Over the past three decades, the energy density of lithium-ion batteries (LIB) has steadily increased to meet the growing demand for portable energy storage. This increase in energy density has made LIBs a potential safety hazard. In the worst case, a so-called thermal runaway in the LIB can lead to the destruction of the entire battery pack and its surroundings.
It is our goal to develop a virtual model of a LIB cell, that is capable of reproducing the LIB cells’ mechanical and thermal response under impact and crushing loads. Such a model would improve the prediction of safety and performance of LIB modules in a cost-effective manner. To more accurately model and predict how LIB cells fail when subjected to mechanical stresses and loads, a detailed, heterogenous finite element (FE) model will be developed. This model will capture the individual components of a single LIB cell with high resolution.
To achieve the required level of detail in the FE model, the material behavior of the jellyroll components, e.g. anode, cathode, and separator, was characterized. Both the anode and cathode foils are three-layer structures, with a metallic layer sandwiched between two active material layers, while the separator, typically made of a single- or multilayer polymer, prevents short circuits between the anode and cathode. The metallic layer in the anode foil is typically made of copper, while the active material is usually a graphite-based compound. In the cathode foil, the metallic layer is usually made of aluminum, while the active material is generally made of metal oxides.
A series of experimental tests were performed to determine the mechanical properties of the individual foils within the jellyroll. These tests, including macro- and nanoindentation as well as tensile tests, were performed on cleaned foils extracted from commercially available LIB cells. The focus was on indentation tests to evaluate the compressive behavior of the active material layers in the anode and cathode foils, as this was considered to be a critical factor influencing the force response of the battery to impact and crushing. FE models were then developed to replicate the experimental setups, such as simulating macro-indentation using an axisymmetric model. Macro-indentation was performed using a cylindrical indenter with a flat tip. During the simulations, the foils were compressed between a flat indenter head and a bed to replicate the experimental setup. The force-displacement curves resulting from the simulations were then compared with the experimental data to evaluate their accuracy and consistency. The material model parameters in the FE models were iteratively refined until the simulation results matched the experimental data within an acceptable threshold. This process was repeated for all test types to capture the responses of the material models to different load types. Finally, a new patch of experimental data was compared to the simulation results to validate the models and ensure the accurate representation of the mechanical behavior of the foils.
The next phase involves building and training a surrogate neural network model to replicate the behavior of individual jellyroll components. These models will be trained using data generated from the validated FE models. In addition to serving as a preparatory step for developing a surrogate model of the entire LIB cell, this approach will also verify, whether the overall behavior of the cell can be accurately reproduced by integrating the surrogate models of its individual components. This would provide valuable practice for implementing surrogate models in simulated battery modules.