TWO APPROACHES TO MULTIDISCIPLINARY OPTIMIZATION PROBLEMS

This paper discusses two approaches to solve multidisciplinary optimization problems regarding complex engineering systems. These approaches have been developed in order to reduce computing time expenditures required for solution of such problems. The first approach is based on utilization of parallel computations not only for object function computing, but also for an "internal" operation of the algorithm that is parallelized. This allows for a significantly higher acceleration of the problem solution process then a trivial usage of parallel CPUs for optimization criteria calculation. The results of numerical testing for the new algorithm are presented. The second approach consists of using multiple fidelity (multilevel) analysis algorithms. The results of a real-life stochastic multiobjective optimization problem solution are presented. The usage of multilevel approach for such optimization problems results in a significant reduction of computing time.