Prior Knowledge on Shear Strength and Compressibility of Glaciolacustrine Sediments in Northern Germany

A major challenge of most geotechnical engineering projects is soil data scarcity. This paper aims at extending prior knowledge on shear strength and compressibility of Glaciolacustrine sediments of Northern Germany. Based on triaxial, incremental loading oedometer and complementary laboratory tests on specimens from 13 different locations, the inherent variability of shear strength and compressibility is analyzed; typical ranges and coefficients of variation are established. Prior to variability analysis, k-means clustering, a simple machine learning algorithm, is applied to distinguish soil types by their descriptive properties. This data-driven methodology serves the multivariate character of soil data and allows to provide data on the variability of soil strength and compressibility more accurately. It was found that plasticity index and clay content can be considered to distinguish different soil types. Moreover, it can be shown that mean and variability of shear strength and compressibility are clearly affected by the dominant soil type.

Keywords: soil variability, glaciolacustrine sediments, k-means clustering, multivariate data analysis.