Towards Faster Design Cycles Through Gradient-Based Optimization

Abstract

As design tasks are significantly complexifying over time, engineers seek more often assistance by optimization tools in their daily design task. Those cost-effective tools propose innovative solutions which could not be found intuitively by human designers. However, as more and more freedom is given to the design problem, classical optimization algorithms tend to require a significant computational cost. This can be reduced remarkably by using gradient-based optimization algorithms, but these require the computation of the gradient. The present paper discusses the advantages of gradient-based methods and illustrates their capabilities on some turbomachinery design problems compared to gradient-free methods.