Bachelor Assignment

Automated design for topology optimization of truss structures is researched through reinforcement learning based design. Topology optimization finds applications in material science, civil engineering and biomedical engineering. The main focus of this thesis is to define AI design guidelines for automated design of truss structures by use of reinforcement learning. The work includes a discussion on the general set-up of the design method, customization of the algorithm and application related requirements, with the goal of making this method accessible to students. An implementation of a reinforcement learning based topology optimization method for truss structures is described as a concrete example. The thesis explores a model-based approach in which a finite element method of a truss structure is used. The accuracy and size of the finite element model can presumably decrease performance, but improve closeness to reality.