Adaptive vibration-based bearing fault diagnosis for mud pumps using MSESSA-optimized variational mode decomposition

Mud pumps operate under harsh conditions, where strong background noise, load fluctuations, and complex excitations result in highly non-stationary vibration signals, posing significant challenges for reliable bearing fault diagnosis. To address these challenges, this study proposes an adaptive vibration-based fault diagnosis framework for mud pump bearings, utilizing variational mode decomposition optimized by a multi-strategy enhanced sparrow search algorithm (MSESSA-VMD). The proposed MSESSA incorporates sine–cosine perturbation, Cauchy mutation, and adaptive weight updating mechanisms to enhance global exploration and convergence stability. In contrast to conventional SSA-based optimization approaches, this strategy enables automatic and robust optimization of key VMD parameters, including the number of decomposition modes and the penalty factor, thereby improving decomposition quality under complex operating conditions. Fault-relevant intrinsic mode functions (IMFs) are subsequently selected based on energy-based criteria for multi-dimensional feature extraction. Intelligent fault classification is then performed using a Light Gradient Boosting Machine (LightGBM) classifier. The effectiveness of the proposed framework is first verified using the benchmark Case Western Reserve University (CWRU) bearing dataset and further validated on simulated mud pump bearing vibration signals to assess robustness under industrial-like operating conditions. Experimental results demonstrate that the proposed method achieves an average diagnostic accuracy of 96.14 %, outperforming conventional VMD-based and SSA-VMD approaches in terms of accuracy and robustness against noise and signal non-stationarity. Overall, this study presents a novel framework that integrates a multi-strategy enhanced sparrow search mechanism with adaptive VMD parameter optimization for mud pump bearing fault diagnosis, providing a robust and generalizable solution for vibration-based machinery health monitoring in complex industrial environments.