Enhancing Designer Understanding by Combining Multiple Dominance Relations and Tabu Search

摘要

In this paper a new tabu search (TS) implementation of the optimization using multiple dominance relations (MDR) approach is presented. Unlike previous population-based implementations, this new algorithm is able to provide the user with information regarding the development history of the final designs produced. This helps the designer gain physical insights into how the observed performance improvements have been achieved by the optimizer. The ability to provide the physical reasoning behind predicted performance improvements is essential if optimized designs are to be adopted in industry, particularly when safety is critical. The new algorithm is shown to outperform previous implementations when applied to both analytic and airfoil design test cases, with the tracking of design development illustrated using the latter problem. The additional information provided to the user comes at no additional computational cost, increasing the value of each optimization run. The more robust design space searching capabilities of the TS upon which the new implementation is built also facilitates the application of the MDR approach to a new class of more complex design problems.