A key opportunity lies in improving how we gather and integrate large amounts of geological, geotechnical, and geophysical data. Smarter data solutions will lead to better designs and lower costs.
On Certain Ground is creating a smarter way to characterize and model ground conditions — especially offshore, where it is demanding and expensive to gather data. By combining (geo-)statistics, machine learning, and physics, the team at NGI is developing tools that will help engineers make better decisions when designing offshore infrastructure.
The core part of the project is figuring out how to reliably handle uncertainty — the unknowns that come with complex data, from data acquisition to final engineering solutions. Creating detailed ground models that properly integrate geotechnical, geophysical, and geological data is possible, but the process is often slow, incomplete, and has limited capacity for acknowledging or quantifying the impact from the various sources of uncertainty.
NGI is developing a workflow to identify, better understand, and organize these uncertainties, helping bring together different types of qualitative and quantitative information. To bring this vision to life, the NGI project On Certain Ground will focus on three key areas, smarter data, fast and reliable modelling tools, and finally creating a unified framework for decision-making. The project will investigate:
WP1: Understanding where uncertainty comes from in ground modelling, from data acquisition to design, and how the different types of uncertainty are connected.
WP2: Using advanced Machine Learning to help understand the ground more quickly through the use of advanced AI to characterize the soil and rock formations and their boundaries and predict soil behaviour.
WP3: Developing robust probabilistic ML model frameworks to transform indirect measurements (e.g., probabilistic seismic inversion results) into engineering parameters for soil properties.
WP4: Developing efficient techniques for analysing and interpolating 3D seismic and soil property data across large areas in a probabilistic framework using novel Machine Learning techniques in conjunction with more traditional frameworks such as Gaussian Processes.
WP5: Building a consistent and logical framework that tracks uncertainty in a clear and transparent way throughout the project workflow, from start to finish.
WP6: Communication and dissemination of the results to both technical and non-technical audiences.
This project looks to significantly reduce the time and cost of offshore site characterization, enhancing the reliability of foundation design, while gaining fundamental insights into the physics of the subsurface.
More information about each WP is available under the individual Work Package tabs above.
