Developement of an optimization method for self-supporting lattice structure design under uncertainties arising from additive manufacturing

The lightweight and high strength characteristics of lattice structures, coupled with their easily modellable topological features, are increasingly becoming preferred choices in engineering applications. Due to their complex geometries, additive manufacturing is preferred for the fabrication of strut-based lattice structures. These structures are primarily composed of strut elements. Considering the current capabilities of additive manufacturing, variations and uncertainties at the micro and millimeter levels often arise during layer-by-layer production of strut elements typically designed at the millimeter scale. These changes cause a difference in mechanical properties between the designed and manufactured structure. The application of optimizations aimed at enhancing the strength of lattice structures is feasible through additive manufacturing, resulting in the formation of complex geometries in optimization outcomes. Optimized designs may necessitate the use of support structures due to the resulting topology optimization. However, generating these support structures without affecting existing strut elements or removing them from the complex topology post-production without causing damage to the structure is not straightforward. In this study, a two-step lattice optimization procedure is proposed as a design approach to account for uncertainties arising from the additive manufacturing of lattice structures, impacting their geometry and material properties. This approach allows to produce self-supporting optimization results without the need for additional support structures during the design process. For material extrusion, structure-property relationships were established using artificial neural networks between the parameters governing the modelling of strut elements, including diameter and angle variations, and the homogenized properties characterized in a previous study under uncertainties arising from the additive manufacturing by material extrusion. The artificial neural network model has been integrated into the lattice optimization process, which involves modelling with strut elements. Initially, an algorithm was created in MATLAB to convert a meshed model to lattice cells. Using the topological information obtained, a two-step optimization algorithm is initiated. The first step of the two-step optimization is the classical layout optimization where strut elements with diameters close to zero are removed from the topology. The second size optimization is performed to determine the optimized diameters of the strut elements in the topology with the specified minimum manufacturing constraint. Between the two optimization stages, a self-support structure algorithm has been developed to identify strut elements in the optimized topology requiring support. This algorithm adds supports by incorporating strut elements into the structure before topology optimization, enabling self-supporting structure fabrication. Additionally, within the scope of this study, a surface generation algorithm has been developed to create a manufacturable STL model representing the final optimized geometry. The proposed method's effectiveness is showed through benchmark examples in literature. Fabrication of the optimized designs was carried out using material extrusion technique, followed by testing to validate the efficacy of the proposed approach.