Geometry generation challenges for modelling and analysis of micro-structured materials

Abstract

Engineers evaluating the performance of a component at the design stage will typically convert Computer Aided Design (CAD) geometry into a Finite Element model, and run a Finite Element Analysis (FEA) to determine deformations and stress levels as a result of applied loads or displacements. The analysis results would then be interpreted by comparing them with the required duty of the component. For metallic components, homogeneous and isotropic material properties are generally assumed - "macro-scale" modelling. For components to be manufactured from composite materials, models may represent heterogeneity at the ply level, and orthotropic material properties applied with appropriate directionality. This ply-level modelling is often termed "meso-scale" modelling. Engineering interpretation of failure in materials is often based on empirical understanding of experimental data. This approach is generally robust: safety critical components would always be subject to validation by means of a suitable programme of testing. The aspect that is missing is the opportunity to improve understanding of the material performance by investigating the material performance at the "micro-scale". This paper describes computational algorithms for generating random geometries exhibiting similar characteristics to those seen at the "micro-scale" in real materials, and the use of these to predict the influence of the "micro-scale" structure on the "macro-scale" material performance.