Experimental Testing and Numerical Modelling of Thermal Runaway in 18650 Lithium-Ion Batteries

Lithium-ion batteries (LIBs) are essential for commercial energy storage due to their high energy density and efficiency, supporting the global shift toward greener energy. However, damaged or abused LIBs can undergo thermal runaway (TR), a rapid exothermic process causing high temperatures in seconds, and often resistant to conventional fire suppression systems. A deeper understanding of TR is vital for developing effective mitigation strategies. This research investigates TR behaviour in commercially available 18650 cylindrical LIBs and develops numerical models using ANSYS Fluent, validated through experimental data. As the first step in such testing in South Africa, this study establishes a foundation for local LIB research. Preliminary experiments tested six LIBs from two brands (LASA and Samsung) at varying capacities, with three at 50 % state of charge (SOC) and three at 100 % SOC per brand. Results revealed manufacturer influence on failure modes: LASA cells typically vented jet flames through positive cap safety vents, while Samsung cells ruptured, ejecting the combined internal layers and electrolyte (often referred to as the “jelly roll”). Capacity differences showed no clear impact. SOC affected TR intensity, referring to nature and fire behaviour of the LIB, with 100% SOC cells causing more structural damage than 50 % SOC. The sporadic nature of TR highlighted challenges in obtaining repeatable data. Refined tests, using improved methods and Samsung LIBs at 100 % SOC, included three cell-level and five pack-level (4x3 arrangement) experiments. Cell-level results showed maximum temperatures of 787 to 847 °C, ignition temperatures of 209 to 240 °C, and ignition times of 296 to 394 seconds, providing sufficient results for calibration of the numerical models. Pack-level tests indicated a 20 % propagation probability, with unmitigated TR likely consuming the entire pack. Numerical 2D modelling achieved comparable agreement with experimental data, serving as an initial tool for surface temperature prediction and risk assessment, while laying groundwork for 3D propagation models. Combined experimental and numerical insights provide a calibrated framework for characterising TR at cell and pack levels, reducing testing costs and risks, and advancing safer energy storage systems.

Grantor dc:publisher

Stellenbosch : Stellenbosch University

Year dc:date.issued

2026

Author dc:creator

• Pretorius, Lourens Badenhorst

Advisors dc:contributor.advisor

• Walls, Richard Shaun
• Devine, Courtney

Language dc:language.iso

en

Repository record dc:identifier.uri

https://scholar.sun.ac.za/handle/10019.1/135925

OAI identifier oai:identifier

oai:scholar.sun.ac.za:10019.1/135925