End-to-end discriminative representation learning for fault diagnosis in safety-critical time series

Dong, F.*; Xiao, Y.*; Chen, S.*; 出町 和之*; 高屋 茂; 吉川 雅紀

Advanced Engineering Informatics, 69(Part D), p.104094_1 - 104094_23, 2026/01

https://doi.org/10.1016/j.aei.2025.104094

Ensuring safe and stable industrial plant operations requires accurate and timely fault diagnosis from multivariate time-series (MTS) sensor data. Conventional methods struggle with the complexity, high dimensionality, and limited feature extraction. To address this, we propose a novel end-to-end fault diagnosis framework that enhances class separability by leveraging instance-wise global and timestamp-wise local views of MTS representations. The dataset is augmented from the dual-view, and a complementary contrastive loss function captures both global and local contextual information. Unlike previous representation learning approaches, the diagnosis model's backbone and classifier are jointly optimized in an end-to-end scheme, ensuring aligned representations with the classification objective. Experimental results on simulated nuclear power plant fault datasets demonstrate the framework's effectiveness and robustness.

Enhancing mechanical properties of medium Mn steel by warm rolling based on laminated elemental segregation

Chen, H. F.*; Liu, B. X.*; 徐 平光; Fang, W.*; Tong, H. C.*; Yin, F. X.*

Journal of Materials Research and Technology, 32, p.3060 - 3069, 2024/09

https://doi.org/10.1016/j.jmrt.2024.08.060

The hot-rolled microstructure of medium Mn steel has coarse grains and severe elemental segregation, resulting in low strength and plasticity. Constructing a multiphase structure, refining the microstructure, and regulating elemental segregation enhance the mechanical properties. In this study, liquid nitrogen treatment created a layered distribution of austenite and martensite. Warm rolling was then used to reduce layer thickness and refine grain structure. After liquid nitrogen and warm rolling treatments, the strength and plasticity of medium Mn steel increased to 1270 MPa and 23.3%, respectively, far exceeding the hot-rolled state (724 MPa, 12.8%). Warm rolling also triggers austenite reverted transformation (ART) and introduces high-density dislocations, further improving austenite stability. This strengthening effect is higher than that from intercritical annealing alone. Improved austenite stability delays the transformation induced plasticity (TRIP) effect, preventing brittle fracture and enhancing deformation coordination between layers, significantly increasing the plastic deformation capacity of medium Mn steel.

Gradient residual strain determination of surface impacted railway S38C axles by neutron Bragg-edge transmission imaging

Hu, F. F.*; Qin, T. Y.*; Ao, N.*; Su, Y. H.; Zhou, L.*; 徐 平光; Parker, J. D.*; 篠原 武尚; Chen, J.*; Wu, S. C.*

Engineering Fracture Mechanics, 306, p.110267_1 - 110267_18, 2024/08

https://doi.org/10.1016/j.engfracmech.2024.110267

Non-destructive and quantitative mapping of gradient residual strain distribution in surface-hardened railway S38C axles could provide a positive reference for determining service lifetime and maintenance strategy. To tackle this concern, time-of-flight neutron Bragg-edge transmission imaging was employed by real axle samples with and without impacted crater. A novel and simple procedure to formulate the residual strain field was also developed in this work, with the transmission batch code in Appendix A. By mapping the global two- dimensional residual strains, it can be verified that the residual strains into the axle are uniformly distributed in the hoop direction. Subsequently, it was revealed that the axial and hoop residual strains, respectively in the cylinder and the long strip samples prepared from a real S38C hollow axle, indicated a gradient evolution distribution with a depth of 8 mm, covering a range of -5500 1000 for axial strains and -6500 1000 for hoop strains. More importantly, the maximum compressive lattice strain of the cylinder sample was increased by 15.61%, and 22.35% at the impacting speeds of 100, and 125 m/s, respectively; and that of the long strip sample increased by 29.17%, and 43.70%, respectively. It can thus be concluded that lattice strains have redistributed around the impact crater, demonstrating the local alteration of the residual strain field. These new findings suggest the localized variation in residual strains should be taken into account while evaluating the service damage evolution of railway axles, especially those affected by high-speed impacts during operation.

Strongly coupled spin waves and surface acoustic waves at room temperature

Hwang, Y.*; Puebla, J.*; 近藤 浩太*; Gonzalez-Ballestero, C.*; 一色 弘成*; Snchez Muoz, C.*; Liao, L.*; Chen, F.*; Luo, W.*; 前川 禎通*; et al.

