Combination of geophysical and geotechnical data using belief functions: Assessment with numerical laboratory data

The identification of the subsoil constitutive materials, as well as the detection of possible interfaces and anomalies, are crucial for many site characterization applications. During investigation campaigns, complementary geophysical and geotechnical methods are usually used. These two sets of methods yield data with very different spatial scales and different levels of incompleteness, uncertainty and inaccuracy. In this work, a mathematical combination of geophysical and geotechnical information is proposed in order to produce a better subsoil characterization. It is shown that belief functions can be used for such a fusion process. A specific methodology is developed in order to manage conflictual information and different levels of uncertainties and inaccuracies from different investigation methods. In order to test and validate this methodology, we focus on the use of two selected methods, Electrical Resistivity Tomography (ERT) and Cone Penetration Test . First, a synthetic model with artificial data is considered, taking advantage of the results obtained to conduct a comparative study (effect of parameters and noise level). Then, an experimental test bench is considered, in which a two-layered model is placed (plaster and saturated sands) and geophysical and geotechnical data are generated, using a mini-ERT device and insertion depth values. This work also aims at providing a better graphical representation of a subsoil section with associated degrees of belief. The results highlight the ability of this fusion methodology to correctly characterize the considered materials as well as to specify the positions of the interfaces (both vertical and horizontal) and the associated levels of confidence.