NaN appearing in eso files on MJW WinNT, Results changing in 4th figure on MJW Win95 (CR #797)

[Myoldmopar]

Added on 1999-08-20 08:24 by @mjwitte

Description

Executed GeometryTest.idf example file and got "NaN"s in the eso file, e.g.:

Line 563 StarTeam ExampleFiles: 25,19.15081
Line 563 GARD Run: 25,NaN

Same thing happend with TermReheat

Using files from StarTeam:EnergyPlus:LicensedDevelopers:ExampleFiles
GeometryTest.idf [Version #1, 03 Jun 1999 13:27:34
TermReheat.idf [Version #1, 29 May 1999 12:20:48

EnergyPlus from LicensedDevelopers:ProgramFiles
EnergyPlus.exe [Version #3, 26 Jul 1999 12:50:26

**NOTE the build number assigned here is wrong, but I had to choose something . . .

MJWitte 99/09/03
Tested on two machines:

A. WinNT 4.00.1381 on a Dell Dimension XPS R400, Pentium II 400 MHz, 64MB ram

B. Win95 on a Gateway 2000 P5-133, Pentium 133 MHz, 32MB ram

On the WinNT machine, the results were still as noted above with NaNs in the eso file for both TermReheat and GeometryTest.

On the Win95 machine:
The results for GeometryTest were identical to the sample files.
The results for TermReheat had no NaNs, but some values were different in the 4th figure:

 Line 579 StarTeam ExampleFiles:  61,21.50875
 Line 579 GARD:  61,21.50869
 Line 6366 StarTeam ExampleFiles:  61,19.54455
 Line 6366 GARD: 61,19.56093

The eso file for this run (done on 99/08/27) is in the defect files directory as 021-GeometryTest Win95.eso

Tested on the WinNT machine using the new 99.08.27 build and got the same results.

Also, per Linda's suggestion, tried installing VFRUN60I.exe runtime library on the NT machine. This made no difference in the results.

MJWitte 99/09/08
Tested suite of idfs from Test Files:Input Files and compared to results
in Test Files:Output Files:Real DVF

Results seem to point to something solar or shading related . . .

1zoneuncontrolled: identical
ctftestspart1: NaNs in summer d-day, winter d-day identical
ctftestspart2: NaNs in summer d-day, winter d-day identical
dualduct: NaNs in summer d-day, differences in winter d-day
geometrytest: NaNs in summer d-day, winter d-day and annual identical (5/1-5/10)
materialtest: identical
solarshadingtest: NaNs in "PHOENIXWINWSUM" d-day
solarshadingtestann: NaNs throughout annual (7/1-7/7)
surfacetest: identical
termreheat: NaNs in summer d-day, differences in winter d-day
vavsimple: NaNs in summer d-day, differences in winter d-day
windowtests: NaNs in summer d-day, winter d-day identical

MJWitte 99/09/09
Tested with new 99.09.09 exe on both WinNT and Win95. All looks OK now.

Workaround

Did not reproduce (see conditions on Defect 22)

Fix

This was caused inadvertantly by Fred W and new shading calcs. Initializing CTHETA, CosIncAng, and new AOSurf1 and AOSurf2 seems to have solved the problem for the moment. We should look at these calculations and make sure the proper things are being set.

External Ref: 21
Addressed in 99.09.16 Build of EnergyPlus
Last build tested: 99.08.27 Build of EnergyPlus