sfv

Small library for predicting variations in the scalar neutron flux using perturbation theory

This repository holds the code necessary to build the sfv Python package. The functionality is quite small, providing three simple functions:

• sfv.applySFV: Apply the prediction to obtain normalized scalar flux
• sfv.getAdjFwdEig: Extract k modes of the foward and adjoint fission source and associated $k$ eigenvalues from the fission matrix
• sfv.lib.predict_spatial_flux: Direct access to the SFV compiled library with a similar API as sfv.applySFV, but with little to no safety checks.

Installation

Installing sfv requires numpy and f2py in order to compile the FORTRAN library. The preferred way to install is using a virtual environment, as this produces a system that is easier to upgrade in the future. Using the venv module, the steps to create a virtual environment are

$ python -m venv /path/to/venv
$ source /path/to/venv/bin/activate
(venv) $ which python
/path/to/venv/bin/python

Note that python >= 3.5 is supported for this project.

Dependencies can be installed with

(venv) $ pip install -r requirements.txt

and the entire package can be built and installed with

(venv) $ pip install .

Alternatively, the following two commands will build the compiled library into the build directory and then install the package

(venv) $ python setup.py build
(venv) $ python setup.py install

During the build stage, numpy and the f2py backend will attempt to find existing FORTRAN compilers, as well as the Intel Math Kernel Library (MKL) if it exists. The implementation is not so demanding that MKL is required, but it could provide some benefits.

Upgrading / uninstalling

Unforuntately, the linking of the compiled library to the python library requires distutils rather than setuptools meaning that the entire package must be deleted by hand when re-installing or upgrading. For further discussion, see this issue in the PYPA GitHub

Testing

Test coverage is not currently great here in this repository. However, the code has been tested on practical problems with good results. Eventually, these data should be added to this repository to ensure the robustness of the package.

Testing is handled with pytest, which is the only dependency listed in requirements-test.txt. It can then be installed with

(venv) $ pip install pytest

or

(venv) $ pip install -r requirements-test.txt

Tests should be run from the tests directory

(venv) $ cd tests
(venv) $ pytest

since the python interpreter will look for the compiled library in the sfv directory, not in the virtual environment.

FORTAN compatibility

The implementation of the SFV prediction is written in a FORTRAN-90 file contained in lib/sfv.f90. For transparency, the library was written to interface with python first and (maybe) FORTRAN second. If you have issues with linking this library in a FORTRAN program, or have advice on this, we welcome that insight.

Citing

If you use this package, please cite the above Nuclear Science and Engineering

@Article{<your bibtex key here>
   author  = {Johnson, Andrew E. and Kotlyar, Dan},
   title   = {A Transport-Free Method for Predicting the Post-Depletion Spatial Neutron Flux Distribution},
   doi     = {10.1080/00295639.2019.1661171},
   number  = {2},
   pages   = {120-137},
   url     = {https://doi.org/10.1080/00295639.2019.1661171},
   volume  = {194},
   journal = {Nuclear Science and Engineering},
   year    = {2020},
}

If you'd like, let us know about your publication and we'll add it to a (to be created) publication list.