Software
Filippo Masi
Research software
Open-source implementations and research code accompanying my work in data-driven constitutive modelling, computational mechanics and geomechanics (GitHub profile)
Constitutive transport learning
Learning inelastic constitutive models directly from stress–strain data under hard thermodynamic constraints. The framework combines neural free-energy potentials with state-dependent generalised transport operators, enforcing thermodynamic consistency by construction.
NICE — Neural Integration for Constitutive Equations
A deep-learning framework for the automatic discovery of constitutive equations from small, sparse and incomplete material-state observations. NICE formulates constitutive learning as an initial-value problem and uses neural differential equations to identify the evolution of material state variables.
Thermodynamics-based Artificial Neural Networks (TANN)
Reference implementation of thermodynamics-based neural architectures for constitutive modelling of inelastic materials. Thermodynamic relations are embedded directly in the neural architecture through automatic differentiation.
TANN-multiscale
Micromechanical simulations, training workflows and finite-element examples for multiscale modelling of inelastic materials with TANN. The repository includes tools for generating microscale data and examples coupling learned constitutive models with finite-element analyses.
Terraquake
Research code for the constitutive and hydrodynamic modelling of fault gouges and frictional dynamics, connecting material-scale behaviour with earthquake-like instabilities.
Teaching and hands-on material
Machine Learning for Physics (UGA)
Course material for Machine Learning for Physics in the M2 MSIAM programme at Université Grenoble Alpes, co-taught with Vincent Acary, Michael Arbel and Henri Leroy. The repository contains lecture slides and hands-on material on physics-informed learning, operator learning, structured models and differentiable solvers.
ALERT Geomaterials Doctoral School 2023 — Machine Learning in Geomechanics
Shared course repository for the 2023 ALERT Geomaterials Doctoral School on Machine Learning in Geomechanics. It collects lecture material and hands-on notebooks from the teaching team.
ALERT Olek Zienkiewicz School — Constitutive modelling meets Machine Learning
Hands-on notebooks and material for learning constitutive equations from experimental geomaterial data, prepared for the ALERT Olek Zienkiewicz Doctoral School.
CSMA Workshop — Deep Learning and Constitutive Modelling
Course material, notebooks and practical exercises on physics-informed data-driven constitutive modelling and Thermodynamics-based Artificial Neural Networks.