Breaking: Bridge Collapse in Minneapolis

KSTP TV 5/ME | 8/1/07 | Me

[ButThreeLeftsDo]

Just turned on the news. 35W bridge collapsed in the Mississippi River. Cars, trucks, semis.....

Fires burning, tanker trucks, at least one school bus, more than ten cars......

Just now breaking.......

[sarasota]

I’m the one that posted this and I haven’t heard another word on that subject since the early aftermath of the collapse. Would be great to hear from those two guys again.

[Issaquahking]

Read your comments on the bridge, (profile page included) and wonder how much total neglect happens in regard to bridges, because of environmental EIS, and other cash wasting items, the greens create? Maintenance is hard to explain to people, the concept of a stitch in time, saves nine goes riht out the window, till something like this bridge happens. Then they are the first to point a finger.

Thanks for an injection of reality from someone familiar with this issue.

[Don Joe]

---

thanks to Sarbanes-Oxley overkill also to meet legal requirements and keep Jim out of trouble

---

This is the first I've heard of this. Can you tell me about it? (Am I supposed to track the comings and goings of visitors to MY website too? Papers, please?)

---

If you want to secure your own system get a copy of G-Zapper ^. It cleans/prevents Google et el tracking. It put the following in my 'hosts' file under the systems 'Window' folder:

# Begin GZapper
0.0.0.0 www.google-analytics.com
# End GZapper

That just consigns any reference for that site to the bit-bucket. I'm sure there's others out there like GZ. Bottom line is don't rely on others to secure your system from all the script-kiddies and their corporate devils.

---

I have the googlepest blocked via a Firefox add-in, but have been planning on adding it to my HOSTS file (which is currently obscenely immense, and not nearly big enough. Ah, the paradox of "life" in an invasion-centric age...)

BTW I would suggest null-routing to 127.0.0.1 rather than 0.0.0.0 -- the former is your own machine ("localhost"), whereas the latter is something that *could* be routable, should something unlikely, but possible happen upstream from your box.

As a postscript, I really don't like what Google has become. I will be polite, and say that they've become a bit too big for my britches. Their "don't be evil" crap has become a cynical joke, as they dive into their goal of owning everything about everyone, everywhere. It just plain sucks, and it's evil with bells on -- and a low-hanging fruit for any "civil authority" OR "civil litigant" looking to pluck "everything on anyone."

I do use their toolbar on IE -- but, I have it locked down in "spayed and neutered" mode, AND, I have furthermore gone to the Trouble (with a kapital T) to REALLY lock it down at Ver. 3.0.131.0 -- the last version before it went uber-invasive-on-crack (and the last version with a usable UI). Still, the @#$%^ tries (in vain) to "upgrade" itself to what THEY want to have shoved onto MY machine. It can't, of course -- and I expect that some day, they'll probably tell their servers to stop honoring queries from older toolbar versions, at which point I'll just delete it, and they can go piss in the wind.

It's MY machine, not *theirs*, and I will decide what software runs on it, not them, or anyone else. If I want to run an application, I'll be damned if I will allow someone to take it away from me when I'm not looking, and replace it with some invasive POS with a totally different (and unusable) interface, and snoopy crap I'll have to start all over again trying to castrate.

I guess that business about "power corrupts" is pretty much on the mark.

As as post-postscript, I'd really question the wisdom of handing the Googlekreepz "the keys to the kingdom", even if it is for a "free lunch." Handing those guys a complete detailed log of "everything on everyone" on a site like this -- with them being as politically "progressive" as they are... it can't be that different from giving George Soros the keys to the back door with a welcome mat laid out on the doorstep.

TANSTAAFL!

And, "Cui bono?" (WHY do they offer such a neato-keeno "service" -- part of which involves THEM obtaining full traffic/profiling data?)

Finally, if this logging is being done in response to some kind of statutory obligation, ISTM that it's fallen far short of the mark, since it can be defeated with such trivial ease (whereas "in-house" logging cannot be defeated, and in fact, does not even need to be visible to the clients).

[Don Joe]

I was watching CNN this evening (I suffered Fox "News" overload when they brought in Whorealdo and then segued to their skanky sex-therapist-cum-pundit to pontificate on why men look at naked women, while interviewing some Internet sleaze merchant as if he were some head of state. Surreal dignification of the guy.

