Analysis of bearing failure modes from time-frequency techniques and machine learning on vibration signals

The article addresses the importance of predictive maintenance in the industry, highlighting the need to handle data to improve productivity and prevent unscheduled failures. The study analyzes the performance of different machine learning methods for diagnosing failures in rotational equipment, such as spherical bearings, and compares the impact of different vibration signal processing methodologies on the models' results. Techniques such as statistical feature extraction and Empirical Mode Decomposition (EMD) are used to feed the learning models. Jiagnan University's database is employed to evaluate the performance of models, with a focus on identifying and diagnosing failure modes in vibration bearings. The study highlights the effectiveness of time-frequency processing and machine learning techniques in identifying faults, aiming at more effective asset management and ensuring the continuity of operations with safety and productivity.