Bridge Collapse Caused By Design Flaw, Not Maintenance

Captain's Quarters ^ | Jan. 15, 2008 | Ed Morrissey

[Attention Surplus Disorder]

“Look at the edges of the paired hexagonal gusset plates midway between the diagonal and the upper box beam. Those edges (both plates in both photos) are “kinked” about halfway between the diagonal and the upper beam.

And they were bent at as least as far in the past as ca 2003...”

I’m not an engineer (nor did I sleep at a Holiday Inn) but that kinky gusset picture is nothing if not third-degree alarm bells going off. As you say, those plates are clearly showing forces acting both in vector and in amount the gussets are/were inadequate to counter. It doesn’t get any simpler.

[TXnMA]

Here is a rude, crude little sketch based on what I see on page 4 of the NTSB PDF.

The deflections appear to me to be all in the same direction -- as if the deck and upper span beam had shifted somewhat to the east relative to the verticals.

OTOH, after re-examining the overhead view, I now think that the roadway curve was too far south to have imparted any live lateral thrust to the L10/U10 section...

At any rate, jeffers, your suggestion that downward vertical deflection in mid-river could have caused these gusset plate "bends" doesn't seem to fit: That would have put these particular edges in tension.

So far, the deflection looks lateral to me -- but don't ask me how it could have happened... ;-)

[Abigail Adams]

You are welcome! There was heavy construction material piled up on the south end on the west side. I’m guessing that was the proverbial “straw” that broke the camel’s back.

[XR7]

Thanks for the photos and the ping.
Do you suppose the fresh paint, may be from grafitti eradication efforts?

[brityank]

Thank you all for the pings and info and explanations; definitely enlightening. Something I noticed on the pic of the bridge in AA’s doc from the lock is some type of addition at the bottom of the Position 10 posts on both sides. Note that there are ladders on both sides going up to a catwalk beneath the roadway, and the same assembly is noted on the north side (Position 18?). There are visible (strain guages ?) ‘patches’ on members attached to those of the south end , and what may be a signal light on the lower chord member. Could that addition have added some weakness to the structure?

If the Piers have shifted from scour, could there be enough ‘drop’ on the chords to effectively ‘close’ the angle between the horizontal and diagonal upper struts making those bends?

I find it telling that, from what little my inexperienced eye can see, they had strain gauges attached at that specific area before the collapse; someone at MDOT knew.

[TXnMA]

