Design optimization and analysis using high performance computing

Abstract

The integration of the advances made in the development of sophisticated computational methods for analysis of complex engineering systems using more exact mathematical models, numerical optimization methods and advanced computer architectures have motivated considerable research activity in design optimization in recent times. This technology enables engineers to arrive at optimal design configurations of complex components and systems in a cost-effective manner and shorter turn-around times. One key element in this integration concerns the choice of the numerical optimization methods. In view of the enormous costs associated with computing the objective function using advanced computational engineering analysis methods, deterministic optimization methods such as the calculus dominated gradient-based methods have been favoured by many researchers. However while these gradient-based methods allow the generation of an improved design, these methods do not enable reaching a global optimum and often restrict the design space to conventional designs. Gradient-based methods also cannot be used for spaces with discrete variables. An alternative to overcome these limitations is to use stochastic optimization methods.