BSSUnfold: A python package for neutron spectrum unfolding from Bonner sphere spectrometer measurements

摘要: Unfolding neutron spectra from Bonner Sphere Spectrometer (BSS) measurements is an ill-posed inverse problem: a continuous neutron fluence must be recovered from several integral readings. This work presents BSSUnfold, an open-source Python library that unifies 107 unfolding algorithms under a single Detector class interface and covers all major paradigms: iterative, regularization, Bayesian, parametric, metaheuristics, maximum entropy, interval analysis and reimplementation of classical codes. The package provides 7 built-in response function sets, ICRP-116 dose coefficients, 41 comparison metrics, per-method convergence diagnostics, and Monte Carlo uncertainty estimation. A systematic grid search over 9,632 configurations of 107 methods on 271 reference spectra (2.6 million runs) identifies the quadratic-programming method qpsolvers, with L₂-norm and first-order smoothing, as the best default algorithm, while the interval regularization method attains the highest composite score on the benchmark, showing that the interval setting competes with the best Tikhonov programs. The aggregated method evaluation rests on a composite score averaging nine normalized spectral and dosimetric components. Optimal parameter tuning improves scores by a median of +12.9 points, a gain comparable in importance to the choice of method itself. Together these results give the quantitative, tuning-aware basis for selecting an unfolding method in BSS practice, and make the package a reproducible reference implementation for the radiation safety specialists.