摘要
滚动轴承在机械设备中至关重要,其健康状态直接关系到机械设备安全运行和整体性能,然而,实际运行中获取足够的故障样本进行研究是一项挑战。因此,针对实际工况下故障样本数量缺少、与正常样本数量相比形成类不平衡的情形,提出一种基于变分模态分解(VMD)重构数据增强的故障识别模型。首先,通过VMD分解和滤波调整将轴承故障信号重构为平衡数据集。其次,建立各故障类型样本特征参数与不同故障尺寸间关联性,实现生成样本特征评估。最后,通过深度学习YOLOv8算法对各不平衡比例数据集进行深入分析。分析实验结果表明,所提方法能有效扩充少样本场景下的轴承故障数据,提高故障识别精度,从数据层面解决类不平衡问题,对于轴承不平衡样本故障识别具有可行性和有效性。
Rolling bearings are crucial in mechanical equipment,and their health state directly influences the safe operation and overall performance of mechanical equipment.However,it is a challenge to obtain enough failure samples for research in actual operation.Therefore,addressing the imbalance in the number of fault samples compared to normal samples under practical operating conditions,a fault recognition model was proposed based on Variational Mode Decomposition(VMD)reconstructed data augmentation.The bearing fault signals were reconstructed into a balanced dataset by VMD decomposition and filtering adjustment.The correlation between sample feature parameters of various fault types and different fault dimensions was established,achieving generated sample feature assessment.Finally,in-depth analysis of imbalanced datasets was conducted using the deep learning YOLOv8 algorithm.Experimental results indicated that the proposed method could effectively augment bearing fault data in scenarios with few samples,enhance fault recognition accuracy and addresse the class imbalance issue at the data level.The method proved to be feasible and effective for recognition bearing imbalance sample faults.
作者
张锐
赵锦钰
郭洪飞
王燕
杨思妍
刘婷婷
周卫斌
游国栋
ZHANG Rui;ZHAO Jinyu;GUO Hongfei;WANG Yan;YANG Siyan;LIU Tingting;ZHOU Weibin;YOU Guodong(College of Electronic Information and Automation,Tianjin University of Science&Technology,Tianjin 300222,China;Inner Mongolia University of Technology,Hohhot 010051,China;School of Computer and Information Engineering,Tianjin Chengjian University,Tianjin 300384,China)
出处
《计算机集成制造系统》
北大核心
2026年第1期339-354,共16页
Computer Integrated Manufacturing Systems
基金
内蒙古自治区重点研发和成果转化计划资助项目(2023YFJM0007)
内蒙古自治区自然科学基金资助项目(2024ZD26)。
关键词
故障识别
不平衡样本
变分模态分解
数据增强
滚动轴承
fault recognition
imbalanced samples
variational modal decomposition
data augmentation
rolling bearings