"If the Piers have shifted from scour, could there be enough ‘drop’ on the chords to effectively ‘close’ the angle between the horizontal and diagonal upper struts making those bends?"

~~~~~~~~~

Scour -- again... :-|

Pier(s) 6 -- on the south side where the failure initiated -- are on dry land, well back from the river, and are, AFAIK, unaffected by scour.

Pier(s) 7, on the north side of the river, may have been affected by scour -- but I don't see how that could have affected gusset plates on the opposite side of the river. Pier(s) 7 did tilt toward the river during the collapse -- and I address that displacement in this thread in comment #72, and in comments immediately preceding #72.

My reading was that the visible displacement of Pier(s) 7 -- which may have been weakened by scour occurred fifteen seconds or so after the collapse began -- and after the river span that had been opposing their movement was long since in the river.

U10 gusset plates bent due to scour on the other side of the river? Sorry, I can't envision it happening...

[TXnMA]

"There was heavy construction material piled up on the south end on the west side. I’m guessing that was the proverbial “straw” that broke the camel’s back."

Indeed...

At this point in the analysis, I believe it is worthwhile for us to re-visit the graphic showing the position of construction load that you posted earlier in this thread.

[jeffers]

Strain gages were installed during a joint (sorta) U of M and MnDOT study, and are well documented on the MnDOT documents page. The conclusion was that live loads stressed those points nowhere near design limits. Their engineering and methodology looked sound to me.

If I bolt two pieces of steel together, with an extra long bolt, and the protruding end gets bent, that doesn’t affect the fastening, at least not directly.

I feel the same way about the bent gusset plates shown in the images. The bent areas do not affect the gusset’s prime design function, which is to fix the ends of the connecting members.

However...

The bent areas tell me two things, indirectly.

1. U9-U10, at least once since construction, changed its angle with respect to L9-U10, in response to a load.

2. The change was of a magnitude that exceeded the gusset’s elastic limits, the gusset did not return to its original shape.

Item 2 above touches on a critical property of bridge steel, once you change its shape permanently, you have at least approached rupture or fracture levels of stress.

Item 1, and the strain gauges, and the preponderance of interest in the four similar points on the main span, and my long standing interest in the U10 point all stem from the same source. This is a critical area of the bridge. These four points are where stresses reverse, and the tensile stresses, more likely to fracture steel than compressive stresses, max out there.

Most scenarios fall into one of two categories. One, a single max load event sometime before those images were taken in 2002-2003, or two, repetitive flexing over an extended period.

Both could be bad news. One max load event could approach, and possibly exceed design limits, damaging the molecular bonds that give steel its strength, making eventual catastrophic failure a given unless detected. Repetitive stresses can induce metal fatigue, with similar result, and can also be just as difficult to spot.

Any competent bridge inspector should have been able to understand the implications of the bent gussets, without worrying about the direct effect of the bent gussets. The clue lies in what bent the gussets, and why they stayed bent, not the fact that they are bent. A bend right where member meets member would be sigificant, but a bend in the free plate of the gusset is not, not directly.

The free area of the gusset plate DOES contribute to the overall strength of the connection, BUT...if the condition of THAT free area was to have become a problem, it wouldn’t have bent under compression, it would have parted under tension. As seen in the images, it is not under enough tension to straighten the bend, and therefore clearly not under enough tension to fracture there. Gussets plates generally are not used to resist compressive forces, the steel members do that themselves. The gussets plates connect members via tension, and to a much smaller degree, hold alignment through complex internal stresses.

In my opinion, those bent gusset plates indicate something significant happened somewhere besides in the gusset itself, and the visible bending itself is not a primary concern.

Further, again in my opinion, the bent gusset plates indicate to me that the U10 nexus, on both trusses, indicate that the U10 centroid moved enough to change the angle between U9-U10 and L9-U10, permanently. The stresses that caused that and the stresses that resulted from that probably, (certainly, in my opinion) set the gusset up for fracture under tensile stress.

This is a very fine point, and I’m not comfortable with my explanation so far. In simple, non engineering terms, the bent area of the plates is or was in compression, no problem.

However, whatever bent those plates atered either the members and/or the gusset, such that it failed in tension. The deflection of the plates is an indicator that something went wrong, but the failure did not originate in those deflected areas. Instead, the failure originated in a different area of the gusset contradictory to the compressive force that bent them originally, since the gussets failed in tension. There are bolt line failures evident in the imagery, but there are others as well. It would take an expert some time, in close proximity to the plates themselves, to determine which fracture led the pack.

I still stand by earlier statements. U10E failed under tension and precipitated the collapse. The new images reinforce this belief, though I have not yet studied them exhaustively.

I am still amazed at the similarity of fractures between the east gusset of U10 west and the east gusset of U10 east. Its like finding two broken pieces of glass, from two different broken windows, with the exact same shape and dimensions. I understand how similarly the two different plates were stressed, over 40 years, but I’ve never seen two different fracture configurations that similar. Once steel starts ripping, molecular level considerations define the topology of the fracture, and when I post side by side images you will see just how similarly they failed.

