Something went wrong with the same card on different computers

Jacob · December 17, 2022, 5:45am

Hello!

I just run the same input card on the different computers, and the version of the OpenMC is 0.13.0 and 0.13.2, respectively. But the combined Keff result of the version 0.13.0 is 0.21117 and the other one is 0.27375. So this is why, is there anything related to the calculating performance of the computer?

paulromano · December 31, 2022, 4:57am

@Jacob Without knowing anything about you problem, it’s hard to say what this could be. My first question to you is: have you looked at the uncertainties on the reported k-eff values? That is, are the differences statistically significant? If you have a model/script you can share that demonstrates this problem, that would allow someone here to look into it further.

Jacob · January 7, 2023, 9:05am

Hello! @paulromano

This is the source of the input card:

github.com

openmc-dev/openmc-notebooks/blob/main/unstructured-mesh-part-i.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Unstructured Mesh: Introduction\n",
    "\n",
    "In this example we'll look at how to setup and use unstructured mesh tallies in OpenMC. Unstructured meshes are able to provide results over spatial regions of a problem while conforming to a specific geometric features -- something that is often difficult to do using the regular and rectilinear meshes in OpenMC.\n",
    "\n",
    "Here, we'll apply an unstructured mesh tally to the PWR assembly model from the OpenMC examples.\n",
    "\n",
    "**_NOTE: This notebook will not run successfully if OpenMC has not been built with DAGMC or libMesh support enabled._**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],

This file has been truncated. show original

The result of the input card on the Github is here:

Reading settings XML file…
Reading cross sections XML file…
Reading materials XML file…
Reading geometry XML file…
Reading U234 from /home/pshriwise/data/xs/openmc/nndc_hdf5/U234.h5
Reading U235 from /home/pshriwise/data/xs/openmc/nndc_hdf5/U235.h5
Reading U238 from /home/pshriwise/data/xs/openmc/nndc_hdf5/U238.h5
Reading O16 from /home/pshriwise/data/xs/openmc/nndc_hdf5/O16.h5
Reading Zr90 from /home/pshriwise/data/xs/openmc/nndc_hdf5/Zr90.h5
Reading Zr91 from /home/pshriwise/data/xs/openmc/nndc_hdf5/Zr91.h5
Reading Zr92 from /home/pshriwise/data/xs/openmc/nndc_hdf5/Zr92.h5
Reading Zr94 from /home/pshriwise/data/xs/openmc/nndc_hdf5/Zr94.h5
Reading Zr96 from /home/pshriwise/data/xs/openmc/nndc_hdf5/Zr96.h5
Reading H1 from /home/pshriwise/data/xs/openmc/nndc_hdf5/H1.h5
Reading B10 from /home/pshriwise/data/xs/openmc/nndc_hdf5/B10.h5
Reading B11 from /home/pshriwise/data/xs/openmc/nndc_hdf5/B11.h5
Reading c_H_in_H2O from /home/pshriwise/data/xs/openmc/nndc_hdf5/c_H_in_H2O.h5
Minimum neutron data temperature: 294 K
Maximum neutron data temperature: 294 K
Preparing distributed cell instances…
Reading plot XML file…
Writing summary.h5 file…
Maximum neutron transport energy: 20000000 eV for U235
Initializing source particles…

====================> K EIGENVALUE SIMULATION <====================

Bat./Gen. k Average k
========= ======== ====================
1/1 0.30539
2/1 0.26134
3/1 0.22212
4/1 0.22179
5/1 0.14131
6/1 0.21170
7/1 0.16877 0.19024 +/- 0.02146
8/1 0.19222 0.19090 +/- 0.01241
9/1 0.17863 0.18783 +/- 0.00930
10/1 0.21289 0.19284 +/- 0.00877
Creating state point statepoint.10.h5…

=======================> TIMING STATISTICS <=======================

Total time for initialization = 3.3074e-01 seconds
Reading cross sections = 3.2297e-01 seconds
Total time in simulation = 1.0950e-02 seconds
Time in transport only = 1.0091e-02 seconds
Time in inactive batches = 5.4269e-03 seconds
Time in active batches = 5.5232e-03 seconds
Time synchronizing fission bank = 5.0570e-05 seconds
Sampling source sites = 3.8890e-05 seconds
SEND/RECV source sites = 6.5900e-06 seconds
Time accumulating tallies = 1.0400e-06 seconds
Time writing statepoints = 7.5274e-04 seconds
Total time for finalization = 1.7600e-06 seconds
Total time elapsed = 3.4390e-01 seconds
Calculation Rate (inactive) = 92133.3 particles/second
Calculation Rate (active) = 90527.9 particles/second

============================> RESULTS <============================

k-effective (Collision) = 0.19269 +/- 0.01294
k-effective (Track-length) = 0.19284 +/- 0.00877
k-effective (Absorption) = 0.24164 +/- 0.01373
Combined k-effective = 0.21117 +/- 0.00443
Leakage Fraction = 0.81400 +/- 0.01122

