Detecting electrolyte wetting defects in lithium-ion batteries based on electrochemical impedance spectroscopy

Electrolyte wetting defects generated during the wetting process compromise the safety and consistency of lithium-ion batteries (LIBs), making reliable detection at the manufacturing stage essential. Here, we report a data-driven detection method for wetting defects based on electrochemical impedance spectroscopy (EIS). We perform EIS measurements on pouch cells after injection but before formation, using ultrasonic C-scan results as non-destructive labels. Analysis of a 100-cell dataset identifies physics-informed EIS response features, including high-frequency ohmic resistance shifts, and mid-frequency charge-transfer perturbations. A few-shot learning algorithm with an asymmetric prototypical network classifies these defects. Cross-validation demonstrates that this method achieves a detection accuracy of 98.0% ± 2.0% under minute-scale measurements, confirming feasibility under limited data conditions. Compared with imaging techniques, the EIS-driven approach offers superior throughput and lower cost. This method provides a practical pathway for online full inspection in large-scale manufacturing, enhancing yield, lifetime, and safety.