Multi-criteria optimisation model for glass facade element with 3D printed structure

Abstract

This thesis investigates the usability and creation of a multi-criteria optimisation tool for a facade element consisting of two outer glass layers and a 3D-printed core. The chosen constraint is related to the structural requirements laid down by the NEN codes based on the ULS and SLS. The criteria are: the transparency criterion, which maximises the transparency of the panel; the insulation criterion, which minimises the U-value; and the daylight criterion, which maximises the indirect sunlight at the summer solstice and the total sunlight during the winter solstice.

Topology optimisation is the first considered optimisation algorithm that allows for complete freedom of the core design through discritisation of the core volume with finite elements. Nevertheless, topology optimisation is not considered usable in this tool as a result of the inability to add relevant criteria to topology optimisation integrated into software, the complexity of developing an optimisation tool with topology optimisation in Python or Matlab, and the intricate optimal core designs, which are not easily printable.

Next, a hexagon structure is used as the core design with some related variables, and it is created in Rhino 7. The daylight criterion is performed with the Ladybug library, and a mesh sensitivity study is conducted. The structural constraint uses Karamba3D, and Karamba3D’s accuracy is investigated with validation with Diana and stiffness experiments. The experiments revealed a large discrepancy between the stiffness observed in the experiments and the predicted stiffness in the Karamba3D and Diana models. The insulation and transparency criteria are added, and the genetic algorithm is used to determine the near-optimal solution. Finally, a case study based on the Echo building is used to compare the near-optimal sandwich panels with regular windows. Although the computational demand inhibits quick results, the sandwich panel performs well compared to the original window in relation to the three criteria.