Investigation of model updating techniques and their applications to aircraft structures

Thesis Type: Postgraduate

Institution Of The Thesis: Middle East Technical University, Faculty of Engineering, Department of Mechanical Engineering, Turkey

Approval Date: 2006

Thesis Language: English

Student: Mustafa Tuğrul Kozak



Mathematical models that are built in order to simulate the behavior of structures, most often, tend to respond differently than the actual structures in their initial state. In order to use the mathematical models and their computational outputs instead of testing the real structure under every possible case, it is mandatory to have a mathematical model that reflects the characteristics of the actual structure in the best possible way. In this thesis, the so called model updating techniques used for updating the mathematical models in order to make them respond in the way the actual structures do are investigated. Case studies using computationally generated test data are performed using the direct and indirect modal updating techniques with the software developed for each method investigated. After investigating the direct and indirect modal updating techniques, two of them, one using frequency response functions and the other using modal sensitivities, are determined to be the most suitable ones for aircraft structures. A generic software is developed for the technique using modal sensitivities. A modal test is carried out on a scaled aircraft model. The test data is used for updating of the finite element model of the scaled aircraft using the modal sensitivities and the usability of the method is thus evaluated. The finite element model of a real aircraft using the modal test data is also updated using the modal sensitivities. A new error localization technique and a model updating routine are also proposed in this thesis. This modal updating routine is used with several case studies using computationally generated test data and it is concluded that it is capable of updating the mathematical models even with incomplete measured data.