On the potential of bearing strain signals for defect severity estimation: modeling and experiments

The detection of bearing defects by means of condition monitoring using acceleration signals has become a commodity. However, robust in-situ bearing defect severity estimation (e.g. spall size) using acceleration signals has been proven difficult in field conditions. To overcome the limitations of acceleration-based methods relying on an indirect response, local strain measurements more directly related to the rolling element contact forces are investigated. In this work is assessed if the strain signal measured on the ring of a bearing contains features related to the defect that are largely independent of speed, and transfer paths. Dynamic modelling of the strain measurement response to over-rolling a bearing defect and experiments on a dedicated setup for testing bearings with artificial spalls reveal that the defect features in the strain signals are clearly observable and independent of running speed. It is shown that a single strain recording could be sufficient to robustly quantify the spall size.