arXiv subjects

Georg Schnabel

Publications and source records attributed to Georg Schnabel.

9 recordsLinked to original sources

arXiv2026

Model bias and parameter optimisation with the example of INCL/ABLA

The accuracy and precision of high-energy spallation models play a crucial role in the design and development of new applications and experiments, as well as in data analysis. We discuss the complementarity between parameter optimisation and model bias estimation approaches within a Bayesian framework. This is illustrated using the IntraNuclear Cascade model of Li\`ege (INCL) together with the Ablation model (ABLA), for which these two approaches for model bias estimation have been applied independently in previous works.

nucl-th↗

arXiv2023

How to explain ENDF-6 to computers: A formal ENDF format description language

The ENDF-6 format, widely used worldwide for storing and disseminating nuclear data, is managed by the Cross Sections Evaluation Working Group (CSEWG) and fully documented in the ENDF-6 formats manual. This manual employs a combination of formal and natural language, introducing the possibility of ambiguity in certain parts of the format specification. To eliminate the possibility of ambiguity, this contribution proposes a generalization of the formal language used in the ENDF-6 formats manual, which is precisely defined in extended Backus-Naur form (EBNF). This formalization also offers several other advantages, notably reducing the complexity of creating and updating parsing and validation programs for ENDF-6 formatted files. Moreover, the availability of a formal description enables the automatic mapping of the low-level ENDF-6 representation, provided by a sequence of numbers, to a more human-friendly hierarchical representation with variable names. Demonstrating these advantages, almost the entire ENDF-6 format description has been translated into the proposed formal language, accompanied by a reference implementation of a parser/translator. This implementation leverages the formal format specification to enable reading, writing, and validating ENDF-6 formatted files.

nucl-th↗

arXiv2023

Parameter optimisation using Bayesian inference for spallation models

The accuracy and precision of high-energy spallation models are key issues for the design and development of new applications and experiments. We present a method to estimate model parameters and associated uncertainties by leveraging the Bayesian version of the Generalised Least Squares method, which enables us to incorporate prior knowledge on the parameter values. This approach is designed to adjust parameters based on experimental data, accounting for experimental uncertainty information, and providing uncertainties for all adjusted parameters. This approach is designed in order both to improve the accuracy of models through the modification of free parameters of these models, which results in a better reproduction of experimental data, and to estimate the uncertainties of these parameters and, by extension, their impacts on the model output. We aim at demonstrating the Generalised Least Square method can be applied in the case of Monte Carlo models. We present a proof-of-concept for Monte Carlo models in the specific case of nuclear physics with the model combination INCL/ABLA. We discuss the challenges in the application of this method to high-energy spallation models, notably the large runtime and the stochasticity of the models. Our results indicate this framework can also be applied to analogous situations where parameters of a computationally expensive Monte Carlo code should be inferred/improved.

hep-ph↗

arXiv2021

Nuclear data evaluation with Bayesian networks

Bayesian networks are graphical models to represent the probabilistic relationships between variables in the Bayesian framework. The knowledge of all variables can be updated using new information about some of the variables. We show that relying on the Bayesian network interpretation enables large scale inference and gives flexibility in incorporating prior assumptions and constraints into the nuclear data evaluation process, such as sum rules and the non-negativity of cross sections. The latter constraint is accounted for by a non-linear transformation and therefore we also discuss inference in Bayesian networks with non-linear relationships. Using Bayesian networks, the evaluation process yields detailed information, such as posterior estimates and uncertainties of all statistical and systematic errors. We also elaborate on a sparse Gaussian process construction compatible with the Bayesian network framework that can for instance be used as prior on energy-dependent model parameters, model deficiencies and energy-dependent systematic errors of experiments. We present three proof-of-concept examples that emerged in the context of the neutron data standards project and in the ongoing international evaluation efforts of $^{56}$Fe. In the first example we demonstrate the modelization and explicit estimation of relative energy-dependent error components of experimental datasets. Then we show an example evaluation using the outlined Gaussian process construction in an evaluation of $^{56}$Fe in the energy range between one and two MeV, where R-Matrix and nuclear model fits are difficult. Finally, we present a model-based evaluation of $^{56}$Fe between 5 MeV and 30 MeV with a sound treatment of model deficiencies. The R scripts to reproduce the Bayesian network examples and the nucdataBaynet package for Bayesian network modeling and inference have been made publicly available.

