Search NASA⌕ Search

Engineering topics

Neudecker, Denise

Publications and source records attributed to Neudecker, Denise.

120 records · Page 7
All sourcesNASA NTRSDOE OSTI
DOE OSTI2021

Example on how to (intelligently) augment the nuclear-data pipeline with machine learning [Slides]

The presentation discusses how machine learning has helped the Los Alamos National Laboratory (LANL) nuclear-data pipeline. It also discusses the strengths of machine learning as it finds trends in large amounts of data where human brains are overwhelmed and that this information may be crucial to improve our nuclear data. It does stress, however, that machine learning is no "silver bullet" and that it is critical to feed it expert knowledge and use physics intuition to interpret the results. The presentation discusses the need to develop infrastructure and tools to provide data in an easily readable and unambiguously interpretable format (e.g., EXFOR format), to develop experimental data and theory to solve physics questions, and that statisticians and nuclear-data experts must be brought together to correctly interpret the results. The presentation concludes by stating that machine learning is a great tool and that LANL needs to use the algorithms along with developing physics data, tools and infrastructure.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2020

Enhancing nuclear data validation analysis by using machine learning

In this work, we demonstrate how machine learning (ML) techniques can be used as an auxiliary tool for nuclear data validation analysis. The output of the ML analysis can inform evaluators and validators on the quality, or lack thereof, of specific nuclear data and benchmark experiments with respect to simulating these experimental benchmark values. To this end, measured and simulated effective neutron multiplication factors, k eff , of 875 selected ICSBEP critical assemblies and the latter's sensitivities with respect to nuclear data as well as benchmarks' features (e.g., material nuclide, core geometry) are used as input for a random forest (RF) regression model. The RF is used to encode the complex inter-dependencies between thousands of nuclear data features (i.e., sensitivity profiles and aspects describing the measurements) and 875 simulated and experimental k eff values in order to understand what nuclear data features are most informative for predicting bias. The complexity of relationships and high-dimensional space is difficult-to-impossible to search fully using simply expert judgment. As a first proof-of-concept—step, it is demonstrated that this technique is able to correctly trace large discrepancies between simulated and experimental k eff back to fabricated shortcomings in nuclear data that were perturbed to simulated k eff values. In a second, real-case scenario, step, the RF algorithm is used to validate the ENDF/B-VIII.0 library in comparison to ENDF/B-VII.1 nuclear data. One case is showcased where the chosen ML algorithms highlighted nuclear data (the 19 F(n,inl) cross section from 0.4–0.9 MeV) that are shown to be problematic by comparing them to associated differential experimental data and nuclear data from other libraries. In addition to that, it is shown that the RF results point towards poor benchmark experiments and associated underestimated uncertainties (e.g., the PU-SOL-THERM-028 series). However, using the RF algorithm for validating nuclear data with respect to k eff is currently limited to pinpointing groups of questionable nuclear data due to the inherent correlations between features introduced by the nuclear data themselves and how k eff is simulated. Due to this, we recommend that the ML methods presented be used to augment—rather than replace—the expert knowledge of evaluators and validators.

96 KNOWLEDGE MANAGEMENT AND PRESERVATION↗
DOE OSTI2020

Evaluating 239 Pu(n,f) cross sections via machine learning using experimental data, covariances, and measurement features

In this paper, the neutron-induced 239 Pu fission cross section, 239 Pu(n,f), is evaluated from 1–20 MeV using experimental data and associated covariances while also considering information on the measurement, termed features here. For instance, methods to determine the background, sample backing material, or impurities in the sample, are explicitly taken into account in the evaluation process. To this end, outliers in the experimental data are identified with a modified version of the Hybrid Robust Support Vector Machine. In a second step, two machine learning methods (logistic regression with elastic net regularization and random forest regression with SHAP feature importance metric) are used to highlight measurement features that are common among many of the outlying data points. Based on this analysis, penalty uncertainties are added to the experimental covariances of outlying data points that have outlier measurement features and are put through the generalized-least-squares evaluation. The resulting evaluated mean values and covariances differ distinctly from those data evaluated without the penalty uncertainties. These results highlight that certain measurement features should be more closely examined.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2020

Validating Nuclear Data Uncertainties Obtained from a Statistical Analysis of Experimental Data with the “Physical Uncertainty Bounds” Method

