Classification and probabilistic description of a sand site based on CPTU K. Lauridsen, S. Andersen, L. Ibsen, B. Nielsen 1
Agenda Objective of the project The test site Probabilistic interpretation of a cone penetration test Correlation at distance 2
Objective of the project Presentation and assessment of present methods of probabilistic site description Utilizing the present methods on a set of cone penetration tests Comparison of methods 3
Literature study Interpretation of cone penetration tests Error detection Methods of deriving soil properties and types Division into data sets (layers) Presenting data as statistical, not deterministic Incorporating a distance into the soil parameters 4
The test site Located in northern Jutland Test site for windturbine foundations Reclaimed seabed 5
The test site Located in northern Jutland Test site for windturbine foundations Reclaimed seabed Comprised mainly of sand and silty sand. 6
The test site Process of site description Removal of erroneous data Description of site through soil behaviour charts Division into statistically homogenises layers 7
Error filter Measurement errors Measurements not representative of intrinsic soil properties Setup Changes in measurements due to halts stones or small layers of silt 8
Division into layers Traditional methods using classificational diagrams such as: Robertson 1990, 2009 Eslami 2004 Schiender 2008 Statistical division Modified Bartlett similarity test RI-index test 9
Division into layers Bartlett test Create two windows and compare the variances and mean. 10
Division into layers Having divided the cone penetration tests into layers, the next step is to assess if the layers can be assumed to be weakly stationary. 11
Division into layers Modified Bartlett test Test of homogeneity Calculate an upper value of B crit 12
Division into layers Modified Bartlett test Test of homogeneity Calculate an upper value of B crit 13
Division into layers Top layer plotted into the soil behaviour chart by Robertson 1990 14
Presenting the layer statistically The layer is presented as a random field Mean, variance and spatial variation Spatial variation Different methods of representing a spatial variation 15
Presenting the layer statistically The layer is presented as a random field Mean, variance and spatial variation Spatial variation Different methods of representing a spatial variation 16
Spatial variation Spatial variation Different methods of representing a spatial variation Semi variogram Autocorrelation function Due to insufficient amount of data a model could not be fitted 17
Spatial variation Correlation over distances Comparable correlations over 10 m, 50 m and 125 m 18
Summery Division into layers based on statistical methods shows good agreement with borehole logs Layers found through statistical division fall well with in the same soil type behaviour Statistical representation of a soil volume A considerable amount of data is needed Intrinsic spatial variations difficult to asses 19