A completely novel three stages topology-sizing design optimization methodology will be developed, in which the capability of performing Linear Buckling and GMNI Analyses as well as treating multiple loading cases in each search iteration will be incorporated. In addition, in order to reduce the required computational effort machine learning techniques will be integrated.

• Task 1.1 Nonlinear three-stage topology-sizing design optimization methodology: The results of the TO approaches consist of massive continuous media and need to be interpreted into non-solid structural members used in structural engineering design. A Nonlinear based Topology-Sizing Design Optimization (NTSDO) methodology composed of three stages will be developed. This approach has obvious advantages, as the final optimized shape will require no further interpretation or additional checks to be manufactured (Lead by NTUA)
• Task 1.2 Machine learning assisted TO: In this Task the computational demands of the TO stage of the NTSDO methodology will be treated; more specifically machine learning (ML) will be integrated into the first stage of the NTSDO methodology. The predicted design by ML will then be passed back to the TO approach in order to perform a fine-tuning, (Lead by NTUA).
• Task 1.3 Design of members, connections and structures with non-linear FEM analyses: In order to design metal members, connections and structures with non-linear FEM analyses, a Linearized Buckling Analysis (LBA) will first be performed, in order to obtain buckling mode shapes that will then be adopted as initial imperfections for the subsequent Geometry and Material Nonlinear Analysis with Imperfections (GMNIA), (Lead by NTUA).