Interpolation Methods and Neural Networks Applied to Geotechnical Mapping of a Brazilian Port Site

摘要

Soil proprieties are typically determined from a limited number of in situ or laboratory tests performed at a specific site. The results of these tests may be used to estimate soil properties of the whole site by means interpolation techniques such as linear, polynomial, statistical or machine learning methods. In this paper, a Standard Penetration Test (SPT) database from the port of Navegantes, south Brazil, is used for analysis. Geotechnical mapping of SPT blow counts is made using classical interpolation methods and the neural networks technique, and their performance compared.