An Adaptive-Backstepping Digital Twin-Based Approach for Bearing Crack Size Identification Using Acoustic Emission Signals

Abstract

Bearings are used to reduce inertia in numerous utilizations. Lately, anomaly detection and identification in the bearing using acoustic emission signals has received attention. In this work, the combination of the machine learning and adaptive-backstepping digital twin approach is recommended for bearing anomaly size identification. The proposed adaptive-backstepping digital twin has two main ingredients. First, the acoustic emission signal in healthy conditions is modeled using the fuzzy Gaussian process regression procedure. After that, the acoustic emission signals in unknown conditions are observed using the adaptive-backstepping approach. Furthermore, the combination of adaptive-backstepping digital twin and support vector machine is proposed for the decision-making portion. The Ulsan Industrial Artificial Intelligence (UIAI) Lab dataset is used to test the effectiveness of the proposed scheme. The result shows the accuracy of the fault diagnosis by the proposed adaptive-backstepping digital twin approach is 96.85%.