Mass Optimization Of 3D-Printed Composites Using Topology Optimization And Artificial Neural Network

Abstract

—Additive manufacturing is a crucial new trend that is steadily taking over traditional methods. Despite its many advantages, the anisotropic nature of the produced parts of most additive manufacturing methods is a significant disadvantage. Of the methods that suffers from this anisotropy drawback is the fused filament fabrication (also known as fused deposition modeling). As a result of this anisotropy in the mechanical properties, a need arises to define the optimum direction of printing to be used for a certain loading condition. Topology optimization is a great numerical design tool for weight and material savings. It’s basically used to determine where to put material to optimize a certain objective function under specific constraints. The design variables in a topology optimization are typically chosen as the densities of the finite elements. Adding the printing direction as an additional design variable complicates the problem further. This eventually gives rise to a huge selection of local minima and further increases in the computational costs. In this work, we attempt to utilize artificial neural networks to tackle this problem. Selected results of mass minimization problems run in ANSYS are used as input data for the neural network model, which is used to predict the fiber angle that has the minimum mass under specific stress constraints. Results so far are promising with small errors considering the computational savings achieved.