So, I swallowed my bile, and punched "202" into the remote for the DreckTV box, and watched CNN. Amazingly, they actually reported on some news other than the Minnesota bridge failure. (For having allowed Fox to drone on during the day, I had NO idea that there was an incident in North Carolina, in which some "middle eastern men" were arrested driving a van packed full of explosives. I guess that's not big enough of a story to push the sleaze and bridge mix off the permanent front age at Fox. Nor was the near-fistfight at Congress, or the stolen vote.)

But I digress.

CNN interviewed an engineer at the bridge -- a guy who knew all the ins and outs of the industry. He gave them a quick but fairly in-depth tutorial on metal fatigue and how it affects bridges, and, how it's treated (i.e., drill a hole at the end of a split to keep it from growing, IF you can find the split).

He said that these metal fatigue cracks are usually impossible to see, they are that tiny.

When questioned further, he was perplexed -- he said he could not understand how the bridge could have failed due to metal fatigue. He said that if a crack developed, it would cause the bridge to tilt, to bend to one side, etc., but for the bridge to collapse the way it did, it would take multiple IDENTICAL splits, failing at the IDENTICAL time, on opposite sides of the bridge.

I can think of one potential cause for that sort of failure, but the media has been running from it like mad, as have quite a few Freepers.

What I cannot for the life of me comprehend is why the government keeps telling us to be on the lookout for attacks of this sort, and then, when they occur, IMMEDIATELY tells us that even though they have no idea what caused it, that there's NO reason to think there's terrorism involved, and then they engage in a full court press -- with media working hand-in-hand -- to "examine" ANY other cause, all the while ridiculing anyone who would even suggest the possibility of terrorist involvement.

Actually, I lied. It doen't confuse me at all. It makes perfect sense that there'd be a "no terrorism involvement" rule when the primary goal is to "avoid panic" -- especially at a time when the stock market is, shall we say, "nervous", and the economy -- while being touted as the best ever, is poised for some rather nasty inflation.

And in line with that, there's the curious removal of the "M3" reporting a few months ago. Without knowing the M3, we don't know the amount of "money" (i.e., the number of dollars) in existance at any given time.

We are constantly told that "inflation" means "the rising prices of 'stuff'", but that's BS. In reality, inflation is simply the amount by which the currency has been INFLATED, i.e., how many dollars are in existence at any given time.

When you INFLATE the size of the currency, each dollar is worth a bit less than it was before you inflated the size of the currency. In the stock market, it's called "dilution" -- when a company starts issuing and selling new shares, printed out of thin air, to generate some working money to burn -- the existing shares aren't worth as much, because the actual value of the company hasn't "inflated". In other words, if the company has a million shares out in the market, each share is worth one millionth of the company. If they suddenly print up and sell off another million shares, then each share is suddenly worth one HALF millionth of the company. (Please, no one jump me on this, these are stupid numbers, but they illustrate the principle.)

So, the economy is, ahem, "fragile", the market is at the point where high rollers are looking for an easy path to the window, and in that environment, the one thing that the government does NOT want is "panic", or anything that would cause "lack of confidence."

The MAIN thing is to try to keep the economy going. And if people were to believe that there was a fifth column inside our borders, doing to us what the French Resistance did to the Germans in the 1940s, then "the economy would suffer" in their minds.

So, I do not expect that ANY actual terrorist attack would be acknowledged as such unless there were absolutely NO way for them to deny it.

I've typed more than I intended, so I'll close by simply reiterating the main thing I intended to convey -- the engineer on CNN was very very persuasive. If you haven't seen the interview, then I'd suggest tuning in and seeing if the replay it, or, checking to see if it's on their website.

[jeffers]

Ok, I've looked a little deeper into this, and found something for everybody.

In a silly mistake that I've paid for many times over the years, I sold my "Handbook for Civil Engineers" during tight times in my younger days. "Vector Mechanics for Engineers; Dynamics" and "Applied Physics" are no help, they don't address any but simple trusses.

"Vector Mechanics for Engineers; Statics and Dynamics" moves beyond simple trusses, but finding the exact case under examination here is problematic too.

