Matching objective and subjective information in geotechnical inverse analysis based on entrophy minimization

Abstract

This chapter describes the Extended Bayesian Method, an inverse analysis procedure, in which Akaike Bayesian Information Criterion (ABIC) is introduced to realize the best matching of the observed data and prior information. In order to overcome the ill-posedness of geotechnical inverse analysis, it is inevitable to introduce prior information of some form, and thus Bayesian statistics. There is, however, a fundamental problem embedded in this problem: the optimum matching between the objective information and the subjective information, so that the model performs best for the prediction purpose. In this study, this problem is solved by taking entropy as a quantitative measure of information: the available information is most effectively used to maximize the relative entropy, thus the best matching of subjective and objective information is accomplished.. In this study, an actual embankment construction record is analyzed to obtain the best set of parameters for the ground deformation model. Since it is requested to simultaneously estimate 11 parameters, the problem is essentially ill-posed. It is shown that the methodology proposed performs rather effectively in such a condition.