Automatic design of truss structures using evolutionary algorithms

A method for the automatic design of truss structures using evolutionary algorithms is described. The underlying approach relies on a problem-dependent genotype-phenotype mapping which mimics the “cellular automata” approach found in the embryonic development of higher individuals. The mapping is performed by a recursive application of simple construction operators and restricts the configuration space to a subclass of “interesting” static truss structures. The mapping is controlled by the individuals´ chromosomes. The individuals are optimised by standard genetic algorithms. Genetic algorithms produce only the topology, i.e. the scheme of how the poles are connected. This structure is further optimised using evolution strategies. As result, it is shown that, using this approach, the weight of the socket of a standard pylon can be reduced by more than 30%