流形结构化半监督扩展字典学习的旋转设备故障诊断

Manifold structured semi-supervised extended dictionary learning for rotating machinery fault diagnosis

针对有标签数据不足及传统故障诊断模型判别性差的问题,本文提出一种流形结构化半监督扩展字典学习(MS-SSEDL)的故障诊断方法.首先,为改善缺少有标签数据而导致模型的识别性能较差问题,在MS-SSEDL模型中提出无标签数据重构误差项,利用无标签数据学习置信度矩阵,从而学习得到扩展字典以增强字典学习的表示性.然后,为增强MS-SSEDL模型的判别性,通过保存数据的流形结构,学习数据中内在几何信息的稀疏表示,增强信号表示能力及字典判别性.最后,在数字图像、轴承故障及齿轮故障公共数据集的实验表明所提MS-SSEDL方法比其他先进方法的识别性能更优越.

Aiming at the problems of insufficient labeled data and poor discriminability of traditional fault diagnosis models,a manifold structured semi-supervised extended dictionary learning for rotating machinery fault diagnosis(MSSSEDL)is proposed.First,in order to improve the problem of poor recognition performance due to the lack of labeled data,an unlabeled data reconstruction error term is proposed in the MS-SSEDL model,and the unlabeled data is used to learn the confidence matrix,so as to learn the extended dictionary to enhance the representation of dictionary learning.Then,in order to enhance the discriminativeness of the MS-SSEDL model,by preserving the manifold structure of the data,a sparse representation of the intrinsic geometric information in the data is learned,and the signal representation ability and dictionary discrimination ability are enhanced.Finally,experiments on public datasets of digital images,bearing faults and gear faults show that the proposed MS-SSEDL method outperforms other state-of-the-art methods.