生物工学および生物医学科学ジャーナル

Deep Stop Model Correlations

Abstract

Wienke BR

Data correlations of computer, table, and software real world implementations of useful and popular diving models are warranted for surety, testing, reproducibilty, and safety. Model correlations are of broad interest across the diving community and the focus of this communication. Permissible supersaturation is a fundamental model element for correlation and we analyze four popular ones, namely the USN, ZHL16, VPM, RGBM model permissible supersaturations within model dynamical constraints. Correlations are obtained in statistical likelihood analysis from computer profile records with DCS outcomes in the Los Alamos National Laboratory Data Bank (LANL DB). Permissible supersaturations, limited by model staging constraints varying across depths, times, and gas mixtures, are quantified for the four models. Parameters and risk functions useful to estimate profile risk are also obtained. To correlate and fit data, a modified Weibull-Levenberg-Marquardt routine is employed across 2994 computer downloaded (only) profile records with 23 cases of DCS in nitrox, trimix, and heliox deep and decompression diving. The routine is useful for low probability (low−p) data usually encountered in the diving accident arena. Model agreement with data is χ2 significant as follows, using the logarithmic likelihood ratio of data set to fit set:

USN-(χ2=0.081)

ZHL16-(χ2=0.131)

VPM-(χ2=0.717)

RGBM-(χ2=0.861)

LANL DB computer profiles exhibit very low DCS prevalence and correlate well with the deep stop models, VPM and RGBM, and further manned testing is always welcome. This correlation suggests that dive computers, software and tables based on deep stop models like VPM and RGBM can safely be used by sport and technical divers. The shallow stop models, USN and ZHL, have, of course, been used safely in computers, tables, and software for decades while deep stop models are fairly new on the diving scene.