Abstract
Mathematical models (Finite Element Model (FEM)) of spacecraft structures are validated via a correlation process with results from dynamic test campaigns. Typically the Modal Assurance Criterion (MAC) and Normalised Cross Orthogonality (NCO) are used to quantify the quality of the match between models predictions and test results, and appropriate thresholds are set to ensure that a model that passes the test gives sufficiently good predictions of the behaviour of the real structure. However these criteria focus on comparing mode shapes, and the results of this correlation can be misleading. If an excellent correlation is achieved (e.g. MAC very close to unit value), the modal model is an excellent representation of the real structural behaviour; but most often there are models where MAC or NCO is just below the threshold and these are deemed not fit for purpose and requiring some FEM updating to push them above the validation threshold. The main contribution of this work is to show with practical examples of real spacecraft FEMs that a model that passes MAC and NCO checks is not necessarily better at predicting some of the relevant structural responses (e.g. reactions at the base of the craft) than a model that does not pass the checks. The examples show that some of the responses are highly uncorrected with MAC and NCO results and therefore the temptation to extrapolate the results of these validation checks to the "quality" of the model - quality intended in terms of capability of model to predict relevant responses - should be rejected. Using Monte-Carlo simulations, it is also shown that, in the System Equivalent Reduction Expansion Process (SEREP), it is better to include only the target modes in the reduction process. Copyright © (2012) by the International Astronautical Federation.