断层(地质)
电池(电)
过程(计算)
聚类分析
电压
动态时间归整
故障检测与隔离
计算机科学
工程类
特征(语言学)
可靠性工程
基质(化学分析)
模式识别(心理学)
区间(图论)
实时计算
控制理论(社会学)
作者
Yue Wang,Jiahui Ren,Lefu Sun,Hankun Liu,Zeyu Cheng,Xiangjun Li,Yunlong Shang
标识
DOI:10.1016/j.xcrp.2026.103654
摘要
Effective fault diagnosis is crucial for the safe use of lithium-ion batteries. However, existing methods often overlook stable and controllable conditions during the charging process, potentially degrading diagnostic performance. Here, a two-stage multifault diagnosis method integrated in charging process is proposed. First, micro pulses are injected during charging. Then, an electrical-based charging process diagnosis model (CPDM) is constructed to achieve rapid diagnosis in the first stage. For minor faults difficult to identify through rapid diagnosis, the voltage features are calculated based on the dynamic time warping distance, and fault identification and severity quantification are achieved using density estimation clustering and gated recurrent unit network, thereby enabling the refined diagnosis in the second stage. Experimental results demonstrate that the mean quantifiable error for poor contact faults is 7.2%, while the quantifiable accuracy for short-circuit faults reaches 99.7%. This provides an effective and practical fault diagnosis method for battery packs.
科研通智能强力驱动
Strongly Powered by AbleSci AI