This thesis is about adjoint sensitivity analysis for discrete systems in structural mechanics on the basis of finite elements. It is shown how the analysis can be processed into the method of generalized influence functions. Thereby it is possible to generalize the classical and well known method of influence functions into a technique, which can be used to compute and visualize sensitivities in an intuitive way. Like this, a better understanding of the behavior of the system can be gained. The coupling of the classical method with the modern procedure of adjoint sensitivity analysis is possible since the influence function of a certain response can be identified as a subtotal of the sensitivity analysis. This thesis gives a description of the method of generalized influence functions next to an introduction to adjoint sensitivity analysis and the classical method of influence functions, in order to get a better understanding for the coupling of the two techniques. Furthermore it will be shown how the classical influence functions for stress resultants and displacements can be derived within the adjoint sensitivity analysis. Additionally, strain-energy will be treated as response, which is not established in civil-engineering practice so far. Sensitivity relations will be derived for all considered types of responses and illustrated with civil-engineering related examples. Also a way to implement the adjoint sensitivity analysis within a finite element environment will be shown. This is the second main emphasis of the thesis next to the theoretical aspects. On the one hand a program structure will there presented in order to run the analysis and on the other hand it will be shown how to realize the theoretically considered response functions in a numerical way.     «

This thesis is about adjoint sensitivity analysis for discrete systems in structural mechanics on the basis of finite elements. It is shown how the analysis can be processed into the method of generalized influence functions. Thereby it is possible to generalize the classical and well known method of influence functions into a technique, which can be used to compute and visualize sensitivities in an intuitive way. Like this, a better understanding of the behavior of the system can be gained. The...     »