[PATCH] gas-model: add R compressibility script

Lubomir I. Ivanov neolit123 at gmail.com
Wed Mar 9 17:32:58 PST 2016


On 10 March 2016 at 02:40, Linus Torvalds <torvalds at linux-foundation.org> wrote:
>
>
>
> On Wed, Mar 9, 2016 at 4:20 PM, Linus Torvalds <torvalds at linux-foundation.org> wrote:
>>
>> and the plot doesn't look that far off what we have now.
>
>
> To verify that, I just plotted our old helium curve on top of the same plot:
>
>
> there *are* actually two lines there if you look closely, although it really just looks like the line gets a bit fatter up at roughly 550 bar.
>
> So I think my R math is correct
>

that's certainly an alternative to the linear mixing of the two polynomials.

in terms of which one is the better fit i have one observation. if add
the data from bauer.com for helium (which pretty much looks to be for
a higher T than 300K) and plot @ google:
1.0 + 4.87320026468e-04*x - 8.83632921053e-08*x^2 +
5.33304543646e-11*x^3, 1.0 + 0.0004796109868797936*x -
0.00000004077670019935*x^2 + 0.00000000000077707035*x^3,
0.9998885418429937 + 0.0005435539837400252*x - 2.3960530861159e-7*x^2
+ 1.999279624e-10*x^3 + 6.760135e-14*x^4

^ 1) LS fit from the data of both temperatures, 2) linear mix of the
two individual polynomials, 3) the bauer.com polynomial

i observe a tendency for Z to rise more with the upper pressures (we
need to look in the < 300bar range, due to the bauer.com cap), and
that remains true for the linear mix of polynomials, while with the LS
fit of the two temperatures there seems to be a less steeper rise for
Z at higher pressures.

i'm pretty sure that i can think of a data set that will render both
methods non-optimal, so i wonder which one is the better fit in the
1-500bar case.
we can't really be sure unless we have the actual experimental data @ 300K.

it's probably a good idea to update the .C file coefficients, so that
any viewers can trace them back to the .R file which holds "the
proof", essentially.

lubomir
--


More information about the subsurface mailing list