A generative approach to a virtual material testing laboratory

Abstract

<p> This thesis presents a virtual material testing laboratory that is highly generic and flexible in terms of both the material behaviour and experiments that it supports. Generic and flexible material behaviour was accomplished via symbolic computation, generative programming techniques and an abstraction layer that effectively hides the material model specific portions of the numerical algorithms. To specify a given member of the family of material models a domain specific language (DSL) was created. A compiler, which uses the Maple computer algebra system, transforms the DSL into an abstract material class. Three different numerical algorithms, including a return map algorithm, are presented in the thesis to illustrate the advantage of the abstract material model. To accomplish the goal of generic and flexible experiments the finite element method was employed and an API that supports both load and displacement controlled experiments, as well as the capability for the experiments to modify their state over time, was developed. The virtual laboratory provides a family of material models with the following behaviours: elastic, viscous, shear-thinning, shear-thickening, strain hardening, viscoelastic, viscoplastic and plastic. As well, the developed framework, by using the Ruby programming language, provides support for a wide variety of programmable experiments, including: uniaxial, biaxial, multiaxial extension and compression, shear and triaxial. </p>