Acoustic detection of boiling states by deep convolutional neural networks; Visual explanation of identification basis

植木 祥高*; 平子 樹*; 手塚 晃輔*; 相澤 康介; 荒 邦章*

AI Thermal Fluids (Internet), 4, p.100021_1 - 100021_12, 2025/12

https://doi.org/10.1016/j.aitf.2025.100021

With a final goal of early detection and understanding of the transition of coolant boiling events in the core of sodium-cooled fast reactors, our present aim is to obtain and maintain the basic knowledge necessary for developing anomaly detection technology associated with local anomalies in the core and to demonstrate basic feasibility. We constructed a deep learning method and evaluated its performance to detect the occurrence and understand the transition of subcooled boiling using acoustic identification. In this research, we aim to acquire acoustic data during subcooled boiling of ultrapure water and learn feature quantities of the boiling in time-frequency expression. A deep learning model of a convolutional neural network for label classification was constructed. In addition to being able to identify the occurrence of boiling with high accuracy, the visualization of the identification basis using the gradient-weighted class activation mapping (Grad-CAM) method revealed the acoustic frequency bands that the deep learning model determined to be of high importance. We also constructed a regression analysis-type deep learning model and demonstrated that boiling heat flux values can be predicted with high accuracy.

Ensemble deep learning model for leak detection from multi-channel acoustic signals

三上 奈生; 相澤 康介; 植木 祥高*; Michel, F.*; Fache, J.*

Proceedings of 2025 International Congress on Advances in Nuclear Power Plants (ICAPP 2025) (Internet), 10 Pages, 2025/09

The present study evaluates the basic feasibility of an ensemble deep learning model to detect leakage from multi-channel acoustic signals in a steam generator (SG) of a sodium-cooled fast reactor (SFR). The acoustic signals from the bubbling and the gas blowout are measured by an array sensor in a basic experimental apparatus for SG to simulate noise and leak signals. Time-frequency representations (TFRs) are produced from these acoustic signals as the inputs of convolutional neural networks (CNNs). Three typical CNNs are introduced as candidates for the base model of ensemble deep learning. The proposed ensemble deep learning model reaches an accuracy of 95.43%, improved by 4.90% from the solo deep learning model. This result indicates that the proposed ensemble deep learning model has the potential to detect leakage more precisely in an actual SG of SFR.

Study on performance evaluation of self-actuated shutdown system for sodium-cooled fast reactor; Flow field measurement around a curie point electromagnet

山崎 僚太; 相澤 康介; 小林 順; 栗原 成計

Proceedings of 13th Korea-Japan Symposium on Nuclear Thermal Hydraulics and Safety (NTHAS13) (Internet), 7 Pages, 2024/11

A self-actuated shutdown system (SASS) is one of the innovative technologies for a sodium-cooled fast reactor. The SASS is a passive reactor-shutdown system that utilizes a Curie point electromagnet (CPEM), which features the characteristic of losing magnetism when the magnet temperature reaches the Curie point. The control rod with the CPEM is inserted into the core by gravity without recourse to any active shutdown system. To allow the effective function of the SASS, it is important ensure a quick response of the CPEM following the coolant temperature increase. Therefore, the CPEM is designed to have a radial fin-like structure to improve temperature response, and a fluid is guided into the narrow gaps between fins. Accordingly, it is important to understand the flow field around the CPEM, including the gaps between fins. In this study, a water experiment was performed to understand the characteristic flow field around the CPEM and to develop the SASS performance evaluation method. The test section modeled the structure from the outlets of fuel subassemblies to the CPEM. The velocity conditions at the fuel subassembly outlet were varied between 0.26m/s (Re=25,000) and 1.55m/s (Re=150,000). The velocity in the gaps between fins was directly measured using a hot-wire anemometer. In addition, the flow field around the CPEM was measured using particle image velocimetry (PIV). As the results, the velocity distributions around the CPEM and inside the gaps between fins were quantitatively obtained. It was shown that the velocity in the gaps between fins was higher at the upstream side than at the downstream side, indicating that the fluid flowed out from the gap between fins flowing to the downstream. Furthermore, these experimental results will be used as validation data of the SASS performance evaluation method in the future.

