Decompression Oddities

[gcbryan]

After playing around with the Buhlmann ZHL-16A algorithm I find a few odd (to me) things. The "C" version which been tweaked with just a bit in the middle compartments is used in many dive computers. If you look at the NDL for 100 fsw it's rather conservative at 15 minutes. At 60 fsw it's not the most conservative nor the most liberal at 56 minutes.

In deco without any modifications though it doesn't make sense. Even though it's NDL at 100 fsw is 15 minutes if you exceed that and stay for 26 minutes the penalty is only 6 minutes of deco...less than the number of minutes you exceeded deco!

If you compare 200 foot/20 minute dives with Vplanner even on nominal it will have you out of the water twice as fast (as V-planner).

I think this just goes to show that really there isn't much science behind decompression and it's all just tweaked by how many people got bent at what time/depth combination.

I know it's often said that we don't really know everything behind what goes on with decompression but it actually looks like we don't know anything.

Am I looking at this incorrectly for anyone "in the know"?

I'm been playing around with writing a basic decompression program just for the learning experience lately and so have really been working with the numbers of this particular algorithm closely and although interesting it's actually kind of shocking just how little science there appears to be behind all this.

Of course, I'm sure the practical answer is that the algorithm has been tweaked beyond all recognition to make it more sane than it appears but it that case it isn't modeling anything.

 

[Blackwood]

First, I'm not "in the know," so these are just my impressions:

I'm assuming you didn't include gradient factors, so you are essentially running 100/100. In other words your algorithm goes up at Ascent_Rate until any of your 16 compartments reaches its M-Value.

One major difference between bubble models and dissolved gas models is that bubble models (including VPM) assume that you ALWAYS have bubbles, and the ascent profiles are designed with "bubble mechanics" in mind to keep them at reasonable sizes (growth controlled by ambient pressure directly, offgasing from tissues into bubbles, surface tension, etc.).

Dissolved gas models, on the other hand, assume that you develop bubbles only once pressure gradients exceed some critical value, and the game is to drive that gradient as close to level as possible without exceeding it, thus moving up and out of the water ASAP to complete decompression where it's safest (on the surface during the interval).

All that in mind, it doesn't surprise me that bubble models call for dramatically more decompression. Applying gradient factors (which limit both M-Values and how much of the full ascent to an M-Value you are willing to jump without stopping) to your algorithm will help bring them more in line with bubble models (at least in terms of TIME, not necessarily SHAPE).

 

[gcbryan]

Yeah, I get that. It's just surprising that you have to "tweak" the algorithm to the point you might as well not have an algorithm as it's not predicting anything.

Any "model" will work if you fit the end result with reality. A model predicting DSC based on how many hamburgers I had for lunch will work as long as I don't stay longer than 20 minutes at 100 fsw or as long as I keep reducing allowable time as more and more people are bent.

Of course, as you mentioned the big difference in time is due to the deeper stops of V-Planner but most people claim that RGMB models aren't necessarily longer but rather more efficient since they can spend less time shallow.

If you tweak both models to match the same outcome then of course they are both about the same but then there is no true "modeling" going on.

Again, I especially find it troublesome that you can exceed the NDL's by11 minutes and only be called upon to do 6 minutes of deco! Even the model doesn't assume that this is a linear process (except in the one instance that I noticed).

It's kind of flakey (to me) more or less like the "rule of 120"...it works at some depths better than others

 

[Mr Carcharodon]

Yeah, I get that. It's just surprising that you have to "tweak" the algorithm to the point you might as well not have an algorithm as it's not predicting anything.

Any "model" will work if you fit the end result with reality. A model predicting DSC based on how many hamburgers I had for lunch will work as long as I don't stay longer than 20 minutes at 100 fsw or as long as I keep reducing allowable time as more and more people are bent.

Of course, as you mentioned the big difference in time is due to the deeper stops of V-Planner but most people claim that RGMB models aren't necessarily longer but rather more efficient since they can spend less time shallow.

If you tweak both models to match the same outcome then of course they are both about the same but then there is no true "modeling" going on.

Again, I especially find it troublesome that you can exceed the NDL's by11 minutes and only be called upon to do 6 minutes of deco! Even the model doesn't assume that this is a linear process (except in the one instance that I noticed).

It's kind of flakey (to me) more or less like the "rule of 120"...it works at some depths better than others

Observations were made, models built, and adjustments have been made to the models for new observations. Still more observations were made and new models were developed. That’s not science? It seems to work pretty well since most of the people I hear about going to chambers have broken the rules. Something is being modeled. Gas does dissolve in tissues and is subject to diffusion and perfusion. Bubbles do form. It is fair to say that dissolved gas models may not work very well outside of the range they were tested for. Bubble models may work better, and may be more fundamental, but again there has not been a lot of testing to validate that. Between having variable levels of acceptable bends, having a fuzzy definition of what a bend is, having physiological variability and a desire for operational simplicity it is not too surprising that the line will be fuzzy and wide.

The goals have shifted over time. Haldane was mostly concerned with pain, paralysis and death, but now we are doing Pyle stops and gradient factors to feel better and avoid fatigue.

So 1:1 deco was not done, where was that written in stone? That sounds like a ratio deco rule. But RD, and the other tables and meters have elements that make them operationally simple. Operational simplicity, safety and efficiency at getting out of the water are trade offs. Dropping 5 minutes off your deco might increase your chances of DCS from .001 to .01 but your chances are still pretty good for any dive, and you get of the water faster. The right answer might depend on the situation.

 

[rossh]

Hi,

You need to appreciate that none of these models are a true digital representation of the human body - they never will be. If such a thing was ever possible to create, it would be put to more important uses such as accurate medical predictions of health issues.

Deco models take a mathematical theory that is thought to closely match the physiological outcomes. They are calibrated to match repeatable data points, which gives them the ability to predict outcomes for more points in a profile.

Different theories have evolved about how to predict the outcome, which leads to different models. There are also several different approaches to basic deco practices, and these are reflected in the different model designs.

Regards