Deep Stop RGBM Statistical Validation

[Dr Deco]

Hello gcbryan:

Haldane was the chief promulgator of his Haldane Method. Buhlmann wrote about his model. Rodgers and Powell wrote about the PADI tables.

Dr Wienke has the biggest stake in the method – plus the best understanding. The short version of the method for recreational SCUBA is touted by NAUI.

Dr Deco :doctor:

 

[gcbryan]

I'm not a detractor by the way. I'm just trying to get some unbiased info rather than cheerleading info from the gallery or self biased info from the source.

I understand the NAUI connection. I'm sure this is all above board. I'm just asking questions.

At least with Buhlmann and VPM it's all open source and anyone can examine it. With RGBM that's not the case.

I know you can argue that it's the outcome and not the formula that matters but the outcome is empirically observed to a large degree (I think) so any model could result in an outcome that works...they all "work" to a degree.

I think you do need to see the source info to know if it's actually modeling anything.

Buhlmann wrote books that allowed others to use the info in those books to independently run the model. In the case of RGBM many books are written but to my knowledge there is never enough included to allow anyone else to run this model.

I just think it fits the scientific model more if the model is able to be removed from the man and his admirers.

You don't have to admire Haldane or Buhlmann to test the models and get the same results that they report.

That's all.

 

[Dr Deco]

Hello gcbryan:

PADI Tables

I certainly see what you mean. Unfortunately, the basics behind some deco tables are not ever going to be “made public.” When I worked on the PADI tables, they asked me if “the basics” should ever be written/published. I replied, yes, you should allow a complete description. Here is what I explained to them.

“These tables are really quite simple in their starting point and development. If you publish a deco table, it is not very complicated for someone to back calculated and derive the whole thing. The NDLs come from nature, and “What Nature gives to one, she gives to all.” I explained that by publishing, they would be "getting points" from the diving public." [See “References” below]

Decompression Tables

Tables that require decompression stops can be difficult to “reverse engineer.” The US Navy tables were calculated from NDLs and a defined progression of allowed supersaturations [Workman's M-value concept]. If the progression to deeper allowed oversaturations is irregular, you cannot calculate the table yourself but must have the whole, finished table itself. When I was at Ocean Systems Inc., the Schreiner-Kelly model was used for the OSI tables and these tables were kept in a book with a lock.

To make a table requires dive data. This is best is derived from laboratory experiments, but this is not actually feasible because of the high cost involved. There are many depth/time combinations for decompression from four hundred feet for varying bottom times; testing is a daunting task. The parameters in Haldane-type models are the supersaturations allowed for each depth and each half-time compartment. To make a table requires dive data from the field; this is always the only economically feasible scheme.

What Is Behind RGBM Tables?

Decompression algorithms that incorporate preformed nuclei [dual phase] are more complicated because of the added complexity. Here we need diffusion constants, surface tension, size/number distribution of bubble radii, bubble regeneration times/methods, and possibly others. Real dive data collection is difficult and requires assistance from dive agencies. A few dives with DCS are more important than several hundred “clean” dives.

The above-mentioned variables must then be analyzed to give a “best fit” with the model. Fitting algorithms to match variables to the data points are available [maximum likelihood], but you must have the data first. [You have neither the data nor a sufficiently powerful computer.]

Can You Generate RGBM Tables from the Published Model?

You cannot generate the RGBM tables from the published information. This is, I believe, your real question. Without the data to give you the variables, table generation is not possible. You can however look at the values of the parameters as determined by Dr Wienke [published]. Values of the variables are determined [constrained] by the data. If the values of the parameters are not physically reasonable, then you can suspect that the model is being “forced” to accept physically unreasonable values in order to make it work. One possible example of this is the radii of bubbles that can be enlarged by the surrounding supersaturations. There radii are much smaller than would be expected from the Laplace equation for the suspected surface tension. You would expect radii in the 1-micron range while the RGBM uses 1/100 micron ranges.

Conversly, the Michael Gernhardt model introduces the variables with expected values. Within limits, it works.

Thus, you do need the source data but rather only the performance data, the rationale for the fitting constants, and the values of these constants

What is the Outcome of This?

As you state, it is the outcome that matters for a successful table. Tables, however, are derived from empirical data and what is desired is a model that explains and allows extrapolation without testing. The Haldane model was successful for certain ranges but did not perform well when extrapolated to deeper depths and longer bottom times. Additionally, the theory behind the Haldane model has been found to be suspect.

Dr Wienke claims great success for the model. How true is this? As you state, can we see the actual results. Dr Wienke would state that he has supplied it.

Dr Deco :doctor:


References

MR Powell and R Rogers. Gas phase formation as detected by Doppler ultrasound in divers performing multilevel and repetitive dives. The Physiologist, 30 (4), 134, (1987).

MR Powell, MP Spencer, and RE Rogers. Doppler ultrasound monitoring of gas phase formation in divers performing multilevel and repetitive dives. Undersea Biomed. Res. 15, Suppl., 84-85 (1988).

R Rogers and MR Powell. Development of multilevel and repetitive tables for recreational divers. Undersea Biomed. Res., 15, Suppl., 84 (1988).

MR Powell, and RE Rogers. Doppler ultrasound monitoring of gas phase formation and resolution in repetitive diving. Undersea Biomed. Res., 16, Suppl., 69, (1989).

Hamilton, RW, RE Rodgers, MR Powell, and RD Vann. Development And Validation Of No-stop Decompression Procedures For Recreational Diving. Diving Science and Technology. (Pp. 78 + appendix). February 28, (1994)

 

[lowviz]

Thought many of you on SB would be interested in this recently published
statistical validation paper for the LANL deep stop model (RGBM). ...........

