Abstract: To address the challenges of identifying internal defects in lithium-ion battery electrodes and the low efficiency of traditional ultrasonic testing, this study proposes an intelligent electrode defect identification method that integrates nonlinear ultrasonic guided waves with artificial intelligence. Four types of electrode samples are tested using a nonlinear ultrasonic guided wave measurement system. By applying the pulse inversion method, the fundamental wave (FW), second harmonic (SH), and nonlinear ultrasonic (NU) signals are separated to construct three distinct datasets. A one-dimensional convolutional neural network (1D-CNN) is employed for automatic feature extraction; the extracted features are then fed into three classifiers—random forest (RF), support vector machine (SVM), and gradient boosting machine (GBM)—to perform defect identification. Results show that the NU-based feature set achieves the best recognition performance. Specifically, GBM achieves an average recognition accuracy of 94.35% for electrode defects using this feature set. For granular contaminants, the recognition rate reaches 91%, representing an 11-percentage-point improvement over the FW-based feature set. The SH carries unique defect-related information not present in the FW, thereby effectively enhancing the detection capability for microscopic defects. The proposed NU-CNN+GBM algorithm outperforms each individual classifier in recognition accuracy, providing technical support for online, non-destructive testing of lithium-ion battery electrodes.