• condition monitoring
• failure analysis
• micropitting
• rolling bearings

This dissertation was carried out at the Life Testing Department of the SKF Engineering & Research Centre located in Nieuwegein (NL) and deals with roller bearings testing. The aims of the work are to study some early signs of micropitting damage on the rolling surfaces and to determine whether they can be detected by a condition monitoring system.
First of all, the literature about rolling contact fatigue was reviewed, with a focus on micropitting damage. It was found that micropitted areas usually develop around geometric features like furrows, scratches or indentations. A previously tested roller bearing showing this kind of damage was selected and a failure analysis was done with the aid of microscopy. The characteristics of the micropitted areas were investigated, in terms of appearance and size. Moreover, two basic engineering models were implemented to evaluate the risk of micropitting and their applicability to bearings testing was investigated. An experimental activity was carried out to monitor the development of a micropitted spot, both in terms of size and effects on the vibration signal. The micropitted bearing was run under conventional operating conditions and periodically dismounted to see how the damage was proceeding. Its vibration signal was acquired with the condition monitoring system currently in use in the Life Testing Department and also with a pilot system developed in parallel with this dissertation, both of them implementing the enveloping technique. Enveloped vibration data were then processed looking for any evidence of this early damage. Finally, signals from more than 30 undamaged bearings were acquired to determine a baseline behavior to be compared with the micropitted bearing.
On the basis of the results of this research, a growth rate for the micropitted spots was estimated and an initiation and development mechanism proposed. Both basic micropitting models were found representative of a widespread damage distribution, whereas local micropitting around geometric features can be simulated only with advanced models taking into account their exact shape. Finally, it was found that a time domain analysis of the enveloped vibration signal is not suitable for the detection of early signs of micropitting, mainly because those thought to be evidences of the damage were found in undamaged bearings as well.