Diagnostic-free onboard battery health assessment
Yunhong Che, Vivek N. Lam, Jinwook Rhyu, Joachim Schaeffer, Minsu Kim, Martin Z. Bazant, William C. Chueh, Richard D. Braatz · 2025
Read Diagnostic-free onboard battery health assessment by Yunhong Che, Vivek N. Lam, Jinwook Rhyu, Joachim Schaeffer, Minsu Kim, Martin Z. Bazant, William C. Chueh, Richard D. Braatz (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-07383.txt.
Introduction
Yunhong Che and co-authors judge a lithium battery's health from its ordinary driving data, skipping the special test cycles garages now run. Aging patterns vary wildly with use, and diagnostic cycles both take hours and bend the battery's own aging path. Cars need a health read that happens on board, from scraps of normal operation.
Their 2025 paper feeds those scraps into an encoder-decoder model whose hidden space is built to mean something physical: electrode states a battery scientist would name. Mechanistic constraints keep the readout honest, and slight fine-tuning adapts it across uses. On hundreds of cells cycled three different ways, the framework reads health and forecasts decay without history files or test bench time.
Catalog record
| CREATOR | Yunhong Che, Vivek N. Lam, Jinwook Rhyu, Joachim Schaeffer, Minsu Kim, Martin Z. Bazant, William C. Chueh, Richard D. Braatz |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 45 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-07383
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Diagnostic-free onboard battery health assessment Yunhong Che, Vivek N. Lam, Jinwook Rhyu, Joachim Schaeffer, Minsu Kim, Martin Z. Bazant, William C. Chueh, Richard D. Braatz Abstract Diverse usage patterns induce complex and variable aging behaviors in lithium-ion batteries, complicating accurate health diagnosis and prognosis. Separate diagnostic cycles are often used to untangle the battery’s current state of health from prior complex aging patterns. However, these same diagnostic cycles alter the battery’s degradation trajectory, are time-intensive, and cannot be practically performed in onboard applications. In this work, we leverage portions of operational measurements in combination with an interpretable machine learning model to enable rapid, onboard battery health diagnostics and prognostics without offline diagnostic testing and the requirement of historical data. We integrate mechanistic constraints within an encoder-decoder architecture to extract electrode states in a physically interpretable latent space and enable improved reconstruction of the degradation path. The health diagnosis model framework can be flexibly applied across diverse application interests with slight fine-tuning. We demonstrate the versatility of this model framework by applying it to three battery-cycling datasets consisting of 422 cells under different operating conditions, highlighting the utility of an interpretable diagnostic-free, onboard battery diagnosis and prognosis model. keywords Lithium-ion battery, health diagnosis and prognosis, battery aging reconstruction, interpretable machine learning, cycle life prediction † † equal-contributors: These authors contributed equally to this work. † † equal-contributors: These authors contributed equally to this work. 1 Introduction The growing trend of electrification has driven the wide application of lithium-ion batteries in electrified transportation and grid storage Dunn et al. (2011) ; Schaeffer et al. (2024) . One of the primary concerns during battery usage is degradation, leading to reduced capacity and power capabilities. Knowing a battery’s current and future state of health (SOH) is critical in optimizing its usage. However, the complex path-dependent nature of battery degradation results in diverse practical degradation patterns, posing challenges for accurate health diagnosis and prognosis Dubarry et al. (2018) ; Schaeffer et al. (2024) . Typically, a diagnostic cycle is employed to assess the SOH of the battery Edge et al. (2021) ; Lewerenz et al. (2019) ; van Vlijmen et al. (2023) ; Geslin et al. (2024) ; Zhuang et al. (2024) ; Rhyu et al. (2024) . This diagnostic cycle may include pulse tests to check resistance, low-rate capacity checkups, and other application-specific tests necessary to extract useful SOH metrics. One common SOH metric is the low-rate battery capacity. However, this metric alone is insufficient as multiple degradation pathways can lead to the same cell-level SOH, limiting the interpretative power of solely relying on cell-level capacity. To address this, electrode-specific capacities have been proposed and extensively studied as a means to further breakdown degradation into electrode-level phenomena, capturing more intricate aging behavior of lithium-ion batteries van Vlijmen et