Constitutive Kolmogorov-Arnold Networks in Finite Element Model Updating: Closed-Form Hyperelastic Strain Energies from Heterogeneous Experiments
Kian P. Abdolazizi, Kevin Linka, Christian J. Cyron
Abstract
Calibrating a hyperelastic constitutive model for a soft material traditionally poses two problems: an expert must choose its functional form, and several homogeneous tests must determine its parameters. Neural constitutive models solve the first, but either remain black boxes or draw on a library of candidate terms fixed in advance. Full-field measurements, such as those from digital image correlation (DIC), promise to solve the second, replacing those tests by a single heterogeneous one. Two gaps remain: flexibility during training with interpretability afterwards, and a way to train such a model on one heterogeneous test. Here we close both. We embed a constitutive Kolmogorov-Arnold network (CKAN) in finite element model updating (FEMU) and train it by minimizing the mismatch between the predicted and the measured kinematic fields and reaction forces. Pruning and symbolic regression then reduce the trained network to a compact closed-form strain energy that satisfies every constitutive requirement. On perforated natural-rubber membranes, the CKANs reproduce the DIC displacement fields with median errors between one and two percent, and transfer to a hole pattern withheld from training. On a perforated silicone plate, a CKAN trained on one stretch profile at a single load reproduces that profile and the measured reaction force to within a few percent. On an aortic specimen inflated under pressure, an anisotropic CKAN predicts the inflated three-dimensional shape. The discovered closed-form energies match the CKANs they replace on the training data and can be compared directly with established models.
Create a lesson
Related papers
Emergent Fatigue Dynamics from Foundational Interatomic Potentials
Jingye Tan, Benjamin A. Jasperson, Eric Rothchild et al.
Two distinct notions of speed in antiferromagnets
Paul Herrgen, Stephan Wust, Christin Schmitt et al.
Perfumer: A Large Language Model-based Agentic Framework Accelerating Structural Relaxation and Electronic Band Prediction for First-Principle Calculation
Zhiwei Li, Yuhao Zhou, Yikai Yang et al.
Unusual increase in electrical and thermal conductivity of severely deformed CrMnFeCoNi high-entropy alloy: Roles of disordering and nanostructuring in electron and phonon transport
Jacqueline Hidalgo-Jiménez, Payam Edalati, Md Amirul Islam et al.
Reward misalignment as an opportunity for materials discovery
Ryotaro Okabe, Shoki Kishida, Abhijatmedhi Chotrattanapituk et al.
Lattice dynamics study of (Bi1-xSbx)2Te3 compounds: Comparative compositional study of force constants and Grüneisen parameters
Baihong Sun, Chenxin Wei, Elissaios Stavrou