Concerns within the nuclear data community led to substantial increases of Neutron Data Standards (NDS) uncertainties from its previous to the current version. For example, those associated with the NDS reference cross section 239 Pu(n,f) increased from 0.6–1.6% to 1.3–1.7% from 0.1–20 MeV. These cross sections, among others, were adopted, e.g., by ENDF/B-VII.1 (previous NDS) and ENDF/B-VIII.0 (current NDS). There has been a strong desire to be able to validate these increases based on objective criteria given their impact on our understanding of various application uncertainties. Here, the “Physical Uncertainty Bounds” method (PUBs) by Vaughan et al. is applied to validate evaluated uncertainties obtained by a statistical analysis of experimental data. We investigate with PUBs whether ENDF/B-VII.1 or ENDF/B-VIII.0 239 Pu(n,f) cross-section uncertainties are more realistic given the information content used for the actual evaluation. It is shown that the associated conservative (1.5–1.8%) and minimal realistic (1.1–1.3%) uncertainty bounds obtained by PUBs enclose ENDF/B-VIII.0 uncertainties and indicate that ENDF/B-VII.1 uncertainties are underestimated.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2020

Unrecognized Sources of Uncertainties ( USU ) in Experimental Nuclear Data

Evaluated nuclear data uncertainties reported in the literature or archived in data libraries are often perceived as unrealistic, most often because they are thought to be too small. The impact of this issue in applied nuclear science has been discussed widely in recent years. Commonly suggested causes are: poor estimates of specific error components, neglect of uncertainty correlations, and overlooked known error sources. However, instances have been reported where very careful, objective assessments of all known error sources have been made with realistic error magnitudes and correlations provided, yet the resulting evaluated uncertainties still appear to be inconsistent with observed scatter of predicted mean values. These discrepancies might be attributed to significant unrecognized sources of uncertainty ( USU ) that limit the accuracy to which these physical quantities can be determined. The objective of our work reported herein has been to develop qualitative and quantitative procedures for revealing and including USU estimates in nuclear data evaluations involving experimental input data. This paper identifies several specific clues that can be explored by evaluators in identifying the existence of USU . It then describes numerical procedures we have introduced to generate quantitative estimates of USU magnitudes. Key requirements for these procedures to be viable are that sufficient numbers of data points be available, for statistical reasons, and that additional supporting information about the measurements be provided by the experimenters. Several realistic examples are described here to illustrate these procedures and demonstrate their outcomes and limitations. Our work strongly supports the view that USU is an important issue in nuclear data evaluation, with significant consequences for applications, and that this topic warrants further investigation by the nuclear science community.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2020

Applying a Template of Expected Uncertainties to Updating 239 Pu(n,f) Cross-section Covariances in the Neutron Data Standards Database

Templates of uncertainties expected in specific measurement types were recently developed. One aim of these templates is to help evaluators in identifying (1) missing or suspiciously low uncertainties and (2) missing correlations between uncertainties of the same and different experiments, when estimating covariances for experimental data employed in their evaluations. These templates also provide realistic estimates of standard deviations and correlations for a particular uncertainty source and measurement type that can be used by evaluators in situations where they are not supplied by the experimenters. This information allows for a more comprehensive uncertainty analysis across all measurements considered in an evaluation and, thus, more realistic evaluated covariances. Here, in this work, we extend a template that is applicable to uncertainties expected in neutron-induced fission, (n,f), cross-section measurements. It is applied to improving covariances of 239 Pu(n,f) cross-section measurements in the database underlying the Neutron Data Standards evaluations. This particular example was chosen since this evaluation is primarily based on experimental information. Also, some uncertainties of individual 239 Pu(n,f) cross-section experiments in this database were suspected to be underestimated. The evaluated uncertainties obtained after updating the covariances in the database by means of the template indeed do increase compared to their original values. Even more importantly, the evaluated mean values change noticeably. These modified cross sections impact application calculations significantly, as is demonstrated by employing them in simulations of the effective neutron multiplication factor for a few selected critical assemblies. However, this updated evaluated 239 Pu(n,f) cross section should not be interpreted as the final one that should replace values of the current Neutron Data Standards project. Evaluations for the Neutron Data Standards of the 239 Pu(n,f) cross section must be linked to many other observables included in the associated database, most notably to cross sections for 235 U(n,f), but also to those for 10 B(n, α ), 6 Li(n,t), 238 U(n,f), and 238 U(n, γ ), because of included measurements of the 239 Pu(n,f) cross section that appear as ratios to these reactions. Some of these other reactions are correlated to further observables in the database. Hence, updating uncertainties of data sets of any of these observables can potentially impact the 239 Pu(n,f) cross section. Uncertainties for all measurements of these linked physical observables have to be updated before a comprehensive evaluation of the 239 Pu(n,f) cross section and its corresponding uncertainties can be provided.

239Pu Neutron-induced fission cross-section↗