A thermodynamically consistent physics-informed deep learning material model for short fiber/polymer nanocomposites
This provides a thermodynamically consistent material model for nanocomposites, addressing a domain-specific problem in materials science, but it is incremental as it builds on existing physics-informed deep learning approaches.
The paper tackled modeling the viscoelastic-viscoplastic behavior of short fiber-reinforced nanoparticle-filled epoxies under ambient conditions, resulting in a physics-informed deep learning model that accurately predicts mechanical behavior for different volume fractions and hygrothermal conditions.
This work proposes a physics-informed deep learning (PIDL)-based constitutive model for investigating the viscoelastic-viscoplastic behavior of short fiber-reinforced nanoparticle-filled epoxies under various ambient conditions. The deep-learning model is trained to enforce thermodynamic principles, leading to a thermodynamically consistent constitutive model. To accomplish this, a long short-term memory network is combined with a feed-forward neural network to predict internal variables required for characterizing the internal dissipation of the nanocomposite materials. In addition, another feed-forward neural network is used to indicate the free-energy function, which enables defining the thermodynamic state of the entire system. The PIDL model is initially developed for the three-dimensional case by generating synthetic data from a classical constitutive model. The model is then trained by extracting the data directly from cyclic loading-unloading experimental tests. Numerical examples show that the PIDL model can accurately predict the mechanical behavior of epoxy-based nanocomposites for different volume fractions of fibers and nanoparticles under various hygrothermal conditions.