Estimation of continuous distribution of iterated fission probability using an artificial neural network with Monte Carlo-based training data

Tuya, D.; Nagaya, Yasunobu

Journal of Nuclear Engineering (Internet), 4(4), p.691 - 710, 2023/11

https://doi.org/10.3390/jne4040043

The Monte Carlo method is used to accurately estimate various quantities such as k-eigenvalue and integral neutron flux. However, when a distribution of a quantity is desired, the Monte Carlo method does not typically provide continuous distribution. Recently, the functional expansion tally and kernel density estimation methods have been developed to provide continuous distribution. In this paper, we propose a method to estimate a continuous distribution of a quantity using artificial neural network (ANN) model with Monte Carlo-based training data. As a proof of concept, a continuous distribution of iterated fission probability (IFP) is estimated by ANN models in two systems. The IFP distributions by the ANN models were compared with the Monte Carlo-based data and the adjoint angular neutron fluxes by the PARTISN code. The comparisons showed varying degrees of agreement or discrepancy; however, it was observed that the ANN models learned the general trend of the IFP distributions.

A Functional expansion tally method with numerical basis sets generated by singular value decomposition for one-dimensional Monte Carlo calculations

Kondo, Ryoichi; Nagaya, Yasunobu

Proceedings of International Conference on Mathematics and Computational Methods Applied to Nuclear Science and Engineering (M&C 2023) (Internet), 10 Pages, 2023/08

A functional expansion tally (FET) method with numerical basis functions generated by singular value decomposition (SVD) is newly proposed. Traditionally, analytical functions were used for the FET calculations, e.g., Legendre polynomials for a one-dimensional distribution. However, the expansion terms could increase to reconstruct steep or complex distributions with these functions. A basis set that can well represent the target distribution with lower order expansion is desired to achieve high accuracy with the small computational resource. In the present study, a numerical basis set is generated from snapshot data by using SVD. This approach is based on the reduced order modeling (ROM). We applied ROM to the FET method in Monte Carlo calculations. The numerical result showed the applicability of the proposed method, on the other hand, some issues were revealed, e.g., discretization of the snapshot data.

Approximate estimation of iterated fission probability by deep neural network

Tuya, D.; Nagaya, Yasunobu

Proceedings of 30th International Conference on Nuclear Engineering (ICONE30) (Internet), 8 Pages, 2023/05

Adjoint angular neutron flux is used as weighting function in various reactor applications such as calculation of kinetics parameters, importance sampling variance reduction techniques, etc. Iterated fission probability (IFP), which is proportional to a fundamental mode of adjoint angular neutron flux, has increasingly been used as weighting function in Monte Carlo calculations. The Monte Carlo based IFP methods stochastically estimate IFP for a given phase-space location. In this work, we investigated the applicability of a deep neural network for approximating an unknown underlying function, which maps from a phase-space location to an IFP in a given fissile system, from dataset produced by a Monte Carlo based IFP method. The preliminary application has been performed for the Godiva and simplified STACY cores. The comparison showed a varying degree of agreement and discrepancy between the results by the DNN and the reference results by a deterministic neutron transport code PARTISN.

Nuclear data processing code FRENDY version 2

Tada, Kenichi; Yamamoto, Akio*; Kunieda, Satoshi; Nagaya, Yasunobu

JAEA-Data/Code 2022-009, 208 Pages, 2023/02

JAEA-Data-Code-2022-009.pdf:3.87MB

The nuclear data processing code has an important role to connect evaluated nuclear data libraries and neutronics calculation codes. Japan Atomic Energy Agency (JAEA) has developed the nuclear data processing code FRENDY since 2013 to generate cross section files from evaluated nuclear data libraries, such as JENDL, ENDF/B, JEFF, and TENDL. The first version of FRENDY was released in 2019. FRENDY version 1 generates ACE files which are used for continuous energy Monte Carlo codes such as PHITS, Serpent, and MCNP. FRENDY version 2 generates multi-group neutron cross-section files from ACE files. The other major improvements are as follows: (1) uncertainty quantification for the probability tables of the unresolved resonance cross-section; (2) perturbation of the ACE file for the uncertainty quantification using a continuous Monte Carlo code; (3) modification of the ENDF-6 formatted nuclear data file. This report describes an overview of the nuclear data processing methods and input instructions for FRENDY.

Neutron importance estimation via new recursive Monte Carlo method for deep penetration neutron transport

Tuya, D.; Nagaya, Yasunobu

Nuclear Science and Engineering, 15 Pages, 2023/00

https://doi.org/10.1080/00295639.2023.2233850

In Monte Carlo neutron transport calculations for local response or deep penetration problems, some estimation of an importance function is generally required in order to improve their efficiency. In this work, a new recursive Monte Carlo (RMC) method, which is partly based on the original RMC method, for estimating an importance function for local variance reduction (i.e., source-detector type) problems has been developed. The new RMC method has been applied to two sample problems of varying degrees of neutron penetrations, namely a one-dimensional iron slab problem and a three-dimensional concrete-air problem. The biased Monte Carlo calculations with variance reduction parameters based on the obtained importance functions by the new RMC method have been performed to estimate detector responses in these problems. The obtained results are in agreement with those by the reference unbiased Monte Carlo calculations. Furthermore, the biased calculations offered an increase in efficiency on the order of 1 to 10 in terms of the figure of merit (FOM). The results also indicated that the efficiency increased as the neutron penetration became deeper.

