Sensitivity Analysis and Parameter Identification on historical Masonry Arch Bridges

Close to reality computation of the serviceability and load-carrying capacity of old masonry arch bridges for rising live loads is a current problem. In Germany many of such roadand railway bridges exist. Some of the most important features for applicable numerical evaluation of masonry arch bridges are a realistic structural model (very often three dimensional models are necessary), the consideration of the nonlinear load history and in particular the inclusion of a realistic nonlinear material model applicable for historical masonry. Often input parameters necessary for a realistic simulation are unknown. Then e.g. material and geometry data must be determined by measurements. In particular at historical masonry bridges these data can show relatively strong variance. By sensitivity studies can be determined correlation between the input data and output value of the computation model. As basis of measuring program so the most important input values can be determined. Sometimes some input values can be measured only indirectly. In these cases parameter identifications can be used for the determination of the input values and for validating the computation model. The paper presents powerful strategies for sensitivity analysis with stochastic sampling methods and for parameter identification with optimization algorithms. Several numerical examples from a masonry arch bridge demonstrate the application of these methods. The masonry behaviour described by the presented practical applications with threedimensional elasto-plastic continuum models for regular and irregular masonry types. The constitutive models are based on multisurface plasticity theory and includes anisotropic elastic and inelastic behaviour depending on the orientation of the masonry joints. The yield domain, its hardening law and softening law are defined according to experimental results. On this basis it is possible to simulate masonry-specific failure and damage mechanisms. 2 ARCH’07 – 5th International Conference on Arch Bridges Sensitivity analysis are just right for answering these questions The localization of relevant areas for taking the material tests can be orientated at the utilization of the structure on one hand and the necessary evidences on the other hand. Thus, it is important to compute the load flow and the different load states within the building. As shown in Schlegel (2004), this can only succeed at masonry with the help of a realistic material model which is capable to describe the essential load relocations and failure mechanisms of the masonry structure. In this article, the FEM calculations are carried out by the FE program ANSYS using especially for regular and irregular masonry types of implemented nonlinear material models under the application of Schlegel (2004). Furthermore, one has to be aware that the material quality can highly scatter especially at historical masonry. Single material samples can only offer local statements. Sometimes, some parameters can not be determined directly. For the identification of the global system behaviour further measurements are necessary (Zabel & Bucher 2000), for example deformation or modal variables. With the help of modern optimization algorithms, essential global system and material parameters can be identified. The adjustment between measurement and recalculation belongs to the classical tasks of a model validation. If the difference between measurement and recalculation is too big, an optimization task of minimizing the difference can be formulated (Will 2006).Optimization tasks of adjusting measurements and recalculation are often called identification problem (e.g. parameter identification) or inverse problem. For the sensivity analysis and parameter identification below, the software optiSLang (Dynardo 2006) is used which has been developed by Dynardo.