A Data-Driven Model to Predict Constitutive and Failure Behavior of Elastomers Considering the Strain Rate, Temperature, and Filler Ratio

Abstract

Abstract This new machine-learned (ML) constitutive model for elastomers has been developed to capture the dependence of elastomer behavior on loading conditions such as strain rate and temperature, as well as compound morphology factors such as filler percentage and crosslink density. It is based on our recent new generation of machine-learning algorithms known as conditional neural networks (CondNNs) Ghaderi et al. (2020, “A Physics-Informed Assembly of Feed-Forward Neural Network Engines to Predict Inelasticity in Cross-Linked Polymers,” Polymers, 12(11), p. 2628), and uses data-infused knowledge-driven machine-learned surrogate functions to describe the quasi-static response of polymer batches in cross-linked elastomers. The model reduces the 3D stress-strain mapping space into a 1D space, and this order reduction significantly reduces the training cost by minimizing the search space. It is capable of considering the effects of loading conditions such as strain rate, temperature, and filler percentage in different deformation states, as well as enjoying a high training speed and accuracy even in complicated loading scenarios. It can be used for advanced implementations in finite element programs due to its computing efficiency, simplicity, correctness, and interpretability. It is applicable to a variety of soft materials, including soft robotics, soft digital materials (DMs), hydrogels, and adhesives. This model has a distinct advantage over existing phenomenological models as it can capture strain rate and temperature dependency in a much more comprehensive way.