Resinsight will not read results from opm-upscaling

[joakim-hove]

[This issue was originally created by @laods in the Resinsight repository. The problem is not really with ResInsight but rather opm-core / opm-upscaling / opm-porsol and ERT. I have therefor moved the issue here. Joakim]

This issue has been discussed via e-mail for some time (at least @joakim-hove, @hhgs, @magnesj and myself has been involved), but for the sake of openness it really belongs to the GitHub issue tracker. I choose to open it here, but it is not clear to me whether this issue belongs here or in another OPM module, e.g. opm-core or opm-upscaling.

The problem is that not all eclipse output files from the upscaling application steadystate_test_implicit are possible to open in ResInsight. The produced EGRID file opens fine, but the UNRST files are not possible to open. Three issues have been identified:

* The basename (without extension) of all eclipse output files must be identical.
* ERT crashes when time stampes are prior to 1970.
* The number of active cells in the EGRID files does not match the number of parameter values in the UNRST files.

By this issue I do not intend to push those of you already working on it, but it is probably for the best to let the hole community of OPM be aware of it.

[atgeirr]

I think (correct me if I am wrong) there are underlying issues with ert usage and resinsight that have not been worked out yet. I therefore suggest this is postponed until after the release.

[joakim-hove]

I think (correct me if I am wrong) there are underlying issues with ert
usage and resinsight that have not been worked out yet.

No - actually as of: Ensembles/ert#203 the only
remaining issue (as I am aware) is about the organisation of restart files
from opm-upscaling. As far as I understand the issue in ERT/Resinsight was
only on Windows / 32bit POSIX.

Joakim

[atgeirr]

Thank you for the clarification, @joakim-hove. Does this have a quick fix, then? Or do we need some more in-depth changes to make this work? I am now thinking more of a workable solution for now than a long-term, robust and flexible solution.

A note to others with an interest in this: @joakim-hove, @bska, myself and others have discussed how to create a long-term solution.

[joakim-hove]

Thank you for the clarification, @joakim-hovehttps://github.com/joakim-hove.
Does this have a quick fix, then? Or do we need some more in-depth changes
to make this work? I am now thinking more of a workable solution for now
than a long-term, robust and flexible solution.

As I see it this only has one solution, which is to reorganize the output
of restart files from the upscaling program. The problem is that the
ECLIPSE convention is quite strict.

All related files (i.e. INIT file, restart files, .DATA file, symmary
files, EGRID file) must be located in the same directory and follow naming
conventions which amounts to the same basename and correct extension.

Below is a e-mail i sent to Lars Odsæter (in Norwegian) with a suggestion
for 100% manual fix:

Bra;

1.   Som du sier er grid file fortsatt I ASCII format, og det er ikke
   

   noen INIT fil. Vi får lage den PR’en en gang til. INIT filen kan vi jo ikke
   trylle frem, men du kan i hvert fall konvertere FEGRID filen til en EGRID
   med f.eks. convert.x:

convert.x testModeld.FEGRID

1.   ECLIPSE har ganske sterke navnekonvensjoner, og jeg tror ikke det
   

   resultatsettet du lister opp her vil fungere:

a. Filene produsert av ECLIPSE skal ha uniform case oppførsel, enten
all lower case eller all upper case. Jeg tror det vil være tryggest å bare
gå all-upper-case; det er i hvert fall det vanligste.

b. Resinsight (og annen slik programvare) er typisk basert på filnavn.
Du velger EGRID filen i en «Open File» dialog, og ResInsight vil loade alle
INIT/RESTART filer med matchende navn. I eksempelet nedenfor så vil i
praksis ikke Resinsight finne noen ting utover griddet.

En skikkelig løsning på 2B krever at de forskjellige
oppskaleringskjøringene havner i hver sin katalog, eller har hvert sitt
base navn og en symlink (med riktig navn) til en felles EGRID fil, eller
lignende. For å teste foreslår jeg følgende:

1.   Lag en katalog results med underkataloger 0.100, 0.3125, 0.525, ….
   

2.   Kopier testModel.EGRID (etter at du har kjørt konvertering FEGRID
   

   -> EGRID) til results.

3.   Kopier ecldump-steadystate-0.1000-x-100000….UNRST ->
   

   results/0.1000/UPSCALE.UNRST

4.   Kopier ecldump-steadystate-0.3125-x-100000….UNRST ->
   

   results/0.3125/UPSCALE.UNRST

5.   ….
   

I hver av mappene lager du en symlink UPSCALE.EGRID -> testModel.EGRID.
Dermed skulle du ende opp med fem mapper som alle inneholder en
UPSCALE.EGRID (link) og en UPSCALE.UNRST resultatfil; hver av disse mappene
skulle så kunne åpnes i Resinsight, eller for den saks skyld
Floviz/Petrel/RMS/…. På relativt kort sikt bør opm-upscale endres litt slik
at resultatene kommer ut i en form som er funksjonelt ~ekivalent med det
som er beskrevet over.

Joakim

[atgeirr]

Thank you very much for your explanation. I guess we should make this an issue for the next milestone since it is important, but requires some work and (not least) quite a bit of testing -- maybe also against other tools than ResInsight.