Structural design and optimisation by organisation of stiffness - Development of a finite element package for generative design

Abstract

Structural design optimisation by organisation of stiffness Development of a finite element package for generative design CARL HOFF & ISAK NÄSLUND Department of Applied Mechanics Chalmers University of Technology Abstract When a structure is designed, the designer has a choice to make their design in any number of ways. What drives the choices could be structural efficiency (oftentimes measured by weight), economy, architectural qualities, rationality, environmental impact or some other factor. Since these factors often can be hard to measure, especially at an early state of design when the level of uncertainty is high, a need for quick tools for generating acceptable designs is identified. This thesis work has focused on developing a structural design tool that can be used for this purpose. It is implemented as a finite element package for the Rhinoceros plugin Grasshopper, a parametric design tool which also is well suited for the early stages of design. The package, called CIGull, contains a set of customisable optimisation methods for the generation of design, and includes simplified section checks. The methods were all focused on the redistribution of internal forces by organisation of stiffness. The methods included automated iterative section sizing,section alignment rotation, canonical stiffness (statical eigenmode) analysis and connection stiffness (discrete springs) redistribution. The methods were tested on a small set of test structures of varying complexity, as well as on a real project: KAFD Metro Station by Zaha Hadid Architects and Buro Happold Engineering. The iterative section sizers were found to be quick, but they sometimes produced inefficient load paths.Improvements to these methods were suggested. The section alignment rotation method was shown to be an effective tool to increase structural efficiency, but questions were raised about its level of rationality. The connection stiffness redistribution method was shown to be able to improve structural efficiency,but automation of the method using genetic algorithms was shown to be very time consuming. It was possible to combine statical eigenmode analysis with an iterative section sizer to create new load patterns, which in some cases produced more effective structures. It was concluded that the best results were obtained by combining different methods.The proposed methods were found to be sensitive to the case-specific input parameters.No one method was shown to produce better results in all cases. It could be concluded that the methods were best used as a means of understanding the structure. From the tested examples we conclude that the developed generative design tool is promising for the early conceptual design stages. Keywords: Structural design, Optimisation, Finite element method, FEM, Parametric design, Grasshopper