Abstract：Fault diagnosis of high-voltage circuit breakers has conventionally relied on supervised learning approaches, requiring a substantial volume of labeled fault data for model training. However, this approach encounters significant challenges in practical applications due to the scarcity of fault labels and the dynamic nature of real-world operating conditions. Moreover, traditional diagnostic models, particularly those based on deep learning, often function as “black boxes”, lacking interpretability in their decision-making processes. It is challenging to elucidate the rationale behind their predictions, thereby restricting their applicability in safety-critical power systems where transparency and reliability are paramount. This study proposes a novel, unsupervised fault diagnosis method based on an explainable artificial intelligence (XAI) framework. The method aims to mitigate the dependency on labeled data while enhancing model interpretability, offering an efficient, transparent, and robust solution for fault diagnosis of electromagnetic mechanism circuit breakers. The proposed method begins by constructing logarithmic Mel spectrograms with mixed features through the fusion of acoustic and vibration signals. Acoustic signals that excel in capturing high-frequency anomalies and vibration signals that are sensitive to low-frequency mechanical faults are combined via a weighted fusion process (with weights of 0.9 for acoustic signals and 0.1 for vibration signals). This fusion strategy optimizes the complementary strengths of both signal types, and spectrograms that comprehensively represent the operational state of the circuit breaker. A convolutional autoencoder (CAE) is then employed for feature extraction, trained exclusively on logarithmic Mel spectrograms from normal operating conditions. The reconstruction error between the input and the reconstructed output is calculated and compared to a dynamic threshold derived from the mean and standard deviation of healthy data. It is shown that the CAE effectively distinguishes between normal and faulty states. This unsupervised fault detection approach eliminates the need for labeled fault data, demonstrating strong generalization capabilities even in the presence of noise and varying operational conditions. Building on the fault detection phase, this study further integrates DBSCAN clustering and the integrated gradients (IG) method to achieve precise fault segmentation and interpretable fault diagnosis. The latent space features extracted by the CAE are clustered using DBSCAN, a density-based algorithm that autonomously identifies fault patterns without predefined cluster numbers or labeled data. This step generates pseudo-labels for fault segmentation to categorize fault types. The IG method quantifies the contributions of latent space features, using the normal state spectrogram as a baseline. By mapping these contributions back to the input spectrograms, the method reveals the relationships between fault types and specific signal characteristics, such as frequency bands and temporal distributions. Finally, a classifier is introduced and trained using the pseudo-labels and latent space features, completing the fault diagnosis pipeline with high accuracy. Experimental validation was conducted on an electromagnetic mechanism circuit breaker under six common fault conditions, including insufficient drive voltage, loose bolts, inadequate lubrication, and other similar scenarios. The results demonstrate that under conditions of missing fault labels, the fault detection accuracy is 99.2%, and the fault diagnosis accuracy is 100%. Compared to traditional clustering methods (e.g., K-means, GMM) and advanced deep learning models (e.g., CBAM, DRSN, WDCNN), the proposed approach outperforms in terms of clustering performance, diagnostic accuracy, and computational efficiency. This paper offers substantial support for the safe and stable operation of power systems.
