Structural Safety Assessment of Reinforced Concrete Structures with Nonlinear Finite Element Analyses and the Significance of the Modelling Uncertainty - Application to Structural Walls

摘要

Nonlinear finite element analyses (NLFEA) allow for simulation of the expected real nonlinear structural behaviour of reinforced concrete structures. NLFEA in structural safety assessment does however introduce potentially significant uncertainties to the design procedure due to complex numerical modelling, which requires comprehension, and management by suitable safety formats. The modelling uncertainty comprises the uncertainties introduced by the solution strategy, the finite element analysis (FEA) software and the user to the design procedure. Solution strategy is used as a collective term for the finite element model and the analysis procedure. In this master s thesis, a structural safety assessment of a reinforced concrete structural wall is performed, with emphasis on assessing and evaluating the modelling uncertainty. The nonlinear FEA software DIANA, version 9.6, is used for all the finite element analyses, and a previously experimental test study of structural walls is used as reference case. Validation of a solution strategy based on recommendations by the Dutch guidelines (DG) for use on structural walls is focused on, since validated guidelines for NLFEA may help minimize the modelling uncertainty and improve the efficiency of the design method. The actual modelling uncertainty is estimated by a statistical approach to multiple structural walls, and relevant global safety formats are applied in the safety assessment, and evaluated with emphasis on the incorporated value of the modelling uncertainty and the impact on the design capacity. The design capacity is also assessed by an analytical method of strut-and-tie modelling. Deficiencies and sources of modelling uncertainty are highlighted in the discussions. The results should be relevant for further studies on this subject and possibly also for later users of NLFEA in assessment of concrete structures for a safer and more efficient use. The estimated modelling uncertainty of a mean ratio of experimental to predicted strength θm=1.21 and a coefficient of variation of the modelling Vθ=6.6% reflects the observed similar behaviour of multiple walls, though at low applied load levels compared to the experimental tests. The constitutive modelling indicates to be the main contributor to the systematic underestimation of the load capacity. The evaluated safety formats provide design capacities greater than by the analytical method, where the safety format by Schlune et.al and ECOV provide the highest design capacity. Significant values of the modelling uncertainty are observed in this study. Until the observed limitations in DG and the FEA software DIANA have been addressed, the selected solution strategy should not be considered as validated for use on structural walls in general, based only on this study. Prescribed, low values of the modelling uncertainty and no correction of bias in the model in the safety formats may be improper for many problems. The difficulty of handling bias, and the modelling uncertainty s dependency on a selected solution strategy and FEA software, is clarified during this evaluation. Model validation and a conscious inclusion of the modelling uncertainty into the safety formats is confirmed as essential for a reliable and profitable use of NLFEA in structural safety assessment.