Investigation of Experimental Statistical Energy Analysis Implementation for Complex Structures

Abstract

In this work, the implications of implementing and using experimental data in the statistical energy analysis (SEA) framework was investigated, especially when dealing with complex structures, such as entire vehicles. Having experimental results on the real structure becomes always more important for quick and accurate assessment of the vibrational behavior of complex assemblies, especially at higher frequency, where SEA is optimally used. Such data are particularly important in the aerospace and automotive industry nowadays, with the increasing use of new materials such as CFRP to reduce structural weight. Measurements were carried out on structures with increasing complexity, to end with a complete regional train section. Thanks to the gathered data, various parameters key to SEA were computed, so that they could either be implemented in numerical models or used as validation for these models. The growing complexity of the analyzed assemblies allowed to define the experimental setups and parameters correctly and establish a proper computation routine for the post-processing of the measurements. Furthermore, basic SEA assumptions were verified by using a relatively simple assembly, which provided a good platform for a first evaluation of the challenges encountered in experimental statistical energy analysis. A section of a regional train was measured with accelerometers and impulse hammer, the whole section being divided in subsystems. Parallely, numerical models of the train were created in the software VA One, with different configurations of materials and subsystems. The experimental and numerical data were then used to observe the impact of various key parameters on either the calculation of SEA variables with the measured data or on the results obtained from the numerical models. While the vibration levels of the subsystems had significant discrepancies between the experiment and the models for this assembly, the various investigations led to good understanding of the system behavior, such as to pave the way for further analysis. Overall, parameters of interest in SEA were obtained from various experiments, and their sensitivity to the experimental parameters and setups was assessed. Similar work was carried out on numerical models, by varying the properties and configurations of the models. This thesis therefore provides a good overview of the experimental practices in the context of statistical energy analysis, the modeling and computation of numerical SEA models, as well as the bridge between experiments and numerical analysis, with concluding remarks on the possible further investigations in this critical field.