An Enhancement and Comparison of Flexible Structure System Identification Methods for Control

摘要

The need for highly accurate pointing control of large flexible space structures has led to the development of many system identification techniques to provide models for modem control methods. Specifically, Q-Markov Covariance Equivalent Realization (COVER) and the Eigensystem Realization Algorithm (ERA) have received considerable attention. Both have been applied to large flexible structure testbeds such as the ACES facility at NASA Marshall. These methods were developed independently and little work has been done on comparing the two methods. In Section 2 of this paper, a brief description of the methods and theoretical comparisons are made. Emphasis is placed on the implications of the comparisons to the practical implementation of the methods. While studying the two methods, it was noted that a quality measure, Degree of Modal Purity (DMP), used in ERA serves two key functions that were lacking in the Q-Markov COVER algorithm. One key function of system identification algorithms is that they provide not only a model of the system, but also information about the limitations/uncertainty in that model. This requirement is especially important when using robust control techniques such as H(sub infinity), l(sub 1) or Maximum Entropy/Optimal Projection. A second key function of system identification algorithms is to eliminate spurious modes induced by noisy measurements in the identification experiment thus determining the proper system order. DMP provides these functions in ERA. In Section 3 of this paper, DMP is developed for the Q-Markov COVER algorithm. Finally, the two methods are applied to a two degree-of-freedom spring-mass-damper system to provide a demonstration of some of the method comparisons and the usefulness of the newly developed DMP for the Q-Markov COVER algorithm. Further method comparisons and implementation issues are also discussed. A summary of the application is given in Section 4.