Abstract
The identification of element parameters to be used in the analysis of bodies where internal sampling is limited is considered. The parameters are defined as the expected values of the properties, and the probabilities of the properties are obtained by Jaynes' maximum entropy principle with Bayesian updating of regions around sample locations and fuzzy updating where likelihood information is semantic. Loading on a part of the surface and associated displacement measurements are predicted from the analysis, where the unknown parameters are defined in terms of an effective homogeneous material by means of a least squares fit.