In this paper two multi-input/multi-output state space methods suitable for the in-orbit system identification of space structures are considered. The ERA (eigensystem realization algorithm) method, which has already been used successfully on the Solar Array Flight Experiment, is compared with the ERA/DC (ERA using data correlations) method in terms of the formulation and computational effort. The methods are applied to simulated data from a space station type model. The ERA/DC method requires less model overspecification in the presence of noise than does the ERA method for similar quality of results. A revision of the ERA/DC method is suggested. UTURE large space structures, such as the Space Station, will require in-orbit system identification of their struc- tural dynamic behavior to validate theoretical models and to insure that control systems do not give rise to any instabilities associated with control-structure interaction. 1 This validation is particularly important as structures become larger and more flexible since the control bandwidth can approach the first structural frequency. Typically, an identified model for con- trol synthesis should represent the modal behavior of the structure in a frequency range up to 10 times the control bandwidth. The need for in-orbit testing arises for two reasons. First, the structures will be so complex that theoretical finite element type models will not be sufficiently accurate to be relied on in the absence of test data; in particular the modeling of joints between substructures and of damping is extremely difficult. Second, even substructures will usually be so large and flexible that they will require multiple supports if they are to be tested in a Ig environment. Therefore, in most cases ground-based modal test results for substructures will be unlikely to predict in-orbit behavior with sufficient accuracy, given also the un- certainty in the stiffness and damping of the joints between substructures. The use of scale models, while valuable in gain- ing understanding of the behavior of space structures, will not obviate the need for in-orbit identification. The in-orbit identification of future structures like the Space Station will prove extremely challenging due to the large number of low-frequency , lightly damped, and very closely spaced modes of vibration. Any test will not only require multiple response measurements but also multiple excitation force locations and multiple sets of inputs (or references) to allow identification of the very close modes. Thus multi-in- put/multi-output (MIMO) identification methods are re- quired. The use of multi-point excitation is well established in the aerospace industry for ground vibration testing of aircraft. However, the use of such steady state sinusoidal excitation techniques in space is considered impractical because test times will be excessive and the energy expended may be pro- hibitive. Instead the application of various patterns of multi- ple simultaneous impulses would seem more appropriate, the sequence of patterns designed to minimize the net rigid body response of the space structure. The test philosophy would depend upon whether the modes of a component (such as the
Search from 250 M+ research papersSearch from 250 M+ research papers
Dec 3, 2007
A.M Si Mohammed + 4
Open Access