Vibration Dynamic Analysis of the Bearing Parameters in Steam Turbine Bearing Systems in Sugar Refinery

Abstract Rolling element bearings are primary components of the rotating machines. Conventional vibration-based condition monitoring methods are widely adopted to maintain the reliability and safety of machine operations. Identifying weak fault features in rolling bearings is challenging due to complex dynamics and random noise. Mathematical modeling and machine learning can synergistically enhance fault diagnosis across various operational scenarios. This article proposes a novel approach for monitoring and diagnosing the bearings clearance fault technique, combining the rotor–bearing system's dynamics model, dimension principle (DP), and Elman neural network (ENN). Using accelerometer data transformed by fast Fourier techniques, our approach eliminates manual feature extraction, leveraging the ENN for fault identification. Vibrational responses from diverse industrial machines are analyzed across varying conditions. Bearing clearance faults are detected with an average error of 2% using the DP model and classified using the ENN. Experiments demonstrate that the proposed method achieves 99.10% accuracy in diagnosing bearing clearance faults across various conditions. Its robust learning capabilities extend to applications in sugar refining-steam turbines, offering swift and precise fault diagnostics.