Home Publications Research Data Research Grants Equipment Inventions & Patents Awards Additional Research Activities Faculties & Institutions Research Areas

Validating a battery fault model for synthetic fault data generation for lithium-ion battery systems

Reiter A, Rosenmüller C, Lehner S, Sauer DU, Bohlen O (2026)

---

Publication Language: English

Publication Type: Journal article, Original article

Publication year: 2026

Journal

Book Volume: Part C

Journal Issue: 181

DOI: 10.1016/j.est.2026.124140.

Open Access Link: https://www.sciencedirect.com/science/article/pii/S2352152X26038041

Abstract

The availability of fault data is a key enabler for the development, selection and adaption of fault detection algorithms for lithium-ion battery systems. Model-based approaches to fault data generation offer a fast and accessible alternative to laboratory or field data but require thorough validation to ensure correct replication of the behavior during a fault condition. This study introduces a publicly available experimental dataset containing fault data obtained from measurements on a 6s3p fault emulation test bench for four different fault scenarios (cell connection fault, micro short circuit, inhomogeneous cell temperature fault and inhomogeneous accelerated degradation) as well as multi-fault scenarios. The fault emulation test bench was designed to optimize repeatability and mitigation of parasitic influences during the emulation of fault mechanisms. The experimental dataset was further used to parameterize and validate a battery fault model for synthetic fault data generation. The validation results show that the behavior of the system being subject to the different fault scenarios can be replicated qualitatively. Quantitatively, cell connection faults and micro short circuit scenarios were replicated with acceptable accuracy. Inhomogeneous cell temperature faults induced comparably small deviations of the electrical behavior, becoming minor versus the observed variation of the connecting resistances. The replication of the inhomogeneous accelerated degradation scenarios proved challenging and requires further enhancements of the model. Across all scenarios, significant connector degradation and increasing variations between assembly cycles over the duration of the campaign proved the main challenge to the validation of the fault model. In the last step, the application of the fault model for the calibration of a correlation-based fault detection algorithm and the generation of training data for a machine learning-based approach was shown. These exemplary applications underline the relevance and usability of the introduced process.

Authors with CRIS profile

Involved external institutions

How to cite

APA:

Reiter, A., Rosenmüller, C., Lehner, S., Sauer, D.U., & Bohlen, O. (2026). Validating a battery fault model for synthetic fault data generation for lithium-ion battery systems. Journal of Energy Storage, Part C(181). https://doi.org/10.1016/j.est.2026.124140.

MLA:

Reiter, Alexander, et al. "Validating a battery fault model for synthetic fault data generation for lithium-ion battery systems." Journal of Energy Storage Part C.181 (2026).

BibTeX: Download