Large-format prismatic lithium-ion cells have high surface-to-volume ratios, and the positioning of the tabs can make the temperature distribution highly non-uniform across the cell surface [1]. When cells are interconnected, the wiring configuration can induce non-uniform pack-level current distributions, leading to varied cell usage [2]. Non-uniformity of either type is a problem because local thermal runaway can occur at ‘hot spots’ and spread to other regions of a cell, or to other cells in a pack [3]. Even in cases of non-catastrophic failure, non-uniformity leads both to ineffective usage of active materials at the cell level and to an unknown distribution of health across cells salvaged from retired packs. These inefficiencies can be leveraged in a positive way, however, because they allow the correlation of external measurements such as temperature distribution and voltage with internal material properties.

A 3-D physics-based model based on the Newman-Tobias porous-electrode theory [4] was developed, augmented with a local energy balance, and implemented in COMSOL Multiphysics^® to predict electrical and thermal behaviour. Various material properties corresponding to relevant observable lock-in thermography signatures can be extracted by fitting the model to experimental data.

Square-wave current signals with amplitudes ranging from 1 C to 4 C and periods ranging from 20 s to 100 s were applied to A123 20 Ah LiFePO₄ cells. These signals keep the average state of charge (SOC) relatively constant, eliminating concerns about how material properties change with SOC [5]. Temperature-distribution and voltage data are gathered together as the battery response reaches a periodic steady state. Parameter sets are extracted by fitting model output to the lock-in thermography data from the most strenuous test (i.e. highest C-rate and longest period). The model has been validated by using parameters extracted from the most strenuous test to simulate tests with different amplitude and period.

Characterization of parameter vectors by the above technique across a cell’s available charge states produces a database that can be compared to data from other batteries to diagnose their states. Figures 1A and 1B show lock-in thermography data and model best-fits at the most strenuous testing condition (4 C amplitude and 100 s period about a mean SOC of 50%). Figures 1C and 1D show similar information from the same cell around a different mean SOC of 30%. Table 1 shows the two parameter sets used to simulate these results, giving us insight into SOC-dependent electrical and thermal parameter values.

State-of-health analysis of a cell can be performed by gathering a broader property database from lock-in-thermography, encompassing parameter vectors from various points in a battery’s cycling history. This ‘health fingerprint’ could be used to quickly diagnose cells salvaged from larger packs with unknown cycling history.

1. Robinson, J.B., et al., Detection of Internal Defects in Lithium-Ion Batteries Using Lock-in Thermography. ECS Electrochemistry Letters, 2015. 4(9): p. A106-109.
2. Offer, G., et al., Module design and fault diagnosis in electric vehicle batteries. Journal of Power Sources, 2012. 206: p. 383-392.
3. Bandhauer, T.M., S. Garimella, and T.F. Fuller, A Critical Review of Thermal Issues in Lithium-Ion Batteries. Journal of The Electrochemical Society, 2011. 158(3): p. R1-R25.
4. Newman, J.S. and C.W. Tobias, Theoretical Analysis of Current Distribution in Porous Electrodes. Journal of the Electrochemical Society, 1962. 109(12): p. 1183-1191.
5. Maleki, H., et al., Thermal Stability Studies of Li-Ion Cells and Components. Journal of The Electrochemical Society, 1999. 146(9): p. 3223-3229.