Propagation of Spacecraft Free-Interface Substructure Uncertainty Into System Test-Analysis Correlation

As space structures become larger and more complex, ground based vibration tests of the entire spacecraft become either problematic or impossible. Instead, the spacecraft is tested and validated only at the substructure level. The substructure tests are usually performed in a simulated free-free configuration for simplicity and accuracy. A methodology is presented for studying the effects of uncertainty on metrics used for test-analysis correlation of complex spacecraft that are validated on a substructure-by-substructure basis. The objective is to quantify the level of accuracy required at the substructure level to produce acceptable accuracy at the system level. This is done by propagating uncertainty in test-analysis correlation metrics from free-free substructure vibration tests into uncertainty in synthesized system correlation metrics. The correlation uncertainty in each substructure can either be prescribed, for the purpose of numerical experimentation, or it may be available from existing substructure test data. Linear covariance propagation is used to propagate the substructure modal matrix uncertainty into the uncertainty in the corresponding Craig-Bampton substructure representations. Linear covariance propagation is then used again to propagate the substructure uncertainties into the full system modal matrices. The uncertainties in the system correlation metrics are then extracted from the modal matrix uncertainties to determine the impact of uncertainty at the substructure level. Organizations, such as NASA and the Air Force, make critical decisions on spacecraft performance and survivability based on the results of test-analysis correlation metrics. In order to ensure the success of finite element model validation where there is no system level test, uncertainty in the substructures must be propagated into the system level correlation metrics. It is believed that the method presented in this paper offers a unique and efficient approach for the required uncertainty propagation. [DOI: 10.1115/1.4006644]