An inverse derivation framework for modelling extreme dynamic behaviour of materials

Material models are used in finite element (FE) solvers to describe the response of materials subjected to various types of loading. These models are typically phenomenological or semi-empirical, with several empirical constants that are determined based on fitting to mechanical characterisation test data, e.g., quasi-static tension, dynamic compression, etc. An alternative approach to derive material model constants, known as inverse modelling, uses an optimisation framework to adjust model constants in FE simulations of the characterisation experiments. This can enable model constants to be derived with fewer experiments than the conventional approach. Previous work has also demonstrated that an inverse modelling approach can be used to derive material model constants through the simulation of “functional” experiments, i.e., experiments that had utility beyond just material characterisation (e.g., ballistic penetration, near field blast, etc.). Although only demonstrated for visco-plastic behaviour, this approach shows promise.;;This project will investigate the utility of inverse modelling for viscoplastic and damage segments of a material model, primarily for application with metals (e.g., aluminium alloy, high-strength steel, etc.). The focus will be on using a range of “functional” experiments that have utility beyond purely mechanical material characterisation and induce a wide range of stress states, strain rates, and failure modes on the material of interest. Such an approach would provide an improvement over the conventional methodology by (i) minimising the number (and therefore cost) of mechanical characterisation experiments, (ii) providing inherent validation of the derived material parameters over the range of loading conditions encompassed by the applied experiments, and (iii) potentially improve the accuracy of FE simulations using the derived material model constants.;;During conduct of the project students will be expected to perform a mixture of experimental tests (quasi static and high strain rate), numerical simulations (explicit FEM) and machine learning/AI modelling activities.