Wednesday, June 9, 2010
Book Review: Normal Accidents
Normal Accidents: Living with High-Risk Technologies
Recently, while talking about the Deepwater Horizon, someone recommended "Normal Accidents" by Charles Perrow. It's all about Normal Accident Theory [PDF]. The book is filled with all sorts of illustrative examples from a variety of industries (nuclear power, marine transport, aviation, etc.).
There's a huge database of incidents and near-misses. Here's an example from Fix the Pumps on why near misses are important. The most ground breaking part of the book is when he talks about "non-collision course collisions" where two vessels with radar in good weather will come close to each other, but are well away from colliding. Both bridge crews react poorly and the vessels collide, sometimes with deadly consequences (one example was from the Mississippi River near New Orleans - NTSB Report [PDF]).
He also constantly harps on the fact that "operator error", while a constant scapegoat, is actually usually a factor of design flaws or production pressure or something else. Operators get blamed for entirely too much, in Dr. Perrow's opinion. Also, lots of "safety systems" can serve to actually make a complex system more dangerous. Remember, Chernobyl was testing a new safety system the day it blew up. More does not necessarily mean better. If it just adds to the complexity or it requires too much maintenance or it leads to nuisance alarms that cause the operators to miss the real flaw, it's bad. Getting the RIGHT safety system is an important balancing act.
The book has its drawbacks, though. Dr. Perrow is a sociologist, so he gets on some things that, if you have a science or engineering background, you just shake your head at. For example, this edition was last updated in 1999 and he recommends buying a generator to prepare for Y2K. Oops. He also occasionally lets his personal political views push for conclusions beyond strictly what the data (at least as he's presented) supports (ex.- he advocates completely abandoning nuclear power).
There's a lot more to the book, but suffice to say, the book made me think. In the future, I think it will be good for my development as an engineer to ensure that I read at least one engineering disaster book per year.