The Quebec Bridge Failure

Progress, Volume XII, Issue 5, 1 January 1917, Page 839

The Quebec Bridge Failure

For the second time in the short history of the Quebec Bridge project, disaster has come at the eleventh hour. It will be remembered that nine years ago, when the southern half of the old bridge was completed nearly to the centre of the suspended span as a cantilever structure, the entire half bridge collapsed, this failure being due to insufficient lacing to a main compression member.

In the present structure, investigation has revealed a remarkably complete and exact record of how and why the suspended span fell from its supports on September 11th. This record is found in the condition of the parts left at the truss seats on the lifting girders, and affords a remarkable picture of the whole sequence of failure.

The work of making a detailed examination of the structure to detect any evidence as to the cause of the failure was started within a few hours after the

CAUSES OF THE DISASTER ANALYSED . The collapse of the centre span of the great Quebec Bridge when it was being placed in position is discussed in this article by Mr. Arthur E. Evans. Resident New Zealand Engineer of the Trussed Concrete Steel Co. Ltd., and the cause of the failure clearly shown. - - — D

collapse. Engineers of the St. Lawrence Bridge Co., and outside engineers retained by them, as well as members of the Board of Engineers, took part in these examinations. The results have been unofficially communicated to the American technical press, and an invitation was extended to them for a separate survey. An accurate and fully detailed description of what happened is not

possible, as events followed each other so rapidly in the few seconds from the time the span started until it had pulled loose from the last support. There was no lack of expert eye-witnesses, as the disaster was observed by over a hundred of the most eminent structural engineers of both Canada and the United States. To summarise concisely the rather

voluminous evidence at present at our disposal: A steel rocker casting, by which the weight or the south upstream corner of the suspended span was trans-

ferred to the lifting girder, broke in such a manner that the girder kicked back in a south-westerly-direction from under it. This corner of the span dropped into the water, starting transversal rotation of the whole south end of the span. For an instant there was heavy extra load on the down-stream corner of the span, but the hanger held, and when the tipping had progressed far enough the south-east corner also left its hanging support and dropped into the water. Owing to the weakness of the laterals, the rotation of the south end of the truss was not very largely communicated to the north end, the trusses crumpling at the pin-connected joints. Excepting

Fig. 3. A. —Just before the accident. B. — Southwest corner gone. G. —Failure of east truss. D. —Final plunge. for a slight reflex action at the north-east corner, both corners at the north end hung on until the other end of the span had disappeared under the water; then they dropped practically simultaneously. To understand fully what occurred it is necessary to comprehend clearly the detail of the rocker-joint hearing between the lifting girder and the suspended span. Reference is made to the accompanying photographs Figs. 4 and 5, and to the drawing, Fig. 6. The main features of the bearings are throe low-carbon steel castings and two forged pins. The pins are at right angles to each other and with the castings form

a universal joint. It should be noted that the bearing as shown in photograph Fig. 4 does not represent the actual hoisting condition. Two hitch connections carried the lifting girder while the span was being floated up the river on scows from Sillery Cove, where it had been erected. Then centreing plates were added to ensure perfect centreing of the lifting girder on the bearing when the load was transferred to the hoisting chains. These plates were shop-fitted and bore against a chipped surface on the pin-bracket C of the intermediate casting, Fig. 4. A black streak at D shows plainly where the chipping was done in one of the corners of the casting. The plates and eastings were match-marked, and therefore it was impossible to put them in place until the girders. were actually centred on the intermediate casting. Once in place they prevented movement in a direction parallel with the length of the span. Their use permits one to immediately dismiss any theory that the load was placed eccentrically on the girder and caused it to cant. The condition of the south-west lifting girder after the span had fallen is clearly shown in Fig. 5. The photograph was taken looking in a south-westerly direction. The face of the girder shown is toward the channel, and the hanger is the upstream or westerly one of the pair. It will be noted that the centreing plate which was bolted on at M has been sheared off, and the plate attached to N ripped loose, and lies twisted in front of its former position. The bolts of both N and M were sheared downward and forward at 45°. The plate at 0 has disappeared, the holts having been sheared off vertically while the remaining plate P lies crushed down vertically against the pin, and of four bolts, two sheared vertically and two are intact. The hitch connection lies over the crushed plate. The pin is scored in a diagonal direction, and has been rotated eastward ( i.e. towards the reader) a circumferential distance of 11 in. It is certain from this photograph that the span did not “slip” off the pin. If slipping had taken place, the plate P would have disappeared, the scoring of the pin would not have been diagonal, the pin would not have been rotated and the damage to the angles at Q would not have occurred. The key to the whole evidence on this hanger lies in the condition of the two centreing plates 0 and P (i.e. the west centreing plates). These plates are the only points on any of the hangers where direct vertical action is indicated; all other details show a combination of turning, twisting and sliding. The vertical injury to these two plates must have preceded all other effects. The condition of the plates and bolts of N and M indicate a backward movement of the girder occurring on the east side of the pin simultaneously with the dropping of the truss shoe. The condition of the lower pin also indicates backward movement of the girder and also a crosswise movement of some superimposed burden. The condition of the hitch connection angles at Q indicates the same movement as shown by the pin. The initial steps of the accident are made apparent by the above conditions at the south-west hanger. Something must have broken in the northeast ouarter of the shoe detail (i.e. above Q Fig. 5). It could have been only the intermediate rocker casting (see photograph Fig. 4 and view Fig.. 6). The

fracture most probably occurred near the root of the front lower pin-bracket of this rocker, putting the bearing on the lower pin out of service; and also it is

most probable that the fracture entered the upper pin seat and one of the upper brackets. The enormous load of 1,200 tons concentrated on the edges of the fracture must have caused crushing, tipping of what was left of the rocker, and some backward movement of the lower shoe and lifting girder.

