Taylor Series Expansion for Statistical Analysis of Existing Concrete Bridge

The paper is focused on semi-probabilistic assessment of existing prestressed concrete bridge using Taylor series expansion. Probabilistic assessment of existing structures represented by non-linear finite element models is computationally highly demanding and thus semi-probabilistic approach gets a growing attention in a recent decade. In semi-probabilistic approach, design value of resistance is estimated in a simplified way. For estimation of design value, it is crucial to estimate coefficient of variation of structural resistance with limited number of simulations. In this paper, recently proposed Taylor series expansion with adapted differentiation is employed and obtained results are compared to a reference solution obtained by latin hypercube sampling. Practical example is represented by an existing bridge structure failing in bending. It can be seen that results obtained by Taylor series expansion are in good agreement with reference solution but its efficiency is significantly higher. Moreover, it is shown that an accuracy of estimated variance can be easily improved by few additional samples in specific coordinates.