Cheap ultrasound sensors costing just dollars or pennies could reveal defects or damage in the lithium-ion batteries used by smartphones, laptops and electric vehicles 鈥� before those issues cause safety problems or lead to expensive recalls.
鈥淚t鈥檚 a really powerful tool to see what鈥檚 inside these lithium-ion cells in a low-cost and low-computational way,鈥� says at the University of Sheffield in the UK.
Lithium-ion batteries power many modern technologies, but they occasionally catch fire. The lithium-ion batteries in e-bikes offer a high-profile example:聽in New York alone . Structural defects inside the batteries may trigger such safety problems, but X-ray scanning every battery to search for imperfections is prohibitively expensive.
Advertisement
Commercially available ultrasound sensors offer a cheaper alternative. They send ultrasonic waves through objects and receive the reflected wave signals, which can be used to reconstruct the object鈥檚 internal structure. However, the reflected ultrasonic signals are typically so complex that it is difficult to interpret them.
Copley and his colleagues solved this problem using a computer algorithm known as a genetic algorithm. This began by generating a random guess of the battery鈥檚 internal structure and predicting what ultrasonic waveforms the design would produce. The algorithm then continued to 鈥渆volve鈥� its idea of the battery鈥檚 structure until the predicted waveform matched the waveform from the actual battery.
The researchers assessed how accurately the algorithm had replicated the battery鈥檚 internal structure by X-ray scanning the battery. This confirmed the algorithm is sufficiently reliable that manufacturers could use it 鈥� rather than X-rays 鈥� to check their batteries.
Experiments showed that the algorithm could make especially accurate predictions when given some prior knowledge of the battery鈥檚 probable structure and material properties. The best predictive performance estimated the speeds of waves 鈥� an indication of the type of material they are passing through 鈥� to within 3 or 4 per cent of actual speed and the layer thicknesses to within 7 or 8 per cent of the actual thicknesses.
This technique鈥檚 predictive power could prove even more potent when feeding into an AI system based on machine learning, says Copley.
Inexpensive ultrasound sensors could prove transformative for battery monitoring because they are 鈥渧ery fast, potentially very low cost and can be integrated into a battery production line鈥�, says at University College London, who was not involved in the research. By comparison, X-ray scans can provide certain details but are relatively slow and more expensive.
Next, the researchers plan to test how well the technique can detect swelling or other structural changes that could indicate battery damage, says , also at the University of Sheffield and part of the team. 鈥淲e鈥檇 like to use it on real deformed, degraded batteries,鈥� he says.
Journal of Energy Storage
