AI-based Gmax prediction from seismic data with improved confidence by improved inputs
conference paper
High costs and efforts of de-risking by CPT probing of the seafloor and the immense amount of data coming from Ultra High Resolution 2D and 3D seismic surveys has driven AI-based methods for geotechnical ground model development. AI methods however still face limitations in their input of geophysical data towards output geotechnical data. We must bridge the gap between geophysical and geotechnical data by means of innovations. Small strain shear modulus Gmax is an important design driver of foundations of fixed offshore structures. It is currently lacking in state-of-the-art AI predictors. Besides the additional output parameter Gmax, the way forward is also to employ improved input parameters as a basis for the AI predictions and improved uncertainties of AI predictions. We succeeded to have the AI prediction process of Gmax to provide an approximation of relationships between seismic reflection data and new, highly relevant geotechnical data. Also we improved the AI predictive power of geotechnical data from geophysical data by the consistent and objective generation of watertight guiding soil units. Lastly, we produced a confidence interval methodology which finally honors and quantifies geophysical, geotechnical, geological and neural network uncertainties altogether.
Topics
TNO Identifier
1023091
Source title
5th EAGE Global Energy Transition Conference & Exhibition
Pages
1-5
Files
To receive the publication files, please send an e-mail request to TNO Repository.