Multimodal fusion-based anomaly detection in the friction stir welding process

As a solid-state joining technology, Friction Stir Welding (FSW) has found widespread application in industries such as aerospace, automotive, and shipbuilding. However, fluctuations in process parameters, variations in equipment conditions, and material property differences in the welding process can lead to anomalies. To address this, this paper proposes an anomaly detection method based on multimodal data fusion, utilizing Principal Component Analysis (PCA) for dimensionality reduction, and combining Density-Based Spatial Clustering of Applications with Noise (DBSCAN) with Local Outlier Factor (LOF) for anomaly identification. Experimental results show that this method not only significantly improves anomaly detection accuracy but also exhibits strong robustness, enabling real-time monitoring of welding quality. This study provides an effective solution for anomaly detection in complex industrial processes and lays a foundation for quality assurance in intelligent manufacturing.