25th INTERNATIONAL CONGRESS OF THE AERONAUTICAL SCIENCES A GLOBAL-LOCAL OPTIMIZATION METHOD FOR PROBLEMS IN STRUCTURAL DYNAMICS

The optimization of complex structures involving many design variables and constraints can be performed using a multi-level approach: a structure consisting of several components is optimized as a whole (global) and on the component level (local). Earlier work [1], [2], [3], described a multilevel technique developed for the optimization the Airbus A380 vertical tail plane. In this application, a global model is used to calculate the loads on each of the components. These components are then optimized using the prescribed loads, followed by a new global calculation to update the loads. The component optimization strategy is based on Neural Networks (NN) and Genetic Algorithms (GA). This paper describes a strategy that makes this global-local optimization method possible for problems in structural dynamics. It is established that a parametrization of the component interactions (e.g. component loads) is problematic due to frequency dependence. Hence, a modified method is proposed in which the speed of Component Mode Synthesis (CMS) is used to avoid this parametrization. The effectiveness of this method is demonstrated in a test case concerning the placement of sensor and actuator locations in Active Structural Acoustic Control (ASAC). Special attention is paid to the behavior of the optimization strategy.