Hotfix generating 0K data from NJOY #905

liangjg · 2017-07-28T15:01:39Z

A tiny bug in generating 0K data from NJOY

import urllib.request
import openmc.data

url = 'https://t2.lanl.gov/nis/data/data/ENDFB-VII.1-neutron/Fe/56'
filename, headers = urllib.request.urlretrieve(url, 'fe56.endf')
fe56_njoy = openmc.data.IncidentNeutron.from_njoy('fe56.endf', temperatures=[0.])
fe56_njoy.export_to_hdf5('fe56.h5')
fe56_hdf5 = openmc.data.IncidentNeutron.from_hdf5('fe56.h5')
fe56_hdf5[1].xs['0K']([1e6])
IndexError: index 3334 is out of bounds for axis 1 with size 2869

It is caused by the action to add 0K elastic scattering cross section from PENDF after running njoy, which overwrites the existing '0K' energy grid. This can be fixed by adding a line above the block

# Add 0K elastic scattering cross section
if '0K' not in data.energy:
  ...

[Not implemented] The error tolerance in reconstruction is currently fixed as 0.001. The users may want a different error in some cases. Should we add a parameter in the APIs to make it possible? I mean to set the err card in input file. For the tempr and errmax, we may remove them to use the default values in njoy. The same changes can be made on broadr. The thnmax (upper limit for broadening) 1.0e6 in the template is recommended to be removed too as the default upper limit for broadening has been changed since NJOY2012.75. But I'm not sure why the values like thnmax and errmax are different in sab case so the changes are not made yet. I'll push the new commit after @paulromano gives more suggestions.

liangjg · 2017-07-28T16:03:16Z

Well, I pushed the changes to make error tolerance a parameter for generating data from njoy. Let me know if there is any problem. @paulromano

paulromano · 2017-07-28T16:48:17Z

Thanks @liangjg, I'll take a look soon!

liangjg · 2017-07-30T15:07:00Z

I found another issue on generating 0K data from NJOY:

nuc = openmc.data.IncidentNeutron.from_njoy('n-018_Ar_036.endf', temperatures=[0.], stdout=True)
"""
 njoy 2016.15  08May17                                       07/30/17 10:45:53
 *****************************************************************************

 reconr...                                                                0.0s

 broadr...                                                                0.1s

 heatr...                                                                 0.1s

 purr...                                                                  0.2s

 mat = 1825                                                               0.2s
      1 of    22 loops done for all temps & sig0s.                       25.6s
      2 of    22 loops done for all temps & sig0s.                       50.2s
      3 of    22 loops done for all temps & sig0s.                       74.6s
      4 of    22 loops done for all temps & sig0s.                       99.0s
      5 of    22 loops done for all temps & sig0s.                      123.3s
      6 of    22 loops done for all temps & sig0s.                      147.7s
      7 of    22 loops done for all temps & sig0s.                      172.1s
      8 of    22 loops done for all temps & sig0s.                      196.5s
      9 of    22 loops done for all temps & sig0s.                      220.9s
     10 of    22 loops done for all temps & sig0s.                      245.4s
     11 of    22 loops done for all temps & sig0s.                      269.8s
     12 of    22 loops done for all temps & sig0s.                      294.3s
     13 of    22 loops done for all temps & sig0s.                      318.7s
     14 of    22 loops done for all temps & sig0s.                      343.2s
     15 of    22 loops done for all temps & sig0s.                      367.6s
     16 of    22 loops done for all temps & sig0s.                      392.1s
     17 of    22 loops done for all temps & sig0s.                      416.5s
     18 of    22 loops done for all temps & sig0s.                      441.0s
     19 of    22 loops done for all temps & sig0s.                      465.4s
     20 of    22 loops done for all temps & sig0s.                      489.8s
     21 of    22 loops done for all temps & sig0s.                      514.3s
     22 of    22 loops done for all temps & sig0s.                      538.8s

 acer...                                                                538.8s
At line 852 of file /home/liangjg/codes/NJOY2016/src/endf.f90 (unit = 24, file = 'tape24')
Fortran runtime error: Bad value during floating point read
"""

The error appears for nuclides with unresolved resonance range. There are a lot of 'NaN' in the output of purr module. I've no idea why NJOY fails on treating probability table at 0K. Any thoughts?

paulromano

Thanks for the PR @liangjg. I also had in mind that it would be useful to specify the error tolerance but then forgot to implement that -- glad you did!

paulromano · 2017-07-31T14:40:49Z

openmc/data/neutron.py

@@ -824,7 +824,7 @@ def from_endf(cls, ev_or_filename):
        return data

    @classmethod
-    def from_njoy(cls, filename, temperatures=None, **kwargs):
+    def from_njoy(cls, filename, temperatures=None, error=0.001, **kwargs):


Let's put this argument on njoy.make_ace instead of here. It will still be possible for a user to specify it as a keyword argument since **kwargs gets passed to make_ace.

