Initialising ...
Initialising ...
Initialising ...
Initialising ...
Initialising ...
Initialising ...
Initialising ...
Tuya, D.; Nagaya, Yasunobu
Journal of Nuclear Engineering (Internet), 4(4), p.691 - 710, 2023/11
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.
Asahi, Yuichi; Onodera, Naoyuki; Hasegawa, Yuta; Shimokawabe, Takashi*; Shiba, Hayato*; Idomura, Yasuhiro
Boundary-Layer Meteorology, 186(3), p.659 - 692, 2023/03
Times Cited Count：2 Percentile：47.50(Meteorology & Atmospheric Sciences)We develop a Transformer-based deep learning model to predict the plume concentrations in the urban area under uniform flow conditions. Our model has two distinct input layers: Transformer layers for sequential data and convolutional layers in convolutional neural networks (CNNs) for image-like data. Our model can predict the plume concentration from realistically available data such as the time series monitoring data at a few observation stations and the building shapes and the source location. It is shown that the model can give reasonably accurate prediction with orders of magnitude faster than CFD simulations. It is also shown that the exactly same model can be applied to predict the source location, which also gives reasonable prediction accuracy.
Nomura, Masahiro; Okita, Hidefumi; Shimada, Taihei; Tamura, Fumihiko; Yamamoto, Masanobu; Furusawa, Masashi*; Sugiyama, Yasuyuki*; Hasegawa, Katsushi*; Hara, Keigo*; Omori, Chihiro*; et al.
Proceedings of 18th Annual Meeting of Particle Accelerator Society of Japan (Internet), p.80 - 82, 2021/10
no abstracts in English
Asahi, Yuichi; Hatayama, Sora*; Shimokawabe, Takashi*; Onodera, Naoyuki; Hasegawa, Yuta; Idomura, Yasuhiro
Proceedings of 2021 IEEE International Conference on Cluster Computing (IEEE Cluster 2021) (Internet), p.686 - 691, 2021/10
Times Cited Count：2 Percentile：64.54(Computer Science, Hardware & Architecture)We develop a convolutional neural network model to predict the multi-resolution steady flow. Based on the state-of-the-art image-to-image translation model pix2pixHD, our model can predict the high resolution flow field from the set of patched signed distance functions. By patching the high resolution data, the memory requirements in our model is suppressed compared to pix2pixHD.
Koyamada, Koji*; Yu, L.*; Kawamura, Takuma; Konishi, Katsumi*
International Journal of Modeling, Simulation, and Scientific Computing, 12(2), p.2140001_1 - 2140001_19, 2021/04
With the improvement of sensors technologies in various fields such as fluid dynamics, meteorology, and space observation, it is an important issue to derive explanatory models using partial differential equations (PDEs) for the big data obtained from them. In this paper, we propose a technique for estimating linear PDEs with higher-order derivatives for spatiotemporally discrete point cloud data. The technique calculates the time and space derivatives from a neural network (NN) trained on the point cloud data, and estimates the derivative term of the PDE using regression analysis techniques. In the experiment, we computed the error of the estimated PDEs for various meta-parameters for the PDEs with exact solutions. As a result, we found that increasing the hierarchy of NNs to match the order of the derivative terms in the exact solution PDEs and adopting L1 regularization with LASSO as the method of regression analysis increased the accuracy of the model.
Subekti, M.*; Kudo, Kazuhiko*; Nabeshima, Kunihiko; Takamatsu, Kuniyoshi
Atom Indonesia, 43(2), p.93 - 102, 2017/08
Reactor kinetics based on point kinetic model have been generally applied as the standard method for neutronics codes. As the central control rod (C-CR) withdrawal test has demonstrated in a prismatic core of HTTR, the transient calculation of kinetic parameter, such as reactivity and neutron fluxes, requires a new method to shorten calculation-process time. Development of neural network method was applied to point kinetic model as the necessity of real-time calculation that could work in parallel with the digital reactivity meter. The combination of TDNN and Jordan RNN, such as TD-Jordan RNN, was the result of the modeling approach. The application of TD-Jordan RNN with adequate learning, tested offline, determined results accurately even when signal inputs were noisy. Furthermore, the preprocessing for neural network input utilized noise reduction as one of the equations to transform two of twelve time-delayed inputs into power corrected inputs.