[1] 李辰辉, 褚继峰, 龙潇, 等. 基于弧触头接触振动特征分析的高压SF6断路器电寿命在线监测方法[J]. 电工技术学报, 2024, 39(15): 4883-4895. Li Chenhui, Chu Jifeng, Long Xiao, et al.Online monitoring method of electrical life of high voltage SF6 circuit breaker based on analysis of arcing contact vibration characteristics[J]. Transactions of China Electrotechnical Society, 2024, 39(15): 4883-4895. [2] 陶彩霞, 王文博, 李泰国, 等. 一种双电容换流型混合式直流断路器[J]. 电气工程学报, 2024, 19(3): 315-324. Tao Caixia, Wang Wenbo, Li Taiguo, et al.A double capacitor commutated hybrid DC circuit breaker[J]. Journal of Electrical Engineering, 2024, 19(3): 315-324. [3] 王奕飞, 林莘, 徐建源, 等. 高压断路器操动机构驱动电机及其控制技术研究[J]. 电机与控制学报, 2024, 28(1): 95-104. Wang Yifei, Lin Xin, Xu Jianyuan, et al.Motor design and control technology of high-voltage circuit breaker operating device[J]. Electric Machines and Control, 2024, 28(1): 95-104. [4] 于晨晖, 王浩名, 李一, 等. 基于深度森林的高压断路器弹簧机构状态识别[J]. 高压电器, 2023, 59(2): 37-43, 51. Yu Chenhui, Wang Haoming, Li Yi, et al.State identification of spring operating mechanism for high voltage circuit breaker based on deep forest[J]. High Voltage Apparatus, 2023, 59(2): 37-43, 51. [5] 李天辉, 庞先海, 范辉, 等. 基于IEMD和GA-WNN的断路器分合闸线圈故障诊断方法[J]. 中国电力, 2022, 55(5): 111-121. Li Tianhui, Pang Xianhai, Fan Hui, et al.Fault diagnosis method for circuit breaker opening and closing coil based on IEMD and GA-WNN[J]. Electric Power, 2022, 55(5): 111-121. [6] 范兴明, 许洪华, 张思舜, 等. 基于IDBO-DHKELM的高压断路器故障诊断研究[J].电工技术学报, 2025, 40(12): 3994-4003. Fan Xingming, Xu Honghua, Zhang Sishun, et al.Fault diagnosis of high voltage circuit breaker based on improved dung beetle optimizer algorithm deep hybrid kernel extreme learning machine[J]. Transa- ctions of China Electrotechnical Society, 2025, 40(12): 3994-4003. [7] 林琳, 陈志英. 基于粗糙集神经网络和振动信号的高压断路器机械故障诊断[J]. 电工技术学报, 2020, 35(增刊1): 277-283. Lin Lin, Chen Zhiying.Mechanical fault diagnosis of high voltage circuit breakers based on rough set neural networks and vibration signals[J]. Transactions of China Electrotechnical Society, 2020, 35(S1): 277-283. [8] 杨元威, 关永刚, 陈士刚, 等. 基于声音信号的高压断路器机械故障诊断方法[J]. 中国电机工程学报, 2018, 38(22): 6730-6737. Yang Yuanwei, Guan Yonggang, Chen Shigang, et al.Mechanical fault diagnosis method of high voltage circuit breaker based on sound signal[J]. Proceedings of the CSEE, 2018, 38(22): 6730-6737. [9] 孙玉伟, 罗林根, 陈敬德, 等. 含噪背景下基于盲源分离与NSVDD的断路器机械故障诊断方法[J]. 高电压技术, 2022, 48(3): 1104-1113. Sun Yuwei, Luo Lingen, Chen Jingde, et al.Mechanical fault diagnosis method of circuit breaker based on blind source separation and NSVDD under noisy background[J]. High Voltage Engineering, 2022, 48(3): 1104-1113. [10] 刘亚魁, 李红运, 林天然, 等. 基于机器视觉的高压断路器机械特性测量方法[J]. 电工技术学报, 2023, 38(增刊1): 222-230. Liu Yakui, Li Hongyun, Lin Tianran, et al.Measure- ment method of mechanical characteristics of high voltage circuit breaker based on machine vision[J]. Transactions of China Electrotechnical Society, 2023, 38(S1): 222-230. [11] 万书亭, 郭胡森, 豆龙江, 等. 基于电流-振动信号联合特征的高压断路器故障诊断方法[J]. 电子测量技术, 2024, 47(20): 140-149. Wan Shuting, Guo Husen, Dou Longjiang, et al.Fault diagnosis method of high voltage circuit breaker based on the combined characteristics of current and vibration signal[J]. Electronic Measurement Techno- logy, 2024, 47(20): 140-149. [12] 马莉, 霍耀佳, 吴杨, 等. 基于VMD和KFCM-SVM的高压断路器声振联合故障诊断方法[J]. 高压电器, 2024, 60(8): 53-62. Ma Li, Huo Yaojia, Wu Yang, et al.Acoustic and vibration joint fault diagnosis method of high voltage circuit breaker based on VMD and KFCM-SVM[J]. High Voltage Apparatus, 2024, 60(8): 53-62. [13] 李建鹏, 赵冀宁, 孟延辉, 等. 高压断路器操动机构非侵入式多信息检测及故障诊断研究[J]. 高压电器, 2024, 60(7): 128-137. Li Jianpeng, Zhao Jining, Meng Yanhui, et al.Research on non-invasive multi-information detection and fault diagnosis for operating mechanism of high voltage circuit breaker[J]. High Voltage Apparatus, 2024, 60(7): 128-137. [14] 任明, 陈荣发, 夏昌杰, 等. 基于三元图谱的真空断路器机电故障诊断研究[J]. 中国电机工程学报, 2022, 42(增刊1): 332-342. Ren Ming, Chen Rongfa, Xia Changjie, et al.Research on electromechanical fault diagnosis of vacuum circuit breaker based on ternary atlas[J]. Proceedings of the CSEE, 2022, 42(S1): 332-342. [15] 王昱皓, 武建文, 马速良, 等. 基于核主成分分析- SoftMax的高压断路器机械故障诊断技术研究[J]. 电工技术学报, 2020, 35(增刊1): 267-276. Wang Yuhao, Wu Jianwen, Ma Suliang, et al.Mechanical fault diagnosis research of high voltage circuit breaker based on kernel principal component analysis and SoftMax[J]. Transactions of China Electrotechnical Society, 2020, 35(S1): 267-276. [16] 司江宽, 吐松江·卡日, 范想, 等. 