Analysis of data identification and generation methods for data-driven approaches in solid mechanics
Speaker: Nour Hachem
Date: Thursday 3rd of December 2026, 1:15pm
Abstract:
The Data-Driven Identification (DDI) method estimates the mechanical response of material samples without assuming a predefined constitutive model. DDI solves an inverse problem based on the principle that similar strains should correspond to similar stresses, while ensuring mechanical equilibrium. This work focuses on the development and theoretical validation of the DDI method. First, a rigorous mathematical and numerical analysis is provided, including the study of the uniqueness criterion, a new implementation with a proof of convergence, and an assessment of conditioning and convergence rates. Second, the case of ill-posed DDI problems is addressed, and corrective strategies are introduced to obtain valid solutions. Finally, essential tools and practical guidelines for reliable implementation are developed, providing a reference for future research and applications. These contributions establish a rigorous theoretical foundation for the DDI method, enhancing its robustness and applicability while enabling safer use through essential tools and a clearer understanding of its practical limitations.
- Details
- Category: Seminar
Speaker: Pedram Massoudi
Date: Monday 23rd of November 2026, 1:15pm
Abstract:
Prise en compte de la forte proportion de valeurs nulles dans les données horaires de précipitations
Prise en compte de la non-stationnarité latitudinale des précipitations et des températures
Corégionalisation avec l’altitude pour les températures et avec les données satellitaires pour les précipitations
- Details
- Category: Seminar
Speaker: Clémence Pavageau
Date: Thursday 8th of October 2026, 1:15pm
Abstract:
Low-enthalpy geothermal energy, characterized by temperatures below 90°C and exploitation at shallow depths (i.e. normally less than 500m), represents a sustainable energy alternative for residential and industrial heating and cooling, while offering a reduced environmental impact. However, the sustainable exploitation of these resources requires a thorough understanding of reservoir behavior and accurate monitoring of the thermal evolution of aquifers. Electrical geophysics provides valuable insights into subsurface thermal processes; therefore, understanding the coupled evolution of heat and electrical charge across multiple scales is of significant interest.
This study, first, presents an experimental approach to establish the relationship between the electrical resistivity of water-saturated rocks and temperature variations with simple yet realistic assumptions and then presents a numerical counterpart. The experimental study was conducted in the laboratory using a cylindrical water-saturated sandstone sample containing a central axial conduit through which water was circulated in a closed loop. The experimental protocol consisted of injecting either hot or cold water through the conduit to induce heating and/or cooling scenarios within the sample. Temperature was monitored using thermocouples positioned at the inlet and outlet of the water circuit, as well as at two radial locations within the rock sample. Electrical resistivity measurements were acquired using a Syscal Junior resistivity meter with a Wenner–Schlumberger electrode array. Measurements were recorded over roughly 30-minute periods, enabling the analysis of the coupled evolution of fluid temperature and the electrical conductivity of the saturated porous medium during both heating and cooling experiments. Through this, bulk resistivity evolution of the quadrupole was measured directly and verified later with the model.
The experimental scenarios were simulated using direct numerical simulation of coupled flow and heat transfer in porous media with electrical current density distribution. The petrophysical model of Waxman and Smits (1968), applied to saturated porous media, in which temperature primarily affects the electrical conductivity of the pore water is used to link temperature effects to electrical conductivity of the DC current injected. We assume and show that the temperature dependence primarily resulted from temperature-induced changes in ionic mobility and fluid viscosity, resulting in the evolution of water electrical conductivity measured directly in the experimental phase. Within the temperature range characteristic of low-enthalpy geothermal systems, this relationship can be considered approximately linear. The surface conductivity effects have secondary importance and can be expressed through the temperature dependence of equivalent conductance of exchange cations (i.e. parameter B).
The experimental and numerical results validate the assumption on the evolution of the electrical conductivity of the saturated Vosges sandstone as a function of temperature. Laboratory validation and calibration of this model provide the basis for its integration into a numerical framework capable of reproducing the experimental observations. At a larger scale, this approach could be extended to Electrical Resistivity Tomography (ERT) surveys, providing a non-invasive tool for in situ thermal monitoring of geothermal reservoirs.
- Details
- Category: Seminar
Modélisation et interdisciplinarité : application aux systèmes énergétiques en Afrique subsaharienne
Speaker: Théo Chamarande
Date: Thursday 19th of November 2026, 1:15pm
- Details
- Category: Seminar
Speaker: Hasnain Khan
Date: Friday 6th of November 2026, 1:15pm
Abstract:
Artificial intelligence (AI) and machine learning are increasingly used to model complex
geotechnical and subsurface systems, where the subsurface material behaviour depends on multiple
interacting factors and available experimental data are often limited or heterogeneous. This
presentation will introduce my research at the intersection of AI, data-driven modelling, and
geomechanics, highlighting applications to the prediction and interpretation of soil behaviour and
subsurface properties.
My doctoral research focused on predicting and analyzing the strength development of cement-
treated soils using machine learning and explainable AI approaches. My current postdoctoral
research extends these approaches towards the transfer properties of compacted sand–clay
mixtures, particularly hydraulic conductivity and gas permeability in the context of deep geological
waste disposal. These studies investigate how data-driven models can predict material behaviour
from experimental data, while addressing challenges related to data availability and model
interpretability.
Building on this work, I will discuss future research perspectives involving interpretable and
physically informed machine learning, uncertainty assessment, and the integration of data-driven
approaches with numerical modelling. I will also explore potential synergies between these
directions and the research activities of the RING team and GeoRessources.
- Details
- Category: Seminar
