Partitioned model reduction for large space structural control problem

Abstract

A model reduction method for control system design of large space structures is presented. The control model of the structure is partitioned into several subdomains, and a generalized Craig-Bampton representation is derived. In considering the placement of actuators/sensors that may affect the selection of the candidate modes, the representative degrees of freedom (DOF) of the actuators/sensors are retained as physical DOF in the final coordinate vector. A reduced model for the controller design is obtained based on this representation. As an example, a space truss subject to four actuator forces is analyzed. Comparison with a conventional reduction method based on a complete model of the truss has been made via eigenpairs and dynamic responses. Good agreement is seen to be achieved, and improved accuracy is obtained with the proposed method. The importance of the actuator/sensor related DOF and the effectiveness of the proposed approach in dealing with the very localized DOF are investigated under the extreme condition where only these four actuator/sensor related DOF are considered as the dimension for the model reduction.