In our recent work for RUMO Logistica in Brazil, this exact methodology was utilized.
“We generated 3D volumetric models of the sediment types along the length of the railway sections. We were able to give a prediction of different materials that included information about the certainty of each of our predictions. These aided RUMO in, among other things, planning their excavation methods”, says Anne Have Rasmussen, Director of Operations at EMerald Geomodelling.
Volumetric analysis in practice
The following sections will, based on the RUMO project, demonstrate briefly how volumetric analysis can be utilized in your project.
The first phase is called the training phase or learning phase. In this phase, ground-truth data are used to calibrate the model. Once a large enough volume of labelled sediment type data from drillings is collected, the supervised machine learning can commence. The dataset of labelled sediment type, overlapping with the area covered by airborne geophysical data, is referred to as the training data. In the RUMO project, five main categories of material were identified: clay, silt, sand, rock, and laterite.
Then the algorithm uses electrical resistivity, spatial gradients of resistivity,and spatial coordinates as input variables and learns how they correlate to the different types of sediment identified at drilling locations.
In the second phase, the prediction phase, the algorithm makes a prediction at all remaining locations where geophysical data are available. The output of the algorithm is a 3D model of material types with probabilities ranging between 0 and 1 of each material and where the sum of probabilities for each materials material is 1 at any given point in space. Where the maximum value at a point is below 0.5,this indicates that the prediction is quite uncertain.
Contributing to saving both time, resources, and the environment
The figure to the right is one example of the results from the project, and it shows an oblique view of the 3D model of material classification, showing the most likely material at surface (upper panel) and the probability of the most likely material occurring (lower panel). This example shows a gradual transition from mostly sand to mostly clay between over a 2 km-long transition zone along the investigated route.
Based on this analysis and other EMerald generated geophysical models, recommendations of areas in need of follow-up ground investigations are presented.
“In the case of RUMO, they estimated that all our deliveries in total saved them 33% of the drillings they would otherwise have done. In other words, volumetric analysis along with other outputs, saves both time, resources, and the environment”, Have Rasmussen says.
Want to learn more?
For more information about volumetric analysis and how it can contribute to increased results for your infrastructure project, please get in touch with us at firstname.lastname@example.org.
For more information about the RUMO project referred to in this article, please visit: