Condition monitoring of machines in non-stationary operations can be considered as one of the major challenges for the research in the field of rotating machinery diagnostics. The applications are ubiquitous: transport systems, energy systems, vehicles, production plants. On these grounds, the present work is devoted to measurement techniques and signal processing methods for the condition monitoring of bearings undergoing non-stationary operation conditions. Several types of experimental set

more »

... s are included in the present study and the advantages and drawbacks of each are discussed. For example, on one side an ad-hoc test rig for precision measurements is developed and utilized; on the other side, real scale measurement campaigns at operating non-stationary energy conversion systems (wind turbines) are performed. A special focus on energy systems is important because often in this kind of devices fault detection becomes much more challenging due to the interplay with electromechanical couplings. To face this drawback, the most advanced post-processing techniques need to be used. The application in the real field constitute an important part of this study because the fault diagnosis, and especially its interpretation, are much more challenging with respect to controlled laboratory conditions. The collected measurements are analysed through the most appropriate post-processing techniques employed for non-stationary signals. In the time domain, the statistical features of the signals are addressed through novelty indexes and principal component analysis. The results support that the Mahalanobis distance is an effective index in order to monitor the level of severity of the fault on the actual machine operation condition. Abstract Condition monitoring of machines in non-stationary operations can be considered as one of the major challenges for the research in the field of rotating machinery diagnostics. The applications are ubiquitous: transport systems, energy systems, vehicles, production plants. On these grounds, the present work is devoted to measurement techniques and signal processing methods for the condition monitoring of bearings undergoing non-stationary operation conditions. Several types of experimental set ups are included in the present study and the advantages and drawbacks of each are discussed. For example, on one side an ad-hoc test rig for precision measurements is developed and utilized; on the other side, real scale measurement campaigns at operating non-stationary energy conversion systems (wind turbines) are performed. A special focus on energy systems is important because often in this kind of devices fault detection becomes much more challenging due to the interplay with electromechanical couplings. To face this drawback, the most advanced post-processing techniques need to be used. The application in the real field constitute an important part of this study because the fault diagnosis, and especially its interpretation, are much more challenging with respect to controlled laboratory conditions. The collected measurements are analysed through the most appropriate post-processing techniques employed for non-stationary signals. In the time domain, the statistical features of the signals are addressed through novelty indexes and principal component analysis. The results support that the Mahalanobis distance is an effective index in order to monitor the level of severity of the fault on the actual machine operation condition.