Batteries, Vol. 8, Pages thirty-five: Method for In-Operando Contaminants of Lithium Ion Batteries for Prediction of Impurity-Induced Non-Obvious Cell Damage
Batteries doi: 10. 3390/batteries8040035
Writers: Patrick Höschele Simon Franz Heindl Bernd Schneider Wolfgang Sinz Christian Ellersdorfer
The safety of lithium-ion batteries within electrified automobiles plays an important function. Hazards can arise through contaminated batteries resulting through non-obvious damages or insufficient production processes. A organized examination requires experimental strategies to provoke a defined contamination. Two prerequisites were necessary: First, the extent and type of contamination should be determinable to exclude randomness. Second, specimens should work properly before the contaminants, enabling realistic behavior. In this study, two fresh methods were developed to allow for the first time a controlled and reproducible application of water or oxygen into 11 single-layer full cells (Li4Ti5O12/LiCoO2) utilized as specimens during electrical cycling. Electrochemical impedance spectroscopy was used to continuously monitor the specimens and in order to fit the parameters of an equivalent circuit model (ECM). For the 1st time, these parameters had been used to calibrate the machine-learning algorithm which was able to predict the contamination state. A decision tree was calibrated with the ECM parameters of eight specimens (training data) and was validated by predicting the contamination state associated with the three remaining individuals (test data). The prediction quality proved the functionality of classification algorithms to monitor for contaminations or non-obvious battery damage after production and during use. It could be an integral part of battery management systems that increases vehicle safety.