Abstract:
The flash welding mooring chain has become the mainstream of the high-quality mooring chain market in the shipbuilding industry. In this present work, we proposed a new methodology for the online quality assessment of flash welding. First, the various sensor signals in the real-world welding process are collected in this work. Second, a novel spatiotemporal warping algorithm is proposed to quantify the spatiotemporal dissimilarity of current and electrode position signals during flash welding processes. Third, the dissimilar distance matrix is embedded into the feature vector of the low dimensional space while the original dissimilar distance between the signals were preserved. Finally, the KNN classification algorithm and K-fold cross-validation are proposed to classify the embedded feature vectors (coordinates in low-dimensional space). Experimental results show that the proposed method in this paper can not only use the collected sensor signals for real-time welding quality assessment, but also effectively reveal the difference between the flash welding processes.