Summary Homogenization and finding the constitutive model of jellyroll in cylindrical lithium‐ion batteries can be challenging because of their form factor. Taking samples out of the original jellyroll wounding or compressing cell assembly in its cylindrical coordinates are two possibilities for measuring the homogenized lateral strength of the cell. However, the former causes loss of accuracy due to changing constraints and electrolyte environment, and the latter requires complex fixtures that are not readily available or even practical to manufacture. Various approaches have been suggested by researchers to circumvent the above difficulties and allow the extraction of hardening curves. However, the precision of those approaches diminishes when the cells are under global compression vs local punch deformations. In this study, an updated homogenization method is established, using a lateral compression test on the jellyroll. The homogenization method is based on the assumption that the circular cross‐section of the jellyroll under compression is deformed in an elliptical shape. Then the principle of virtual work is used to extract the hardening curve. To validate the above characterization model, isotropic and anisotropic finite element models were developed using crushable foam and modified honeycomb material models from the LS‐DYNA library. Four sets of cell‐level experiments were performed on cylindrical batteries using custom‐designed fixtures, including flat lateral compression, rod indentation, hemispherical punch, and three‐point bending. The voltage and surface temperature of the batteries were measured to capture the onset of short circuit during the tests. Comparison of the simulation results confirmed that the proposed homogenization method and the FE models can predict the behavior of cylindrical lithium‐ion batteries with much higher accuracy compared to the currently available methods presented in the literature.

Li‐ion batteries are widely used in electric vehicles (EVs) propulsion. Therefore, ensuring their safety under mechanical abuse and accidental loads is a major challenge for the industry. To get a better understanding of the battery behavior in such cases, material calibration and computational modeling of the battery cells are essential. This paper aims to develop a universal homogenized model for an 18,650 cell that can predict cell behavior under both axial and lateral loading cases as well as three‐point bending. Previous homogenized models presented in the literature have covered one or two of these cases, but none have been validated in all these three major loading scenarios. To achieve this, precise shell casing and jellyroll material calibrations were performed. The features included in this universal model are (I) uncoupled calibration of axial and lateral properties for the cylindrical jellyroll from experiments performed in these two loading directions and employment of an anisotropic crushable foam model to simulate these features, (II) using Hill's anisotropic yield criteria and modified Mohr–Coulomb fracture criteria for the shell casing. The universal model developed here was able to predict the response of the cell in all lateral, axial, and bending loading scenarios. A comparison of this model with the previously developed isotropic models shows the special advantage of the new model in cases of axial loading and bending. However, for lateral compression cases, even the isotropic model provides a very close prediction. The experiments used for this study were all performed on fresh discharged cells under quasi‐static loading.

Numerical simulations of heterogeneous structures like battery modules of electric vehicles are challenging due to the various length scales involved in it. Even with the latest computing technology, it is impossible to simulate the crash scene of the full vehicle resolving all length scales. Such hurdles have prevented manufacturers to understand the mechanical response of battery packs in vehicle crash scenarios. In this work, the problem of multiple length scales was solved using the RVE technique based on homogenization theory. An appropriate representative volume element was identified, and a 3D FE model was developed. Classical first-order boundary conditions were used in this research work. The RVE was subjected to several macroscopic deformations, and its response was obtained. The homogenized material properties were computed from the obtained responses, and a material model available in LS-Dyna’s material library was selected and calibrated to describe the nonlinear multiaxial behavior of the homogenized battery module at the macroscale. For validation, the USABC Crush and Drop Tests were simulated for the detailed and homogenized battery modules. The main output of this research is a robust and computationally efficient tool enabling satisfactory integration of a battery pack model to the vehicle for crash simulations, eliminating the need to simulate micro details at macro length scales. This approach significantly reduced the computational cost. For example, for a Drop Test simulation, the homogenized model reduced the simulation time from 40 hours (detailed model) to about 3 minutes, while maintaining a high precision ( R 2 = 0.9871 ) in predicting the load-displacement response. The system level modeling will enable the stakeholders to perform efficient optimization and safety evaluation for full-scale crashworthiness of electric vehicles.