Evan on Engineering Software

If you lived in the San Francisco Bay area, I wouldn't need to tell you that the Bay Bridge is closed.  A  tie rod in a repair brace failed dramatically, dropping a several thousand pound steel saddle, with tie rods still attached, onto the bridge deck, hitting several cars (though, fortunately, no one seems to have been seriously injured.)

The Labor Day repair.

This is a story is about how the Bay Bridge was repaired last Labor Day weekend.  It is based on interviews of several people directly involved in the repair of the bridge.  I spoke to these people literally hours after the bridge was reopened.

The back story is that the Bay Bridge is due to be replaced with a new bridge.  While the construction of the new bridge won't be finished for quite some time, preparations for that construction have already started. 

One of the first steps in that preparation was the installation of a bypass section at the East end of the existing bridge.  The plan was to close the Bay Bridge for four days, over the Labor Day weekend, and literally cut out an existing section, float-in the replacement section, and bolt it into place.  This planned closure was to be the longest in recent history for the bridge, and Caltrans (the California Department of Transportation, which is responsible for the bridge) planned a biennial structural inspection of the bridge to coincide with the closure.

Though the previous structural inspection revealed no problems, this inspection was different.  A large crack was discovered in the end of a eye-bar.  The Bay Bridge is largely structural steel, and uses tension links, each made up of four eye-bars, attached to pins at each end.  These links are similar to the links you'd find in a bicycle or motorcycle drive chain—but a lot bigger.  The eye-bars are cut out of thick plate steel, with holes at each end for the pins.

The failed eye-bar discovered on Labor Day weekend had a crack running from the outside edge, nearly through to the pin.  Caltrans said, at the time, that they suspected the crack was caused by corrosion.  If you look at pictures of the failed link, it seems pretty clear that the crack started from a stress riser.  Whether that stress riser was caused by corrosion, or by a steel worker pounding on the eye-bar with a sledgehammer when it was installed 73 years ago probably isn't important.

When Caltrans found the bad eye-bar, there was good news, and not so good news.  The good news was that the bridge was already closed, and several of the companies they would have otherwise called to help fix it were already there—working on installing the bypass section.  The not so good news was that none of those companies could afford to endanger their contractual commitments to finish the bypass section on-time, and it was Labor Day weekend—so it wasn't like Caltrans could call up a steel fabricator, and have anyone actually answer the phone.

Caltrans put one of their most senior bridge engineers in charge of getting the bridge opened up by Tuesday morning, at the end of the holiday weekend.  Working with him were a team of people from C.C. Myers, Inc. (a construction company known for its ability to handle emergency bridge repairs), Danny's Construction (steel erection specialists), and T.Y. Lin, International (bridge engineers).

The first step was to come up with a conceptual design for the repair.  Since the repair would only need to last until the replacement bridge is built, it didn't need to be elegant.  Caltrans' engineer drew a sketch of what he had in mind:  A brace, consisting of two saddles, connected by tie rods.  The next step was to figure out who could build the parts for the the brace.  And that's an interesting story in its own right.

A history of success.

In 2007, a gasoline tanker truck overturned in the MacArthur Maze, a freeway interchange in Oakland.  The resulting inferno was so hot that it melted the steel in the roadway deck above the tanker, resulting in its collapse.  C.C. Myers was hired to repair the Maze, and managed to get the job done just 26 days after the collapse, beating the deadline by over a month.  This feat was significant enough that it was not only memorialized in Wikipedia, it was the subject of a television special.  C.C. Myers gives a lot of credit  to Stinger Welding, in Coolidge Arizona, for fabricating and delivering replacement girders for the job in just nine days.Girders for the MacArthur Maze. The truck actually had a blowout on the way.

A call to the wild.

The legendary C.C.Myers (L), talking with Carl DouglasLast Labor Day weekend, when the failed eye-bar was discovered, Carl Douglas, president of Stinger Welding, was in the wilds of Montana on a fishing trip.  Where he was, there was no email, and no cell phone service.  About 1:00 on Saturday afternoon, he headed into town (where there was cell phone coverage) to get some supplies.  And coincidentally, that's the moment that his phone rang.

On the other end of the line was a person saying they had an emergency, and needed some parts fabricated right away.  Initially, Douglas thought it was a joke. But the words "C.C. Myers" and "Bay Bridge" got his attention.

