Model Updating of Modal Parameters from Experimental Data and Applications in Aerospace

Kurzfassung

The research in this thesis is associated with different aspects of experimental analyses of structural dynamic systems and the correction of the corresponding mathematical models using the results of experimental investigations as a reference. A comprehensive finite-element model updating software technology is assembled and various novel features are implemented. The software technology is integrated into an experimental test facility for structural dynamic identification and used in a number of real life aerospace applications which illustrate the advantages of the new features. To improve the quality of the experimental reference data a novel non-iterative method for the computation of optimised multi-point excitation force vectors for Phase Resonance Testing is introduced. The method is unique in that it is based entirely on experimental data, allows to determine both the locations and force components resulting in the highest phase purity, and enable to predict the corresponding mode indicator. A minimisation criterion for the real-part response of the test structure with respect to the total response is utilised and, unlike with the application of other methods, no further information such as a mass matrix from a finite-element model or assumptions on the structure's damping characteristics is required. Performance in comparison to existing methods is assessed in a numerical study using an analytical eleven-degrees-of-freedom model. Successful applications to a simple laboratory satellite structure and under realistic test conditions during the Ground Vibration Test on the European Space Agency's Polar Platform are described. Considerable improvements are achieved with respect to the phase purity of the identified mode shapes as compared to other methods or manual tuning strategies as well as the time and effort involved in the application during Ground Vibration Testing. Various aspects regarding the application of iterative model updating methods to aerospace-related test structures and live experimental data are discussed. A new iterative correction parameter selection technique enabling to create a physically correct updated analytical model and a novel approach for the correction of structural components with viscous material properties are proposed. A finite-element model of the GARTEUR SM-AG19 laboratory test structure is updated using experimental modal data from a Ground Vibration Test. In order to assess the accuracy and physical consistency of the updated model a novel approach is applied where only a fraction of the mode shapes and natural frequencies from the experimental data base is used in the model correction process and analytical and experimental modal data beyond the range utilised for updating are correlated. To evaluate the influence of experimental errors on the accuracy of finite-element model corrections a numerical simulation procedure is developed. The effects of measurement uncertainties on the substructure correction factors, natural frequency deviations, and mode shape correlation are investigated using simulated experimental modal data. Various numerical models are generated to study the effects of modelling error magnitudes and locations. As a result, the correction parameter uncertainty increases with the magnitude of the experimental errors and decreases with the number of modes involved in the updating process. Frequency errors, however, since they are not averaged during updating, must be measured with an adequately high precision. Next, the updating procedure is applied to an authentic industrial aerospace structure. The finite-element model of the EC~135 helicopter is utilised and a novel technique for the parameterisation of substructures with non-isotropic material properties is suggested. Experimental modal parameters are extracted from frequency responses recorded during a Shake Test on the EC~135-S01 prototype. In this test case, the correction process involves the handling of a high degree of modal and spatial incompleteness in the experimental reference data. Accordingly, new effective strategies for the selection of updating parameters are developed which are both physically significant and likewise have a sufficient sensitivity with respect to the analytical modal parameters. Finally, possible advantages of model updating in association with a model-based method for the identification and localisation of structural damage are investigated. A new technique for identifying and locating delamination damages in carbon fibre reinforced polymers is introduced. The method is based on a correlation of damage-induced modal damping variations from an elasto-mechanic structure to the corresponding data from a numerical model in order to derive information on the damage location. Using a numerical model enables the location of damages in a three-dimensional structure from experimental data obtained with only a single response sensor. To acquire sufficiently accurate experimental data a novel criterion for the determination of most appropriate actuator and sensor positions and a polynomial curve fitting technique are suggested. It will be shown that in order to achieve a good location precision the numerical model must retain a high degree of accuracy and physical consistency.