A Data-Driven Model for Prediction of Diffusion-Limited Oxidation in Elastomers With the Presence of Antioxidant

Abstract The impact of oxygen on elastomers, commonly referred to as oxidation, stands out as a critical aging mechanism. The interplay between thermal influences and antioxidants complicates the oxidation of polymers, rendering it a multifaceted process. In many cases, oxygen’s impact extends beyond the samples’ surface layer. This in-depth aging phenomenon is commonly known as diffusion-limited oxidation (DLO) and is mainly associated with oxygen’s diffusion-reaction behavior with the elastomer network. This leads to an irreversible change in the material properties and reduced usability. Research has demonstrated that stabilizers play a crucial role in influencing oxidation depth profiles. To accurately predict the behavior of aged elastomer components, a detailed understanding of oxygen absorption and diffusion into the material during thermo-oxidative aging is required. Here, we develop a physics-informed machine-learned engine capable of representing damage in the polymer matrix and simulating the heterogeneous oxidation processes in elastomers. On the one hand, oxygen uptake is simulated by a neural network model. On the other hand, further diffusion of oxygen into the material is described by a model based on Fick’s law, in which the reaction rates of oxygen with polymers and antioxidants are simulated by the neural network. The machine-learned (ML) model of oxygen uptake and DLO are benchmarked against the experimental data of various elastomers, which are able to predict heterogeneous degradation, permeation, and oxygen consumption in the course of aging.