Hierarchical Bayesian physics-augmented neural networks for populations of hyperelastic laws: specimen variability and uncertainty from scarce, noisy data
Organizations: Technical University of Munich, Professorship of Data-driven Materials Modeling, School of Engineering and Design, Boltzmannstr. 15, 85748 Garching, Germany · Munich Data Science Institute (MDSI), Garching, Germany
Abstract
The hyperelastic law of a material such as soft tissue differs between specimens, and often only a few specimens can be tested. Physics-augmented neural networks (PANNs) learn such laws from data without a prescribed functional form while keeping them physically admissible, and Bayesian versions of related models quantify the uncertainty of the law of one specimen or of a pooled group of specimens. A pooled fit, however, attributes the differences between specimens to measurement noise or to uncertainty about a mean law, and neither describes a new specimen. We first develop a Bayesian PANN for one specimen, whose priors keep the learned law plausible for few, noisy data. We then extend it to a hierarchical Bayesian PANN, a nonlinear mixed-effects model in which all specimens share learned polyconvex features and each specimen is a draw from a population of laws. Features, specimens, and population are inferred jointly by variational inference, with a separate model of the measurement noise, so that the variability between specimens is not mistaken for noise, the population carries its own uncertainty, and every specimen drawn from it is polyconvex. The population predicts the response of new specimens with credible intervals, without data or from a partial test, and assigns them to cohorts with quantified uncertainty, all without a prescribed form of the law. On synthetic populations, it recovers the population from few specimens and predicts new specimens more reliably than three published methods. On Treloar's rubber and on healthy and aneurysmal aortas, it reaches the accuracy of a deterministic PANN and assigns no aorta specimen to the wrong cohort with confidence, although it underestimates the uncertainty of untested curves.
Figures & tables
| model class | calibration | use | |||||
| M1 no law | M2 admissible | C1 affordable | C2 noise | C3 population | U1 partial | U2 cohort | |
| method | in advance | draws | data | separated | uncertainty | test | assignment |
| Bayesian-EUCLID [ 20 ] | (✓) | mostly | – | – | – | – | – |
| Sparse Bayesian input convex network [ 22 ] | ✓ | ✓ | – | – | – | – | – |
| Bayesian CANN [ 21 ] | (✓) | – | – | – | – | – | – |
| Mixed effects, fixed law [ 27 , 28 ] | – | parameters only | ✓ | ✓ | ✓ | ✓ | ✓ |
| section | task | data | load cases | material | reference |
|---|---|---|---|---|---|
| 4.2.1 | one specimen | synthetic: S-shape law ( 30 ) | uniaxial and equibiaxial stretch (PS held out) | compressible, isotropic | deterministic PANN |
| 4.2.2 | one specimen | real: Treloar’s rubber [ 40 ] | UT, ET (PS held out) | incompressible, isotropic | benchmark of [ 41 ] |
| 4.3 | population | synthetic: incompressible upturn family ( 31 ) | UT, ET, PS | incompressible, isotropic | competitors |
| 4.4 | new specimen | synthetic: both upturn families | UT only, partially | both | competitors, no population |
| 4.5 | cohort assignment | real: aortas, 6 healthy and 6 aneurysmal | biaxial: circumferential, axial | incompressible, transversely isotropic | pooled laws (null test) |
| load case | , compressible ( ) | , incompressible | measured |
|---|---|---|---|
| UT | |||
| ET | |||
| PS |
| relative error | coverage | band width | |||
|---|---|---|---|---|---|
| mean | median | mean | median | (kPa) | |
| ours, unlabelled population | 0.12 | 0.08 | 0.61 | 0.62 | 1.6 |
| Bayesian CANN | 0.33 | 0.19 | 0.89 | 1.00 | 11.8 |
| no population | 0.15 | 0.10 | 0.74 | 0.85 | 2.2 |
Appendix figures & tables5 assets
Supplementary material from the paper’s appendix.
Appendix
| Hyperparameter | Value |
|---|---|
| Network and priors (both models) | |
| Hidden layers neurons | |
| Activation | softplus |
| Inputs | standardised invariants ( 4 ) |
| Prior of the hidden layers | fan-in widened by ( 7 ) |
| Prior of the (mean) output layer | ; ( 12 ) |
| Model or procedure | Hardware | Time |
| Bayesian PANN, one fit (incl. MAP starts) | GPU | min |
| Hierarchical Bayesian PANN, one fit ( , steps) | GPU | min |
| Prediction from a partial test (SIR, proposals), tests of one population | GPU | s |
| Cohort assignment, leave-one-out (aortas, specimens, both cohorts) | GPU | min |
| Deterministic PANN, restarts | GPU | min |
| Known-law ceiling (NUTS, chains) | CPU, cores | s |
Physics-Informed ML
Physics-informed machine learning combines data-driven models with physical laws to solve, predict, and infer scientific systems.
See all 906 papers