A Convex Route to Thermomechanics: Learning Internal Energy and Dissipation

Dimensions overall score 7.0

GitHub Code Pulse

No public code linked for this paper yet.

• The proposed framework uses internal energy and a dissipation potential as primary constitutive functions instead of the classical Helmholtz energy formulation.method 95%
• Thermodynamic admissibility of the learned models is guaranteed by construction through the use of input convex neural networks.technical 90%
• The framework enables thermodynamically consistent modeling without requiring entropy data by inferring entropy internally through constitutive relations.method 85%
• Objectivity, material symmetry, and normalization are embedded directly into the neural network architecture.technical 90%
• The learned models accurately capture the underlying constitutive behavior of soft tissues and filled rubbers in fully coupled thermomechanical responses.result 80%
• The framework uses a specific neural network architecture where activation functions must be convex and monotonically non-decreasing to maintain convexity properties.technical 95%
• The method is specifically applied to materials without preferred directions or internal variables.limitation 90%
• An auxiliary loss function enforces consistency between the thermodynamic state law and the entropy predicted by an auxiliary MLP.method 85%

Competitive landscape

Competitor map is still being generated for this paper. Enable generation or check back soon.

Startup potential card