TOPOLOGY OPTIMIZATION FOR ADDITIVE MANUFACTURING

Topology optimization methods solve a material distribution problem to generate an optimal topology. It is usual for each finite element within the design domain to be defined as a design variable, allowing a variation in density (homogenization, SIMP)  or void-solid (bidirectional evolutionary structural optimization (BESO)) . Other methods exist such as genetic algorithms and level set methods but these are still in their infancy with regards to their suitability to real life problems and so are not discussed here. Usually, topology optimization methods are used to tackle practical design problems with traditional manufacturing processes in mind, such as casting and machining. Processes where the part is produced by material removal can be described as subtractive processes and processes where the part is produced by a mold can be described as formative processes. These approaches have significant manufacturing constraints that must be taken into account during the design stage to ensure a feasible design. For example, the need for tool access in the case of machining or the need for part removal from a mold in the case of casting or molding. These constraints limit the physical realization of the optimal topology and a compromise has to be made between optimality and ease of manufacture. Typically these constraints are either included in the actual optimization by limiting the topology to feasible designs, or by subsequent simplification of the unconstrained optimization. The former of these is usually preferable, but not all constraints can be included easily in the optimization process.

Additive manufacturing (AM) contrasts to the two aforementioned process classifications in that the part is built up layer-by-layer. AM is a development from rapid prototyping (RP) and aims to produce end-use parts rather than prototypes. To this end, significant efforts have been made in recent years to process metals in addition to polymers, and there are now several 348 commercial metal processes able to produce end-use parts. Like RP, AM usually requires a 3D computer-aided design (CAD) model of the part. This is sliced in a single direction into many very thin slices (cross section profiles). These cross section perimeters are traced either by a laser, electron beam, extrusion nozzle or jetting nozzle and the area contained by the perimeters filled with a hatching pattern. Once a layer has been deposited/melted/cured, the next layer is added. This is repeated until the whole part has been generated.

Due to this layer manufacturing approach, parts of significantly greater complexity can be produced compared with traditional processes and this increased complexity generally does not have a significant effect on the cost of the process. This provides the designer with significantly greater design freedom and enables the built part to be closer to the optimum design than is possible with traditional processes. This paper discusses the application of topology optimization to parts designed for AM, highlighting the main practical difficulties and opportunities for optimization. This work is part of an industrially focused project called Atkins which is investigating carbon reduction through the use of AM and component optimization to reduce weight [