Elastomers are one of the most widely used engineering materials that find its applications in varied temperatures. When exposed to prolonged periods of these temperatures, these materials undergo heat-aging, progressively altering their failure behavior. Moreover, as these components endure heat-aging at different temperatures and durations, they are often utilized in various operational conditions, impacting their quintessential applications and failure. To predict elastomer failure under such circumstances, the initial phase of the study involved developing several phenomenological closed-form expressions. These expressions delineate the influence of thermal history and current operational temperature on the reference bulk response and fracture stress (for unaged rubber at room temperature). The proposed model can hence handle various real-world temperature-time profiles captured by sensors installed in engineering components during its application. The project is a collaboration between SnT and the industrial partner, SISTO Armaturen SA.
Team
Dr. Lars Beex
Prof. Stéphane Bordas
Reference
Behnke, R. and Kaliske, M., 2018. Numerical modeling of thermal aging in steady state rolling tires. International Journal of Non-Linear Mechanics, 103, pp.145-153.
Loew, P.J., Peters, B. and Beex, L.A., 2019. Rate-dependent phase-field damage modeling of rubber and its experimental parameter identification. Journal of the Mechanics and Physics of Solids, 127, pp.266-294.
In the past decade, domain-specific languages (DSLs) have emerged as a powerful tool for straightforward and expressive formulation of systems of partial differential equations (PDEs) in a variational setting. Yet, many important problems in solid mechanics involve non-trivial constitutive models that are difficult to express in variational form such as plasticity, multiscale and neural-network-based constitutive models. It is therefore challenging to express these problems in DSLs that work at the variational form level. We introduce a framework for FEniCSx / DOLFINx that allows for the straightforward implementation of a wide range of constitutive models through the application of automatic differentiation technique and Unified Form Language (UFL), the DSL developed as a part of the FEniCSx Project. The application of the framework is demonstrated by solving three elastoplasticity problems for von Mises, Mohr-Coulomb and neural-network-based constitutive models using UFL, Numba, JAX and TensorFlow software.
Key-words: constitutive models, external operator, automated finite element solvers, FEniCSx
Team
Collaborators: Dr. Jack S. Hale (Legato), Prof. Corrado Maurini (Sorbonne Université, France), Dr. Jérémy Bleyer (École des Ponts ParisTech, France)
PhD researcher: Andrey Latyshev (Legato)
Reference
A. Latyshev, J. Bleyer, J.S. Hale, C. Maurini. A framework for expressing general constitutive models in FEniCSx. Submitted to CSMA 2024.
Project 2 description