Limits on the parameters of VPM ?

[Patoux01]

Since the thread on Deep stops has been closed (and this question removed), and I had a relatively simple question about the VPM model, I'll just quote myself. This was a response to a claim that there could be no such thing as VPM-B+7 (+7 being a conservatism setting that is out of range on commercial planning software without fiddling around)

So VPM-B+3 is fake as well, I assume, since it's a fudge adding conservatism?

Edit: If VPM-B +3 is "real and existing and proven", where is the limit in conservatism in terms of parameters of the model, and why are those limits what they are? (since VPM is "internally consistent" or something like that by your words, you should be able to tell us why, physiologically, those limits are set)

Basically, VPM relies on a few parameters, listed below (happy to be corrected if I'm wrong, I don't seem to be able to get the FORTRAN code from decompression.org)

Altitude_Dive_Algorithm='OFF' !Options: ON or OFF
Minimum_Deco_Stop_Time=1.0 !Options: real positive number
Critical_Radius_N2_Microns=0.8 !Adj. Range: 0.2 to 1.35 microns
Critical_Radius_He_Microns=0.7 !Adj. Range: 0.2 to 1.35 microns
Critical_Volume_Algorithm='ON' !Options: ON or OFF
Crit_Volume_Parameter_Lambda=7500.0 !Adj. Range: 6500 to 8300 fsw-min
Gradient_Onset_of_Imperm_Atm=8.2 !Adj. Range: 5.0 to 10.0 atm
Surface_Tension_Gamma=0.0179 !Adj. Range: 0.015 to 0.065 N/m
Skin_Compression_GammaC=0.257 !Adj. Range: 0.160 to 0.290 N/m
Regeneration_Time_Constant=20160.0 !Adj. Range: 10080 to 51840 min
Pressure_Other_Gases_mmHg=102.0 !Constant value for PO2 up to 2 atm

Anyone able to tell me why we would be unable to force the parameters Critical_Radius_N2_Microns, Crit_Volume_Parameter_Lambda, Surface_Tension_Gamma, or any of the others to whatever value we want?

This question was initially meant for rossh after many claims that VPM is consistent while GFs are not, but if anyone is able to answer it, or provide some sort of hypothesis, it'd be quite appreciated.

(And I do hope I'm not violating any rules....)

 

[UWSojourner]

Since the thread on Deep stops has been closed (and this question removed), and I had a relatively simple question about the VPM model, I'll just quote myself. This was a response to a claim that there could be no such thing as VPM-B+7 (+7 being a conservatism setting that is out of range on commercial planning software without fiddling around)




Basically, VPM relies on a few parameters, listed below (happy to be corrected if I'm wrong, I don't seem to be able to get the FORTRAN code from decompression.org)


Anyone able to tell me why we would be unable to force the parameters Critical_Radius_N2_Microns, Crit_Volume_Parameter_Lambda, Surface_Tension_Gamma, or any of the others to whatever value we want?

This question was initially meant for rossh after many claims that VPM is consistent while GFs are not, but if anyone is able to answer it, or provide some sort of hypothesis, it'd be quite appreciated.

(And I do hope I'm not violating any rules....)

The VPM-B+7 conservatism setting is just fine. It actually aligns pretty well with Ross's own VPM-BE+5 for some profiles. See here.

Baker's code for VPM-B has a check of the critical radius:

You can see that Baker checked to ensure the critical radius was below 1.35 microns. The VPM-B+7 setting is just over 1 micron (~1.03), so it would have run just fine in Baker's code.

There's no merit to the claim of VPM-B+7 being fake. It's simply more conservative than Ross's chosen 0-5 conservatism settings, but it's clearly VPM-B.

The only reason for the claims that +7 somehow is not VPM-B, or not a valid VPM-B profile, is because VPM-B+7 has the same run time as the NEDU profiles that were tested. And it's been shown that VPM-B+7 is clearly pretty close (not identical) to the deep stop profile that had higher DCS risk (see here and here). So the "fake" claim is just an attempt to distance VPM-B from the NEDU study.

I can send you the VPM-B code if you PM me an email address.

 

[leadduck]

ftp://decompression.org/Baker/VPM-B%20Fortran%20Source%20Code.txt
Login: downloadfiles@decompression.org
Password: decompression1

 

[leadduck]

Anyone able to tell me why we would be unable to force the parameters Critical_Radius_N2_Microns, Crit_Volume_Parameter_Lambda, Surface_Tension_Gamma, or any of the others to whatever value we want?

This question was initially meant for rossh after many claims that VPM is consistent while GFs are not, but if anyone is able to answer it, or provide some sort of hypothesis, it'd be quite appreciated.

Define "consistent" ...

The usual purpose of a model is: interpolate between the data points that were used to fit the model. There may be mathematical limits on model parameters such as "Critical Radius must be positive", just to avoid trivial divisions by zero. There should be limits on the input variables to the model (max.depth, bottom time, gases), because a model shouldn't be used to extrapolate way beyond the experimental data points that were used to fit it.

But if you didn't use experimental data points to fit the model, what meaning do you expect from limits on model parameter values? Where should they come from other than out of thin air as well?

 

[boulderjohn]

---

---

A ScubaBoard Staff Message...

The reason the other thread was closed was because of the flurry of personal attacks. In some cases, posts that responded to other posts that had been deleted due to personal attacks had to be deleted as well because of the reference to a deleted posts. It was a very difficult process to separate the wheat from the chaff, and I sincerely hope that we can discuss this important topic while staying on the topic and not the individuals involved.

