As many practitioners of chemical process risk analysis know, LOPA is a simple tool for simple problems. In many cases, the simple rule sets that guide LOPA yield results that are patently absurd when you step back and look at them with a Lucid Eye (no reference to the paintings of Verneer intended). As I have noted on several occasions, LOPA is a simple tool for simple problems, and by no means an infallible golden rule for safety. Risk analysis, in general, is an art of estimation. With all of the uncertainty and variance associated with risk analysis, none of its practitioners ever claim to know anything with any degree of certainty. Instead, we use various levels of conservative assumption to attempt to “bound” the upper limit of the risk. If these conservatively bounded assumptions yield an answer that a process plant can live with, we simply implement the recommendations and move on.
The problem comes when one implements a the recommendations of a LOPA that are:
1. Excessively expensive.
2. Inconsistent with standard design practices
3. Inconsistent with the actual operating history of a facility.
Consider an example. If you’re LOPA determines that an event is occurring once per year, and resulting in a fatality – it will probably result in high risk reduction requirements, most likely quite expensive to implement. When this result occurs, it is incumbent upon the team implementing the solution to ensure that the result of the LOPA are realistic. This can usually be done with a simple check of the history of the facility. At Kenexis, we refer to this type of study as a “focused QRA”. In this type of study we direct and concentrate the tools of QRA on a single specific scenario.
Going back to the previous scenario, if a LOPA determines that extensive safeguarding is needed because an event occurs annually that results in a fatality, this is quite easy to check against the actual operating history of the facility. If the facility has been in operation for 10 years, then 10 fatalities should have occurred. If that is not the case, either reality is lying, or the model is. My guess, is that the model is inaccurate, because reality tend to be, in a word, real. At that point if a trained analyst brings out more sophisticated tools, to source of the overly pessimistic assumptions are readily identified. And hopefully, a more realistic design will be implemented.