Session: 13-06-01: Multiscale Models and Experimental Techniques for Composite Materials and Structures I

Paper Number: 173892

Multiscale Reduced-Order Modeling of Lithium-Ion Battery Cell Under Mechanical Loading 

The growing demand for safer lithium-ion batteries (LIBs) under mechanical abuse scenarios necessitates the development of predictive tools that overcome the limitations of costly and time-consuming experimental testing. Recent efforts have shifted toward numerical simulations to evaluate the mechanical response of LIBs in such extreme conditions. Microstructure-based direct numerical simulations (DNS), which account for a fully resolved LIB microstructure together with relevant constitutive laws for each constituent, provide valuable insights into the response of LIB under different loading conditions. The microstructure typically considers the layered separator, cathode, and anode, with current collectors embedded in the electrodes, while the constitutive laws consider relevant deformation and damage mechanisms. However, DNS remains computationally expensive due to the intrinsic multiscale architecture of LIBs and the nonlinear, heterogeneous behavior of their constituent materials, which often limits it to RVE simulations and prevents its efficient upscaling to the cell or module scale.

This study presents an efficient multiscale reduced-order modeling (ROM) framework designed to simulate the mechanical behavior of LIBs subjected to mechanical abuse conditions before thermal and electrochemical failures. The framework is based on the Eigendeformation-based Reduced Order Homogenization Model (EHM), which operates in a two-scale setting and focuses on reducing the computational cost at the microscale. EHM partitions the microstructure into a few sub-domains (also known as parts) and precomputes coefficient tensors, including each part’s localization tensor and the interaction tensors between parts. By assuming a uniform strain response across each part, a reduced-order nonlinear system can be solved to obtain part-wise responses, thereby replacing the DNS RVE simulation and achieving high computational efficiency with moderately low error levels.

The current ROM for LIB accounts for the anisotropic elastoplastic behavior of the metallic current collector, continuum damage in the polymer separator, and pressure-sensitive yielding of the granular materials in the electrodes. To address the pronounced heterogeneity in constitutive responses across different components, we reformulate the eigenstrain-based reduced-order homogenization formulation, allowing all different constituents to be incorporated into a single set of ROM equations. This enhancement enables the model to accurately represent distinct mechanical behaviors within LIB architectures. The model is first verified against the DNS of a LIB RVE, and then adopted to study the response of the LIB cell under 3-point bending and indentation tests. The proposed approach achieves a balance between computational efficiency and physical fidelity, enabling efficient multiscale simulations of LIB cells and modules while preserving critical damage and deformation mechanisms.

Presenting Author: Mohammad Hasaninia University of Wyoming

Presenting Author Biography: Mohammad Hasaninia is a Ph.D. student in the Computations for Advanced Materials and Manufacturing Laboratory (CAMML) at the University of Wyoming, under the supervision of Dr. Xiang Zhang. His research focuses on multiscale modeling and reduced-order simulation techniques, with applications in lithium-ion batteries and composite materials. At CAMML, he is developing efficient computational frameworks to simulate the mechanical behavior of lithium-ion batteries under abuse conditions, integrating advanced constitutive models such as anisotropic plasticity, continuum damage, and phase-field approaches into reduced-order homogenization schemes. He is also involved in the ROM-based multiscale analysis of woven composites to understand the relationship between microstructure and macroscale performance.

Authors:

Mohammad Hasaninia University of Wyoming
Qiang Guo University of Wyoming
Yanyu Chen University of Georgia
Xiang Zhang University of Wyoming