It was incidental to this stage of the disaster that the westerly fragment of the rocker bore down on the two west centreing plates O and P Fig. 5, and in the same action the fragments of the broken rocker were ejected from between its two pins like an orange pip, and, the impulse kicking back the entire swinging girder, the corner of the span fell free, only grazing the lower pin and the cover plate of the lifting girder as it went off.

The breaking of a main truss member —and this possibility has been dismissed by every engineer for obvious reasons, as well as on the basis of the photographic evidence afforded by Fig. 5 and by those who witnessed the fall of the span— have probably carried the end off in a direction parallel with the lower pin. Also the truss continued to hang on at both ends after its members had actually came apart, which proves that the initial failure did not occur in any of the members of the truss itself.

So much for what actually happened. It does not tell us what caused the initial failure. For this we must look to the design of the rocker casting, a sketch of which is given, Fig. 6. What were its loads and stresses? Was it designed with sufficient strength? The entire end support of the suspended

span was designed for a 3,000,000 lb. load at each corner, which figure includes an impact allowance of 20 per cent. At the time of the failure, however, the span was perfectly quiescent, and there can be no question of impact addition at the instant previous to failure, so that the load was really 2,500,000 lbs. There was a possibility, with respect to the load on the rocker at other times, of an increased load due to lifting one corner a trifle ahead of the adjoining corner, but this would not add appreciably to the normal shoe reaction. All the rockers carried the span while on falsework at Sillery Cove, and bore an overload of at least 15 per cent., without failing or showing any signs of failure. This overload was from a 30-ton loco, crane, track and material trains. In addition, it had resisted the longitudinal tipping moment arising from friction on the pin, as the span slid along the lower rocker pin, due to expansion on the falsework. This friction is estimated at possibly 8 per cent.; its overturning lever arm being about 10in. A copy of the St. Lawrence Bridge Co’s stress calculations for the design of the rocker casting shows that it has three parts. The first has reference to the bonding effect on the lower arm of the rocker and assumes that the casting deflects away from the pin just enough to reduce the stress intensity due to bending moment at the root of the bracket to the specified unit stress of 20,000 lbs. per sq. in. The unit stresses for erection material were fixed by the Board of Engineers’ specifications at 20 per cent, higher than those allowed in parts of the permanent structure. The second and third Computations deal with the lower pin and lower shoe casting, using the same method of calculation. This “tempering of the wind to the shorn lamb” might be applied to even smaller

castings and pins with the same results, except that still shorter lengths of loading would be found. As an instance, if we take the width of the rocker bracket

to be 7fin. instead of 9-|in., they would only be three quarters as strong; yet the Bridge Go’s method of calculation would show the same bending stress of 20,000 lbs. per sq. in., i.e., it would show them to be just as safe as the original bracket. This process could evidently be continued up to the point where the bearing stress was the limiting factor, which being viewed superficially, is absurd. These calculations afford a striking example of the question whether a part may be stronger than the whole. Is it possible for a piece to be strengthened by the cutting away of some portion of it? The rocker under consideration would figure out to be strong enough if the longitudinal brackets were cut down to one half their length. Are we justified in saying that the casting, without any such cutting, is -also strong enough? Taking the view that the smaller part is necessarily the weaker, one is tempted to answer yes, but in the case in hand such reasoning is a fatal error. Simply regarded, the “stress calculation” is merely a formal computation to verify minor features of the design. It does not analyse or

approve the design, hut tacitly accepts the design as adequate and deduces a few auxiliary figures from it, and is therefore chiefly significant as being an expression of complete faith in the trained eye and sound judgment of the designer who proportioned the casting. Unfortunately, events have not justified this faith. It will be seen on reference to Fig. 4 that at D the angle fillet had been chipped away to form a hearing for the centreing plate, making this the weakest part of casting as it had to bear the highest stress. The more usual method of calculation, based on the assumption of an uniform distribution of load along the full length of pin gives a bending stress (tension) of 43,000 lbs. per sq. in. at the lower edge (only f in. wide) of the longitudinal bracket. The loss of the span, while it may appear to the lay mind to cast discredit on those responsible for the work, in reality put the remaining parts of the structure to a most extraordinary test and so proved the ability of the designers and builders. The canti-

lever arms deflected between 9 and lOin, under the 5,000 ton load from the suspended span; the application of the load was gradual, its release on the contrary, was sudden. Men on the ends were thrown down, while the vibration lasted long enough for one man to run about 250 ft. toward the anchorage. The known details as to the successive steps in the plunge show that very large overloads must have been borne by the down-stream truss on the south cantilever, and the upstream truss of the north cantilever. Yet the cantilevers stood; and now stand as a silent tribute to the marvellous precision and finish of the shopwork, the ingenious and highly successful erection methods, and to the splendid ability of the responsible engineers. The latest failure can only delay for a short time the completion of the bridge. The suspended span is, after all, of moderate proportions compared with the gigantic cantilevers with which it will connect; and it will not be long before the successful placing of a new suspended span will be recorded. For many of the details, and five of the illustrations we arc indebted to the “ Engineering News” and “Engineering Record.”