CT Image-Based PCA Reconstruction Residual Analysis for Structural Anomaly Diagnosis of Lithium-Ion Batteries

Lithium-ion batteries in electric vehicles operate under complex conditions, including frequent charge–discharge cycling, high-power operation, fast charging, temperature variations, and mechanical vibration, which can contribute to progressive internal structural degradation and associated safety risks. However, conventional electrochemical parameters have limited capability to directly characterize localized structural changes inside batteries. To address this issue, a structural anomaly diagnosis method based on principal component analysis (PCA) reconstruction residuals of X-ray computed tomography (CT) projection images is proposed. A cylindrical lithium iron phosphate cell subjected to accelerated high-rate cycling was used for evaluation, and projection images from a fresh cell of the same type were used to construct the PCA normal-structure model. The central region of each projection image was extracted as the region of interest. Pixel-wise squared reconstruction residuals were further analyzed to characterize localized structural deviations. Among the candidate residual-intensity and spatial-distribution features, the most discriminative feature achieved an absolute standardized mean difference of 9.332 between the pre-cycling and 100-cycle feature-evaluation datasets. When the fixed decision rule was applied without model retraining or threshold adjustment, consistent structural-state classification was observed across the angular projection views acquired after 200, 300, and 400 cycles. Sensitivity analysis and comparison with reference methods further supported the robustness of the observed feature separation under the present experimental setting. The proposed method reduces the dependence on complete three-dimensional CT reconstruction and extensive anomalous-sample annotation, demonstrating the feasibility of an interpretable CT-projection-based approach for nondestructive assessment of internal battery structural states.