When 10:30 AM - 11:30 AM Jan 15, 2016
Where 1670 Beyster Building
Add event to calendar vCal

Improving Understanding and Prediction of Microstructurally Small Crack Growth by Coupling Experimental Observations, Numerical Simulations, and Machine Learning

Ashley Spear
Department of Mechanical Engineering, University of Utah

Advancing the state of materials design and structural prognosis will inevitably require an improved understanding of the mechanisms that govern 3D evolution of microstructurally small fatigue cracks (MSFCs). This talk will first describe recent efforts to couple synchrotron-based measurements of microstructurally small fatigue cracks with multiscale modeling (mesoscale to macroscale) using 3D finite-element simulations. Ex-situ techniques were employed to characterize fatigue-crack propagation within the microstructure of an aluminum alloy. The experimental characterization involved X-ray tomography along with near-field high-energy X-ray diffraction microscopy, which provided a 3D grain map adjacent to fatigue-crack surfaces. The experimental data were then used to digitally reconstruct the measured polycrystalline volume and fatigue-crack morphologies as a way to reproduce, within a computational environment, the observed crack evolution. The numerical reproduction serves as a way to quantify micromechanical fields in local neighborhoods along the observed crack fronts. In the 3D finite-element model, the discontinuity of the observed 3D crack surface is represented explicitly. The second part of the talk will describe a parallel effort to develop a machine-learning framework to post-process and learn from the vast amount of data obtained from both experiment and simulation. It is anticipated that the machine-learning framework will enable a better understanding of the quantitative relationships among local, microstructure-sensitive fields and the rate of 3D crack-shape evolution. The long-term goal will be to utilize the framework to improve predictive capabilities for MSFC evolution.

More information about this event…