al. (2023) ; Geslin et al. (2024) ; Mohtat et al. (2019) ; Dey et al. (2020) ; Che et al. (2023) . This mechanistic model, known as differential voltage analysis (DVA), reconstructs the battery’s open-circuit voltage (OCV) by subtracting matched and rescaled half-cell OCV curves. Using these half-cell curves, DVA reveals key modes of battery aging such as loss of active materials (LAM) and loss of lithium inventory (LLI) Marongiu et al. (2016) ; Dubarry et al. (2012) ; Stadler et al. (2024) . While this analysis provides valuable in situ insights into aging mechanisms, it is typically performed on pseudo-OCV curves at rates of C/20 or lower taken during a diagnostic cycle. Furthermore, even at low rates, this adjustment does not fully capture the pseudo-OCV curve due to changes in the half-cell overpotential, inhomogeneities of lithiation, and other aging effects Fath et al. (2019) ; Schmitt et al. (2022) ; Sieg et al. (2020) . While diagnostic cycles provide valuable insights into the battery’s SOH, such as enabling DVA, they have significant drawbacks. Taking time away from aging to perform disruptive and time-intensive diagnostic cycles can drastically alter the trajectory of battery aging Lewerenz et al. (2019) . From a practical point of view, time-consuming diagnostic tests cannot be performed routinely on consumer electronics, leading to the need for onboard diagnostics that use operational data. In real-world applications, battery charging current rates (C-rates), state-of-charge (SOC) ranges/voltage windows, and load profiles can vary significantly Pozzato et al. (2023) ; Sulzer et al. (2021) ; Schaeffer et al. (2024) , necessitating data-driven models that are robust across diverse operating conditions. Moreover, a significant number of lithium-ion batteries are expected to retire from electric vehicles (EVs) in the near future, but performing a lengthy diagnostic cycle to evaluate their performance will be costly Harper et al. (2019) ; Ciez and Whitacre (2019) ; Xu et al. (2020) . To facilitate more intelligent repackaging and reuse, rapid evaluation of battery performance has garnered increased attention. Finally, different C-rates-based diagnostic tests are required in diverse applications, necessitating the model’s suitability under varying scenarios. In recent years, the advanced development of machine learning models has led to their implementation in battery health estimations and predictions Schaeffer et al. (2024) ; Roman et al. (2021) ; Thelen et al. (2024) ; Severson et al. (2019) ; Nozarijouybari and Fathy (2024) ; Hu et al. (2020) . Data-driven predictions of the OCV curve and incremental capacity (IC) curve have shown potential for onboard health diagnosis by using measured current and voltage Tian et al. (2021) ; Tang et al. (2021) . Deep neural networks are used predominantly in these data-driven models for the prediction of the battery OCV curve or charging curve due to their good nonlinear mapping capabilities Hofmann et al. (2024) ; Tian et al. (2021) ; Su et al. (2023) ; Ko and Chen (2024) . However, despite achieving high prediction accuracy, the input data is usually fixed to a full voltage curve under constant current between the voltage limits. This limits generalization to real applications with partial SOC cycling, or dynamic load profiles. Additionally, the “black box” nature of these models limits their interpretability for a battery engineer to better understand the failure modes of a given battery. Consequently, mechanistic models are still required for subsequent aging mechanism analysis, separate from machine learning. To improve the interpretability of battery health models, promising ways for combining physical models with machine learning have been discussed Aykol et al. (2021) ; Navidi et al. (2024) . One approach involves extracting physically interpretable features from mechanistic models to inform machine learning models used for battery modeling, health estimation, and lifetime prediction van Vlijmen et al. (2023) ; Geslin et al. (2024) ; Weng et al. (2021) ; Tu et al. (2023) . However, extraction of these physical features poses challenges, and typically requires specific and time-consuming diagnostic tests. Physics-informed neural networks (PINNs), which integrate governing equations from mechanistic models to constrain machine learning models, show promise in improving both interpretability and the speed of solving governing equations in complex models Karniadakis et al. (2021) ; Borah et al. (2024) ; Huang et al. (2024) . However, not only is constructing explicit models with partial derivatives challenging, but current models also do not fully capture highly interconnected degradation processes during battery aging Wang et al. (2024) . Here, we propose an interpretable model