Calibration of constitutive parameters by inverse analysis for a geotechnical boundary problem

The finite element method (FEM) has become a standard tool for solving complex problems in geotechnical engi- neering. Many different advanced constitutive models for fine-grained soils have been developed in recent years, which can consider various phenomena of soil. Generally, the number of constitutive parameters increases with phenomena incorpo- rated in the model, and their determination is one of the key issues in numerical modeling in geotechnics. Normally, experi- mental data, experience, and back analyses based on engineering judgment are used to arrive at appropriate input parameters for a particular model. However, this procedure is not always satisfactory, especially when the number of re- quired input parameters is large. In this paper, a population-based algorithm has been used to determine the constitutive pa- rameters for a geotechnical boundary value problem, namely a floating stone column foundation under an embankment. Measurements of surface settlements and excess pore-water pressures at different depths are available for calibration. A sub- sequent statistical assessment of the calibration results is followed to assess the quality of the identified parameters in de- pendency of the different set of measurements evaluated. As a major result of this research, measures of the utility and reliability of the constitutive models for further predictive computations can be estimated.