Few-Shot Mechanical Fault Diagnosis of High-Voltage Circuit Breakers Using Prior Knowledge-Constrained Wavelet-Based Multi-Frequency Feature Fusion

断路器 断层(地质) 计算机科学 融合 保险丝(电气) 特征提取 人工智能 特征(语言学) 模式识别(心理学) 传感器融合 故障检测与隔离 工程类 电子工程 故障指示器 信息融合 状态监测 瞬态分析 计算机视觉 人工神经网络 配电盘 信号处理
作者
Yanxin Wang,Jing Yan,Z H Zhang,Jianhua Wang,Yingsan Geng,Dipti Srinivasan
出处
期刊:IEEE Transactions on Industry Applications [Institute of Electrical and Electronics Engineers]
卷期号:: 1-11 被引量:1
标识
DOI:10.1109/tia.2026.3711620
摘要

Mechanical faults in high-voltage circuit breakers (HVCBs) significantly affect power system reliability and require accurate diagnosis for condition-based maintenance. However, practical applications are limited by scarce fault samples, strong environmental noise, and insufficient incorporation of physical mechanisms in existing data-driven methods, which reduces robustness and interpretability. To address these challenges, this paper proposes a few-shot mechanical fault diagnosis framework for HVCBs based on prior knowledge–constrained wavelet-based multi-frequency feature fusion. First, domain-specific prior knowledge is constructed through excitation–response mechanism analysis, and signal feature extraction principles, providing physically meaningful guidance for feature learning. Second, a wavelet-based multi-frequency feature fusion network is developed to refine backbone features while integrating multi scale frequency-domain information, thereby enhancing the representation of weak and subtle fault characteristics. Third, physically motivated prior knowledge constraints are embedded into the feature fusion process, enabling the learned features to remain consistent with fault mechanisms. Furthermore, a joint loss optimization strategy is adopted to mitigate class imbalance, and metric learning is incorporated to support reliable few-shot classification. Experimental results on HVCB fault datasets demonstrate that the proposed method achieves a diagnostic accuracy of 98.67% in the 5-way 5-shot task. Compared with convolutional neural network and the wavelet multi-frequency feature fusion network without prior knowledge constraints, the proposed method improves the diagnostic accuracy by 8.91% and 8.11%, respectively. In addition, the proposed framework exhibits strong robustness and generalization capability under severe noise interference and sample imbalance, indicating its effectiveness and practical value for real-world HVCB condition monitoring.
最长约 10秒,即可获得该文献文件

科研通智能强力驱动
Strongly Powered by AbleSci AI
科研通是完全免费的文献互助平台,具备全网最快的应助速度,最高的求助完成率。 对每一个文献求助,科研通都将尽心尽力,给求助人一个满意的交代。
实时播报
1秒前
1秒前
yyy完成签到,获得积分10
2秒前
孔wj完成签到,获得积分10
2秒前
2秒前
ili完成签到,获得积分10
3秒前
威武的沂发布了新的文献求助10
4秒前
冷静剑鬼完成签到,获得积分10
4秒前
万能图书馆的应助被uu采纳,获得10
4秒前
4秒前
隐形曼青的应助被uu采纳,获得10
4秒前
香蕉觅云的应助被uu采纳,获得10
4秒前
5秒前
5秒前
无花果的应助被uu采纳,获得10
5秒前
Orange的应助被颖颖不吃酸采纳,获得10
5秒前
molihuakai的应助被uu采纳,获得10
5秒前
打打的应助被uu采纳,获得10
5秒前
Ava的应助被uu采纳,获得10
5秒前
vary发布了新的文献求助10
5秒前
6秒前
6秒前
张l发布了新的文献求助10
7秒前
嘻嘻哈哈完成签到,获得积分20
8秒前
8秒前
8秒前
张原宇完成签到,获得积分10
9秒前
Elanie.zh完成签到,获得积分10
10秒前
Zoe_Zhang发布了新的文献求助10
11秒前
重要砖头完成签到,获得积分10
11秒前
科研小李发布了新的文献求助30
11秒前
wenhan发布了新的文献求助10
11秒前
11秒前
爆米花的应助被GBYYYY采纳,获得10
12秒前
Maestro_S的应助被renn采纳,获得30
12秒前
独特采萱完成签到,获得积分10
12秒前
科研爱溪溪完成签到,获得积分10
13秒前
晓风发布了新的文献求助10
14秒前
14秒前
15秒前
高分求助中
(应助此贴封号)【重要!!请各用户(尤其是新用户)详细阅读】【科研通的精品贴汇总】 10000
拉伸高分子熔体中缠结行为的模拟和小角中子散射研究 1700
The World Atlas of Wine 9th edition 1000
奈特基础组织学:配套组织病理学(第4版)(Netter's Essential Histology With Correlated Histopathology) 800
八旗制度の研究 600
Improving the wear resistance of vehicle gear oil by oil soluble LaF3 nanoparticles and typical additives of gear oil 500
Boundaries and Bodhisattvas 500
热门求助领域 (近24小时)
化学 材料科学 医学 生物 纳米技术 计算机科学 工程类 有机化学 化学工程 生物化学 物理 复合材料 内科学 催化作用 光电子学 人工智能 细胞生物学 无机化学 物理化学 数学
热门帖子
关注 科研通微信公众号,转发送积分 8184061
求助须知:如何正确求助?哪些是违规求助? 9573151
关于积分的说明 21829878
捐赠科研通 7781058
什么是DOI,文献DOI怎么找? 3434216
关于科研通互助平台的介绍 2558337
邀请新用户注册赠送积分活动 2480674