munitor:
Interesting. I'm currently in the process of doing a fault tree analysis for various configurations to determine if there is any significant difference in risk. Considering mechanical reliability as well as human factors.
Resurrecting the issue of reliability, I've finished doing the FTA. I did it twice, once using NPRD-95 data, which is what's used for mil-spec FTA, and once using data from BSAC reports, scaled to match the overall incident rates reported by BSAC and DAN (which agree well). I’ve excluded human error, such as diving with an empty tank, failing to turn on the air, or failing to monitor air supply.
Of course, I take no responsibility for anyone using the information for any purpose whatsoever.
Bottom line is that for a single tank with primary and octopus, the likelihood of getting a rapid failure of air supply (60 seconds or less to no gas) is 1:500 dives to 1:2,500 dives. These numbers are roughly plus or minus 50%, which seems like a lot but is typical in the quantitative risk assessment business.
For a main and backup pony rig, the likelihood is 1:1,000,000 to 1:10,000,000.
In each case the higher
frequencies (edit: lower ratios) are for the BSAC data, which makes sense given their high 1st stage free flow rates (usually due to icing).
As a sanity check, I checked the rates I got doing a component based NPRD analysis against the overall failure rate for oxygen demand regulators in military aircraft service, and they agree pretty well. Unfortunately, the database doesn’t break down the failure modes for the oxygen demand, so had to do the check on overall only.
Just as a information point, if I include the specific ‘failure to monitor gas supply’ (only available from the BSAC information), the likelihood of failure jumps to 1:250 dives.
Since many people consider their buddy to be their backup, I took a look at the data and it appears that the likelihood your buddy will be too far away to help ranges between 0.3% to 4% of the time, depending on whether you use BSAC or DAN incident data. In the end I calculated that the average risk that a diver with a single tank plus octopus rig will be OOA and no buddy available, due to mechanical failures alone, is 1:100,000 dives to 1:50,000 dives. Throw in typical rates for other human error (all those oopsie reasons people end up OOA) and it at least doubles the frequency. With a pony backup, of course, you don't need a buddy for air, so I'm back to the 1:1,000,000 to 1:10,000,000.