very poorly executed andin some cases not done at all.

10.No welding specialistor for thatmatter, any structural steel inspec-torwas used to monitor the (poor)workmanship.As you can see, the list of unaccept-

able fabrication and construction prac-

tices is long. Considering the confluenceof the design and construction errors Ihave described, it should come as no sur-prise that the girder collapsed due to aweld failure.

Spencer Erling is Education Director for theSouth African Institute of Steel Construction(SAISC). This article is reprinted from theFebruary 2005 issue of Steel Construction,the official journal of SAISC.


learning from a structural failure

Documents