A method of fuzzy clustering identification for weld defects by magneto-optical imaging

Weldment of high strength steel (HSS) in laser welding was used as the research object, and a magneto-optical imaging detection method based on Faraday magnetic rotation effect was studied. By applying alternating-current power and changing the size of induced magnetic field of welds, a magneto-optical sensor was used to capture the magneto-optical images. The gray-level occurrence matrix (GLCM) texture features of the specific area of magneto-optical images were extracted and analyzed. For accurately detecting weld defects and classifying the type of defects, a fuzzy clustering identification model was established. Different calculation results of weld defects by adjusting the fuzzy index, the input characteristic numbers and the sample of weld defects of fuzzy c-mean clustering (FCM) were compared and the identification effect of weld defects were analyzed. Experimental results show that the fuzzy C-means clustering is effective for identification of the weld defects, which also has a better identification effect on cracks, incomplete penetrations, sags and different forms of same kinds of weld defects.