Abstract:

Gradient-based multidisciplinary design optimization (MDO) methods cannot be used directly to problems with discrete and integer design variables or with disjointed design space, and they have a propensity to converge to a relative optimum closest to the starting point. This paper introduces coevolutionary algorithms to multidisciplinary design optimization based on their substantial similarity, to sufficiently employ the merit of evolutionary algorithms in MDO. A distributed coevolutionary multidisciplinary design optimization algorithm is proposed. Each discipline conducts optimization using an independent distributed evolutionary algorithm. The distributed coevolution mechanism is used to coordinate disciplinary optimizations. The algorithm is applied to missile aerodynamic/engine/control multidisciplinary design optimization, and the result shows that the algorithm is promising in design optimization of coupled systems.

Key words: multidisciplinary design optimization, evolutionary computation, coevolutionary algorithms, optimization