Geological Strength Index quantitative estimation of flysch rock masses from on-field geomechanical data based on AI tools

[EN] This work proposes a quantitative estimation of the Geological Strength Index (GSI) in flysch rock masses based on data gathered from geomechanical stations, i.e. Rock Quality Designation (RQD), and spacing, persistence, aperture and roughness of the discontinuities. Artificial Intelligence (AI...

ver descrição completa

Detalhes bibliográficos
Autores: Garzón-Roca, Julio, Rodríguez-Peces, Martín J., Ramos, Adrià, Torrijo, F.J.|||0000-0001-6048-6792
Tipo de documento: artigo
Data de publicação:2025
País:España
Recursos:Universitat Politècnica de València (UPV)
Repositório:RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Idioma:inglês
OAI Identifier:oai:riunet.upv.es:10251/230613
Acesso em linha:https://riunet.upv.es/handle/10251/230613
Access Level:Acceso aberto
Palavra-chave:Flysch formations
Geomechanical
Characterization
Discontinuities
Geological Strength Index
Artificial Neural Network
Support Vector Machine
Descrição
Resumo:[EN] This work proposes a quantitative estimation of the Geological Strength Index (GSI) in flysch rock masses based on data gathered from geomechanical stations, i.e. Rock Quality Designation (RQD), and spacing, persistence, aperture and roughness of the discontinuities. Artificial Intelligence (AI) techniques are explored as a tool to set such estimation. Particularly, Artificial Neural Networks (ANN) and Support Vector Machine (SVM) techniques are used, and a total of 40200 AI models are developed, investigating the optimum value of different hyperparameters. Data for training and testing such models comes from an intensive geological-geotechnical investigation conducted on 33 flysch outcrops of Late Cretaceous of age in a 100 km2 area belonging to the Basque Arc in northern Spain. This formation consists of a thick package of more than 700 m of interbedded marls and marly limestones, changing to quartz-rich clastic turbidites towards the top of the succession. The results show that the AI techniques can satisfactorily estimate the GSI of flysch rock masses from the characterization of their discontinuities. ANN show to yield better performance than SVM, and the best ANN models provide a superior performance to estimate the GSI than existing expressions based on the Rock Mass Rating (RMR). These ANN models reach very high values of R2 (very close to 1), and low values of the RMSE, being therefore suitable for their use by practitioners.