After hearing what was needed, Douglas got on the phone immediately, to track down the employees he needed to do the job.  It was not easy.  His shop foreman, for example, was on vacation in Utah, and had to turn-around, and head back to Arizona.  By mid-afternoon, Douglas had managed to assemble a 15 person crew.  He had his people go into the steel yard, and take an inventory to send to the Caltrans team, so they could adapt the design of the saddles to the materials on hand.  
 
Douglas then took on the task of trying to find a plane that was big enough to fly ten tons of steel from Coolidge to Oakland.  Fortunately, he knew people in the air-cargo industry.  Unfortunately, being the holiday weekend, no one was around to answer his phone calls.  Just as a passing notion, he thought to call FedEx.  They not only answered his call, they offered three different planes which could handle the load, and deliver it on time. (If Stinger Welding ever needs to do something like this in the future, it's good to know that FedEx can even charter an Antonov An-124, with a maximum payload of 330,000 pounds.)   With the people and transportation in place, Douglas flew back to Coolidge, arriving at the shop at 11:30 PM, Saturday night.
 
While Douglas and his crew were working in Arizona, the Caltrans team were working in California, taking measurements and refining the design of the saddles.  The two teams communicated back and forth until about 3:00 AM, Sunday—at which point the Stinger Welding crew started cutting and welding.
 
By 9:00 AM Sunday, 18,000 pounds of steel saddles (and Carl Douglas himself) were being loaded onto a FedEx charter plane, which would fly directly to Oakland, where a flatbed truck and a California Highway Patrol escort awaited.
 
While the saddles fabricated by Stinger were critical to the repair, the work could not be completed without the high-tension steel tie rods.  For these, C.C. Myers called on AVAR Construction Systems, a major supplier of post-tensioning systems for bridges and stadiums.  The rods were much easier to fabricate than the saddles, as they only had to be cut to length.  Since AVAR is based in Fremont, transportation was not an issue.
 
With the saddles and tie rods delivered, the steel erection crew got to work.  Hoisting the first saddle up, they found that it fit perfectly.  Except no one had anticipated that the tie rods, if installed directly on the saddles, would interfere with the eye-bars.  What was needed were spreaders, to move the tie rods so they wouldn't interfere. 

Ultimately, American Bridge Fluor, Inc., which had been working on the bridge bypass, was able to fabricate the spreaders in their Oakland yard, and deliver them to the bridge Monday night.  They were quickly installed, and the tie rods were carefully brought up to tension.
 
With no time to spare, Caltrans got all the construction debris cleaned up, and started contacting toll booth operators, to get them to come to work.  After a hastily called 6:00 AM press conference, the bridge was reopened at 6:38 AM, Tuesday morning—just 1-1/2 hours after the originally planned opening.

The failure of the repair.

It's only been 7 weeks since the Bay Bridge was repaired.  The question arises, why did the repair brace fail?

Caltrans' initial statement is that the failure was likely due to a fatigue crack in a tie rod, exacerbated by high winds. But the brace, properly designed and installed, should have been able to handle hurricane force winds without a problem.  There shouldn't have been a fatigue crack in a nearly new tie rod.

So, I'm going to speculate, and tell you why I think it failed. (Since I'm not a P.E., you should take this with a grain of salt.)

The basic design of the brace was very robust.  The saddles and spreaders were far stronger than they needed to be.  The only critically stressed components were the tie rods. 

If the brace had been installed across one tension link, the calculation for sizing the tie rods would have been pretty simple.  But the brace wasn't installed across one tension link.  It was was bolted to the top and bottom pins of two separate links, which are joined at a central fixed pin that's part of large fixed girder. Take a look at the picture, and try and wrap your head around the force vectors in this configuration. The brace itself absorbs the load that was previously born by the three girders that converge at the point labled "joint" in the photo below. Rather than reinforcing the weak link, the brace, as installed, has created complex multi-bar linkage.  If you consider statics, dynamics, and kinematics, this makes for an "interesting" system.

When I saw a picture of the actual installed configuration of the brace, my first thought was that it  probably made the design calculations a lot more complicated.  (Acutally, my very first thought was "Kansas City." If you happen to be a structural engineer, and would like to explain to the readers what that means, please post a comment.)

Since this blog is about engineering software, you probably know I'm going to tell you how straightforward it would have been to design and analyze the brace, using Tekla Structures, Ansys/CivilFEM, STAAD.Pro, or SAP2000 (among others.) 

So, here's the question:  Was any computer-based analysis run to optimize the design of the brace?

