Multidisciplinary Hybrid Constrained GA Optimization

Realistic engineering problems always involve interaction of several disciplines like fluid dynamics, heat transfer, elasticity, electromagnetism, dynamics, etc. Thus, realistic problems are always multidisciplinary and the geometric space is typically arbitrarily shaped and three-dimensional. Each of the individual disciplines is governed by its own system of governing partial differential equations or integral equations of different degree of non-linearity and based on often widely disparate time scales and length scales. All of these factors make a typical multidisciplinary optimization problem highly non-linear and interconnected. Consequently, an objective function space for a typical multidisciplinary problem could be expected to have a number of local minimums. A typical multidisciplinary optimization problem therefore requires the use of optimization algorithms that can either avoid the local minimums or escape from the local minimums. Non-gradient based optimizers have these capabilities. On the other hand, once the neighborhood of the global minimum has been found, the non-gradient based optimizers have difficulty converging to the global minimum. For this purpose, it is more appropriate to use gradient-based optimizers.