Fault Diagnosis of Multichannel Bearing-Rotor System via Multistructure Collaborative Discriminative Embedding

Rotating machinery with bearing-rotor systems as their basic architecture have been widely used in important fields such as aerospace, rail transportation, and wind power generation. To accurately monitor the operating status of these devices, multiple sensors are usually installed at key cross sections of the system, which gives rise to a new problem of how to fuse multichannel information. Therefore, this article proposes a new framework for multichannel bearing-rotor system fault diagnosis based on multistructure collaborative discriminative embedding (MSCDE). The key of this framework is the feature reduction of the high-dimensional fault data set using the designed MSCDE algorithm, which can synthesize the information collected by multiple sensors and enhance the discriminative performance of the classifier. Specifically, the multichannel signals of the bearing-rotor system are first acquired; then the multichannel signals are feature extracted to construct a high-dimensional fault dataset, and the MSCDE is used for feature dimensionality reduction, and finally different classifiers are used for fault mode identification. The experimental results of multichannel bearing-rotor systems verify the superiority of the proposed method. Compared with the popular fault diagnosis methods, the MSCDE-based fault diagnosis framework for multichannel bearing-rotor systems developed in this article has better interpretability and trustworthiness.