For the last few years, EMerald Geomodelling has proven that this might not be the case in the future. The “Fellesprosjektet Ringeriksbanen og E16” project, also known as FRE16, is a good example of this. With nearly 40 km of road, railways and tunnels to be constructed, it's a big project. The scale of it demands a great deal of geotechnical investigations, which is normally done with traditional, invasive geotechnical drillings. But in this case, beyond drillings, a new technology was tested.
Getting in the air
The EMerald Geomodelling team, which at the time was part of the Norwegian Geotechnical Institute (NGI), was brought into the project in 2016. The initial intention was to get an overview of ground conditions by using helicopter geo-scanning. In June that year, helicopter-based instruments were deployed, scanning the subsurface along the planned route between the cities of Sandvika and Hønefoss. These instruments utilize the principle of electromagnetic induction to detect changes in how well different earth materials conduct electricity.
One of the key aims of the site investigation was to map the depth to bedrock. Expert geophysicists and engineers manually interpreted the geo-scanning results. Bedrock ranged from very shallow to up to 140 m deep, according to the results. These were valuable insights for the engineers as it made it way easier for them to plan follow-up drillings. However, manual interpretation of this data is very time-consuming. The time-consuming process was a barrier that prevented the geophysical data from being reused in later stages of the project.
Implementing machine learning
This inspired us to develop an automated, machine learning-based method to interpret the depth to bedrock. In 2017 and 2018, we developed and refined an artificial neural network (ANN) algorithm that combined geotechnical drillings and geophysical data to model the depth to bedrock. Using boreholes drilled up to that time, we showed that this new method could give a much more accurate model compared to the traditional method of just triangulating boreholes. In other words, a smaller number of boreholes were needed to
make an accurate bedrock model.
In the spring of 2021, using the most recent set of boreholes, we used this machine learning algorithm to generate a new depth to bedrock model, and compared it to the triangulated model in use by Bane NOR and Statens vegvesen at the time (illustrated below). The use of this method gave the customer a greater overview of the ground conditions and confirmed many of their assumptions based on the drilling data.
“The 3D visualization gave us a much better understanding of the ground conditions. This is very useful in many ways, for example if we want to move the planned line”. – Inger Lise Ullnæss, Chief Technical Officer at FRE16, Bane NOR SF
Discovering quick clay occurrences
Since 2019, EMerald Geomodelling has developed a related machine-learning algorithm that models quick clay. We combine lab samples, geotechnical soundings, and geophysical data to identify volumes of probable quick clay. This also made us able to point to areas where quick clay was very unlikely, and to areas where we were uncertain and needed more data. Such discoveries of quick clay help lead not only to a safer and more efficient construction process, but also lead to a lower environmental impact – as excavating in quick clay would have released CO2 emissions.
Using the latest data from spring 2021, we generated such models for Bane NOR and Statens vegvesen (illustrated below). What we discovered was that there were high levels of quick clay near the Storelva and Sogna rivers, and some smaller isolated patches near Sundvollen. Not only did these results correlate closely with previous investigations, the illustrative 3D models made it easier to visualize the extent and geometry of these deposits.
“The 3D mapping confirmed a lot of the assumptions we had made, but also showed us a couple of areas where the 3D mapping could get the overview needed to decide the right line much quicker.” – Inger Lise Ullnæss, Chief Technical Officer FRE16, Bane NOR SF
Detecting weakness zones
Another important result of the geophysical mapping has been the detection of major weakness zones along the planned tunnel alignment. More precisely, we discovered a low-resistivity unit near the northern portal (illustrated below). This zone was first detected in 2016 after the initial survey, but EMerald Geomodelling revisited this anomaly with its improved 3D visualization routines. Core drillings confirmed that this resistivity unit coincided with a transition from igneous rocks (rhomb-porphyry, basalt) to sedimentary rocks (sandstone, siltstone, mudstone), as well as a decrease in drilling resistance (to the left in illustration below) and in RQD (to the right in illustration below). This feature persists perpendicular to the alignment of the tunnel (to the left in illustration below), which means this unit cannot be avoided by a minor alignment adjustment.
In conclusion, helicopter geo-scanning has enormous potential for improving how ground investigations are done for large infrastructure projects. Not only does it give engineers a better overview of ground conditions, EMerald Geomodelling’s technology for creating integrated models helps unlock the value in both geotechnical and geophysical datasets in a way that was not possible even just five years ago. Our analysis of the FRE16 projects shows that the technology offers most value when used early on in a project. When we can scan and map the ground conditions at an early stage, we can save a great amount of time and resources as the amount of drillholes required is drastically reduced.
“EMerald Geomodellings technology, when used early in projects, has huge potential. It gives us the possibility to more efficiently manage projects, and helps us save both time and costs.” – Inger Lise Ullnæss, Chief Technical Officer at FRE16, Bane NOR SF