Deep Learning Architecture for Topological Optimized Mechanical Design Generation with Complex Shape Criterion

Topology optimization is a powerful tool for producing an optimal free-form design from input mechanical constraints. However, in its traditional-density-based approach, it does not feature proper definition the external boundary. Therefore, integration of shape-related constraints remains hard. It requires experts’ intervention to interpret generated designs into parametric shapes; thus, making process time-consuming. With growing role additive manufacturing industry, developing approach considering and geometrical simultaneously becomes interesting way integrate aesthetics design. In this paper, we propose generate mechanically geometrically valid using deep-learning solution trained via dual-discriminator Generative Adversarial Network (GAN) framework. This Deep-learning-geometrical-driven generates very similar traditional topology optimization’s outputs fraction time. Moreover, allows easy shape fine-tuning by simple increase/decrease condition (here total-bar-count), task that cannot achieve.