Abstract | Keywords | Export | Availability | Bookmark

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

In the industry, an unplanned failure can lead to a complete shutdown of the production, which results in non negligible losses. Predictive maintenance aims assessing the machine condition for an early detection of component failures and an optimal planning of overhauls. Vibration analysis is the most used condition monitoring technology in that purpose. The accuracy and the reliability of the diagnostic strongly depend on the way the vibratory signal is processed, from the data acquisition until its analysis. This Master's Thesis is dedicated to these signal processing methods. 

The first part of this work consists into a state of the art of existing processing methods and sets the theoretical background. In this last purpose, Fourier analysis, statistical approach and time-frequency analysis, are covered. 

In the second part of this work a development path is explored with the presentation of two methods, based on the kurtosis, aiming to enhance the bearings faults diagnostic. One is focused on the time waveform while the other is applied in frequency domain. This work is not limited to the theoretical approach and insists on the discussion about the applicability of the method collected in the industry.

Ingénierie, informatique & technologie > Ingénierie mécanique