CNN-based acoustic identification of gas-liquid jet; Evaluation of noise resistance and visual explanation using Grad-CAM

三上 奈生*; 植木 祥高*; 芝原 正彦*; 相澤 康介; 荒 邦章*

International Journal of Multiphase Flow, 171, p.104688_1 - 104688_13, 2024/01

https://doi.org/10.1016/j.ijmultiphaseflow.2023.104688

For the analysis of anomalies in a steam generator (SG) of a sodium-cooled fast reactor (SFR), we evaluate the noise resistance of CNN-based acoustic identification methods of gas-liquid two-phase jets and produce visual explanations for their decisions. First, we introduce the water flow sound and the three types of gas-liquid jet sounds, which simulate the background noise and the anomaly sounds, respectively. Second, we produce time-frequency representations for various signal-to-noise ratios (SNRs) and employ AlexNet, VGG16, and ResNet18 to the identification of the gas-liquid two-phase jets. As a result, the best CNN of ResNet18 achieves more than 0.92 for SNR = 0, -4, -8, and -12 dB and 0.69 for SNR = -16 and -20 dB. This result indicates that our proposed methods can identify the flow states of gas-liquid two-phase jets in low-level noise environments and detect the gas-liquid two-phase jets even in high-level noise environments. Also, Grad-CAM suggests that ResNet18 focuses on one of the spectrum peaks of the water flow sound and all or part of the signal intensity pattern of the gas-liquid jet sounds. Our proposed methods lead to the safe operation and fast, accurate, and accountable analysis of anomalies in SFR.

State sensing of bubble jet flow based on acoustic identification of deep learning

三上 奈生*; 植木 祥高*; 芝原 正彦*; 相澤 康介; 荒 邦章

Proceedings of 17th International Heat Transfer Conference (IHTC-17) (Internet), 9 Pages, 2023/08

To increase the safety of sodium-cooled fast reactors, it is necessary to develop a method to identify the states of bubble jet flow caused when a heat transfer tube is damaged in steam generators (SGs). For this issue, we propose a novel state sensing method with time-frequency representations (TFRs) and convolutional neural networks (CNNs). This study consists of three phases. First, using water and air as simulant fluids to perform the proof of concept, pipe flow sound and bubble jet flow sound are acquired, each of which simulates normal and anomaly sound. Second, three TFRs are extracted from raw signals based on short-time Fourier transform (STFT), continuous wavelet transform (CWT), and synchrosqueezed wavelet transform (SWT). Third, typical CNNs including AlexNet, VGG16, and ResNet18 are introduced for the identification of pipe flow sound and three types of bubble jet flow sound. As a result, the model combining ResNet18 and STFT reaches the highest accuracy and correctly identifies 1984 out of 200 test data. These results demonstrate that our proposed method based on the acoustic identification of deep learning has great potential to sense the states of bubble jet flow in actual SFRs.

Boiling sensing based on acoustic recognition and deep learning

植木 祥高*; 橋本 俊作*; 芝原 正彦*; 相澤 康介; 荒 邦章

Proceedings of 30th International Conference on Nuclear Engineering (ICONE30) (Internet), 5 Pages, 2023/05

In sodium-cooled fast reactors, coolant boiling in reactor cores is one of the important phenomena in the safety assessment. Our final target of the present study is to realize the acoustic anomaly detection of the boiling inception in actual reactors. In the actual environment, various sorts of noises are expectedly superposed on accidental boiling sounds. It is inevitable to distinguish the boiling sounds from the superimposing hostile disturbance with high accuracy. To achieve this, we utilize machine learning techniques and assess the feasibility of boiling sensing based on acoustic recognition and deep learning. In the present study, we employ an autoencoder to denoise boiling sounds, and a convolutional neural network to detect the boiling inception. The boiling acoustics have not been fully understood yet. In the present study, we find that some characteristics of the boiling acoustics are consistent with the resonance vibration of the heating body. This finding contributes to elucidating the physics of boiling acoustics.

State sensing of bubble jet flow based on acoustic recognition and deep learning

三上 奈生*; 植木 祥高*; 芝原 正彦*; 相澤 康介; 荒 邦章

International Journal of Multiphase Flow, 159, p.104340_1 - 104340_8, 2023/02

https://doi.org/10.1016/j.ijmultiphaseflow.2022.104340

この研究では、ナトリウム冷却高速炉(SFR)の蒸気発生器(SG)管の損傷によって引き起こされる液中への気泡噴流の検知に関するものである。本研究の主な目的は、気泡噴流に対して音響認識と深層学習に基づく新しい状態検知手法を開発することである。新しい手法の適用性を評価するために、模擬流体として水と空気を用いた概念検討を実施した。まず、SFRのSG管からの正常音と異常音をシミュレートする配管内流動音と気泡噴流音の取得と分析を行い、音響信号と特徴周波数の整理を行った。得られた試験結果を基に、ディープラーニングモデルを構築し、性能評価を実施した。その結果、提案したすべてのモデルは、ほぼ100.00%の精度で配管内流動音と気泡噴流音を識別できた。最も良いモデルでは、配管内流動音と3種類の気泡噴流音を99.76%の精度で識別できた。この結果は、深層学習による音響認識が実際のSFRにおけるバブルジェットの流れの状態を感知する大きな可能性を秘めていることを示唆している。