Physical Review Letters, 132(5), p.056704_1 - 056704_7, 2024/01

https://doi.org/10.1103/PhysRevLett.132.056704

Here, we report the observation of strong coupling between magnons and surface acoustic wave (SAW) phonons in a thin CoFeB film constructed in an on-chip SAW resonator by analyzing SAW phonon dispersion anticrossings. We employ a nanostructured SAW resonator design that, in contrast to conventional SAW resonators, allows us to enhance shear-horizontal strain. Crucially, this type of strain couples strongly to magnons. Our device design provides the tunability of the film thickness with a fixed phonon wavelength, which is a departure from the conventional approach in strong magnon-phonon coupling research. We detect a monotonic increase in the coupling strength by expanding the film thickness, which agrees with our theoretical model. Our work offers a significant way to advance fundamental research and the development of devices based on magnon-phonon hybrid quasiparticles.

Extreme phonon anharmonicity underpins superionic diffusion and ultralow thermal conductivity in argyrodite AgSnSe

Ren, Q.*; Gupta, M. K.*; Jin, M.*; Ding, J.*; Wu, J.*; Chen, Z.*; Lin, S.*; Fabelo, O.*; Rodriguez-Velamazan, J. A.*; 古府 麻衣子; et al.

Nature Materials, 22(8), p.999 - 1006, 2023/08

https://doi.org/10.1038/s41563-023-01560-x

Ultralow thermal conductivity and fast ionic diffusion endow superionic materials with excellent performance both as thermoelectric converters and as solid-state electrolytes. Yet the correlation and interdependence between these two features remain unclear owing to a limited understanding of their complex atomic dynamics. Here we investigate ionic diffusion and lattice dynamics in argyrodite AgSnSe using synchrotron X-ray and neutron scattering techniques along with machine-learned molecular dynamics. We identify a critical interplay of the vibrational dynamics of mobile Ag and a host framework that controls the overdamping of low-energy Ag-dominated phonons into a quasi-elastic response, enabling superionicity. Concomitantly, the persistence of long-wavelength transverse acoustic phonons across the superionic transition challenges a proposed 'liquid-like thermal conduction' picture. Rather, a striking thermal broadening of low-energy phonons, starting even below 50 K, reveals extreme phonon anharmonicity and weak bonding as underlying features of the potential energy surface responsible for the ultralow thermal conductivity ( 0.5 WmK) and fast diffusion. Our results provide fundamental insights into the complex atomic dynamics in superionic materials for energy conversion and storage.

Attention-based time series analysis for data-driven anomaly detection in nuclear power plants

Dong, F.*; Chen, S.*; 出町 和之*; 吉川 雅紀; 関 暁之; 高屋 茂

Nuclear Engineering and Design, 404, p.112161_1 - 112161_15, 2023/04

https://doi.org/10.1016/j.nucengdes.2023.112161

To ensure nuclear safety, timely and accurate anomaly detection is of utmost importance in the daily condition monitoring of Nuclear Power Plants (NPPs), as any slight anomaly in a plant may result in an irreversible and serious accident, as well as high costs of maintenance and management. Nevertheless, due to the unique inherent attributes of anomalies, the difficulty of automatic detection in NPPs is increased. Previous model-driven anomaly detection approaches required skilled priori knowledge, leading to their limited usability. Commonly adopted deep learning-based data-driven anomaly detection approaches may not easily acquire the most relevant features when dealing with sensor data containing redundant information with uneven distribution of anomalies. To alleviate these issues, this paper propose an attention-based time series model for anomaly detection to ensure safety in NPPs. First, we employ one-dimension convolutional neural network (1D-CNN) backbone for feature extraction to preserve original inherent features of time series inputs. Subsequently, we originally adopt soft-attention mechanism to automatically extract the most relevant temporal features considering the specificity of anomaly detection in NPPs. The performance of the proposed model was experimentally validated on the High Temperature Gas-cooled Reactor (HTGR) anomaly case dataset simulated using the analytical code. The experimental results indicate that the proposed model was capable of detecting anomalies in NPPs with superior performance to the baseline model, while ensuring fast detection at short time steps.

Extended neutron orbital and the shell closure in Ca

Enciu, M.*; Liu, H. N.*; Obertelli, A.*; Doornenbal, P.*; Nowacki, F.*; 緒方 一介*; Poves, A.*; 吉田 数貴; Achouri, N. L.*; 馬場 秀忠*; et al.

