Our objective is the development of partition algorithms for active elements for multiperformance objectives. The approach is based on inner-outer neural network control designs such as the dynamic hill-climbing algorithm that have efficient outer loop that optimized the partition choice and focuses the inner loop to best implement a particular choice of control structure such as, LQG/LTR or PPF. The introduction of distributed array of actuation mechanisms for vibration suppression and drag reduction has been investigated at NASA Langley for the purposes of reduction of interior noise. In Ref. 3 it is poignantly shown that the process electronics dictates the control architecture for designs that use an array of high-bandwidth devices. The increasing number of actuator devices forces is beyond the multichannel capability of existing processors, forcing the partitioning of the actuators into 'ganged' subsets. With the advent of mesoscale technology, with large arrays of active elements, this will become increasingly the central issue. It was also demonstrated in Ref. 3 that when the partition choice is optimized for a particular performance objective (e.g., active control of interior pressure) it many have a deleterious effect on another performance parameter (structural vibration).
展开▼
