A Physics-informed Assembly of Feed-Forward Neural Network Engines to Predict Inelasticity in Cross-Linked Polymers

TL;DR

This paper introduces a physics-informed, assembly-based neural network framework for predicting inelastic behavior in cross-linked polymers, addressing high-dimensional data challenges and improving accuracy, speed, and interpretability over traditional methods.

Contribution

It develops a reduced-order, multi-agent neural network assembly leveraging physics principles to enhance material modeling in complex loading conditions.

Findings

Outperforms traditional constitutive laws in training data efficiency.

Achieves higher accuracy and faster training speeds.

Effectively captures all loading modes with limited experimental data.

Abstract

In solid mechanics, Data-driven approaches are widely considered as the new paradigm that can overcome the classic problems of constitutive models such as limiting hypothesis, complexity, and high dependence on training data. However, implementation of machine-learned approaches in material modeling has been modest due to the high-dimensionality of the data space, significant size of missing data, and limited convergence. This work proposes a framework to hire concepts from polymer science, statistical physics, and continuum mechanics to provide super-constrained machine-learning techniques of reduced-order to overcome many of the existing difficulties. Using a sequential order-reduction, we have simplified the 3D stress-strain tensor mapping problem into a limited number of super-constrained 1D mapping problems. Next, we introduce an assembly of multiple replicated Neural Network agents to systematically classify those mapping problems into a few categories, all of which are replications of a few distinct agent types. By capturing all loading modes through a simplified set of disperse experimental data, the proposed hybrid assembly of agents provides a new generation of machine learned approaches that simply outperforms most constitutive laws in training data volume, training speed, and accuracy even in complicated loading scenarios. Also, it avoids low interpretability of conventional AI-based models.