Abstract
This paper borrows the concept of spectral clustering in the computer vision field, proposes an alternative approach to optimise space frame structure. Spectral clustering was implemented to segment the whole structure into two subclusters. Then genetic algorithm was used to optimise member sizes of each subcluster separately. It is hypothesized that optimizing the structural stability for subassemblies will largely reduce the search space, which allows greater computational efficiency. The program has been developed in MATLAB and tested on differently shaped space frame structure under varied loading conditions. Results show that for a heterogeneous structure with high a level of complexity, the implementation of spectral clustering can separate the enormous search space of GA down to smaller search space, leading to faster convergence with increased the computational efficiency, while providing an equivalent or better optimisation solution.
Highlights
Genetic algorithm has long been used as a global optimum searching tool for structural optimisation
Since space frame structures of unprecedented scale have been used in a wide range from aircraft to architecture and art, the optimisation problems of structures can be computationally inefficient when encountering a complex structure with too many variables to optimise
Building upon Simon, when optimizing a complex structure becomes computationally infeasible for a considerable number of members, one way to improve computational efficiency is to reduce the size of search space according to its hierarchy and decomposability, and optimizing its structural stability for its subassemblies rather than a whole
Summary
Genetic algorithm has long been used as a global optimum searching tool for structural optimisation. When solving the problems with a large number of variables, the search space can be considerably big, such that GA will lose efficiency and require a long time to converge (Rylander and Foster [3]). A nearly-decomposable system has the corresponding genes which can be supposed to operate nearly independently, and hierarchically control the phenotype of specific organs, and the time required for the evolution of a complex form depends critically on the number and distribution of the subassemblies, the systems that evolve by assembly of simpler systems evolves faster than non-hierarchical systems without subassemblies even of comparable size. Building upon Simon, when optimizing a complex structure becomes computationally infeasible for a considerable number of members, one way to improve computational efficiency is to reduce the size of search space according to its hierarchy and decomposability, and optimizing its structural stability for its subassemblies rather than a whole
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have