6533b86efe1ef96bd12cc6a9

RESEARCH PRODUCT

Parallel Genetic Solution for Multiobjective MDO

subject

description

Publisher Summary This chapter reviews a multiobjective, multidisciplinary design optimization of two-dimensional airfoil designs. The control points on leading and trailing edges remain fixed, and the y-coordinates of the other control points are allowed to change during the optimization process. The grid for the Euler solver depends continuously and smoothly on the design parameters. The number of nodes and elements in the mesh might vary according to design because the meshes for the Helmholtz solver are done using the local fitting. The computations are made on an IBM SP2 parallel computer using high-performance switch and the MPICH message-passing library. As gradients are not required and the cost functions do not have to be continuous, it can be used in any standard state solvers for shape optimization. Also, it succeeds in getting better parallel efficiency with standard sequential state solvers. The number of performed cost function evaluations is rather high and therefore, the optimization is computationally expensive. To reduce the amount of computations, the convergence toward the set of Pareto optimal solutions needs to be improved.