Martedì 11 luglio 2023 - Ore 14:30 - Aula 24
Dott. Giuseppe Romano, Massachusetts Institute of Technology (MIT), Cambridge, Boston (USA)
  Seminario  

Additive manufacturing enables the fabrication of materials with voxel-level precision. Topology optimization (TopOpt) aims to find optimal material distributions given desired physical properties and under manufacturable constraints. Combining these two techniques allows fast materials development for various applications, ranging from structural materials to optics. After reviewing the main concepts behind TopOpt, such as the adjoint method, I will describe our recent efforts in shape optimization using TopOpt and reverse-mode automatic differentiation (AD). This technique falls under the broader term differentiable programming, a computing approach that enables the differentiation of computer programs by composing a set of predefined rules. Consequentially, it is possible to explore various cost functions and manufacturable constraints quickly. I will showcase a few examples regarding thermal transport, both at the macro- and nano-scales, also including lengthscale constraints and a novel connectivity constraint. Final remarks and future works will conclude the talk.