A method for the solution of mixed-discrete structural optimization pr
oblems based on a two-level parallel evolution strategy is presented.
On the first level, the optimization problem is divided into two subpr
oblems with discrete and continuous design variables, respectively. Th
e two subproblems are solved simultaneously on a parallel computing ar
chitecture. On the second level, each subproblem is further paralleliz
ed by means of a parallel subevolution-strategy. Periodically, the des
ign variables in the two groups are exchanged. Examples are included t
o demonstrate the implementation of this method on a parallel computer
. (C) 1997 Elsevier Science Ltd.