The concern is the fallacy of the ad hominem attack, in which the character of a person is attacked rather than the argument at issue. That sounds simple enough, but it is actually tricky, because not all personal attacks are fallacies. There are cases in which it is legitimate to question the credentials of an individual making a case, and it is possible to question the motives of a person making a case. For example, when Andrew Wakefield was shown to have a major financial interest in getting people to believe that autism was linked to vaccinations, it severely undermined his credibility by showing a motivation for misreporting data, which he also was shown to have done. With both of those problems clearly demonstrated, ad hominem comments became legitimate matters of discussion. (See the link in this paragraph for further discussion of this.)

In the previous thread, words like fake and phony were red flags because of the implication of an attack on an individual's character. If two people look at data and draw different conclusions, that is a disagreement, and one can question the legitimacy of an opposing interpretation. To say that a position is fake or phony, on the other hand, means that the other's position is a deliberate misrepresentation with the intent to deceive, and it is thus a personal attack. As stated in the last paragraph, if you have clear information showing that the opposing view is in fact a deliberate attempt to deceive, then that is a legitimate point, but you had better be able to prove that intent to the satisfaction of others, especially the moderators. If someone has accused you of such intent, you of course will have the right to respond.

Please be careful in this thread--don't make it personal!

---

 

[ajduplessis]

VPM is limited like every other model as they all have parameters that are based on a lot if's, but's and maybe's. None of the models or approaches is a guarantee of a successful outcome. This is the limiting conundrum IMO

What I do know is that none of the VPM software I own and use produce conservatism setting anything higher that +5 nor do they produce profiles that come close to what the NEDU guys are using.

I have dived both VPM, GF, RGBM, GAP and a few more with the same levels of success for same profile dives.

 

[leadduck]

VPM is limited like every other model as they all have parameters that are based on a lot if's, but's and maybe's. None of the models or approaches is a guarantee of a successful outcome.

VPM does not play in the same league with other models but is special because its parameters were not fitted and validated experimentally. Although no model can guarantee a successful outcome, it's reasonable to use models with an experimentally validated high probability of a successful outcome.

 

[Patoux01]

Thanks all for the reply, I was a bit busy doing "homework" like some call it.

The VPM-B+7 setting is just over 1 micron (~1.03), so it would have run just fine in Baker's code.

Correct if I'm wrong (This also means that you @ajduplessis are allowed to tell me why this would be wrong), as I'm combining info from multiple sources here to make my reasoning.
From MultiDeco VPM & VPM-B & VPM-B/E & ZHL GF dive decompression software for technical divers I get

Inside the program, the conservatism setting increases the Critical Radii of N2/He in the VPM algorithms. From these base values, by this increase: 1 = 5%, 2 = 12%, 3 = 22%, 4 = 35%, 5 = 50%. The Critical Volume setting is on by default.

From the same source I get

This means that the nominal critical radius for N2 is 0.55, and for He it is 0.45 in VPM-B.

From ftp://decompression.org/Baker/VPM-B%20Fortran%20Source%20Code.txt
I get

RADIUS MUST BE BETWEEN 0.2 AND 1.35 MICRONS

So now, if I'm not mistaken, by setting the critical radius to 1.03 (I'll have to trust UWSojourner here, I can't find the original post, it's somewhere out there, in a few thousand pages of debate):
- we have an increase of 90% roughly
- we are in the scope defined by Baker (and I don't think we can say Baker made VPM wrong, did he?)

I wild guessed that a setting of +6 would be around 70% increase in critical radius, and plotting the critical radius vs the conservatism settings (including the guessed +6 and the "maybe or maybe not real" +7 conservatism), I get this

While it does show some exponential-like behaviour, it doesn't seem too bad of a guess, adding a few "intermediate" conservatism settings (eg making +6 = 65%, +7 = 80% and +8 = 90%) would make it look like a linear increase (not that it matters much).


So, until now, there's nothing wrong with VPM-B+7, except maybe that it could have been called +8 if we wanted to make the increase of radius vs conservatism more linear (would we want that, that's another debate).

(More investigation coming as soon as I'm able to run the original code and see what happens when running the NEDU dive at maximum conservatism of VPM-B and see what comes out)
Thanks @leadduck for the correct login info, I couldn't find the right ones. I guess I don't need your version then UWSojourner.

Define consistent

I guess I mean something in the sense: "If I want more conservatism, I get more conservatism", not "If I want more conservatism, at some point I get a massive spike in % of hits".

Basically in the sense of what rossh said:

When we add conservatism, it lowers both tissue supersaturation pressures in mid ascent, and surfacing tissue pressures. Both of those are considered to be leading causes of stress and risk, and conservatism lowers both of those. But total decompression stress and risk comprises of many more parts than just tissue gas pressure stress.

Your favorite gradient factors also work in precisely the same manner - by lowering the tissue supersaturation pressures for both ascent and surfacing. i.e. changing the allowable M-values lower.

Increased critical radii == reduced supersaturation pressures == lowered gas pressure stresses.

so increased critical radii == increased conservatism, without a breaking point (or at least, the breaking point should be the limit of the algorithm, not something well within the defined limits)


Basically, @ajduplessis , assuming VPM-B+7 has only set a critical radius of 1.03 (I have read 1.01 somewhere else as well), why would it not be a possible value?


I guess that's it for my homework for today...