Machine Learning-based Design Optimization of Steel Connections

The use of Machine Learning Algorithms and Artificial Neural Networks (ANNs) entered civil engineering applications with increased velocity throughout the last decade. Nowadays, a wide range of ML algorithms offers problem solutions and design variations, suitable to increase and explore data. IDEA StatiCa Connection is an established product allowing to design any connection, any topology and any loading by automatically generated component-based finite element models. The goal of MADESCO is the use and development of novel design and optimization tool within the software “Idea StatiCa Connection”, where mechanics-based methods (like the finite element method - FEM) are augmented and combined with various machine learning (ML) methods, i.e. data-driven models that make use of deep-learning and are trained on large datasets of "synthetic structural tests". This strategy allows for an accelerated prediction of the response of structural joints and connections (within certain "template categories") and their acceptability in design.