Fault Detection in Rotating Machinery Based on Machine Learning

This chapter delves into the critical role of rotating machinery in industrial applications and the importance of robust monitoring protocols for gearbox components. It highlights the advancements in statistical process control (SPC) methodologies for gear system fault detection, including the use of EWMA and MEWMA charts. The study proposes a systematic methodology for automatic fault diagnosis using vibration data analysis, leveraging advanced signal processing algorithms and machine learning techniques. The proposed machine learning paradigm involves Principal Component Analysis for dimensional compression, Gaussian Mixture Modeling for pattern extraction, and Statistical Process Control with novelty detection for anomaly identification. The simulation results demonstrate the effectiveness of this approach in detecting progressive gearbox degradation states, from pristine to critical conditions. The study concludes with recommendations for future research, emphasizing the need for validation against experimental data and the development of adaptive thresholding algorithms to enhance diagnostic reliability and early fault detection sensitivity.