Having exhausted local resources, I went to a website maintained by Johns Hopkins University which allows you to design a truss and calculate the forces on each member, here:

http://www.jhu.edu/~virtlab/bridge/truss.htm

To refresh everyone's memory, here's a picture of the bridge before collapse:



The images below show a truss very similar in design of the mainspan of the bridge that failed. Notable differences may be heigth to depth to span ratios of the truss, which will affect the magnitudes of some of the forces in play, but not the types of forces in play, namely tension and compression. The number of panels and the placement of diagonal members precisely matches the prototype truss used on the collapsed bridge.

In the first image, I used the simple case of a point load, applied first to the top chord of the truss, centerspan, then to the bottom chord of the truss, centerspan. On the left, you see the load and where it is applied, on the right you see the resultant forces on each member after calculation.



In the second image, I applied uniform loading across the span of the bridge, first to the topchord, then to both the top and bottom chords. Again, the left panels show the loads, the right panels show force resolution.



In the third image, I applied assymetrical loading, first to the top chord, then to the bottom chord, loads at left, resolved forces at right.



My conclusions are as follows:

1. Spunkets is largely correct in saying that most members of the bottom chord of this type arched truss are usually in tension.

2. Supercat and I are correct saying that the bottom chord members closest to the support piers are either not in tension, are in tension with zero magnitude, or are in compression under some loading conditions.

3. Under none of the tested load conditions were those bottom chord members adjacent to the piers in tension with greater than zero magnitude.

4. Under centerspan point loads, tensile stresses were maximized in top and bottom chord members nearest to center span. Tensile and compressive stresses in diagonal members tended towards maximum at or about the quarterpoints of the span. Tensile stresses in vertical members trended towards maximum at the span quarterpoints in the case of top chord loading, and at misdspan, and, to a lesser extent, the quarterspan points in the case of bottom chord centerspan point loading. Kingposts on the piers carried significant, but not maximum loads.

5. In the case of uniform top chord loading and uniform top and bottom chord loading, tensile stress in the bottom chord and compression stress in the top chord trended towards maximum at center span. Tension and conmpression stresses in diagonal members trended toward maximum nearest the piers. Vertical members under uniform top loading conditions were in compression, with magnitudes equaling the imposed load in the case of struts between downsloping diagonals, and with part of the compressive stress distributed to the diagonals in the case of verticals located at the apex of two adjacent diagonals. Kingposts on the piers carried significant, but not maximum loads.

6. In the case of nonuniform assymetrical loading, top chord compressive and bottom chord tensile forces reached maxima between the imposed load and midspan. Compressive and tension stresses in diagonals reached maxima between the point load and the pier supports, with a secondary maxima at the mirror image point in the span with respect to center. Forces on vertical members were negligible in comparison with the exception of the vertical member actually loaded in the case of top loading and the kingposts. Kingposts on the piers carried significant, but not maximum loads.

7. Maximum stresses per member were tensile, applied to bottom chord memnbers, in the case of symmetric point loads, and uniform top chord loading. Maximum stresses per member were compressive, and applied to top chord members, in the case of top and bottom uniform loading, and in the cases of top and bottom assymetrical loading.

8. Failure of the river side bottom chord member closest to the southeast pier probably was not the initial trigger in the collapse sequence, as it appears to be lightly or not loaded at all under any of the tested loading schemas. Theoretical analysis of the failure sequence is contra-indicated, because we do not know how the structure was loaded at the time of collapse, nor do we know the design stresses on any of the members, either at construction, or after aging, corrosion and weathering. Further, loading considerations at the time of the collapse were dynamic in nature, not static as was used in the above calculations, which opens several cans of worms, including the possibility of resonant vibration effects. Instead we are left with empirical analysis of the collpase, i.e. that which remains, in imagery, video, and in steel and concrete reality. Without access to the bridge debris, we are probably correct in localizing the collapse initiation to the southeast corner of the bridge. We are probably correct in saying that the southeast kingpost buckled before any of the other three. However, attempting to resolve the trigger event further without access to the bridge debris is probably impossible. The failure of any component in the eastern truss assemblies, southern sidespans or mainspan, could have triggered the collapse. With the information currently in hand, we've taken this about as far as we can go. Final conclusions my have to wait for the official reports, which may or may not be conclusive, and which may or may not be controversial. Neither the FEMA nor the NIST reports of the WTC 1 collapse utilized the available oxygen budget for the fires, and neither went very deep into the culpability of the hat truss assmebly's load re-distribution in triggering the progressive collapse. Sometimes even the experts, with access to all the available data, don't report the whole story. As unsatisfying as that can be, that's just the way it is. In this case, we have a good chance of getting answers, but we will have to wait for them. When the final reports come out, I imagine we'll all be back here to go through some of this again.

