Abstract: Wire arc additive manufacturing has become an important method for the overall manufacturing of heterogeneous and heterostructured high-performance materials and complex components. However, during this process, the heat input fluctuates due to the influence of process parameters and arc stability, and the molten pool morphology is highly sensitive to the heat input and process. Abnormal molten pool morphology may lead to defects such as pores, collapse, and discontinuity in the weld bead during the additive manufacturing process. To achieve the effective identification of forming defects in wire arc additive manufacturing, a defect diagnosis method driven by multi-modal data fusion was proposed, which could achieve the effective diagnosis of various defects. Firstly, molten pool images and electrical signals during the wire arc additive manufacturing process were collected through experiments, and key feature parameters were extracted using feature extraction methods. Secondly, a multi-modal data fusion algorithm was proposed to evaluate, screen, and fuse the extracted feature parameters to obtain the optimal feature vector. In addition, a defect diagnosis model based on a support vector machine (SVM) was constructed, and the hyperparameters of the model were optimized via the grey wolf-cuckoo joint optimization algorithm. The model was used to classify and identify the test set samples. The results indicate that the average classification accuracy of the test set reaches 99.38% with a good classification effect. Furthermore, the effectiveness of the method was verified through comparative experiments.