Investigation on natural circulation behavior for decay heat removal in reactor vessel of sodium-cooled fast reactor under severe accident condition, 1; Effect of decay-heat conditions on natural circulation behavior under dipped-type DHX operation conditions

辻 光世; 相澤 康介; 小林 順; 栗原 成計

Proceedings of 12th Japan-Korea Symposium on Nuclear Thermal Hydraulics and Safety (NTHAS12) (Internet), 6 Pages, 2022/10

ナトリウム冷却高速炉(SFR)において、炉心溶融を含むシビアアクシデント時の安全性強化のため、炉内冷却機器の設計と運用を最適化することが重要である。SFRの原子炉容器を模擬した1/10縮尺の水試験装置を用いて、原子炉容器内部の自然循環現象を把握するための水試験を実施している。本報では、炉心燃料とコアキャッチャ上の燃料デブリの発熱割合が原子炉容器内部の自然循環挙動へ与える影響を調査するために、浸漬型DHXを運転した条件で実施した実験結果を示す。全体の発熱量を一定として、全体の発熱量に対するコアキャッチャ上の燃料デブリの発熱割合を20%, 80%とした2条件で原子炉容器内部の温度分布及び流速分布を計測した。炉心部とコアキャッチャ上の燃料デブリの発熱割合による炉容器内の自然循環挙動への影響を定量的に把握した。

Sodium-cooled Fast Reactors

大島 宏之; 森下 正樹*; 相澤 康介; 安藤 勝訓; 芦田 貴志; 近澤 佳隆; 堂田 哲広; 江沼 康弘; 江連 俊樹; 深野 義隆; et al.

Sodium-cooled Fast Reactors; JSME Series in Thermal and Nuclear Power Generation, Vol.3, 631 Pages, 2022/07

https://shop.elsevier.com/books/sodium-cooled-fast-reactors/morishita/978-0-12-824076-2

ナトリウム冷却高速炉(SFR: Sodium-cooled Fast Reactor)の歴史や、利点、課題を踏まえた安全性、設計、運用、メンテナンスなどについて解説する。AIを利用した設計手法など、SFRの実用化に向けた設計や研究開発についても述べる。

ワイヤスペーサ型燃料ピンのサブチャンネル内流動場に関する試験研究; 3本ピンバンドル体系における流動場の特性

檜山 智之; 相澤 康介; 西村 正弘; 栗原 成計

JAEA-Research 2021-009, 29 Pages, 2021/11

JAEA-Research-2021-009.pdf:2.25MB

ナトリウム冷却高速炉では、実用化に向けて燃料の高燃焼度化が求められている。高燃焼度の燃料集合体はスウェリングや熱的な燃料棒の変形によって局所的に除熱能力が低下することが懸念され、燃料集合体での冷却材流動挙動を予測評価することが重要である。本研究では、現象解明および熱流動解析コード検証用のデータベース構築を目的として、3本ピンバンドル体系の試験体を用いた流動場計測試験を実施した。現象解明の着眼点は、以下の(1)ワイヤスペーサ近傍を含めたサブチャンネル内の全体流況、(2)層流領域を含むレイノルズ数と流動場の関係、(3)ワイヤスペーサの有無が流動場に与える影響評価である。試験の結果、PIV計測によりサブチャンネル内の詳細な流動場データを取得し、3本ピンに囲まれたセンターサブチャンネルにワイヤスペーサが交差する際に隣のサブチャンネルへ逃げる流れとワイヤスペーサの巻き方向に対して追従する流れが生じることが分かった。層流領域のレイノルズ数条件では流速分布の傾向が遷移領域および乱流領域と大きく異なることを確認した。ワイヤスペーサのない体系と比較すると、層流領域においてもワイヤスペーサによるミキシングが生じていることを確認した。