Integrated topology optimisation tool for additive manufacturing

Additive Manufacturing (AM) is now in common use for rapid prototyping and, increasingly, for manufacturing permanent metal components. One of AM’s greatest strengths is its ability to produce more complex geometry than traditional manufacturing methods. This can allow greater design freedom for Topology Optimisation (TO) processes – computational methods for optimising distribution of material within a design domain. TO is a powerful engineering design tool, often used alongside AM in aerospace, automotive, defense, and biomedical industries. However, TO is typically not robust to the flaws and limitations of AM. Integration of TO and AM promises to improve performance and reliability of AM-built aerospace components and medical implants. This thesis identified various research gaps relating to integration of TO and AM and investigates the following: ‘Effects of modelling AM-induced material properties during TO’. To address this research gap, this thesis considered the effect of AM-induced stiffness anisotropy on TO in multiple 2D and 3D design cases, of varying complexity. The performance and design of TO solutions where AM-induced stiffness anisotropy was modelled were compared to those where isotropy was assumed. Mesh-dependence studies were performed for each design case. Two objective functions were considered: ‘stiffness-maximisation’ and ‘natural frequency-maximisation’. The TO method used was hard-kill BESO and the AM-built material modelled was SLM-built Ti-6Al-4V. A review of literature data found that this material is transversely isotropic with an anisotropy ratio of 3-8%. Therefore, representative ratios of 1%, 5%, and 10% were considered, over a range of orientations. To accomplish and automate these studies, Fortran programs were created, utilising Abaqus for FEA, which are readily-modifiable for future studies. It was found that modelling 5% anisotropy during ‘stiffness-maximising’ TO could improve some solutions’ performance by up to 4.4% (relative to assuming isotropy). However, in most cases studied, modelling anisotropy slightly decreased performance – an unexpected result of interaction with hard-kill BESO’s parameter-dependent instabilities. Therefore, it is recommended that future TO studies instead use soft-kill BESO, density-based methods, or level-set methods. It was found that modelling AM-induced stiffness anisotropy during ‘natural frequency-maximising’ TO did not consistently improve performance of resulting designs (relative to assuming isotropy). When targeting multiple modes simultaneously, modelling anisotropy benefitted optimisation of some modes at the expense of others. It was hypothesized that this reflected differences in each mode shapes’ sensitivity to stiffness anisotropy. However, upon analysis of the obtained results, this does not appear to be the case. This thesis recommends that similar studies are performed which consider more complex objective functions (e.g. p-norm stress, fatigue) and AM-induced properties (e.g. porosity, surface roughness). These studies should use more robust TO methods such as soft-kill BESO, density-based methods, or level-set methods. Noting the sparsity and variability of post-AM material data in literature, it is also recommended that study of all post-AM properties (and anisotropy thereof) is accelerated and uses greater numbers of repeat samples.