[jeffers]

Thanks for the link.

That’s an interesting quote, but it’s hard to apply to the collapse progression. The road deck gives the bridge rigidity, but only in the plane of the road deck.

It could help keep the middle of the bridge from swaying side to side, for example, but it wouldn’t help hold the bridge up, and removing road deck concrete would’t help bring the bridge down, unless something else was very, very wrong.

That brings us back to the diagonal bracing between the kingposts above the piers on the south end of the mainspan. If they were already compromised, then the road deck would have helped keep the kingposts from swaying side to side.

If that was the case, the problem should have been readily apparant to anyone under the bridge looking at them.

All structures have some give to them. It is very difficult to accurately judge how much give is occurring when you’re standing on the structure. You can say “a little” or “a lot”, but relative comparisons are risky without a transit and repeatible observations and calculations. A human’s inner ear and perception are subjective at best.

In the past, I’ve experienced what felt like excessive sway in some of my structures and the procedure in all such cases is the same. Slowly and carefully get off the structure, then report the sway up the chain of command and go looking for the cause.

In one case, we had a new guy removing braces that should not have been touched. In another case, the sway was significant, but within design tolerances. Both structures survive to this day.

That is an interesting report, to be sure, but I’d be careful giving it more credibility that it deserves. I’ve never seen a construction stiff stay around when he had real concerns about structural integrity. Sometimes that’s the fastest you see them move all day. If they stayed on the bridge, they either weren’t too concerned about stability, or else they were driven by slave-owning tyrants with guns.

[spunkets]

"3. Under none of the tested load conditions were those bottom chord members adjacent to the piers in tension with greater than zero magnitude."

That must be the case in the bridge you have, because you've designed it that way. Your model doesn't represent the same bridge. You have a bearing under the first king post with nothing on the other side, so no forces will ever appear in that beam. You could take it right out and the situation wouldn't change.

Look at the pic of the real bridge. It has another section on the other side of the first king post, so you must ignore that first triangle and only look at the forces that appear at other king post connections. Then you can see that the bar that broke was part of a complete king post truss element, whose base is always in tension.

In a real bridge, the only way to get the lower chord in compression is to remove the bearing and fix the position of the chord ends. The real bridge has a bearing. It could freeze, but that's unlikely here, because it would be obvious to the inspectors that it was approaching such a condition and the fix is simple cleaning. there appears to be rust, but good maintenance and paint. Also, that chord element in the real bridge appears to be torn free, not buckled, or shoved back. The symmetric chord element on the other side of the post did not break. That's why that section torqued over after the other side went straight down. Note that the SW side assy just slid off the post.

In the figs I drew everything has simple supports at the bottom. That's the same as a bearing. I also just gave a simple triangle to show how the forces look in that. The king post complicates things, and so does showing forces in assemblies of more than one element. So I took it out and didn't address anything except the simple triangle and what's above the post on the real bridge. Time...

Thanks for the link. http://www.jhu.edu/~virtlab/bridge/truss.htm . I won't have time to look at it yet. I'll just makes some comments. A real bridge must be more, or less simply supported to allow for expansion/contraction, and has bending moments in both the structure and it's elements. The model uses fixed ends somewhere to show forces.

MAybe you could make the 3rd, or 4th node look like the junction of the 2 bridge halves at the post. What does green mean in the figs? Does the model allow elements to be removed? The chord element I circled in the other post is critical and broke. Take it out and the bridge falls. The above doesn't show that for the reasons I gave. If the model can be changed, it would be possible to see the tension rise in that chord section when adjacent beams are removed.

[brityank]

FReepmail incoming.

