Abstract
Purpose - The purpose of this paper is to assess the suitability of various methods for the reduction of a large finite element model (FEM) of satellites to produce models to be used for correlation of the FEM with test results. The robustness of the cross-orthogonality checks (COC) for the correlation process carried out utilizing the reduced model is investigated, showing its dependence on the number of mode shapes used in the reduction process. Finally the paper investigates the improvement in the robustness of the COC that can be achieved utilizing optimality criteria for the selection of the degrees of freedom (DOF) used for the correlation process. Design/methodology/approach - A Monte Carlo approach has been used to simulate inaccuracies in the mode shapes (analysis and experimental) of a satellite FEM that are compared during the COC. The sensitivity of the COC to the parameters utilized during the reduction process, i.e. mode shapes and DOFs, is then assessed for different levels of inaccuracy in the mode shapes. Findings - The System Equivalent Expansion Reduction Process (SEREP) has been identified as a particularly suitable method, with the advantage that a SEREP reduced model has the same eigenvalues and eigenvector of the whole system therefore automatically meeting the criteria on the quality of the reduced model. The inclusion of a high number of mode shapes in the reduction process makes the check very sensitive to minor experimental or modelling inaccuracies. Finally it was shown that utilizing optimality criteria in the selection of the DOFs to carry out the correlation can significantly improve the probability of meeting the COC criteria. Research limitations/implications - This work is based on the FEM of the satellite IT>Aeolus/IT>, and therefore the numerical values obtained in this study are specific for this application. However, this model represents a typical satellite FEM and therefore the trends identified in this work are expected to be generally valid for this type of structure. Practical implications - The correlation of satellite FEM with test results involves a substantial effort, and it is crucial to avoid failures of the COC due to numerical issues rather than real model inaccuracies. This work shows also how an inappropriate choice of reduction parameters can lead to failure of the COC in cases when there are only very minor differences (e.g. due to minor amount of noise in the results) between analytical and test results. Vice versa, the work also shows how the robustness of the reduced model can be improved. Originality/value - The paper shows how the robustness of the correlation process for a satellite FEM carried out utilising a SEREP reduced model needed to be investigated, to demonstrate the suitability of this method to reduce large FEM of satellites. Copyright © 2012 Emerald Group Publishing Limited. All rights reserved.