A parallel structural optimization method and its implementation on a Transtech Paramid Original Research Article

This paper presents a parallel structural optimization method and its implementation on a distributed memory, multiple-instruction, multiple-data parallel computer. The proposed strategy does not require the formation or solution of global system equations. Instead, each processor of the parallel computer stores only the data relevant to one structural element or substructure and generates its local stiffness matrix. A parallel conjugate gradient procedure using the optimality criterion approach is employed to compute the displacements and modified design for each element (or substructure). An implementation of the parallel structural optimization method on a Transtech Paramid parallel computer is discussed, the speedup and efficiency of this method being demonstrated by numerical examples.