Electromagnetic fields, machine learning and algorithms reduce risks and costs
From a grass-covered field in Stjørdal [a town in Norway], a helicopter lifts off. Below it hangs a hexagonal structure measuring 342 m2 that weighs 700 kg. After a few days of flying a total of 30 km2 of new alignments of the E6 [motorway] in Trøndelag [a county in Norway] will be covered: 20 km2 between Ulsberg and Kvål, and 10 km2 between Kvithammar and Åsen.
The equipment, which the Danish company SkyTEM has hanging below the helicopter, sends electromagnetic waves into the ground. The data it produces gives information about what lies below the surface. For many years, such measurements have been used mainly by the mining industry to explore for mineral deposits. As a spin-off of the University of Aarhus, SkyTEM adapted the technology to be used for groundwater mapping.
New product – New company
NGI is approaching the end of ten years of research and development on new applications of airborne electromagnetics, often referred to as AEM. The result is EMerald Geomodelling, a spin-off of NGI in cooperation with Kjeller Innovasjon, that now offers mapping of the ground conditions to contractors and others with similar needs. It was EMerald Geomodelling, in collaboration with NGI, who performed the job in Trøndelag, with SkyTEM acting as a subcontractor.
When the magnetic field reaches the ground, it induces electric currents in conductive materials. The currents move deeper into the ground and create a secondary magnetic field that is measured by the equipment towed by the helicopter. The strength of the secondary magnetic field reflects the properties of the ground.
Making 3D models
Resistivity gives information about the conductivity of rocks and sediments. These are established relationships that make it possible to use the data from the measurements to assess what is in the ground. Up until now, engineers have received pdf reports with large numbers of figures, data and color charts that require interpretation and a lot of manual work.
“We have developed an algorithm that interprets the data and translates it into understandable 3D models. Instead of delivering large amount of complex data, we give the engineers numbers they can understand and use in the design,” says Guro Huun Skurdal, VP Operations at EMerald Geomodelling.
Better decisions, fewer drillings
Skurdal says the solution should be used early in the design phase, preferably before the route is selected.
“Yes, because an overview of the ground conditions, including depth to the rock and locations of weakness zones, makes it easier to choose the route. Our method does not replace drilling, but it provides a basis for obtaining an overview of where to drill, thus significantly reducing the number of drillings.”
Kari Charlotte Sellgren is the technical lead for tunneling in Nye Veier Trøndelag. She tells Teknisk Ukeblad that the new method offers substantial savings.
“For one of the two routes, we will save 1.5 million [Norwegian kroner] on field investigations and another half million on laboratory tests. The savings on the second road section will come on top of that. There is no doubt that this is economical for us. At the same time, we get unique preliminary data that can be used not only during the design and construction phase, but also for the operation and maintenance phase,” says Sellgren.
She says they now get a more accurate and detailed overview of the rock surface, making it easier to calculate, among other things, mass balance.
Traditionally, geotechnical investigations are carried out to map out ground conditions in major infrastructure projects. This often involves many boreholes, a process that is both time consuming and expensive. EMerald Geomodelling combines geotechnical data and helicopter scanning data using machine learning.
“We use this data as input for the machine learning process. We tested our algorithms with data from an area where over 1000 holes had been drilled and where helicopter data was collected. The result showed that data from helicopter scanning, combined with only a few boreholes, provides a good overview of the main structures in the area,” says Craig Christensen, VP Technology at EMerald Geomodelling.
He adds that with the use of airborne electromagnetic measurements, the uncertainty in the model will be halved compared to if only traditional ground investigations had been done at an early stage of the project.
“What’s just as important is that the model provides information on what locations should be drilled.”
Andreas Pfaffhuber, CEO of EMerald Geomodelling, holds a PhD in this field and has been the main driver for the development of this innovation under the auspices of NGI.
Christensen says that McKinsey and the University of Oxford have done studies that show that 90 percent of all infrastructure projects exceed their budget. This applies globally, and the cost overruns are in the range of 20 to 50 percent. The biggest overruns occur in tunnelling projects, followed by bridges, railroads and roads. With these figures in mind, EMerald Geomodelling does not fear that there is lack of demand.
“The core of what we do is combining geotechnics and helicopter geo-scans using machine learning algorithms, including artificial neural networks. That’s something that nobody else does,” says Skurdal.