| Abstract Scope |
Instrumented indentation is widely used for evaluating elastoplastic material properties, especially when testing standard coupon samples is challenging. Despite significant progress made in recent years, a robust and accurate method is still missing for extracting the full stress-strain behavior from indentation test data. Multi-fidelity deep-learning algorithms are trained to obtain elastoplastic properties of metals and alloys from instrumented indentation, using simulation and experimental datasets with different levels of fidelity/accuracy. The accuracy and advantages of the unique approach have been validated for different commercial alloys, including six additively manufactured titanium alloys.
Characterizing internal structures and defects in materials is another challenging task. A general method based on physics-informed neural networks (PINNs) is developed for obtaining unknown geometric and material parameters, with solid mechanics equations built in and boundary conditions parameterized within a meshless framework. The method is validated for materials with internal voids/inclusions using constitutive models including linear elasticity, hyperelasticity, and plasticity. |