Global structural cost optimization by simulated annealing

This paper discusses the algorithmic requirements far optimising structural designs for overall cost, including material, construction and other costs which can be influenced by the design. Firstly, the characteristics of the resulting design space are described, leading to the formulation of criteria which must be satisfied by the optimisation method The paper then outlines the particular benefits of simulated annealing in this context as a discrete global optimisation method. Finally, the paper reports the results of parametric studies which have been undertaken to calibrate the simulated annealing procedure for global structural cost optimisation. In particular, recommendations are made with regard to the selection of the range of neighbourhood moves, the control parameter schedule, the handling of structural constraints, the Markov chain length and the truncation limit, as well as the initial and final acceptance probabilities and the normalisation constant.