Comparing Calculated and TSUNAMI-IP E for a Set of SDF Files

This example demonstrates the use of tsunami_ip_utils for calculating the similarity index \(E\) for a set of TSUNAMI-IP SDF files and comparing to the TSUNAMI-IP calculated value. This example uses a set of 12 SDF files generated from the highly enriched uranium fast metal (HMF) series of critical experiments [Bess2019].

from tsunami_ip_utils.comparisons import E_calculation_comparison
import pandas as pd
import pickle
from paths import EXAMPLES

# Create a list of application and experiment SDF files (same set of applications and experiments)
applications = [ f"{EXAMPLES}/data/example_sdfs/HMF/HEU-MET-FAST-003-00{i}.sdf" for i in range(1, 3) ]
experiments = applications
data = E_calculation_comparison(applications, experiments)
E_total_comparison = data['total']

Done with calculations
Creating pandas dataframes

The output is a pandas dataframe. The application is the first index, and can be accessed via passing the application number as an index. For example, to access the first application:

print(E_total_comparison[1])

Attribute                    Calculated Manual Uncertainty  ... Relative Difference in Mean Relative Difference in Manual Uncertainty
Experiment Number                                           ...
1                  1.000E+00+/-1.71E-03           1.71E-03  ...                  3.3307E-15                                2.8798E-01
2                  9.983E-01+/-1.72E-03           1.72E-03  ...                  2.8934E-04                                2.8323E-01

[2 rows x 5 columns]

From the output, you can see that each application has a few attributes that are calculated. The attributes are:

print(E_total_comparison[1].keys())

Index(['Calculated', 'Manual Uncertainty', 'TSUNAMI-IP',
       'Relative Difference in Mean',
       'Relative Difference in Manual Uncertainty'],
      dtype='object', name='Attribute')

To access, for example, the calculated E value, you can pass the appropriate key

print(E_total_comparison[1]['Calculated'])

Experiment Number
1    1.000E+00+/-1.71E-03
2    9.983E-01+/-1.72E-03
Name: Calculated, dtype: object

This gives the calculated E value for application 1 compared to all other experiments. To select a specific application experiement pair, you can pass the experiment number as the second index. For example, to access the calculated E value for application 1 compared to experiment 2:

print(E_total_comparison[1]['Calculated'][2])

9.983E-01+/-1.72E-03

Now the pandas dataframe comparison for each of the \(E\) types (i.e. \(E_{\text{total}}\), \(E_{\text{fission}}\), \(E_{\text{capture}}\), and \(E_{\text{scatter}}\)) can be written to an excel file. The gold standard data is also provided in the gold directory.

# Save output data as xlsx
for E_type in data.keys():
    data[E_type].to_excel(f"{E_type}_comparison.xlsx")

# Test that the calculated E comparison is the same as the gold standard
for E_type in data.keys():
    dumped_data = pd.read_excel( f"{E_type}_comparison.xlsx"  )
    gold_data   = pd.read_excel( f"{EXAMPLES}/gold/{E_type}_comparison.xlsx" )

    assert dumped_data.equals( gold_data )

References

[Bess2019]

J. Bess, “International handbook of evaluated criticality safety benchmark experiments (ICSBEP),” Organization for Economic Co-operation and Development-Nuclear Energy Agency Report NEA/NSC/DOC (95) 3 (2019).