Quickstart#

This guide walks through a complete nuclear data assimilation workflow from loading Serpent output files to producing a posterior and exporting adjusted nuclear data.

Overview#

A typical ANDALUS workflow consists of four steps:

  1. Load benchmarks — Serpent sensitivity and results files for experiments with known measurements.

  2. Load applications — Serpent sensitivity and results files for the system you want to predict.

  3. Load covariances — Nuclear data covariance matrices from ERRORR (via SANDY).

  4. Run GLLS — Combine everything in an AssimilationSuite and call .glls().

Step 1: Load a benchmark#

from andalus import Benchmark

hmi = Benchmark.from_serpent(
    title="HMI-001",
    m=1.0,  # measured k-eff
    dm=0.0005,  # measurement uncertainty (1-sigma)
    sens0_path="data/hmi001_sens0.m",
    results_path="data/hmi001_res.m",
    kind="keff",
)

print(f"k_eff discrepancy: {hmi.c - hmi.m:.5f}")

To attach a flux spectrum (needed for the ealf metric), pass flux_det:

hmi = Benchmark.from_serpent(
    title="HMI-001",
    m=1.0,
    dm=0.0005,
    sens0_path="data/hmi001_sens0.m",
    results_path="data/hmi001_res.m",
    flux_det=("flux", "data/hmi001_det0.m"),
)

Step 2: Load a benchmark suite#

from andalus import BenchmarkSuite

benchmark_suite = BenchmarkSuite.from_yaml("benchmarks.yaml")

# Or build from a list
benchmark_suite = BenchmarkSuite.from_list([hmi, hmi2, hmi3])

print(suite.m)  # measured values as pd.Series
print(suite.c)  # calculated values as pd.Series

Step 3: Load covariances#

from andalus import CovarianceSuite

covariances = CovarianceSuite.from_hdf5(
    "covariances.h5",
    zais=[922350, 922380, 942390],  # U-235, U-238, Pu-239
)

Step 4: Load applications#

from andalus import Application, ApplicationSuite

reactor = Application.from_serpent(
    title="my-reactor",
    sens0_path="data/reactor_sens0.m",
    results_path="data/reactor_res.m",
)

application_suite = ApplicationSuite.from_list([reactor])

Step 5: Run GLLS and inspect results#

from andalus import AssimilationSuite

suite = AssimilationSuite(
    benchmarks=benchmark_suite,
    applications=application_suite,
    covariances=covariances,
)

posterior = suite.glls()

posterior.summarize()
print(f"Chi-squared: {posterior.chi_squared():.3f}")
print(posterior.applications.c)  # adjusted calculated values

Step 6: Export to ACE (optional)#

If you want to use the adjusted nuclear data in a Monte Carlo transport code:

posterior.to_ace_direct(
    out_dir="adjusted_ace",
    xsdata_path="path/to/jeff40.xsdata",
    temperature=300,
    create_xsdata=True,
)

Using a YAML configuration#

For reproducible workflows, describe everything in a single YAML file and load it in one call:

suite = AssimilationSuite.from_yaml("assimilation.yaml")
posterior = suite.glls()

See the YAML Configuration page for the full schema.

Filtering benchmarks#

Remove outliers before running GLLS using composable filter objects:

from andalus import Chi2Filter

# Keep only benchmarks with chi-squared < 3
filtered = suite.filter(Chi2Filter(threshold=3.0))
posterior = filtered.glls()

Filters support &, |, and ~ operators:

from andalus import Chi2Filter, Chi2NuclearDataFilter

f = Chi2Filter(3.0) & Chi2NuclearDataFilter(3.0, covariances.matrix)
posterior = suite.filter(f).glls()