Time and again, ebike batteries that catch fire make the headlines. They cause fires in bike cellars, in bike shops and on the open road. Sometimes people are injured or even die as a result. In future, new types of fire detectors could warn of an acutely fire-prone battery earlier than before – by triggering an alarm when a very specific noise is heard.
Nowadays, burning ebike batteries are a worldwide problem. Particular attention has been focussed on the city of New York in recent years. In 2022, the New York Fire Department was called out a total of 216 times to extinguish lithium-ion batteries. In the end, there were 147 injuries and six fatalities. The following year, 2023, the figures rose again. In 268 such fires, 150 people suffered injuries and 18 people died.
Extremely dangerous dynamics
One reason for this devastating record lies in a specific aspect of fire development. If a lithium-ion battery catches fire, it reaches a temperature of up to 1,100 degrees Celsius within around one second, which is released in the form of a jet of flame during an explosion. Until the fire actually ignites, however, hardly any smoke escapes from the battery. Often the amount is not sufficient for a conventional smoke detector to register it. However, if no alarm is triggered, the warning only reaches people when the fire can hardly be contained.
Researchers analyse the phase before the fire
As is so often the case, good and bad – in this case aggravating and facilitating – are close together. When a lithium-ion battery catches fire, it is not only the fire development that differs from other fires, but also the associated noises. Or rather, certain noises triggered before the fire starts. Researchers from Xi’an University of Science and Technology (XUST) in China and the National Institute of Standards and Technology (NIST) in the USA have been working on this together.
They took a close look at what happens when a lithium-ion battery heats up too much or is damaged by the effects of other forces. This is because the individual substances inside the battery react chemically with each other. The resulting pressure build-up causes the battery to swell at the level of the battery cells. However, the material cannot escape anywhere. The battery cells themselves are usually enclosed in a housing that prevents this swelling process. In principle, a safety valve integrated into the housing is intended to equalise the pressure. In short, it breaks and the pressure that has built up inside escapes. However, this safety solution also has its limits, meaning that the entire battery may explode at any given time.
But as already mentioned, the safety valve of the battery cells breaks first. And the research team’s project revolves around this noise. It reliably signals that a lithium-ion battery will soon catch fire.
AI assists with detailed acoustic work
In a corresponding scientific publication, the team reports that it is unfortunately not a very distinctive sound. Rather, it could be relatively easily confused with other sounds from our daily lives. Therefore, a mere detector calibrated to this sound would be useless.
The researchers use special software to reliably isolate and analyse the sound. This recognises it and can successfully distinguish it from other, very similar-sounding noises. To achieve this result, the team worked with artificial intelligence. A specially created machine learning algorithm had to respond to the breaking safety valve. Over and over again. In experiments at the XUST, single-cell lithium-ion batteries were repeatedly caused to explode. A total of 1,330 acoustic samples were obtained during the thermal runaway in the early stages. These samples were compared with 1,128 acoustic data samples that sound similar, but do not reproduce breaking safety valves. As targeted ‘interference signals’, so to speak. In the end, a microphone placed close enough to an explosion was able to correctly detect the signal from the battery indicating its imminent thermal runaway in 94 per cent of cases.
New warning detector targeted for practical implementation
What sounds like a great success to us, the researchers see as a first milestone at best. They want to continue their work with tests on other battery types and use different microphones. In addition to the noise, there is also a time factor to verify. So far, the safety valves have given way after around two minutes, followed by the battery exploding. The aim now is to check whether this value also applies to other lithium-ion batteries.
In the foreseeable future, this research could lead to the development of a detector that clearly recognises when a lithium-ion battery is about to catch fire within a few minutes. This would be a significant benefit for bike shops, ebike fleet operators and private individuals. People could get to safety in good time. And from there to a safety system integrated into the battery, which sends a warning message to the ebike app, the road suddenly no longer seems so far.
Study:
Tam, W., Chen, J., Tang, W., Tong, Q., FANG, H. and Putorti Jr, A. (2024), Development of a Robust Early-Stage Thermal Runaway Detection Model for Lithium-ion Batteries, 13th Asia-Oceania Symposium on Fire Science and Technology, Daegu, KR, [online], https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=958183 (Accessed December 6, 2024)
Protective bag for ebike batteries
Pictures: Institut für Schadenverhütung und Schadenforschung; Tavo Romann; Fahrer Berlin GmbH