Motivation and Goals

Traditional structural-design education generally requires engineering students to carry out in a fixed sequence of tasks. This limits students to single-strategy and single-solution mentalities, restricting adaptability and creativity. The goal of this project is to enhance structural-engineering education of decision-making with multi-criteria in building design. A primary emphasis is to incorporate explainability into the design procedure, empowering students with a deeper understanding of the rationale behind design solutions. More specifically, this research involves exploring the possibilities of novel AI approaches to shape the conceptual design thinking of novices, this helps to prepare them for complex engineering challenges.

Recent Results 

An explainable fast-analysis framework has been proposed and a preliminary investigation was conducted to assess the decision priority of design parameters. To improve the extrapolation of the ML analysis model in building design, a graph neural network (GNN) model is being developed to represent building structure topology through node and edge interactions, simulating force transmission across the structure. An investigation into the "Tragwerksentwurf" class at TUM in winter season 2024/2025 is underway to understand the current teaching environment. Additionally, study of a design tool (developed in Rhino) is in progress and will incorporate GNN training results, with classroom implementation expected next year.