[RLM]

Jeffers,
Thank you for the analysis and link to the John Hopkins bridge simulation software for truss bridges. However your values and conclusions are counter-intuitive if we consider the bridge consists of two more or less independent cantilevered sections. Using that assumption, the bottom chord members would always be in compression, and the upper chord members always in tension. Also, the vertical values at the king post would be double the assumptions. The horizontal stress values adjacent to the king post (both tension and compression) would be highest of the entire structure. Again, there are a myriad of unknowns, such as the tension or compression between the two cantilevered sections where they join over the river, same for the points of tie-in to the concrete sections and, of course, the dynamic stresses at the time of failure. Nonetheless, the analysis was interesting.

[concrete is my business]

I have not had much time to watch the videos or see the news but heard some concrete guys here speculate that the train might play a part in the final report, if the bridge failed at that point. The live load of all the stopped traffic, the steep slope cut into the bank next to the tracks, combine with shaking from the train, and all that silt on the bottom, no bedrock there, could have made for ‘a whole lot of shaking going on’.

[B-Cause]

Minnesota Department of Transportation has just posted a new web site containing substantial information on the history of the I-35W bridge inspections.

Check here.

Also there is a very interesting inspection report dated June 0f 2006.

Check here .

[XR7]

Ever see the frame of a 40 year-old car from Minnesota?
Or Chicago? Or New York?
Or anywhere they use salt on the roads?

It should be no surprise this bridge rotted out.
The only surprise will be if any other old bridges actually don't cave in too.

[SilvieWaldorfMD]

I was watching the NTSB spokesman earlier today stating that they will look at all daily weather records since the bridge started being operational in the 1960’s to see if they could find a root cause.

Read: The blame will be placed on global warming.

[TXnMA]

Bump to placemark a most informative post...

[supercat]

You have a bearing under the first king post with nothing on the other side, so no forces will ever appear in that beam. You could take it right out and the situation wouldn't change.

You are correct in noting that the model does not match the bridge. However... In a real bridge, the only way to get the lower chord in compression is to remove the bearing and fix the position of the chord ends.

That is just plain wrong. In the real bridge design, that chord will always be under substantial compression (unless something else on the bridge has already failed catastrophically). Imagine that the king post over one of the piers were split in two along its length, with one half attached to each half of the bridge. What would happen? The side of the bridge away from the other pier would want to fall away from it. To prevent that from happening, the top of the bridge there must be under tension. To balance that, the bottom must be under compression.

Cantilevered truss arrangements are frequently used in building bridges because they substantially reduce the tension and compression loads found in the center of a span. Since the top center of the span is under compression, any tensile force applied to the top outside of the bridge will reduce the tension by an amount equal to such tensile force.

[supercat]

Out of curiosity, what is the purpose of the centermost kingpost? In general, kingposts with a five-point connection at the top won't carry any significant force, but if the bottom strut is under compression they may still be useful to prevent buckling. But when the bottom member is and always will be under tension, what's the point of the kingpost? Tensile members never buckle anyway.

Incidentally, with regard to 'redundancy', if yield strength were not a factor, adding a single strut almost anywhere to a non-redundant structure would allow it to withstand the failure of any single element. There are three caveats:

1. The failure of certain elements could cause the loads on certain others to be increased by many orders of magnitude, so their yield strength would almost certainly become a factor.
2. In a rigid redundant structure, thermal expansion may create very large internal stresses which if not dealt with can cause failures by themselves.
3. In a non-redundant structure, it's fairly easy to predict the loads on individual structure members. In redundant structures, load-sharing may be hard to predict.

I would not be surprised if the shift toward redundant structures was a result, at least in part, of increases in computational horsepower allowing reasonable modeling.

[jeffers]

spunkets wrote:

"...Look at the pic of the real bridge. It has another section on the other side of the first king post, so you must ignore that first triangle and only look at the forces that appear at other king post connections. Then you can see that the bar that broke was part of a complete king post truss element, whose base is always in tension.

In a real bridge, the only way to get the lower chord in compression is to remove the bearing and fix the position of the chord ends. The real bridge has a bearing."

******************

I understand your point, and it is a valid question. You note I only used "half a triangle" above the piers, and this was by concious choice.