paulromano · 2017-07-31T14:41:27Z

openmc/data/thermal.py

@@ -588,7 +588,8 @@ def from_ace(cls, ace_or_filename, name=None):
        return table

    @classmethod
-    def from_njoy(cls, filename, filename_thermal, temperatures=None, **kwargs):
+    def from_njoy(cls, filename, filename_thermal, temperatures=None,
+                  error=0.001, **kwargs):


Same thing here -- move this argument to njoy.make_ace_thermal and let **kwargs pass it if needed.

added several arguments for the modules

liangjg · 2017-08-01T14:14:20Z

@paulromano The argument error has been removed from the upper-level function from_njoy.
Additionally, I tried to add several other arguments to allow omitting certain modules when running NJOY, like:

openmc.data.IncidentNeutron.from_njoy("n-092-U-235.endf", [0.], heatr=False, purr=False)

will skip calling heatr and purr modules. I found this is useful for me to parse pure 0K ce xs for vector fitting.

paulromano · 2017-08-02T14:30:40Z

openmc/data/njoy.py

@@ -50,84 +50,66 @@
    75: ThermalTuple('ouo2', [8016, 8017, 8018], 1),
 }

-
-_PENDF_TEMPLATE = """
+_ACE_TEMPLATE_RECONR = """


We might as well rename all these to just _TEMPLATE_* since they are not specific to ACE now.

paulromano · 2017-08-02T14:34:23Z

openmc/data/njoy.py

@@ -147,6 +130,8 @@ def run(commands, tapein, tapeout, stdout=False, njoy_exec='njoy'):
        Dictionary mapping tape numbers to paths for any input files
    tapeout : dict
        Dictionary mapping tape numbers to paths for any output files
+    print_commands : bool, optional


I think it would be more useful to have the inputs saved to a file. Why don't we have an argument run(..., input_filename=None, ...) that if it's set to a string will save the inputs to a file with that name. So below, you'd have something like:

if input_filename is not None: with open(input_filename, 'w') as input: input.write(commands)

paulromano · 2017-08-02T14:37:20Z

Really nice set of additions @liangjg. These would have been useful to me as I was debugging this module actually! Should be ready to merge after those last two comments of mine are addressed.

By the way, if you just need 0 K cross sections, you could do something like:

import openmc.data

openmc.data.njoy.make_pendf(endf_filename)
data = openmc.data.IncidentNeutron.from_endf('pendf')
elastic_scatter = data[2].xs['0K']

write NJOY input into file rather than printing on screen _ACE_TEMPLATE -> _TEMPLATE

liangjg · 2017-08-02T15:16:02Z

@paulromano Thanks for the suggestion to use make_pendf() and from_endf() .

liangjg added 2 commits July 28, 2017 10:08

skip adding 0K elastic xs if it already exists

35d24a9

made error tolerance a parameter for generating data from njoy

7988bad

liangjg force-pushed the njoy_0K branch from 6d2af9b to 7988bad Compare July 28, 2017 16:00

paulromano self-requested a review July 28, 2017 16:48

paulromano added Data Python API labels Jul 28, 2017

paulromano requested changes Jul 31, 2017

View reviewed changes

liangjg added 3 commits July 31, 2017 21:51

remove the argument "error" out of "from_njoy" function

8bb990b

allow to omit specific modules in generating data from njoy

ac4b8d7

added several arguments for the modules

add argument to support printing NJOY input commands

3f1416a

paulromano reviewed Aug 2, 2017

View reviewed changes

address @paul's comments

18f7ae3

write NJOY input into file rather than printing on screen _ACE_TEMPLATE -> _TEMPLATE

paulromano approved these changes Aug 2, 2017

View reviewed changes

paulromano merged commit f124bec into openmc-dev:develop Aug 2, 2017

liangjg deleted the njoy_0K branch February 28, 2019 17:15

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Hotfix generating 0K data from NJOY #905

Hotfix generating 0K data from NJOY #905

liangjg commented Jul 28, 2017

liangjg commented Jul 28, 2017

paulromano commented Jul 28, 2017

liangjg commented Jul 30, 2017

paulromano left a comment

paulromano Jul 31, 2017

paulromano Jul 31, 2017

liangjg commented Aug 1, 2017

paulromano Aug 2, 2017

paulromano Aug 2, 2017

paulromano commented Aug 2, 2017

liangjg commented Aug 2, 2017

Hotfix generating 0K data from NJOY #905

Hotfix generating 0K data from NJOY #905

Conversation

liangjg commented Jul 28, 2017

liangjg commented Jul 28, 2017

paulromano commented Jul 28, 2017

liangjg commented Jul 30, 2017

paulromano left a comment

Choose a reason for hiding this comment

paulromano Jul 31, 2017

Choose a reason for hiding this comment

paulromano Jul 31, 2017

Choose a reason for hiding this comment

liangjg commented Aug 1, 2017

paulromano Aug 2, 2017

Choose a reason for hiding this comment

paulromano Aug 2, 2017

Choose a reason for hiding this comment

paulromano commented Aug 2, 2017

liangjg commented Aug 2, 2017