If you already know where to look, there are numerous clues to this failure sequence, including the west truss bottom chord, originally shoreward of pier six, which briefly withstood a MASSIVE bending moment, as if the entire mainspan was loaded onto it. Also of interest are the difference between the UxLx verticals on either side of the U10 point. South of U10, the verticals fell over sideways. north of U 10 the verticals endured a sudden maximal compressive load and pretzyled when they absorbed the ground impact of the superstructure.

After I finish my studies, I will post some relevant images with labels, but so far, U10 east still looks like the culprit to me.

Remember, government PDFs should be open source, and you can always screen cap them, convert to jpg, and post them here for reference.

[brityank]

Pier(s) 6 -- on the south side where the failure initiated -- are on dry land, well back from the river, and are, AFAIK, unaffected by scour.

Maybe I shouldn't have used the term 'scour'. :^(
If you look at a full area view, you'll notice that both of the piers were set right at the edge of the riverbanks; it is only encumbered around the base of Pier 6 because of the dam/weir and lock. Which came first? I am betting the bridge was placed after the dam; if you look the water outlet for the lock, it appears it runs behind the pier, so that pier is subject to seepage from both sides, as against only the river side on the other bank.

The principle behind the tension/compression forces along the structure could shift the stresses from the settling of Pier 6, the further weakening of Pier 7, and the noted load imbalance evidenced by the strain gauges; to cause the bend/shearing at U10.

I agree that Pier 7 moved as a result of the collapse, not as the main cause. Thank you for your expertise.

[TXnMA]

If I understand you correctly, the bends in the gusset plates were not a source of failure, but were symptomatic of forces and displacements that acted elsewhere. If that understanding is correct, then we are in agreement: the bends were symptoms -- not causes.

I really wish that the photos on page 18 (two views of the same object[s], BTW) had something to indicate scale. Even though the article referred to the "thin" plates as being 1/2 inch thick, I just can't envision the thickness of that torn gusset plate (on p18) as being anything over 1/4 inch -- if that much...

If that plate is 1/2 inch thick, then those rivet heads must be three inches or so in diameter...

Do you see anything there that would be a good indicator of scale?

[SouthTexas]

Thank you both for your diligence and for sharing your expertise.

[jeffers]

Yes, you understand correctly.

The force which bent the gussets was compressive, because the angle between top chord and diagonal lessened. If U10 had returned to original position, it would have pulled the gussets back straight, or at least straighter. But it didn’t.

Whatever moved U10 (downward) in the first place, causing the bent gussets, whether a single event or prolonged abuse, also transmitted significant tension up the U10-L11 diagonal, and thence along the U10-U9 top chord back to...well eventually back to concrete on shore.

That may or may not be the cause of the fatal fracture later on. The triggering failure, in my opinion, was the vertical fracture in both U10E gussets, just south of the U10 vertical strut.

Those plates look like half inch to me. Three quarters on some boxes and H beams, perhaps on the pier gussets. I’ve cut both and the pics look right to me. Those rivet and bolts are pretty big, bigger than your fist is not uncommon at all. Grapefruit sized hex-nuts aren’t rare either.

[TXnMA]

Wow! Thanks for the calibration. With my limited experience with large structural members, my mind just wasn't ready to accept "fist-sized" rivet heads. Now I'll have to go re-study those photos with my new "eyeball calibration"...

Of course, most of us aren't prepared to see half-inch steel torn like toilet paper -- even along "perforations" (rows of rivet-holes)...

Thanks again!

BTW, did you follow my reasoning re my "lateral displacement" sketch? With all four gusset plates bent in the same direction, I perceived that as evidence of a lateral bias -- even if most of the distorting force was vertical (compressive). Am I wrong?

[brityank]

Of course, most of us aren't prepared to see half-inch steel torn like toilet paper -- even along "perforations" (rows of rivet-holes)...

That is something that has always amazed and confused me. I got to see a lot of steel structures while in the Navy, but never gave it much thought. Rivet-heads and bolts/nuts three to four inches in diameter were pretty common. Other than the roller-nests on the piers, I don't see any other strain relief from one side to the other. While the overall expansion would only amount to an inch or two, could that enhance the 'buckle' in the faces of the gussets? I also looked through the photos to see if there was anything to give some type of scale; any that have men in them aren't in close enough proximity for a definitive comparison, but there are some to give an idea.

Something else in the back of my mind from seeing the very large dimensions of the girder structure -- anyone seen any info on braided components? One of the promises (ha!) that moving out into space and setting up manufacturing there could give us was enhanced structural members, and braided I-beams would be one third the size needed of solid rolled for the same carry capacity.

[UCANSEE2]

“There was heavy construction material piled up on the south end on the west side. I’m guessing that was the proverbial “straw” that broke the camel’s back.”

Hmmmmmm....

That is exactly what I said the day it happened.

All the ‘experts’ on the thread at the time told me I was wrong.

[UCANSEE2]

This is a good link for pictures

http://www.freerepublic.com/focus/f-news/1884493/posts

[UCANSEE2]

And here is the latest info with pictures of the bridge gusset in question, from pre-disaster photos.

Pictures show buckling, article discusses falsified inspections.

http://www.freerepublic.com/focus/f-news/1990455/posts