VBN Forskningsportal (Aalborg Universitet) · 2012 · 13 citations · 3 references
ArchitectureComputational EngineeringComputational GeometryDesign OptimizationOptimizationEvolutionary OptimizationGenetic AlgorithmsGeometric OptimizationGlobal OptimizationMetaheuristic OptimizationStructural OptimizationComputational OptimizationGlobal Search AlgorithmsGradient Method OptimizationHybrid Genetic AlgorithmsLocal SearchLocal Search AlgorithmsStructural Topology OptimizationTopological Design Optimization
The paper presents a method to generate the geometry of reciprocal structures by means of a hybrid optimization procedure. The geometry of reciprocal structures where elements are sitting on the top or in the bottom of each other is extremely difficult to predict because of the non-hierarchical nature of such assemblies and the intrinsic three dimensionality of the joint. Such assemblies develop naturally in out-of-plane configurations because elements axis are not aligned, and the position of a single element depends on the thickness and the position of each other element in the assemblies. In this paper it is shown that the geometrically compatible position of the elements could be determined by local search algorithm gradient-based (GB). However the control on which bar sit on the top or in the bottom at each connection can be regarded as a topological problem and require the use of algorithms that explore the global domain of solutions as genetic algorithms (GAs). The benchmark tests show that when the control on the topology is required the best result is obtained by a hybrid approach that combines the global search of the GA with the local search of a GB algorithm. The optimization method is then applied to a recent example of free-form reciprocal structure.
3