Fault diagnosis of electric drill turntable bearings based on CNCEEMDAN and DE-GWO-LSSVM

This paper introduces a fault diagnosis method for petroleum electric drill turntable bearings, addressing the challenge of low diagnostic accuracy due to difficulties in feature extraction. The proposed approach utilizes the Adaptive Cauchy Noise Ensemble Empirical Mode Decomposition with Adaptive Noise (CNCEEMDAN) and the Differential Evolution Grey Wolf Optimization Algorithm (DE-GWO) to optimize hyperparameters for the Least Squares Support Vector Machine (LSSVM). In this paper, CNCEEMDAN is an improved signal decomposition method designed for processing vibration signals. This improves the efficiency of extracting both time and frequency-domain features. The feature datasets are then constructed based on the energy entropy criterion. LSSVM serves as the core algorithm of the classification model, while DE-GWO is a hybrid optimization strategy used to search for the optimal combination of LSSVM hyperparameters. The optimized LSSVM is subsequently used for fault diagnosis on the feature data sets. Comparative experiments demonstrate the method’s excellent performance under actual operating conditions, achieving a fault diagnosis accuracy of 97.33%. The method enables early fault detection and enhances equipment reliability in the oil industry, providing a dependable solution for diagnosing faults in electric drill turntable bearings.