« FATIGUE DISCUSSION | Main | Spots Still Open for Company Meeting Before/During/After the TapRooT(R) Summit (April 5-8) in Gatlinburg »
September 15, 2005
What's Wrong with Fault Trees
I've seen many people suggest that Fault Trees are an excellent too for root cause analysis. I learned to use Fault Trees when I was getting my Masters Degree in Nuclear Engineering. I could see how they could be used proactively to have a systematic analysis of potential problems. The logic in the tree appeals to engineers.
But I've seen Fault Trees misused by engineers and others when analyzing the root causes of accidents. How were they misused? The investigators used the Fault Tree to "prove" that the cause they understood (they had seen before - they had experience with) was the cause of the accident.
I've seen many people suggest that Fault Trees are an excellent too for root cause analysis. I learned to use Fault Trees when I was getting my Masters Degree in Nuclear Engineering. I could see how they could be used proactively to have a systematic analysis of potential problems. The logic in the tree appeals to engineers.
But I've seen Fault Trees misused by engineers and others when analyzing the root causes of accidents. How were they misused? The investigators used the Fault Tree to "prove" that the cause they understood (they had seen before - they had experience with) was the cause of the accident.
Recently a reader sent me a link to a Japanese web site that was filled with Fault Trees. To me it once again proved why Fault Trees don't make a very good accident investigation tool.
See the site at: http://shippai.jst.go.jp/en/index.html and see if you agree.
Posted by Mark at September 15, 2005 01:18 PM
Comments
Hey Mark,
Thanks for the link. Interesting looking site, have to read the article on these "failure mandalas" before I come up with a judgment, though. One thing I noticed right ahead in the examples on the frontpage:
1) at places a clear mixing up of cause and effect - a common mistake one alas sees often in these trees (e.g. disregard of procedure will not lead to ignorance, but rather the other way around)
2) lack of a proper taxonomy (no categorization, things from various "levels" thrown together)
3) apparently no proper stopping rule applied
Anyway, these critiques aside... I don't think that tells so much about why the method of fault trees (in case of an investigation I prefer causal trees, by the way) isn't any good. Sure the methods has its limitations, but also huge advantages like the structure and logics.
The problem with fault trees lies (I think) most of the time within the people that build them (as you say: "misuse") in a wrong or incomplete or biased way. It's sad to say that I've observed this also with so called "better" methods like for example Tripod (no Taproot experiences, so far on my part, but I guess that happens just as well). So I've come to the conclusion that there is not one "perfect" method, it's mostly about good competence of the tool you use, critical peer-testing and keeping your mind open.
Just my 2%
C.
Posted by: Carsten Busch at February 2, 2007 03:29 AM