A General Probabilistic Framework Combining Experiments and Simulations to Identify the Small Crack Driving Force

Identifying the small fatigue crack (SFC) driving force of polycrystalline engineering alloys is instrumental in correlating the inherent microstructure variability and the scatter exhibited by SFC during the early stage of propagation. By utilizing synchrotron images of a SFC propagating through a beta-metastable titanium alloy a general framework to identify the SFC driving force is presented. FFT-based crystal plasticity simulations are then used to computed micromechanical quantities not available from the experiment. The experimental and simulation results are consolidated into a multimodal dataset which is sampled using physically based non- local data mining techniques. Sampled data are analyzed via a machine learning Bayesian Network framework to identify statistically relevant correlations between micromechanical fields and the SFC propagation direction and rate. Statistically relevant correlations are further analyzed and critical variables are selected to formulate a data driven SFC driving force. The predictive capabilities of the identified SFC driving force are evaluated by comparing experimental data and simulations. Furthermore, a comparison between the proposed SFC driving force and the ones available in the literature is also presented. Results show the stronger quantitative behavior of the identified SFC driving force compared to most commonly used in literatur