Validation of a physics-based model for rolling element bearings with diagnosis purposes

The use of rolling element bearings is widely extended to many fields such as wind energy systems, transportation and machine tools, among others. This broad use makes their performance analysis an interesting field of research. There are techniques to determine the life of a bearing and the on-going failure, if any, under some assumptions with some values of reliability. However, the unfulfilment of those hypothesis or other effects that affect the standard operation of rolling element bearings (e.g. current leakage, overloading, corrosion, etc.) leads to a higher probability of the appearance of failure. The monitoring of the condition of rolling element bearings has two main goals, the diagnosis and the prognosis of the item. Indeed, diagnosis, i.e. damage detection, localization and identification, has a great interest on the knowledge of the state of rolling element bearings in order to prevent faulty situations that may cause risky or costly situations, identifying those adverse situations and trying to mitigate the undesired effects. Therefore, risky situations due to failures need additional knowledge about the dynamics of a system (rolling element bearings in this case) and physics-based models can be used in order to represent it. They have an interesting potential due to the fact that they are able to simulate situations that may arise in some damaged conditions that might be either difficult, costly or insecure to reproduce in a real system. However, there is a need to validate the physics-based models to assure that it follows the real response of the system. This work presents the validation process of a model already developed by the authors. Experimental tests have been done in a test rig and the vibration measurements taken from these tests have been used to validate the model. Damage on the surface of the outer race has been induced to one of the rolling element bearings of the test rig. Thus, frequency-domain and order-domain analysis have been performed and the experimental results have been compared to the results obtained from the simulations. Differences lower than 2.5 % have been found for a wide range of constant and variable speeds and, hence, the model is validated.