基于CEEMD和迁移学习的滚动轴承故障诊断研究

Research on Rolling Bearing Fault Diagnosis Based on CEEMD and Transfer Learning

【目的】在实际生产环境中,由于机器特征复杂和工况变化,智能诊断模型在跨机组迁移时需要重复训练,这不仅增加了时间成本,还加大了算力资源的消耗。为了解决这些问题,需要开发出一种能适应复杂机器特征并在不同工况下保持高准确度的轴承故障诊断方法,同时,减少模型迁移时所需的重复训练,以便实现更高效的故障识别和预测。【方法】研究提出基于互补集合经验模态分解(Complementary Ensemble Empirical Mode Decomposition,CEEMD)和迁移学习的滚动轴承故障诊断方法。首先,采用CEEMD法对原始信号进行分解,并计算出对应分量的峭度值。其次,采用多核最大均值差异法对源域数据与目标域数据进行域适应处理。最后,在凯斯西储大学轴承数据集和美国机械故障预防技术学会轴承数据集之间进行迁移故障诊断试验及对比分析。【结果】研究结果表明,与直接迁移模型算法相比,基于CEEMD改进的迁移学习网络在不同数据集上的迁移效果更好,其故障诊断的准确率最高。【结论】经试验验证,研究所提的方法表现出良好的变工况跨机组适配能力,具有较高的故障诊断精度,为研究复杂工况下多机组相似故障诊断场景提供了非常有价值的参考。

[Purposes]In the actual production environment,due to the complexity of machine features and the change of working conditions,the intelligent diagnosis model needs repeated training when migrating across units,which not only increases the time cost,but also increases the consumption of computing resources.In order to solve these problems,it is necessary to develop a bearing fault diagnosis method that can adapt to complex machine features and maintain high accuracy under different working conditions.At the same time,the repeated training required for model migration is reduced to achieve more efficient fault identification and prediction.[Methods]The study proposes a rolling bearing fault diagnosis method based on CEEMD and transfer learning.First,the CEEMD decomposition method is used to decompose the original signal and the kurtosis value of the corresponding component is calculated.Then,the multi-core maximum mean difference method is used for the source domain data and the target domain data.Domain adaptation processing,and finally a migration fault diagnosis test and comparative analysis between the Case Western Reserve University dataset and the American Society for Mechanical Failure Prevention Technology dataset.[Findings]The research results show that compared with the existing direct transfer model algorithm,the improved transfer learning network based on CEEMD has a better transfer effect on different data sets,and its fault diagnosis accuracy is the highest.[Conclusions]It is verified by experiments that the method proposed in the study shows good cross-unit adaptability under variable working conditions,and has high fault diagnosis accuracy,which provides a valuable reference for studying similar fault diagnosis scenarios of multiple units under complex working conditions.