RUL Estimation of Rolling Element Bearings Using a Hybrid Wavelet Packet Decomposition–Recursive Feature Elimination–Adaptive Neuro Fuzzy Inference System Framework

Rolling element bearings are critical components in rotating machinery, and their unexpected failures can cause severe downtime and economic losses. Therefore, accurate estimation of remaining useful life (RUL) is essential to ensure system reliability and enable predictive maintenance strategies. This paper presents a novel hybrid framework that integrates Wavelet Packet Decomposition (WPD), Recursive Feature Elimination (RFE), and Adaptive Neuro-Fuzzy Inference System (ANFIS) for intelligent RUL estimation of bearings. First, vibration signals from the well-known IMS dataset are acquired and decomposed using WPD to capture multi-resolution information. A comprehensive set of health indicators is then computed from each decomposition level, reflecting the degradation dynamics of bearings. To reduce redundancy and enhance discriminative power, the most relevant features are selected using the RFE algorithm. Finally, the refined features are fed into an ANFIS model to estimate the RUL. Comparative analyses with multiple Artificial Neural Network (ANN) based models are conducted to assess the effectiveness of theproposed approach. Experimental results demonstrate that the hybrid WPD–RFE–ANFIS framework achieves outstanding predictive performance, reaching an accuracy of 99.98%, thereby outperforming traditional ANN architectures. This study highlights the potential of hybrid intelligent models for advancing prognostics and health management (PHM) in industrial applications.