Systems far from equilibrium respond to probes in a history-dependent manner. The prediction of the system response depends on either knowing the details of that history or being able to characterize all the current system properties. In crystal plasticity, various processing routes contribute to a history dependence that may manifest itself through complex microstructural deformation features with large strain gradients. However, the complete spatial strain correlations may provide further predictive information. In this paper, we demonstrate an explicit example where spatial strain correlations can be used in a statistical manner to infer and classify prior deformation history at various strain levels. The statistical inference is provided by machine-learning techniques. As source data, we consider uniaxially compressed crystalline thin films generated by two dimensional discrete dislocation plasticity simulations, after prior compression at various levels. Crystalline thin films at the nanoscale demonstrate yield-strength size effects with very noisy mechanical responses that produce a serious challenge to learning techniques. We discuss the influence of size effects and structural uncertainty to the ability of our approach to distinguish different plasticity regimes.

Experimental and theoretical solutions have shown that imperfections in wide-flanged structural columns may reduce the failure load of the column by as much as 30% with respect to that of a perfect column. Therefore, the early detection and prevention of such imperfections, which would likely reduce the load capacity of a structure, are critical for avoiding catastrophic failure. In the present article, we show how machine learning may be used to detect imperfection sensitivity in pultruded columns using observable column deformations occurring at loads as low as 30% of the design load. Abaqus simulations were used to capture the behavior of such columns of various lengths under service load. The deformations found from the simulations were used to train the machine learning algorithm. Similar deformations could be easily collected from in-service columns using inexpensive instrumentation. With over 3000 test cases, 95% accuracy in the correct detection of imperfection sensitivity was found. We anticipate that the proposed machine learning pipeline will enhance structural health monitoring, providing timely warning for potentially compromised structures.

Nanoindentation, a common technique for probing the mechanical properties of crystalline materials, exhibits both surface and bulk-dominated responses that are linked to the parent crystal's elasticity and plasticity. For FCC crystals, the nanoindentation response is primarily controlled by the indented-grain's crystalline orientation and dislocation density. However, the indentation properties, such as hardness, have ambiguous significance at the nanoscale due to tip-specific size effects. Through extensive discrete dislocation dynamics simulations of flat-punch nanoindentation on FCC single crystals, we show that a one-to-one correspondence between an indented location and the pre-existing dislocation density is possible with the help of statistical, machine-learning algorithms. We cluster and classify various experimentally plausible ensembles with varying dislocation density and/or crystalline orientation, using nanoindentation force-depth curves or post-indent surface displacement images. We use this classification to statistically predict the nanoindentation response of a given location, through the average response of classified load-displacement curves. In addition, we demonstrate that our simulation results qualitatively agree with experimental data on common FCC single crystals. Ultimately, we propose a novel pathway to uncover the depth dependence of local surface material properties for more complete, non-destructive, microstructural characterization, through nanoindentation.

The present thesis makes a connection between spatially resolved strain correlations and material processing history. Such correlations can be used to infer and classify prior deformation history of a sample at various strain levels with the use of Machine Learning approaches. A simple and concrete example of uniaxially compressed crystalline thin films of various sizes, generated by two-dimensional discrete dislocation plasticity simulations is examined. At the nanoscale, thin films exhibit yield-strength size effects with noisy mechanical responses which create an interesting challenge for the application of Machine Learning techniques. Moreover, this thesis demonstrates the prediction of the average mechanical responses of thin films based on the classified prior deformation history and discusses the possible ramifications for modelling crystal plasticity behavior in extreme settings.