Additively manufactured composite lattice structures combine the stiffness of metallic lattice structures and the high damping behavior of polymeric filler materials. The design variations of the volume fractions offer promising material properties for the integration in dynamically loaded parts, such as milling tools. The ideal distribution of the material in the parts can be calculated by computationally intensive topology optimization algorithms. The homogenization method enables to assign the material properties of a composite lattice structures to the material properties of an finite element according to its relative density. This supports the integration of the composite lattice structures into optimization methods and reduces the computational effort. On this basis, Artificial Intelligence (AI) models for image generation have the ability to speed up this complex optimization process.

In this work, a methodology for an efficient design of tools with an optimal distribution of composite lattice structures under dynamic loads is derived. For this purpose, a homogenized multi-material function for the integration of composite lattice structures in a dynamic topology optimization was simulated. This material curve was used to generate data sets for the training of three AI models: a conditional Generative Adversarial Network (cGAN), a Topology Generative Adversarial Network (TopologyGAN), and a Generative Adversarial Network with Transfer Learning (GANTL). The hyperparameters of these models were tuned and various performance metrics were compared. The GANTL model reached a minimum mean squared error of 0.54%. In the end, the design options for generating a machining tool based on the two-dimensional structures were discussed.