Efficient conditional modeling for geotechnical uncertainty evaluation

Andrew J. Graettinger

doi:10.1002/nag.197

Efficient conditional modeling for geotechnical uncertainty evaluation

2001-12-17

7

PRE

AI

A

Andrew J. Graettinger

DOI：10.1002/nag.197

原文求助

摘要

En 中文

A first-order Taylor series method including direct derivative coding (DDC) is presented as a computationally efficient method for producing the probability distribution associated with calculated geotechnical performance. The probability distribution is employed in reliability analyses to calculate the probability of failure. valuable information that is not typically associated with deterministic analyses. The probability distribution also is used to identify important input parameters and to direct sampling efforts. Another approach to generate the probability distribution is the Monte Carlo (MC) method, however, Taylor series results generally are calculated in less time than the MC approach. One key to the implementation of the Taylor series approach is efficient approximation of the sensitivities required by the Taylor series calculation. DDC provides the technique to produce an efficient Taylor series algorithm. Directly coding the sensitivity analysis into the engineering model is accomplished by automatic and hand programming of derivatives. ADIFOR 2.0 was employed to automatically add derivatives to an existing engineering analysis model. For this paper a meshing program and 3D FEM for soil deformation is used to demonstrate the DDC approach. Although DDC requires a large up-front programming effort. it is not site or data specific. Therefore. once the derivative programming has been performed, the numerical model can be applied to a wide variety of problems without additional user intervention. Copyright (C) 2001 John Wiley Sons. Ltd.

Keyword:

first-order Taylor series

Monte Carlo

sensitivity

reliability

Efficient conditional modeling for geotechnical uncertainty evaluation

摘要

期刊

机构

引用论文