physics.data-an↗

arXiv2021

Conception and software implementation of a nuclear data evaluation pipeline

We discuss the design and software implementation of a nuclear data evaluation pipeline applied for a fully reproducible evaluation of neutron-induced cross sections of $^{56}$Fe above the resolved resonance region using the nuclear model code TALYS combined with relevant experimental data. The emphasis is on the mathematical and technical aspects of the pipeline and not on the evaluation of $^{56}$Fe, which is tentative. The mathematical building blocks combined and employed in the pipeline are discussed in detail. A unified representation of experimental data, systematic and statistical errors, model parameters and defects enables the application of the Generalized Least Squares (GLS) and its natural extension, the Levenberg-Marquardt (LM) algorithm, on a large collection of experimental data. The LM algorithm tailored to nuclear data evaluation accounts for the exact non-linear physics model to determine best estimates of nuclear quantities. Associated uncertainty information is derived from a Taylor expansion at the maximum of the posterior distribution. We also discuss the pipeline in terms of its IT (=information technology) building blocks, such as those to efficiently manage and retrieve experimental data of the EXFOR library and to distribute computations on a scientific cluster. Relying on the mathematical and IT building blocks, we elaborate on the sequence of steps in the pipeline to perform the evaluation, such as the retrieval of experimental data, the correction of experimental uncertainties using marginal likelihood optimization (MLO) and after a screening of thousand TALYS parameters -- including Gaussian process priors on energy dependent parameters -- the fitting of about 150 parameters using the LM algorithm. The code of the pipeline including a manual and a Dockerfile for a simplified installation is available at www.nucleardata.com.

physics.data-an↗

arXiv2020

A computational EXFOR database

The EXFOR library is a useful resource for many people in the field of nuclear physics. In particular, the experimental data in the EXFOR library serves as a starting point for nuclear data evaluations. There is an ongoing discussion about how to make evaluations more transparent and reproducible. One important ingredient may be convenient programmatic access to the data in the EXFOR library from high-level languages. To this end, the complete EXFOR library can be converted to a MongoDB database. This database can be conveniently searched and accessed from a wide variety of programming languages, such as C++, Python, Java, Matlab, and R. This contribution provides some details about the successful conversion of the EXFOR library to a MongoDB database and shows simple usage examples to underline its merits. All codes required for the conversion have been made available online and are open-source. In addition, a Dockerfile has been created to facilitate the installation process.

cs.DL↗

arXiv2018

A first sketch: Construction of model defect priors inspired by dynamic time warping

Model defects are known to cause biased nuclear data evaluations if they are not taken into account in the evaluation procedure. We suggest a method to construct prior distributions for model defects for reaction models using neighboring isotopes of $^{56}$Fe as an example. A model defect is usually a function of energy and describes the difference between the model prediction and the truth. Of course, neither the truth nor the model defect are accessible. A Gaussian process (GP) enables to define a probability distribution on possible shapes of a model defect by referring to intuitively understandable concepts such as smoothness and the expected magnitude of the defect. Standard specifications of GPs impose a typical length-scale and amplitude valid for the whole energy range, which is often not justified, e.g., when the model covers both the resonance and statistical range. In this contribution, we show how a GP with energy-dependent length-scales and amplitudes can be constructed from available experimental data. The proposed construction is inspired by a technique called dynamic time warping used, e.g., for speech recognition. We demonstrate the feasibility of the data-driven determination of model defects by inferring a model defect of the nuclear models code TALYS for (n,p) reactions of isotopes with charge number between 20 and 30. The newly introduced GP parametrization besides its potential to improve evaluations for reactor relevant isotopes, such as $^{56}$Fe, may also help to better understand the performance of nuclear models in the future.

nucl-th↗

arXiv2018

Fitting and Analysis Technique for Inconsistent Nuclear Data

Consistent experiment data are crucial to adjust parameters of physics models and to determine best estimates of observables. However, often experiment data are not consistent due to unrecognized systematic errors. Standard methods of statistics such as $χ^2$-fitting cannot deal with this case. Their predictions become doubtful and associated uncertainties too small. A human has then to figure out the problem, apply corrections to the data, and repeat the fitting procedure. This takes time and potentially costs money. Therefore, a Bayesian method is introduced to fit and analyze inconsistent experiment data. It automatically detects and resolves inconsistencies. Furthermore, it allows to extract consistent subsets from the data. Finally, it provides an overall prediction with associated uncertainties and correlations less prone to the common problem of too small uncertainties. The method is foreseen to function with a large corpus of data and hence may be used in nuclear databases to deal with inconsistencies in an automated fashion.

nucl-th↗

arXiv2018

Estimating model bias over the complete nuclide chart with sparse Gaussian processes at the example of INCL/ABLA and double-differential neutron spectra

Predictions of nuclear models guide the design of nuclear facilities to ensure their safe and efficient operation. Because nuclear models often do not perfectly reproduce available experimental data, decisions based on their predictions may not be optimal. Awareness about systematic deviations between models and experimental data helps to alleviate this problem. This paper shows how a sparse approximation to Gaussian processes can be used to estimate the model bias over the complete nuclide chart at the example of inclusive double-differential neutron spectra for incident protons above 100\,MeV. A powerful feature of the presented approach is the ability to predict the model bias for energies, angles, and isotopes where data are missing. The number of experimental data points that can be taken into account is at least in the order of magnitude of~$10^4$ thanks to the sparse approximation. The approach is applied to the Liège Intranuclear Cascade Model (INCL) coupled to the evaporation code ABLA. The results suggest that sparse Gaussian process regression is a viable candidate to perform global and quantitative assessments of models. Limitations of a philosophical nature of this (and any other) approach are also discussed.

nucl-th↗