Cell finishing represents a time-critical bottleneck in lithium-ion battery cell manufacturing, yet decisive quality information is typically confirmed only after extended aging and end-of-line testing. This delay ties up production resources and limits early defect detection. To address this limitation, this study proposes a multimodal feature fusion approach combining incremental capacity analysis (ICA) and electrochemical impedance spectroscopy (EIS) for early-stage capacity prediction. Physically interpretable features, including ICA peak characteristics and EIS-derived impedance parameters, are extracted from formation data acquired under production-representative conditions. Ensemble-based regression models are trained and evaluated, with XGBoost yielding the lowest prediction error. A complementarity analysis demonstrates that combining ICA and EIS features improves prediction accuracy compared to unimodal approaches. Feature importance analysis confirms that both modalities contribute substantially, with ohmic resistance and ICA peak height among the most influential predictors. The results validate the complementary nature of ICA and EIS for early quality assessment and support inline decision-making for cell grading and scrap reduction in battery cell manufacturing.
Li et al. (Thu,) studied this question.