Yoshino, Ryuji
Nuclear Fusion, 45(11), p.1232 - 1246, 2005/11
Times Cited Count：41 Percentile：76.40(Physics, Fluids & Plasmas)Prediction of major disruptions observed at the 
-limit for tokamak plasmas has been investigated in JT-60U with developing neural networks. A sub-neural network is trained to output a value of the 
limit every 2 ms. The target limit is artificially set by the operator in the first step training and is modified in the second step training using the output limit from the trained network. To improve the prediction performance further, the difference between the estimated limit and the measured and the other 11 parameters are inputted to a main neural network to calculate the stability level. Major disruptions have been predicted with a prediction success rate of 80% at 10 ms prior to the disruption while the false alarm rate is lower than 4%. This 80% is much higher than about 10% previously obtained. A prediction success rate of 90% has been also obtained with a false alarm rate of 12% at 10 ms prior to the disruption. This 12% is about a half of previously obtained one.
Shu, Y.; Nakajima, Norihiro
Proceedings of 11th International Conference on Human-Computer Interaction (HCI International 2005) (CD-ROM), 9 Pages, 2005/07
To understand the behavior of NPP (nuclear power plant) under different operating environment, JAERI is carrying out full-scaled plant simulation. As one part of full scaled plant simulation, our ongoing work is to develop an information inference system to manage and interpret NPP quake-proof data. In this paper, we proposed a hybrid data mining approach, which integrates human cognitive model in a data mining loop. Rule-based mining control agent emulated human analysts directly interacts with the data miner, analyzing and verifying the output of data miner and controlling data mining process. In additional, artificial neural network method, which is adopted as a core component of the proposed hybrid data mining method, is evolved by adding the retraining facility and explaining function for handling complicated nuclear power plant quake-proof data. To demonstrate how the method can be used as a powerful tool for extracting information relevant to plant safety and reliability, plant quake-proof testing data have been applied to the inference system.
Ono, Tomio*; Subekti, M.*; Kudo, Kazuhiko*; Takamatsu, Kuniyoshi; Nakagawa, Shigeaki; Nabeshima, Kunihiko
Nihon Genshiryoku Gakkai Wabun Rombunshi, 4(2), p.115 - 126, 2005/06
Control-rod withdrawal tests simulating reactivity insertion are carried out in the HTTR to verify the inherent safety features of HTGRs. This paper describes pre-test analysis method using artificial neural networks to predict the changes of reactor power and reactivity. The network model applied in this study is based on recurrent neural networks. The inputs of the network are the changes of the central control rods position and other significant core parameters, and the outputs are the changes of reactor power and reactivity. Furthermore, Time Synchronizing Signal(TSS) is added to input to improve the modeling of time series data. The actual tests data, which were previously carried out in the HTTR, were used for learning the model of the plant dynamics. After the learning, the network can predict the changes of reactor power and reactivity in the following tests.
Shu, Y.; Nakajima, Norihiro
Proceedings of 1st International Workshop on Risk Management System with Intelligent Data Analysis (RMDA 2005) in Conjunction with 19th Annual Conference of the Japanese Society for Artificial Intelligence (JSAI 2005), p.35 - 44, 2005/06
This paper presents an intelligent information inference system based on a hybrid data mining approach, which integrates human cognitive model in a data mining loop. In the proposed system, the mining control agent emulated human analysts interacts directly with the data miner, analyzing and verifying the output of the data miner and controlling the data mining process. In additional, the neural network method, which is adopted as a core component of the proposed hybrid data mining method, is evolved by adding the retraining facility and explaining function for handling complicated quake-proof data of nuclear power plant. To demonstrate how the method can be used as a powerful tool for extracting information relevant to plant safety and reliability, plant quake-proof testing data have been applied to the inference system.
