Fractile Based Sampling Procedure for the Effective Analysis of Engineering Structures

The non-linear analysis of the performance of engineering structures requires in general a huge computational effort. Moreover, in some cases a model updating procedure is needed. In this contribution, a model updating procedure has been applied for the simulation of pre-stressed reinforced concrete (RC) beams. The combined ultimate shear and flexure capacity of the beams is affected by many complex phenomena, such as the multi-axial state of stress, the anisotropy induced by diagonal concrete cracking, the interaction between concrete and reinforcement (bond), and the brittleness of the failure mode. Spatial distribution of material properties may be considered by random fields. Furthermore, statistical and energetic size effects may influence the analysis. To incorporate all the mentioned affects within a probabilistic analysis by using Monte Carlo simulation, feasibility limits are achieved quickly. Therefore, the aim was to improve the sampling technique for the generation of the realizations of the basic variables for, a general, computationally complex analysis tasks. The target was to develop a method similar to a simplified probabilistic method e.g. Estimation of Coefficient of Variation (ECoV). Therefore the so-called fractile based sampling procedure (FBSP) by using Latin Hypercube Sampling (LHS) has been developed. It allows a drastic reduction in the computational effort and allows the consideration of correlations between the individual basic variables (BV). However, fundamental aspect of the presented procedure is the appropriate selection of a leading basic variable (LBV). The appropriate choice of the LBV among the defined BVs is essential for mapping the correct correlation. Three methods for the determination of the LBV were investigated in this paper.