基于SO-PAA- GAF和AdaBoost集成学习的高压断路器故障诊断[J]. 电力系统保护与控制, 2024, 52(3): 152-160. Si Jiangkuan, Tusongjiang·Kari, Fan Xiang, et al. Fault diagnosis of high-voltage circuit breaker based on SO-PAA-GAF and AdaBoost ensemble learning[J]. Power System Protection and Control, 2024, 52(3): 152-160. [17] Sun Shuguang, Sun Liang, Wang Jingqin, et al.Fault diagnosis of conventional circuit breaker accessories based on grayscale image of current signal and improved ZFNet-DRN[J]. IEEE Sensors Journal, 2023, 23(2): 1343-1356. [18] 王艳新, 闫静, 王建华, 等. 基于特征融合度量学习的高压断路器机械故障诊断[J]. 仪器仪表学报, 2022, 43(9): 98-105. Wang Yanxin, Yan Jing, Wang Jianhua, et al.Mechanical fault diagnosis for high voltage circuit breaker via a novel feature fusion metric learning[J]. Chinese Journal of Scientific Instrument, 2022, 43(9): 98-105. [19] Lu Yang, Li Yongli.A novel fault diagnosis method for circuit breakers based on optimized affinity pro- pagation clustering[J]. International Journal of Elec- trical Power & Energy Systems, 2020, 118: 105651. [20] Ye Xinyu, Yan Jing, Wang Yanxin, et al.A novel U-Net and capsule network for few-shot high-voltage circuit breaker mechanical fault diagnosis[J]. Measure- ment, 2022, 199: 111527. [21] Li Xiaofeng, Chen Hantao, Xie Fuqi, et al.Hybrid model of multiple echo state network integrated by evidence fusion for fault diagnosis of a high-voltage circuit breaker[J]. IEEE Transactions on Consumer Electronics, 2024, 70(3): 5269-5277. [22] 严如强, 周峥, 杨远贵, 等. 可解释人工智能在工业智能诊断中的挑战和机遇: 归因解释[J]. 机械工程学报, 2024, 60(12): 21-40. Yan Ruqiang, Zhou Zheng, Yang Yuangui, et al.Challenges and opportunities of XAI in industrial intelligent diagnosis: attribution interpretation[J]. Journal of Mechanical Engineering, 2024, 60(12): 21-40. [23] 严如强, 商佐港, 王志颖, 等. 可解释人工智能在工业智能诊断中的挑战和机遇: 先验赋能[J]. 机械工程学报, 2024, 60(12): 1-20. Yan Ruqiang, Shang Zuogang, Wang Zhiying, et al.Challenges and opportunities of XAI in industrial intelligent diagnosis: priori-empowered[J]. Journal of Mechanical Engineering, 2024, 60(12): 1-20. [24] 房佳姝, 刘崇茹, 苏晨博, 等. 基于自注意力Trans- former编码器的多阶段电力系统暂态稳定评估方法[J]. 中国电机工程学报, 2023, 43(15): 5745-5759. Fang Jiashu, Liu Chongru, Su Chenbo, et al.Multi- stage transient stability assessment of power system based on self-attention Transformer encoder[J]. Pro- ceedings of the CSEE, 2023, 43(15): 5745-5759. [25] Xie Tianming, Xu Qifa, Jiang Cuixia, et al.The fault frequency priors fusion deep learning framework with application to fault diagnosis of offshore wind turbines[J]. Renewable Energy, 2023, 202: 143-153. [26] 吕云龙, 胡琴, 胡紫园, 等. 考虑样本不平衡条件下风机叶片覆冰诊断及其可解释性研究[J]. 电工技术学报, 2025, 40(11): 3667-3679. 27 Lü Yunlong, Hu Qin, Hu Ziyuan, et al.Diagnosis and interpretability study of wind turbine blade icing under consideration of sample imbalance condi- tions[J]. Transactions of China Electrotechnical Society, 2025, 40(11): 3667-3679. [27] 刘会兰, 常庚垚, 赵书涛, 等. 溯源弹簧形变过程的断路器振动信号递归量化分析辨识方法[J]. 电工技术学报, 2024, 39(8): 2567-2577. Liu Huilan, Chang Gengyao, Zhao Shutao, et al.Circuit breaker state identification method based on tracing spring deformation process and recurrence quantitative analysis of vibration signals[J]. Transa- ctions of China Electrotechnical Society, 2024, 39(8): 2567-2577. [28] 田敬越, 王昕, 唐浩哲, 等. 基于跨域特征与深度对比学习的断路器故障诊断方法研究[J/OL]. 华北电力大学学报(自然科学版), 1-10[2025-09-30]. https://link.cnki.net/urlid/13.1212.TM.20241111.1353.002. Tian Jingyue, Wang Xin, Tang Haozhe, et al. Research on circuit breaker fault diagnosis method based on cross-domain features and deep contrastive learning[J/OL]. Journal of North China Electric Power University, 1-10[2025-09-30]. https://link.cnki.net/urlid/13.1212.TM.20241111.1353.002. [29] 齐子豪, 仝杰, 张中浩, 等. 基于多粒度知识特征和Transformer网络的电力变压器故障声纹辨识方法[J]. 中国电机工程学报, 2025, 45(4): 1311-1323. Qi Zihao, Tong Jie, Zhang Zhonghao, et al.A voiceprint classification method for power trans- former fault identification based on multi-granularity knowledge features and transformer network[J]. Proceedings of the CSEE, 2025, 45(4): 1311-1323.
2026, Vol. 41 
