Data Fusion and Modeling for Fatigue Crack Growth Prediction

A methodology is proposed to reduce the effects of uncertainty in fatigue crack growth investigations, especially for very high cycle fatigue. The approach integrates experimental data with modeling in order to manage the uncertainty with minimal amounts of data. An extensive set of very high cycle fatigue data collected on SUJ2 bearing strength steel will be used to demonstrate the procedure. The fatigue cracks nucleate from internal particles as well as surface damage, both of which can have fatigue lives in excess of 10⁸ cycles. Consequently, it would be advantageous to have a methodology that would predict long term fatigue life with multiple modes of damage growth by infusing limited data with fatigue modeling.