Indeed. Good data are hard to come by as a DCS hit on a dive profile is much more informative than a safe dive.

RGBM thinking has a wider acceptance than what is immediately obvious. I know of a Haldane-RGBM hybrid model, the "folded Haldane algorithm", that might even be in my VR3 in some form. The manual states: "VR dive computers use a derivative of the Buhlmann ZHL 16 algorithm modified to assimilate the latest thinking in practical micro-bubble avoidance for enhanced decompression safety." RGBM?

As I understand it, Haldane models a single phase, gas dissolved in liquid. The RGBM model uses two phases, gas dissolved in liquid and free gas (bubbles). RGBM places limits on the acceptable total volume of the gas phase across all bubble radii. Thus, the pure RGBM model is more computationally intense.

RGBM strongly discourages reverse profile diving, which I take as the model being able to predict the hazards of this type of behavior. Reverse profiles are key to my interest in this model as I can't seem to find much about this topic other than "don't do it". -advice that is not always followed. The published description of the phase function restriction for reverse profile diving is difficult to apply to practice. Can reverse profiles be done 'safely' if one further decreases bottom times or suitably alters the deep stops?

Again, thanks for sharing this information. Looking forward to more.

 

[BRW]

Funky556, gkndivebum, lowviz, Randy43068, Saturation;

Many thanks and you are welcome.

gc;

Sorry you didn't connect dots -- they are all out there
in RGBM published papers and books. Maybe try asking
specific questions? BTW, some 25-35 "folks" out
there have successfully implemented RGBM in
codes, tables, software both commercially
and privately using pubs and advice. We get
beaucoup questions weekly. Spare me your
vitriol if you ask.

Mike (Dr Deco);

Mucho thanks for the comments on table building --
plus hands on experiences on same with Rogers and
PADI. Nice work for rec diving for sure. On bub
sizes below a micron, will post some info when
I get back -- monomolecular layers have been
shown to be sufficient for skins, and there
are classes of "bubble-like" junk in submicron
ranges (anti-bubbles that don't float, weird
wraps, etc). Of course, we DON'T know
anything substantive about bubs, sizes,
lifetimes in body, and particularly divers,
so nothing really changes when developing
parameter fits to bubble models on that
side of house.

lowviz;

Good question.. On reverse profiles, the amount
of bubbles excited in excess of the shallower exposure
is added to the bubble phase volume at that point in the
dive. Remember too that the bubble phase volume excited
on the first (shallower) dive has already gone through all kinds
possible changes in size (Boyle expansion-contraction, gas
diffusion into or out of bubble assembly, pressure changes,
etc, etc), so the excited initial phase volume and the
present phase volume (before addition of the excess on the
second dive) are not generally the same. Gets messy.

On the road, back here when I have time to get to it.

BW

 

[hg frogman]

Thanks Dr Wienke for sharing this info.

Concerning deep stops and 170/200 fsw air diving (NEDU and French Navy deep stops experiments), did you experimentally test the LANL staging (I mean, with actual dives) that you present ? If yes, the results would be extremely interesting.

This range of depth is still very casual in Southern Europe for air diving, and not only by the Navies. 200 fsw is the limit (by French law) for commercial diving using air in France, for exemple.

 

[Jrock1062]

great thread excellent information

 

[BRW]

hqfrogman,

You are certainly welcome.

Yes, if you read the paper, you'll note that we have 2879 actual (computer
downloaded) profiles across mixed gas OC and RB systems. Our conclusions
and risk analyses are based on that collection -- not a single air dive to 170/30
(NEDU) without real deep stops, too much time in the mid zone, and then inadequate
deco in the shallow zone, and not a deep stop for a short time juxataposed
on an otherwise shallow stop schedule (200/30 French Navy). You need a consistent,
full up, bottom to top, deep stop staging paradigm in the latter case, and a real
deep stop profile to test in the former case. Check paper for comparative
risk analyses too. In our LANL Data Bank, 80% of profiles come from our
C&C Team ops, and with 23 cases of DCS (chamber treated), including yours
truly. Comments above are of course based on our data and analyses.

It seems that (mostly only) world Navies continue to dive deep air and shallow stops.
Although (see paper) you can make deep stops and shallow stops at the SAME
relative risk levels, deep stop staging is always shorter overall -- so why shallow
stop staging unless you like hanging on a deco line or lift bag, or, of course, miss
the requisite deep stops and are forced to make the shallow stops in penalty
mode (sort of). On deep air, data (COMEX I believe, plus others) underscores
a 10 - 20 times risk increase below 150 fsw compared to shallower air exposures -- not too good for diving deep air according to anybody's staging
protocol?

BW

Thanks Dr Wienke for sharing this info.

Concerning deep stops and 170/200 fsw air diving (NEDU and French Navy deep stops experiments), did you experimentally test the LANL staging (I mean, with actual dives) that you present ? If yes, the results would be extremely interesting.

This range of depth is still very casual in Southern Europe for air diving, and not only by the Navies. 200 fsw is the limit (by French law) for commercial diving using air in France, for exemple.

 

[gjaws]

Perhaps I have my reference wrong, but doesn't the Comex AB model, source of the MT92 table take into account some sort of gas phase (arterial bubbles if I recall)? It doesn't seem to translate into deep stops (I'm no specialist though).

Greg

 

[BRW]

Kevrumbo, Kern, Carlisle, and Jrock

Thanks for the pointers and comments too.

Gjaws

COMEX tables were not originally deep stop (maybe changing somewhat),
and don't recall that arterial model was/is used in their tables (maybe also
changing).

BW

Perhaps I have my reference wrong, but doesn't the Comex AB model, source of the MT92 table take into account some sort of gas phase (arterial bubbles if I recall)? It doesn't seem to translate into deep stops (I'm no specialist though).

Greg