Please wait...
×

Error

Posted Thu, 04 Oct 2012 14:54:33 GMT by Magnus Rohde
Hi

I am trying to set up a inverse calibration of a transient problem with Pest(standalone). I run Feflow with the following command
  [font=courier]feflow61c -work ..\  -steps steps.pow -dar base_ascii.dar base_ascii.fem[/font]
and I use the -steps option to control the output times in the .dar-file. This works fine 99.9 % of the time, but eventually Feflow will miss an output time with a very small amount e.g. use 79.99999 days instead of 80.00000 days. This will cause Pest to stop, since it is looking for an observation at 80.00000 days.

My Pest instruction file looks like this:
[font=courier]pif %
%RESULTS% %TIME = 8.000000e+01%
%----------%
%----------%
l6 [obswell1_1]4:18[/font]

And Feflow produces a .dar-file like this:
[font=courier]***************************************************
RESULTS AT STEP =  64  AND TIME = 7.999999e+01 [d]:
---------------------------------------------------
Obs    H [m]      Vx [m2/d]      Vy [m2/d]
---------------------------------------------
  1 -1.452183e-01 -1.798366e-02  1.786912e-02
  2 -3.418914e-01 -6.729764e-02  5.468340e-02
  3 -5.286261e-01 -1.310629e-01 -8.419374e-02
  4 -5.296815e-01 -1.313571e-01 -8.475472e-02[/font]

Time steps is supplied like this

[font=courier]# 1
! Timesteps
.
.
66.36    0.01
80    13.64
80.01    0.01
.
END
END[/font]

I cannot use step number to identify output at a specific time since I use automatic time step control and the number of steps can change from model run to model run.

Is there a way to force Feflow to use the timesteps specified with -steps? Alternatively can the output times in the .dar file be written with fewer decimals?



Best regards
Magnus
Posted Fri, 12 Oct 2012 01:29:21 GMT by psinton@aquageo.us
write a preprocessor to fix the dar before sending to pest?
Posted Tue, 16 Oct 2012 13:45:22 GMT by Magnus Rohde
Yes, that turned out to be the solution.
Posted Thu, 18 Oct 2012 11:19:04 GMT by Giovanni Formentin
It just sounds strange to me that PEST compares two real numbers by simple equality, not considering a tolerance around the desired value.

You must be signed in to post in this forum.