What I've found out so far is that the steel erection people made measurements for the brace by cutting  foamcore patterns to fit.  (This isn't a bad way to do it, but it's not as accurate as using a laser scanner.)  Design information and dimensions for the saddles were sent back and forth from California to Arizona via fax.  And, finally, the images provided by Caltrans that show the saddles are not 3D models, but are rather photoshopped pictures that don't even accurately represent the design. 

All of these things, taken together, lend credence to the possibility that the brace was designed without the benefit of engineering software.

Though it's too early to know for certain, today's newspaper articles hint that the modified/repaired brace now being installed on the bridge hasn't been analyzed with engineering software either.

Is there a moral to the story?

It's easy to be a Monday morning quarterback.  Overall, I think everyone involved in the Labor Day weekend repair deserves tremendous credit for working their asses off to get the bridge opened on time. 

Yet, I think the original decision to design the brace using the “back of an envelope” method was short-sighted.

It's pretty likely that either Caltrans or T.Y. Lin has a structural model of the Bay Bridge already.  I'm making an assumption here, but given the retrofits done to the bridge since the Loma Prieta earthquake in '89, I think it's a reasonable one.  So, why not use that model?

Not pretty, is it?It's probably old-school thinking.  Not that many years ago, the software tools for doing this kind of work were a bit, ahem, user-hostile.  Faced with the need to design, fabricate, and install a brace within an impossibly short deadline, and having had some experience with the vagaries of engineering software, a reasonable manager might say “to hell with using computers.”

Looking at the progression of events over the Labor Day weekend (for example, discovering the need for the spreaders late in the process), and the failure of the brace yesterday (which will have a major economic impact on the bay area, despite the likelihood that the Federal Highway Administration will pay for the actual repair), the notion of using CAD and CAE, even given the tight deadlines, seems, in retrospect, a lot more interesting.

What can engineering software vendors do better?

I'd like to see companies in the engineering software industry put some skin in the game.  All too often, it is their customers who are left holding the bag when software problems result in missed deadlines and failed projects.  While I recognize that engineering software companies can't jump every time a customer has a short deadline and doesn't know how to read the manual, I do think it's reasonable for the major players to have emergency response plans—If only to give their customers the confidence that, when failure is not an option, someone's got their back.

I can't help but wonder if this story would have turned out differently if the engineers responsible for the job had more confidence in their engineering software—and in their engineering software vendors?

There is a demo version of a CAD program that I'd like to take a look at.  It fits into a category that I'm very enthusiastic about:  Tools that empower non-CAD-gurus.  I wanted to talk about it here, in my blog.

Yet, I can't install it.

It's not because my computer is incompatible. It's because, after having read the license agreement, I can't, in good conscience, click "accept".

It sucks, because this program looks really interesting.

At first, It occurred to me to "out" the vendor, and tell everyone why the license agreement is so egregious. But, after some thought, I figured I'd rather ask you some questions, about how you respond to license agreements.

So, let's assume that you recently installed a demo version of a CAD program:

• Did you read the license agreement?
• If you didn't, why not? (seriously.)
  • Were you just being lazy?
  • Did you even scan it a little bit?
  • Did you think it's not an enforceable contract?
  • Did you think it's full of meaningless boilerplate?
  • Did you think the vendor would never actually enforce its terms?
• If you did read it,
  • Did you understand it?
  • Did you grasp its implications?
  • If you showed it to your lawyer, what would he say?
• Why did you click "accept?"

I recently received an email, asking me to recommend a CAD program.

I thought it would be interesting to post a copy of the email here, and ask what you think.

I am looking for a new 3D CAD program that is easy and intuitive to use like a Tool driven direct modeler that can also produce accurately dimensioned 2D drawings to be plotted with title blocks and BOM lists and all for fabrication and construction.  I generally work by myself and don’t have CAD projects often but occasionally need to work with other engineers and Naval Architects on projects.  It seems we all use different CAD programs so the program would need to be able to import and export the various standard file exchange formats.

I mostly design small Submersibles and underwater equipment and do some Naval Architectural work.  I also like to design furniture and house and shop additions.    

I made the switch from the drafting board making engineering drawings using AutoCAD V 1 on a Wang with a 12” monochrome monitor.  I upgraded through all the versions to AutoCAD 2000 which I still use to produce 3D models to get the 2D dimensioned views.  Due to the complexity of this method I can’t find others that still work like this to consult with.  I would like something easier to use.  I just purchased [redacted], which I don’t like.  I am also just getting started with Google Sketchup Pro which is fast and intuitive but I am having problems with accuracy.  I also don’t know yet how the BOM and 2D layouts will work.  Can I plot out professional looking 2D drawings?

