Juan Pablo Gaviria-Cardona

Diagnosing lithium‐ion battery degradation is a crucial part of managing energy storage systems. Recent research has explored ultrasonic testing for non‐invasive health assessment as an alternative to traditional, time‐consuming, electrical‐only methods. Assessing the state of health is vital for determining quality at end of ‘first’ life, with retired batteries at 70–80 % health still holding value for secondlife applications. Over the coming years, tens of GWh of salvaged batteries will hit the market, requiring rapid noninvasive methods to classify retired batteries according to their state of health. This study uses a 64 – element ultrasonic array to obtain mid‐band quantitative ultrasound spectroscopy parameters – including mid‐band fit, spectral slope, and intercept – from circumferential waves around cylindrical batteries. Thirteen cylindrical cells were used to evaluate the methodology: three pristine and ten retired from the same source. The mid‐band fit showed the ability to track the state of charge and discriminate between the state of health levels in accelerated degradation experiments both with pristine batteries, and also with recovered secondlife batteries with unknown historical use. Linear‐array ultrasonic transducers, coupled with quantitative spectral parameters, show promise for future non‐destructive battery health screening methods, offering valuable insights for the emerging used battery market.

Analysis of Electrochemical Impedance Spectroscopy (EIS) data for electrochemical systems often consists of defining an Equivalent Circuit Model (ECM) using expert knowledge and then optimizing the model parameters to deconvolute various resistance, capacitive, inductive, or diffusion responses. For small data sets, this procedure can be conducted manually; however, it is not feasible to manually define a proper ECM for extensive data sets with a wide range of EIS responses. Automatic identification of an ECM would substantially accelerate the analysis of large sets of EIS data. We showcase machine learning methods to classify the ECMs of 9,300 impedance spectra provided by QuantumScape for the BatteryDEV hackathon. The best-performing approach is a gradient-boosted tree model utilizing a library to automatically generate features, followed by a random forest model using the raw spectral data. A convolutional neural network using boolean images of Nyquist representations is presented as an alternative, although it achieves a lower accuracy. We publish the data and open source the associated code. The approaches described in this article can serve as benchmarks for further studies. A key remaining challenge is the identifiability of the labels, underlined by the model performances and the comparison of misclassified spectra.