An Adaptive Multisignal Framework for Real-Time Fault Diagnosis of Rolling Bearings

The online condition monitoring of rolling bearings is crucial for the prognosis and health management of high-speed equipment. Using traditional deep learning (DL) models, multisensor data are not adapted to train timely online because of the time-consuming of the models. An online health monitoring framework for roll bearings using correlative signals is proposed to address this issue. First, in the offline phase, a copula function-based analysis model is developed to determine the light signals as auxiliary prediction signals for the vibration signal. Second, based on the adaptive representation of the selected signal, the training parameters are dynamically estimated to fulfill individualized learning by adopting a combination of the sample complexity and real-time prediction errors so as to fulfill individualized training, during the online phase. An aggregation learning framework is also, furthermore, presented in the online phase on a cloud-computing platform to determine the optimization target of the model to make the online update adaptive enough to the online diagnosis task. Finally, the proposed framework is verified by multisensor data, including vibration and force signals. Our framework can capture and adapt new patterns in stream data, and the training accuracies approach 99.98%, the training times are less than 5 s, respectively.