Data-driven scheme for two-scale chemo-mechanical coupled problem for heterogeneous materials under transient chemical diffusion

This study developed a data-driven computational scheme of multiscale multiphysics coupled problem for heterogeneous materials under transient chemical diffusion. Firstly, the data-driven scheme decouples the multiscale problem into two stages, i.e., offline generation of material constitutive behavior database by microscopic simulation, and online macroscopic boundary value problem (BVP) based on the data-driven mechanical scheme. In addition, data-driven scheme converts the multiphysics coupled constitutive relationship into the material conjugate quantities in database which avoids the poor convergence in multiscale multiphysics coupled calculation. Then, a two-scale chemo-mechanical model with a simple microstructure is validated in two cases: diffusion-induced mechanical deformation and deformation-enhanced chemical diffusion. The results show that the proposed framework can characterize the chemo-mechanical coupled behavior with good convergence and accuracy, which is more efficient than the full-scale finite element simulation. Finally, this data-driven scheme is applied to analyze the multilayer porosity structures for lithium-ion battery cathodes. The results show that cathodes with gradient porosity design can improve the utilization of active materials and enhance the effective capacity.