Bearings are critical components in most machines that suffer from fatigue damage, and hence need to be monitored. The development of spalls in the bearing races is the prevalent failure mode. In this research, we propose a new diagnostic method for ball bearings monitoring that is based on strain measurement. A standard Fiber Bragg Grating (FBG) sensor has been used for the strain measurement. A physics-based theoretical analysis is used to establish insight and analysis of the strain signals generated by both healthy and faulty bearings. The analysis is based on two complementary models that were integrated into a strain model: a quasi-static Finite element (FE) and a solid bodies dynamic model. The integrated strain model enables simulation of a wide range of spall lengths and describes the behavior of the strain signals as a result of various phenomena. The strain model was validated with measured data. A new signal processing scheme of strain signals for the diagnostics of bearings was established. This scheme enabled the development of a novel automatic algorithm for spall length estimation, which makes it possible, for the first time, to track spall length propagation in bearings outer race, based on strain signals. The algorithm was developed and validated based on seeded fault experiments, wherein different sizes of faults on the outer race at different rotational speeds were examined. The trend of the spall length during a bearing endurance test was generated and verified. The insights from the strain model, combined with the experimental observations, enable generalization of the algorithm and definition of its capabilities and limitations.