Adjoint-weighted correlated sampling for -eigenvalue perturbation in Monte Carlo calculation

Tuya, D.; Nagaya, Yasunobu

Annals of Nuclear Energy, 169, p.108919_1 - 108919_9, 2022/05

https://doi.org/10.1016/j.anucene.2021.108919

Estimating an effect of a perturbation in a fissile system on its -eigenvalue requires special technique called perturbation theory when the considered perturbation is small. In this study, we develop an adjoint-weighted correlated sampling (AWCS) method based on the exact perturbation theory without any approximation by combining the correlated sampling (CS) method with iterated-fission probability (IFP) based adjoint-weighting method. With the advantages of the CS method being good at providing very small uncertainty for small perturbations and the IFP-based adjoint-weighting method being suitable for continuous-energy Monte Carlo calculation, the developed AWCS method based on the exact perturbation theory offers a new rigorous approach for perturbation calculations. The obtained results by the developed AWCS method for verification problems involving Godiva and simplified STACY density perturbations showed good agreement with the reference calculations.

Solomon; A Monte Carlo solver for criticality safety analysis

Nagaya, Yasunobu; Ueki, Taro; Tonoike, Kotaro

Proceedings of 11th International Conference on Nuclear Criticality Safety (ICNC 2019) (Internet), 9 Pages, 2019/09

A new Monte Carlo solver Solomon has been developed for the application to fuel-debris systems. It is designed not only for usual criticality safety analysis but also for criticality calculations of damaged reactor core including fuel debris. This paper describes the current status of Solomon and demonstrates the applications of the randomized Weierstrass function (RWF) model and the RWF model superimposed voxel geometry.

Development of the Unified Cross-section Set ADJ2017

Yokoyama, Kenji; Sugino, Kazuteru; Ishikawa, Makoto; Maruyama, Shuhei; Nagaya, Yasunobu; Numata, Kazuyuki*; Jin, Tomoyuki*

JAEA-Research 2018-011, 556 Pages, 2019/03

JAEA-Research-2018-011.pdf:19.53MB
JAEA-Research-2018-011-appendix1(DVD-ROM).zip:433.07MB
JAEA-Research-2018-011-appendix2(DVD-ROM).zip:580.12MB
JAEA-Research-2018-011-appendix3(DVD-ROM).zip:9.17MB

We have developed a new unified cross-section set ADJ2017, which is an improved version of the unified cross-section set ADJ2010 for fast reactors. The unified cross-section set is used for reflecting information of C/E values (analysis / experiment values) obtained by integral experiment analyses; the values are stored in the standard database for FBR core design via the cross-section adjustment methodology, which integrates with the information such as uncertainty (covariance) of nuclear data, uncertainty of integral experiment / analysis, sensitivity of integral experiment with respect to nuclear data. The ADJ2017 is based on Japan's latest nuclear data library JENDL-4.0 as in the previous version of ADJ2010, and it incorporates more information on integral experimental data sets related to minor actinides (MAs) and degraded plutonium (Pu). In the creation of ADJ2010, a total of 643 integral experimental data sets were analyzed and evaluated, and 488 of the integral experimental data sets were finally selected to be used for the cross-section adjustment. In contrast, we have evaluated a total of 719 data sets, and eventually adopted 620 integral experimental data sets to create ADJ2017. ADJ2017 shows almost the same performance as ADJ2010 for the main neutronic characteristics of conventional sodium-cooled MOX-fuel fast reactors. In addition, for the neutronic characteristics related to MA and degraded Pu, ADJ2017 improves the C/E values of the integral experimental data sets, and reduces the uncertainty induced by the nuclear data. ADJ2017 is expected to be widely used in the analysis and design research of fast reactors. Moreover, it is expected that the integral experimental data sets used for ADJ2017 can be utilized as a standard database of FBR core design.

FRENDY; A New nuclear date processing system being developed at JAEA

Tada, Kenichi; Nagaya, Yasunobu; Kunieda, Satoshi; Suyama, Kenya; Fukahori, Tokio

EPJ Web of Conferences, 146, p.02028_1 - 02028_5, 2017/09

https://doi.org/10.1051/epjconf/201714602028

JAEA has started to develop new nuclear data processing system FRENDY (FRom Evaluated Nuclear Data libralY to any application). In this presentation, the outline of the development of FRENDY is presented. And functions and performances of FRENDY are demonstrated by generation and validation of the continuous energy cross section data libraries for MVP, PHITS and MCNP codes.

General-purpose Monte Carlo codes for neutron and photon transport calculations; MVP version 3

Nagaya, Yasunobu

Hoshasen, 43(2), p.49 - 54, 2017/05

no abstracts in English