Degradation Indicator Construction Using Dual-Class Component Feature Fusion Recalibration for Bearing Performance Evaluation

Degradation indicator (DI) can represent the degradation trend and status of machines, providing decision support in the operational maintenance of machines. However, the accuracy of evaluation is generally influenced by some factors, such as data imbalance and strong noise. Hence, a novel DI is constructed for evaluating the performance degradation of bearing using dual-class component feature fusion recalibration (CFFR). First, a high-dimensional original feature space is constructed from a vibration signal. Subsequently, the feature space is decomposed to construct the trend component space and detrending reconstruction (DR) component space. Then, the two component spaces are fused using trend-CFFR and DR-CFFR. Thereafter, the optimal DI is constructed using the trend component fusion feature and DR component fusion feature. Finally, the effectiveness of the proposed DI is verified through two cases. The results indicate that the proposed DI could improve the prediction performance.

I. Introduction

Degradation prediction of typical transmission components is vital in intelligent condition monitoring and health management of mechanical systems. Rolling bearing is a typical transmission component widely used in mechanical systems that work on variable loads and complex environments. It inevitably and slowly degrades, affecting the operating performance of the mechanical systems, such as accuracy, reliability, and even service life [1], [2], [3]. Thus, it is important to identify the health status of rolling bearing periodically and predict its remaining useful life (RUL) to reduce the downtime of mechanical systems and ensure efficient production [4], [5], [6].