Operator Learning of Microstructure-Dependent Internal-Variable Models for Inelastic Homogenization: Approximation and Complexity
Organizations: Department of Computing and Mathematical Sciences, California Institute of Technology, Pasadena CA, 91125, USA. · Department of Applied Mathematics and Theoretical Physics, University of Cambridge, Wilberforce Road, Cambridge CB3 0WA, United Kingdom.
Abstract
We study the learning of homogenized constitutive laws for inelastic heterogeneous materials, aiming to replace costly concurrent unit-cell simulations with surrogates that generalize across loading histories and microstructures. We focus on the recurrent neural operator framework, motivated by the explicit form of homogenized Kelvin--Voigt linear viscoelasticity. In this framework, internal variables enable a Markovian approximation of loading-history dependence, and neural operators represent the microstructure-dependent constitutive response. For periodic Kelvin--Voigt viscoelasticity, we prove a universal approximation theorem showing that the model can approximate the homogenized constitutive law to any prescribed accuracy, uniformly in time and over a wide set of admissible strain trajectories and microstructural property fields, in two and three dimensions. This comprises a significant extension of the existing, somewhat limited, one-dimensional theory that uses explicit solutions for piecewise constant materials. Instead, in two and three dimensions, we adopt a more versatile approximation based on the spectral representation of the memory kernel that arises from homogenization. This also enables us to give an explicit bound on the internal-variable dimension, sufficient for approximation, exhibiting logarithmic growth in the inverse relative stress error and in material contrast. Numerical experiments on viscoelasticity and crystal plasticity complement the theory, illustrating the roles of internal-variable dimension and sample complexity in learning homogenized responses from unit-cell data. Theoretical results are formally verified in Lean 4.
Figures & tables
| Name | Input | Output | Physical Interpretation | Mathematical Interpretation |
| Strain Corrector Projectors and | Identify purely viscous or elastic microscopic corrector (negative) | Orthogonal Projectors | ||
| Strain Localization and | Determine purely viscous or elastic microscopic strain | Complementary Projectors | ||
| Strain Corrector Mismatch | Difference between purely viscous and elastic microscopic correctors | Difference of Projectors | ||
| Relaxation | Dictates the relaxation timescales | Riesz Representation Map |
| Name | Definition | Interpretation |
| Logarithmic Relaxation Span | Logarithm of the ratio of the largest to the smallest relaxation rate | |
| Viscosity Contrast | Ratio between the maximum and minimum viscosity | |
| Elasticity Contrast | Ratio between the maximum and minimum elasticity | |
| Relative Stress Error | Ratio between the stress error tolerance and the prescribed stress scale |
Operator Learning
Operator learning studies models that map functions or fields to functions, enabling fast prediction across scientific and dynamical systems.
See all 113 papers