Using Machine Learning Clustering Techniques as a Support for Optimizing the Laboratory Testing Campaign of an Offshore Wind Farm

1. Home
2. Publications
3. Online Library
4. Using Machine Learning Clustering Techniques as a Support for Optimizing the Laboratory Testing Campaign of an Offshore Wind Farm

Clustering is a sub-set of machine learning methods that can organize data points with similar characteristics into group or clusters to help better understand the structure of the data. Site investigations for Offshore Wind Farms (OWFs) commonly include drilled boreholes with sampling. Numerous basic laboratory tests are often carried out offshore to allow for geotechnical profiling of the material. However, adequate characterisation of the geotechnical properties for foundation design requires advanced laboratory tests which, for budget and planning reasons, have to be limited in scope. Clustering methods can be harnessed to optimize the onshore laboratory campaign by helping the selection of representative samples for advanced testing.
This paper investigates the use of clustering techniques as screening tools to help identify groups of similar materials within glacial clay formations as typically encountered at North European OWF sites. The input for the clustering algorithm included preliminary basic information available at the end of the fieldworks (e.g. moisture content, bulk density, undrained strength from offshore laboratory tests) and the data was processed using algorithms available from open-source Python libraries. The study investigates the ability of two clustering methods, K-means and Gaussian Mixtures Models, and their associated parameters to identify groups of similar material across investigation point locations and to inform the selection of representative samples for advanced laboratory testing onshore. The effects of various data pre-processing options on the results were also explored. The paper discusses some of the advantages and disadvantages when carrying out such analyses.

5th International Symposium on Frontiers in Offshore Geotechnics (ISFOG2025)

5 - Data Analytics and Machine Learning