Nabeshima, Kunihiko; Ayaz, E.*; Seker, S.*; Barutcu, B.*; T
rkcan, E.*
Proceedings of International Conference on Artificial Neural Networks and the International Conference on Neural Information Processing (ICANN/ICONIP 2003), p.406 - 409, 2003/06
On-line plant monitoring system with neural networks and an expert system has been developed for Borssele Nuclear Power Plant (NPP) in the Netherlands. The feedforward and the recurrent neural networks are utilized for plant modeling and anomaly detection. The rule-based expert system is applied for plant diagnosis with the outputs of the neural networks. The off-line results showed that the neural network could model the plant dynamics precisely. The on-line results indicated that the monitoring system could sufficiently diagnose the plant status in real time.
Nabeshima, Kunihiko; Suzudo, Tomoaki; Ono, Tomio*; Kudo, Kazuhiko*
Mathematics and Computers in Simulation, 60(3-5), p.233 - 244, 2002/09
Times Cited Count：16 Percentile：70.82(Computer Science, Interdisciplinary Applications)This study presents a hybrid monitoring system for nuclear reactor utilizing neural networks and a rule-based real-time expert system. The whole monitoring system including a data acquisition system and the advisory displays has been tested by an on-line simulator of pressurized water reactor. From the testing results, it was shown that the neural network in the monitoring system successfully modeled the plant dynamics and detected the symptoms of anomalies earlier than the conventional alarm system. The real-time expert system also worked satisfactorily in diagnosing and displaying the system status by using the outputs of neural networks and a priori knowledge base.
Tsuji, Hirokazu; Fujii, Hidetoshi*
Proceedings of 10th German-Japanese Workshop on Chemical Information, p.127 - 130, 2002/00
A neural network model within a Bayesian framework was adopted based on the material database constructed by JAERI for prediction of creep rupture properties of irradiated type 304 stainless steel. Stress level was modeled as a function of 18 variables, including rupture life, creep test temperature, chemical compositions; 10 elements, heat treatment temperature, heat treatment duration, neutron irradiation temperature, fast neutron fluence, thermal neutron fluence, irradiation time, based on JAERI material database in which 347 creep rupture data sets of type 304 stainless steels were stored. The Bayesian method puts error bars on the predicted values of the rupture strength and allows the significance of each individual factor to be estimated.
Nabeshima, Kunihiko; Suzudo, Tomoaki; Ono, Tomio*; Kudo, Kazuhiko*
Knowledge-Based Intelligent Information Engineering Systems & Allied Technologies, p.1506 - 1510, 2001/09
no abstracts in English
Kugo, Teruhiko; Nakagawa, Masayuki
Journal of Nuclear Science and Technology, 36(4), p.332 - 343, 1999/04
Times Cited Count：0 Percentile：0.01(Nuclear Science & Technology)no abstracts in English
Nabeshima, Kunihiko; Tuerkcan, E.*; Suzudo, Tomoaki; Nakagawa, Shigeaki; Inoue, K.*; Oono, Tomio*; ; Suzuki, Katsuo
Proc. of Human-Computer Interaction International'99, 2, p.1187 - 1191, 1999/00
no abstracts in English
Nabeshima, Kunihiko; Suzudo, Tomoaki; Suzuki, Katsuo; Tuerkcan, E.*
Journal of Nuclear Science and Technology, 35(2), p.93 - 100, 1998/02
Times Cited Count：36 Percentile：91.39(Nuclear Science & Technology)no abstracts in English
Kugo, Teruhiko; Nakagawa, Masayuki
Proc. of Int. Conf. on the Phys. of Nucl. Sci. and Technol., 1, p.704 - 711, 1998/00
no abstracts in English
Nabeshima, Kunihiko; Suzuki, Katsuo; Nose, Shoichi*;
Monitoring and Diagnosis Systems to Improve Nuclear Power Plant Reliability and Safety, 0, p.17 - 26, 1996/00
no abstracts in English
Kugo, Teruhiko; Nakagawa, Masayuki
PHYSOR 96: Int. Conf. on the Physics of Reactors, 1, p.B73 - B81, 1996/00
no abstracts in English
