Conference Agenda

Failures in industrial machinery represent a significant risk, as they can lead to production losses, reduced competitiveness, and accidents. In this context, vibration measurement has become the primary non-invasive method for detecting and predicting faults in rotating machinery, since failures are typically manifested through progressive increases in vibration levels. However, the effectiveness of this technique largely depends on expert interpretation, which limits its automation and large-scale application.

To overcome this limitation, this study proposes the development of a fault diagnosis model based on machine learning techniques using vibration signals to automatically and accurately classify the condition of a gearbox. Vibration data from the Open Energy Data Initiative (OEDI) repository corresponding to a wind turbine gearbox were used in this work. The methodology included signal preprocessing, feature extraction in the time and frequency domains, and the training of multiple supervised classification models. The results demonstrated that the Support Vector Machine model achieved the best performance, providing a promising solution for optimizing predictive maintenance strategies and preventing failures.