And this is my result:

Reading settings XML file…
Reading cross sections XML file…
Reading materials XML file…
Reading geometry XML file…
Reading U234 from /home/。。/endfb-viii.0-hdf5/neutron/U234.h5
Reading U235 from /home/。。/endfb-viii.0-hdf5/neutron/U235.h5
Reading U238 from /home/。。/endfb-viii.0-hdf5/neutron/U238.h5
Reading O16 from /home/。。/endfb-viii.0-hdf5/neutron/O16.h5
Reading Zr90 from /home/。。/endfb-viii.0-hdf5/neutron/Zr90.h5
Reading Zr91 from /home/。。/endfb-viii.0-hdf5/neutron/Zr91.h5
Reading Zr92 from /home/。。/endfb-viii.0-hdf5/neutron/Zr92.h5
Reading Zr94 from /home/。。/endfb-viii.0-hdf5/neutron/Zr94.h5
Reading Zr96 from /home/。。/endfb-viii.0-hdf5/neutron/Zr96.h5
WARNING: Negative value(s) found on probability table for nuclide Zr96 at 294K
Reading H1 from /home/。。/endfb-viii.0-hdf5/neutron/H1.h5
Reading B10 from /home/。。/endfb-viii.0-hdf5/neutron/B10.h5
Reading B11 from /home/。。/endfb-viii.0-hdf5/neutron/B11.h5
Reading c_H_in_H2O from /home/。。/endfb-viii.0-hdf5/neutron/c_H_in_H2O.h5
Minimum neutron data temperature: 294 K
Maximum neutron data temperature: 294 K
Reading tallies XML file…
Preparing distributed cell instances…
Reading plot XML file…
Writing summary.h5 file…
Maximum neutron transport energy: 20000000 eV for Zr90
Initializing source particles…

====================> K EIGENVALUE SIMULATION <====================

Bat./Gen. k Average k
========= ======== ====================
1/1 0.26116
2/1 0.23484
3/1 0.27700
4/1 0.16372
5/1 0.29024
6/1 0.16335
7/1 0.32487 0.24411 +/- 0.08076
8/1 0.21856 0.23560 +/- 0.04740
9/1 0.16202 0.21720 +/- 0.03823
10/1 0.23567 0.22089 +/- 0.02984
Creating state point statepoint.10.h5…
WARNING: Output for a MOAB mesh (mesh 1) was requested but will not be written.
Please use the Python API to generated the desired VTK tetrahedral
mesh.

=======================> TIMING STATISTICS <=======================

Total time for initialization = 2.5415e+02 seconds
Reading cross sections = 1.0766e+01 seconds
Total time in simulation = 5.0045e-01 seconds
Time in transport only = 5.9375e-02 seconds
Time in inactive batches = 2.6384e-02 seconds
Time in active batches = 4.7406e-01 seconds
Time synchronizing fission bank = 1.4183e-04 seconds
Sampling source sites = 1.0218e-04 seconds
SEND/RECV source sites = 8.8750e-06 seconds
Time accumulating tallies = 3.7546e-03 seconds
Time writing statepoints = 4.3656e-01 seconds
Total time for finalization = 9.6627e-01 seconds
Total time elapsed = 2.5563e+02 seconds
Calculation Rate (inactive) = 18950.6 particles/second
Calculation Rate (active) = 1054.71 particles/second

============================> RESULTS <============================

k-effective (Collision) = 0.23887 +/- 0.02017
k-effective (Track-length) = 0.22089 +/- 0.02984
k-effective (Absorption) = 0.24047 +/- 0.02812
Combined k-effective = 0.27375 +/- 0.01088
Leakage Fraction = 0.79800 +/- 0.03338

Shengli · July 7, 2023, 8:29am

Do you find the reason for such difference?
We met a similar problem. A test case run with 0.13.0 (PC1), 0.13.2 (PC1), 0.13.0-dev (PC2), and 0.13.4-dev (PC2), 0.13.0-dev has keff ~200 pcm (vs ~10 pcm uncertainty) lower than other three.

paulromano · July 10, 2023, 9:41pm

Looking at the @Jacob’s output above, it looks like the k-effective values themselves have very high uncertainties, indicating that more batches/particles are needed in order to get a better comparison.

For your case @Shengli, there have been numerous bug fixes since 0.13.0 so it’s hard to say what exactly caused your results to be changed between that and the more recent versions. If you are getting consistent answers with recent versions of OpenMC, I would probably trust those.

Topic		Replies	Views
Keff results significantly different from MCNP Validation and Verification	3	543	January 6, 2020
Keff values vs expect values User Support	2	250	April 21, 2022
Neutron flux and reaction rates discrepancies with MCNP Validation and Verification	2	185	January 19, 2024
Different k-eff in RESULTS and calculation User Support	14	1278	August 3, 2022
[Need help] Keff results significantly different from SCALE6.1 User Support	2	229	September 5, 2023

Something went wrong with the same card on different computers

Related Topics