Physical Review Letters, 129(26), p.262501_1 - 262501_7, 2022/12

https://doi.org/10.1103/PhysRevLett.129.262501

230MeV/nucleonでのCaからの中性子ノックアウト反応が線分光と行われ、と軌道からの中性子ノックアウト反応の運動量分布が測定された。断面積はの閉殻と整合し、Ca同位体でのと閉殻と同程度に強い閉殻であることが確認された。運動量分布の分析からと軌道の平均二乗根半径の差は0.61(23)fmと決定され、これはmodified-shell-modelによる予言の0.7fmと整合した。これは、中性子過剰なCa同位体での軌道半径が大きいことが、中性子数にしたがって線形的に荷電半径が増える意外な現象の原因であることを示唆している。

A Weakly supervised time series analysis framework for anomaly detection in nuclear power plants

Dong, F.*; Chen, S.*; 出町 和之*; 吉川 雅紀; 関 暁之; 高屋 茂

Proceedings of 29th International Conference on Nuclear Engineering (ICONE 29) (Internet), 7 Pages, 2022/08

https://doi.org/10.1115/ICONE29-91609

Condition monitoring is essential to the management and maintenance of Nuclear Power Plants (NPPs), as anomalies in the condition of components can affect the normal operation state of the entire plant. Therefore, timely and automatic detection of anomalies plays an important role and is in high demand. At present, deep learning is widely used for anomaly detection. Nevertheless, anomalies are difficult to define, sparsely occurring, and are accompanied by variable noise labels, which poses challenges to detection. Moreover, the problems such as loss of temporal features and gradient vanishing that exist in general deep learning models when dealing with time series data also increase the difficulty of anomaly detection. In response to these problems, a weakly supervised time series analysis framework for anomaly detection in NPPs is proposed, constituted of weakly supervised learning (WSL) and attention mechanism. The validation of the proposed framework was performed on the HTTR (High Temperature Engineering Test Reactor) anomaly cases dataset using the analytical code "ACCORD", which distributed anomalies independently across multiple instruments and were recorded from the responding sensors to each anomaly. At this stage, 3 classes of anomalies in higher priority were used as input data for the validation experiments. The experimental results demonstrate the effectiveness and feasibility of the proposed framework on anomaly tasks for the condition monitoring of NPPs.

A First glimpse at the shell structure beyond Ca; Spectroscopy of K, Ca, and Ca

小岩井 拓真*; Wimmer, K.*; Doornenbal, P.*; Obertelli, A.*; Barbieri, C.*; Duguet, T.*; Holt, J. D.*; 宮城 宇志*; Navrtil, P.*; 緒方 一介*; et al.

Physics Letters B, 827, p.136953_1 - 136953_7, 2022/04

AA2021-0775.pdf:0.62MB

https://doi.org/10.1016/j.physletb.2022.136953

中性子過剰核Caでは、新魔法数34が発見されて以来、その構造を知るために多くの実験がなされてきたが、それを超える中性子過剰核の情報は全く知られてこなかった。本論文では、理化学研究所RIBFにてK, Ca, Caの励起状態から脱励起するガンマ線を初めて観測した結果を報告した。それぞれ1つのガンマ線しか得られなかったものの、KおよびCaのデータは、それぞれ、陽子のと軌道間のエネルギー差、中性子のと軌道間のエネルギー差を敏感に反映し、両方とも最新の殻模型計算によって200keV程度の精度で再現できることがわかった。また、1粒子状態の程度を特徴づける分光学的因子を実験データと歪曲波インパルス近似による反応計算から求め、その値も殻模型計算の値と矛盾しないことがわかった。

The Fe(n,)Fe cross section from the surrogate ratio method and its effect on the Fe nucleosynthesis

Yan, S. Q.*; Li, X. Y.*; 西尾 勝久; Lugaro, M.*; Li, Z. H.*; 牧井 宏之; Pignatari, M.*; Wang, Y. B.*; Orlandi, R.; 廣瀬 健太郎; et al.

Astrophysical Journal, 919(2), p.84_1 - 84_7, 2021/10

https://doi.org/10.3847/1538-4357/ac12ce

The long-lived Fe (with a half-life of 2.62 Myr) is a crucial diagnostic of active nucleosynthesis in the Milky Way galaxy and in supernovae near the solar system. The neutron-capture reaction Fe(n,)Fe on Fe (half-life=44.5 days) is the key reaction for the production of Fe in massive stars. This reaction cross section has been previously constrained by the Coulomb dissociation experiment, which offered partial constraint on the E1 -ray strength function but a negligible constraint on the M1 and E2 components. In this work, for the first time, we use the surrogate ratio method to experimentally determine the Fe(n,)Fe cross sections in which all the components are included. We derived a Maxwellian-averaged cross section of 27.53.5 mb at = 30 keV and 13.41.7 mb at = 90 keV, roughly 10%-20% higher than previous estimates. We analyzed the impact of our new reaction rates in nucleosynthesis models of massive stars and found that uncertainties in the production of Fe from the Fe(n,)Fe rate are at most 25. We conclude that stellar physics uncertainties now play a major role in the accurate evaluation of the stellar production of Fe.