Anne van der Pauw, a master's student in Civil Engineering at TU Delft with a specialization in geotechnical engineering, recently completed a research internship at CEMS. During her time with us, she explored the use of machine learning to improve the process of generating ground profiles from cone penetration test data.

This year I did a research internship at CEMS in Delft. It was part of my elective space for my master Civil Engineering at the TU Delft, where I specialize in the geotechnical engineering track.

My assignment was to make a first step in improving the process for obtaining ground profiles. Geotechnical engineers use these profiles to gain insight into the underground. Usually, the data for making such a profile is obtained from cone penetration tests (CPT), where the classified CPT layers are connected to form ground layers. It is often difficult to connect neighbouring CPT's completely, since a layer might not be present in an adjacent CPT, causing a disruption in the ground layer. Therefore, a geotechnical engineer often manually connects all these layers, and this can be difficult and time consuming.

Anne van der Pauw, Master Student Geotechnical Engineering - TUD

My research consisted of exploring unsupervised machine learning possibilities in the form of clustering algorithms to connect similar ground layers in neighbouring CPT's. I did this by dividing the raw CPT signals in blocks representing soil layers using the layer division algorithm of CPT-Core. By extracting signal characteristics like the mean cone tip resistance and sleeve friction, the standard deviations of these signals and other parameters based on signal analysis it became possible to cluster similar layers across multiple CPT's. The cluster algorithms were also able to detect a difference in signals, causing it to identify deviating layers. I concluded that clustering algorithms can detect similarity in signals before classification. In the future this can help improve the classification process and make the ground profile easier to obtain.

This internship was my first experience working in a geotechnical software company. I developed my python programming skills considerably and gained an insight in what it is like to work in a professional environment. There was always room for questions, and everyone was very welcoming and open. I learned that the skills that I developed during my study can be applied to real life problems and that research is accompanied by making a lot of mistakes and learning from them. The different backgrounds and knowledge within both CRUX and CEMS made it very interesting to have discussions and come up with new ideas together. I recommend everyone looking for an internship or graduation project to consider joining these amazing people!

In recent years, CEMS has had several internship positions in the field of geotechnical engineering and structures. Students from several study programs have successfully completed their internship at CEMS. Are you interested in the internship opportunities at CEMS and do you have Python programming skills? Please send your CV and motivation to aW5mb0BjZW1zYnYubmw=.

Site by Alsjeblaft!