Forward and backward probabilistic simulations in geotechnical engineering

To account for non-uniformity and uncertainties in soil parameters, in recent years, civil engineering practice and in particular, geotechnical engineering practice has seen an increasing emphasis on probabilistic treatment to data and subsequent simulation/design. Consistent development of a probabilistic framework for geotechnical simulation/design will not only provide a rational way to address our confidence (or lack thereof) in simulated/designed behavior, but also, it will empower engineers to demonstrate the need for more, uniform data of soil properties, to develop novel site characterization techniques, and to design geotechnical systems that will (probably) achieve best performance. This paper discusses the influence of uncertainties in soil properties on seismic ground motions. Both spatial and point-wise uncertainties - natural variability, testing and transformation uncertainty - in soil properties are identified and consistently propagated through the governing equations of geomechanics, in evaluating the complete probabilistic behavior of the response. Recently developed probabilistic elasto-plasticity and stochastic elastic-plastic finite element method are used for this purpose. Also discussed is setting up of the forward uncertainty propagation problem as an optimization problem in inversely solving for site characterization details from a given probabilistic behavior of the response. Copyright ASCE 2009.