HaomingLuo / AgentFEM-DENIM

AgentFEM DENIM

DENIM means Discrete-Energy Neural Internal-variable Model . It is a compact gray-box constitutive model for path-dependent small-strain J2 plasticity. Elasticity, yield geometry, associative flow, plastic incompressibility, non-negative plastic increments and the implicit return map remain explicit. Neural components represent only the unknown isotropic and kinematic hardening closure.

Why this checkpoint is an incomplete-physics test

The AgentFEM reference material has three kinematic-memory channels and a piecewise-linear isotropic-hardening table. This checkpoint has only two memory channels and receives neither the reference equations nor parameters. The second learned state must close two unresolved reference time scales.

The model has 918 trainable parameters. On two completely held-out non-proportional path families:

Coarse/fine path RMSE was 0.944/0.958 MPa. In 12-element notched-bar tests, cyclic and monotonic reaction relative-L2 errors were 0.636% and 0.500%. A severe cyclic case passed with one global trust-region fallback.

Use

import torch
from src.t2_graybox_discrete_energy import DENIM, rollout

checkpoint = torch.load("denim.pt", map_location="cpu", weights_only=False)
model = DENIM(channels=2)
model.load_state_dict(checkpoint["state_dict"])
model.eval()

# strain: [batch, steps, 6], tensor-shear Voigt order xx,yy,zz,xy,yz,xz
result = rollout(strain, young, poisson, yield_stress, model)
stress = result["stress"]

See inference.py for a runnable example. config.json records the known material constants and architecture contract.

Scope and limitations

• fixed synthetic material; not experimental calibration;
• small-strain, rate-independent, isotropic J2 plasticity;
• trained with high-fidelity internal-state supervision;
• PyTorch material-point implementation, not yet a compiled production UMAT;
• implicit return mapping is slower than direct GRU inference;
• the severe cyclic structure test still needed one global fallback.

The associated data, source trajectories and evidence are available at AgentFEM-Material-Loading-Memory .