Detection de l'usure de roulements par emission acoustique

The work presented in this thesis focuses on the detection of bearing defects by using acoustic emission. The main objective is to detect the fault at an early stage of degradation at a low severity level never reached. On the other hand, efficient signal processing tools have been developed for good detection of the frequencies associated with the default and these harmonics. The thesis is organized by articles. The work is divided over two parts who have been the subject of five journal papers.;The first part of this thesis is focused on the ability of acoustic emission to detect a fault in its early stages of degradation. A comparative study between the acoustic emission and vibration measurements is exposed. Two different approaches are used: • The first approach uses a simple signal processing tools, easy to operate and that are most commonly used in the industry. Exploiting the frequency content of the signal remains the most used tool in the diagnosis of rotating machinery. The quality of the representation is based on the signal processing tool used for this purpose. The best choice is made on the basis of having the best presentation of the spectrum, where the frequencies of the default are clearly identified in terms of amplitude and number of the associated harmonics. A first study was conducted at the ETS (Ecole de Technologie Superieure) on a simple test rig composed by a shaft on which two ball bearings are installed and connected to a motor. The second study was conducted in the laboratories of Pratt & Whitney Canada (the industrial partner). The tests were performed on an aircraft engine PW120. • The second approach is based on a much more sophisticated signal processing tool that focuses on the characterization of the cyclostationary appearance in acoustic emission signals.;In the second part of my thesis, the study is focused on the development of new signal processing tools for better detection of bearing faults. Three hybrid methods have been developed mainly based on the EMD technique (Empirical Mode Decomposition) for the diagnosis of rolling defects, but also on minimum entropy methods, empirical wavelet transform and Teager Kaiser Energy Operator (TKEO). All these methods were validated through simulations and experimental data from acoustic emission signals and vibration measurements.;These studies support consistently the utility of acoustic emission and its ability to detect a bearing fault at an early stage of degradation. Our study allows for deciding between the scientific communities to the fact that the acoustic emission is rather better than the vibration measurements for early detection. The defect used is about 40microm that, to our knowledge, is the smallest default never studied to date by the scientific community.;A comparative study of the proposed methods is exposed with my main recommendations for future researches. The empirical wavelet transform is appeared as the method having the greatest potential for early detection of defects.;Keywords: Bearing defects, early detection, acoustic emission, vibration measurements, signal processing.