Multidisciplinary Design Optimization Supported by Knowledge Based Engineering

11Ways and Means for Control and Reduction of the Optimization Computational Cost and Elapsed Time

11.1 Introduction

In the complex world of multidisciplinary design and optimization (MDO), it is tempting to think that one can rely on the power of the algorithms available in the range of MDO products and the power of modern computing to allow engineers to engage MDO solution systems without any need to consider the nature of the problem being addressed. This temptation should be resisted as the size of the optimization/analysis computational activity can rapidly escalate if not subjected to some form of control. The analysis component can consume more than 90% of the total computational effort, and it may grow at a rate varying from quadratic to cubic with the number of the design variables. Similarly, the optimization computing requirement grows with the number of constraint functions. MDO amplifies that computational effort by introducing additional operations necessary to support decomposed data organization schemes such as those described in Chapter 8.

Before embarking on a commitment to use a specific MDO approach or method, the user is advised to carefully examine the nature of the design problem with a view to controlling the computational effort measured by the metrics introduced in Section 11.2. A careful examination of the degree of nonlinearity and continuity of the mathematical functions used in both the optimization and analysis modules may suggest variable transformations ...