Talking to all the individual sales reps is not helping me come to a fast decision as to the best way to proceed.  I could use some unbiased advice.

Which CAD program would you recommend for this person?  And why?

Deelip Menezes is a CAD plugin developer in India, who writes a blog discussing a variety of CAD related subjects.

He's pretty good at stirring the pot:  he writes long blog posts, with plenty of opinion, and lightly filtered rants.

His most recent rant is about CAD users who think that CAD programmers are not CAD users.  He points out that he has 44 or so CAD programs on his computer, which he uses.  Therefore, he is a CAD user.

Here is my thought: A CAD user is not merely a person who uses CAD. A CAD user is a person who is a domain expert in the use of CAD software to solve a design or engineering problem. It may be that a CAD user speaks only one CAD tool "language" (e.g, Solid Edge, ComputerVision, Anvil 4000), but if they can wield that tool skillfully to solve their design or engineering problems, they are a CAD user.

Deelip claiming to be a CAD user, because has a bunch of CAD programs, and can use them to develop CAD plugins, is like a person claiming to be a mechanic because they have a bunch of wrenches, and can use them to tighten bolts.

While Deelip is a domain expert in the development of certain types of CAD plugins, that expertise doesn't necessarily translate to related domans.  (Though that doesn't stop him from jumping in with both feet.)

In his post, and the related comments, a large number of  CAD-related domains are discussed:  CAD use, CAD programming, CAD development, CAD plugin development, CAD analysis, CAD writing, CAD reviewing, CAD QA, CAD testing, CAD demonstration, CAD usability analysis, CAD failure analysis, CAD UI design, CAD documentation.

If I were looking for a domain expert in CAD plugin development, I might put Deelip pretty high on my list. 

But the other domains? Not so much.  Here are a few possibilities:

SolidCAD use: Matt Lombard.

CAD programming: Mike Riddle or Tim Olsen.

CAD development (big picture): Chuck Grindstaff or Mike Payne.

CAD analysis: Ken Versprille or Mark Halpern.

CAD reviewing: Al Dean.

As for me: I may be a domain expert in some disciplines, but I'm not fool enough to think I could give any of these folks a run for their money in their particular domain. Well... except maybe Al. But, then, he'd still kick my butt in terms of how fast he gets reviews done.

Now, back to Matt Lombard's question, as expressed by Deelip:

"To the best of my very poor memory, all of the proponents of Direct Editing as the New World Order for CAD are non-users, and the proponents of history are users ...[snip]... Any comments on that?"

My answer is that it's not his memory. It's more likely to be confirmation bias.

His question is straw-man: Most serious proponents of direct feature modeling that I know have expressed the sentiment that a two-trick pony is better than a one-trick pony, unless the one-trick pony meets your needs.

Translated: A lot of people will benefit from having both modeling technologies. Some will do fine with just parametric feature-based modeling. And some, especially those for whom feature-based modeling is just too hard, will benefit from having just direct feature modeling.

Parametric feature based modeling isn't going away. At least, not for a long time.

Deelip Menezes has been talking in his blog about solid modeling technologies,  such as parametric feature-based, direct feature, and sometihng that appears to be feature-tree inference.

He seems to be hung up on technology.  User's don't really understand CAD technology.  All they really understand is that the CAD tools they have don't cut it.

One of the reason why I was (and am) so enthusiastic about Synchronous Technology is that it was the first major new technology from a top-tier CAD vendor, in perhaps 20 years, that wasn't mostly about pwning users.

Think about it.  Back in the days when CAD salesmen wore Gucci loafers and drove BMWs, they were selling a technology as addictive as cocaine.  They'd weave a powerful tale of how good it was going to feel once the customer got parametric feature based modeling... and customers, grasping for anything that would make things better, would gladly buy.  It was only after the customer was hooked that they realized the high was not as good as they'd hope for, and the cost to get it was more expensive than they'd ever imagined. 

Synchronous Technology, and all the other variants of direct feature modeling (especially those of Kubotek and SpaceClaim), are more like chocolate than cocaine.  Have you ever heard of a person screwing up their life because they're a chocoholic?

We can argue about technology -- whether feature tree inference is better than feature inference, or whether history-based modeling is better than amnesia-based modeling -- but , in the end, the most important question, for users, is can they get their jobs done better with this technology, or are they just trading one drug for another?