Article,

Polymorphic Uncertainty Quantification of Computational Soil and Earth Structure Simulations via the Variational Sensitivity Analysis

, and .
PAMM, 19 (1): e201900289 (2019)
DOI: \url{10.1002/pamm.201900289}

Abstract

Abstract Due to the high natural variability of mechanical properties in soil, comparatively high safety factors are still used for the dimensioning of earth structures. Back in the 1970s, probabilistic concepts were recommended as the technical standard for geotechnical applications. However, their dissemination failed because of the high computation costs on one hand and, on the other hand, often not enough data being available to carry out such analyses. By means of increasing computing power and enhanced mathematical models, a better understanding of the nature and influence of the uncertainties and errors could help to develop a more reliable risk assessment and to reduce inefficient safety factors. For this reason, we perform a research project within the priority program SPP 1886, installed by the DFG and focused on polymorphic uncertainty quantification. In the present subproject (SP 12), the focus is driven on quantification and assessment of polymorphic uncertainties in computational simulations of earth structures, especially for fluid-saturated soils. To describe the strongly coupled and transient solid-fluid response behavior, the Theory of Porous Media (TPM) is used and solved via the Finite Element Method (FEM). The goal is to overcome the above mentioned two problems by providing suitable numerical methods. Motivated by structural optimization research, the Variational Sensitivity Analysis (VSA) provides detailed prior information about the current equilibrium state for subsequent analyzes. On its basis, the information for which parameters and in which areas of the simulation domain a more detailed investigation is required can be obtained in advance. The method enables an immediate decision support, e.g. for site investigators or researchers. Not only the computational effort but also the applicability to any arbitrary FE-Model are great advantages of this approach.

Tags

Users

  • @timricken
  • @carlahenning

Comments and Reviews