Supervisor of Master's Candidates
Name (Simplified Chinese): 吴慕遥
Name (Pinyin): wumuyao
E-Mail:
Date of Employment: 2022-12-27
School/Department: 车辆工程系
Education Level: With Certificate of Graduation for Doctorate Study
Business Address: 安徽省合肥市屯溪路193号合肥工业大学格物楼515
Gender: Male
Degree: Doctoral Degree in Engineering
Professional Title: Lecturer
Status: Employed
Alma Mater: 中国科学技术大学
Supervisor of Master's Candidates
Discipline: Automobile Engineering
MORE>
基于多尺度特征融合与1D-CNN的电池故障诊断
Hits:
Impact Factor:1.6
DOI number:10.19535/j.1001-1579.2026.05.003
Journal:电池
Key Words:动力电池;故障诊断;卷积神经网络;小波变换;多尺度特征融合;
Abstract:针对动力电池在复杂动态工况下早期故障特征微弱、传统方法难以有效辨识特征与工况耦合的问题，提出基于一维卷积神经网络(1D-CNN)的多级故障诊断方法。获取实车远程监控数据，经时间对齐、异常剔除、缺失修复及两级分段处理，构建电化学一致性较强的准连续样本单元；融合时域统计、小波频域特征与空间一致性指标，构建涵盖宏观分布、局部瞬态与离群偏差的26维多尺度特征向量；设计包含三级卷积块、批归一化及余弦退火策略的1D-CNN诊断架构，实现L0(正常)至L4(极严重)五级故障的端到端识别。基于138 160个实车样本进行验证，发现所提方法的诊断准确率达97.10%,宏平均F1值达92.40%,较传统阈值法提升36.6个百分点。用消融实验验证多尺度特征融合的有效性；5折交叉验证与多工况测试，证实模型在少数类故障识别及工况迁移上的鲁棒性。
Note:北大核心
Co-author:吴仕苇,杨彦辉,曾国建,崔伟,吴慕遥
First Author:王丽
Indexed by:Journal paper
Correspondence Author:吉祥
Discipline:Engineering
Document Type:J
Volume:56
Issue:05
Page Number:1190-1196
ISSN No.:1001-1579
Translation or Not:no
Date of Publication:2026-11-11
Links to published journals:https://kns.cnki.net/kcms2/article/abstract?v=G1GIFP_B-qWRGM2LjWy-KYmG943l0MBCG_IHV5hLWtcrpI9-tyozr2-EmqnXJSDnmPNnYBoEGoOnkUmBI3Ke8-FIRdFYffw6oIpUHCzyCp1Ghdp43e6Q_wrsZtCLRhBoZ5rZ_-ArF5-5LGbNiaG5y-wW3k
The Last Update Time : ..