It does not affect the force resolution solution however, at least, not in your favor.

I used that configuration because of limitations in the model, sizewise, so that I could include all panels in the prototype main span truss.

A simplified (fewer panels) diagram with full panels above and on both side of the piers as you suggest, puts the bottom chord in compression at those points under uniform top and bottom chord loading, and resolves the forces at zero magnitude under uniform topchord only loading schemas.



Note also that in the above and all previously posted stress resolution diagrams, the left truss support represents a fixed bearing surface, while the right truss support represents and is modeled after a roller bearing. This is a limitation of the modeling software and is mandatory prior to executing the force resolution calculations.



For further consideration, here is MnDOT's own stress diagram, from page 50 of the 2006 inspection report, an excellent source posted by B-Cause, in post 2631:



Please forgive the image format and quality, I had to lift if from a PDF document and rotate it 90 degrees to make the salient points visible and legible.

Note that MnDOT shows the bottom chord in compression for the first three panels on either side of the piers.

Further note that the topchord is in tension over two of those three panels either side of the piers, and the combination of these two stress resolutions represents the cantilever element of the truss design I spoke of since early in this thread. This cantilever element is discussed at some length in the official MnDOT report, and probably was instrumental in the collapse initiation of the north sidespan after the mainspan collapse sequence was complete.

I fully agree with your observation that the bottom chord failed just north of the southeast pier. I am not willing to say with certainty that this failure was the trigger for the collapse sequence, but I'm also not willing to say it wasn't the initiating event either. It is a critical component, and its failure would not go...unnoticed...but it does appear to be lightly loaded under normal equilibrium bridge loading conditions.

I will also repeat my earlier observation that the failure of that bottom chord member appeared in the imagery to be consistent with compression failure, i.e. diagonal shearing.

With the information available, this is as far as I am able to go.

The full MnDOT inspection document can be found here, again thanks to B-Cause:

http://www.dot.state.mn.us/i35wbridge/pdfs/06fracture-critical-bridge-inspection_june-2006.pdf

[jeffers]

It is and it isn’t a cantilever. It’s a complex truss. There are cantilever forces in play near the piers, but closer to midspan the superstructure acts as a simple truss. There is no floating truss or pivot pin at mainspan center, and no tiebacks or counterweight on the approach sides (except the cantilever/truss dead loads), so treating the structure as two independant cantilevers does not resolve the forces in play consistently with MnDOT’s judgement.

On the other hand, ignoring the cantilever aspect severely restricts understanding of the northern sidespan collapse progression. In order to properly model all the forces at work here we’d need much more complex software, and then we’d have to plug in all the defects that have accumulated since the bridge was built, and already that’s out of my processing horsepower range, without even looking at fatigue effects, wind loads, settlement, etc., etc., etc.

For what it’s worth, MnDOT shows the pier kingposts in compression, though without indicating magnitude. The Johns Hopkins models indicate magnitude, but as stated earlier, I’m uncomfortable using those except in trends and comparison, since my width to depth to length ratios are essentially random, without relation to the prototype truss.

One of my earlier images, and to a much better degree an image Spunkets posted, shows the southeast kingpost was subjected to severe compression forces at some point in the collapse progression, it folded up like a toothpick with a quarter of the weight of a big bridge resting on it.

[jeffers]

VERY nice find.

Because it will useful to see what others come up with independently, I will refrain from commenting on the contents and relevancy of the 2006 report right now, but I would say there are significant revelations there.

In order to forestall future claims of piggybacking, I will say that I’m referring to probably the most significant series of structural events since the completion of bridge construction, right in the middle of this thread’s long running primary area of interest and leave it at that for now.

More on this after others have a chance to read the report and comment.

[supercat]

Note also that in the above and all previously posted stress resolution diagrams, the left truss support represents a fixed bearing surface, while the right truss support represents and is modeled after a roller bearing. This is a limitation of the modeling software and is mandatory prior to executing the force resolution calculations.

If both support points were fixed, the structure would be redundant unless a member were removed somewhere. One could fix the right support point by adding a member connecting it to the left; one could then remove almost any member and the structure would still be supported (though depending upon the member removed, stress levels on remaining members may become excessive).