Mechanical property tests and physics-informed data-driven modeling of viscoelastic materials subjected to thermal-oxidative aging

Abstract

The purpose of this paper is to investigate and characterize the performance changes of viscoelastic materials (VEMs) in viscoelastic (VE) devices used for structure vibration control due to the influence of aggressive environmental conditions during long-term service. To this end, a series of thermal-oxidative (TO) aging experiments are conducted on the developed VEMs, variation of microstructure and macroscopic mechanical behaviors of TO aged VEMs are tested and analyzed. The molecular chain network of VEMs that can be simplified into elastic chain network and free chain network is described by the architecture of microsphere model, a physics-informed data-driven model is proposed to reflect the mechanical properties of TO aged VEMs, in which the physics-informed neural network (PINN) theory is incorporated to describe the statistical characteristics of molecular chains and the reaction rate of TO aging. Based on the experimental data, implementation and validation of the proposed model are conducted on the mechanical properties of VEMs under varying aging levels. The experimental results suggest that TO aging could alter molecular chain structure of VEMs from the microscopic level, leading to an increase in crosslinking degree and modulus, and the degeneration of deformation ability and energy-dissipation capacity for VEMs. The verifications reveal that the proposed model possesses high accuracy in characterizing the mechanical properties and energy-dissipation capability of VEMs under different aging levels, which preliminarily proofs the potential of PINN theory in reflecting the complex mechanical behaviors of VEMs.