Friday, March 9, 2012

Prima risk management

I recently learned about the risk mitigation built into the new Oakland-San Francisco bay bridge being built to replace the repaired structure damaged in the 1989 earthquake. There are a few problems to deal with:
  • Unlike the rock anchoring the Golden Gate across the west bay, the east bay floor is mud for the most part. This poor foundation contributed in part to the 1989 failure
  • The San Andreas fault runs within a stone's throw, and this is "the big one". It's the same fault line that destroyed SF in 1906.
  • The bridge carries a lot of traffic, all day and night, so failure is not an option

Of course, this is a zillion-billion dollar infrastructure project, so risk management and project management are not taken lightly.

So, what've we got here?
  • First, about the mud: There will be only one tower for the suspension of the road bed, and this tower is anchored on the only rock in the east bay. But, it's actually four towers side by side connected by fusible links designed to fail under high stress loads, allowing each tower segment to go its own way. These links are replaceable after the event

  • Second, the suspension is like a self-eating watermelon. That is, think of the road bed and then think of the suspension as a rubber band around the long axis of the road bed and also hung over the tower. You can't put on the rubber band until the road is in place, and you can't put the road in place until the rubber band is there. Obviously, to break this circular dependency, some temporary structure are needed.

  • And third, the road bed is actually multiple sections that are joined by sliding joints. It's possible for the road to separate by six feet and then come back together like it is on drawer slides. And with this system, the tower can sway five feet and not affect the road.
Now, what about acceptance testing? I recall projects I worked on that were nuclear event hardened communications systems. The army would never pony up a nuclear event as GFE (government furnished event) for a user acceptance test. Area 51 was not available. So, we went with analysis only. Seems like the same conundrum on this as well.

I hope it works; failure is not an option

Are you on LinkedIn?    Share this article with your network by clicking on the link.