In a broad sense, the work described herein addresses either extreme of a traditional da/dN vs. dK plot. The first chapter addresses the uppermost limit of such plot, where tearing represents the limit state of structural failure. Subsequent chapters address the lower limit, where emphasis is placed on nucleation and early propagation of microstructurally small fatigue cracks (MSFCs). A common theme throughout the dissertation is the development of new tools and techniques (be they experimental or numerical) to enable unprecedented interrogation of crack evolution in 3D, with applications to various aluminum-alloy structures. Each chapter represents a separate body of work providing novel contributions in one or more areas involving fracture mechanics (viz. structural prognosis, corrosion science and engineering, and materials characterization). A brief overview of each chapter is described next. In Chapter 1, a methodology is described for predicting in real time the residual strength of structures with discrete-source damage. An artificial neural network (ANN) is trained using linear-elastic fracture mechanics (LEFM)-based data from numerical models; the ANN predicts residual strength given a set of damage parameters. Section 1.3 focuses on augmenting the existing LEFMbased modeling toolset to simulate ductile tearing and thereby improve resid- ual strength values used to train an ANN. Validation results are presented for two ductile-tearing simulations. Chapters 2 through 4 focus on MSFC initiation and propagation in an Al-Mg-Si alloy used to line composite-overwrapped pressure vessels. Chapter 2 describes an experimental study regarding the effect of an alkaline-based chemical milling treatment used to dimensionally reduce the Al-Mg-Si pressure-vessel liners. 3-D scanning electron microscopy is employed to quantify surface pitting caused by the chemical-milling treatment. The 3-D surface characteristics, along with high-magnification fractographs, are used to explain the observed 50% reduction in low-cycle fatigue lives among the chemically-milled specimens compared to a control group. In Chapter 3, an experimental methodology based on post-mortem measurements is developed to quantify 3-D rates of propagation and crack-surface crystallography for a naturally nucleated MSFC in an Al-Mg-Si specimen. The measurements are made possible through recent developments in 3-D characterization methods. Findings from the study demonstrate: 1) the complexity and variability of 3-D MSFC evolution in the Al-Mg-Si alloy and 2) the viability of the post-mortem characterization approach for quantifying 3-D MSFC evolution in polycrystalline alloys. The dissertation culminates with Chapter 4, which, for the first time, demonstrates the 3-D digital reconstruction and numerical simulation of a sequence of directly measured MSFCs, where both MSFC geometry and individual grain morphologies and orientations are explicitly represented at the polycrystalline length scale. The numerical reconstruction is demonstrated using 3-D measurements from Chapter 3. Work from Chapters 1 and 2 is published in [1, 2, 3, 4] and [5], respectively. The combined work from Chapters 3 and 4 is described in [6, 7, 8] and is currently in preparation for additional journal publication.