Automatic Detection and Quantification of Incomplete Penetration in TIG Welding Through Segmentation and Morphological Image Processing of Thermographs

Infrared Thermography is an advanced Non Destructive Evaluation technique based on the detection of infrared radiation. It is best suited for online weld monitoring to produce defect free welds. The most important step involved in online weld monitoring is to develop an automated algorithm for weld defect detection and quantification. Incomplete Penetration occurs when penetration of the weld metal into the joint is insufficient compared to that specified by the welding symbol. It is caused by too low welding current or too high welding speed or incorrect joint geometry. It can contribute to failure as stress raisers and least resistant path. It manifests itself as variation in the hottest region in a thermograph. The paper aims at developing an image-processing algorithm for automatic identification and quantification of Incomplete Penetration in Tungsten Inert Gas (TIG) welding from the acquired thermograph. The developed image processing algorithm acquires the input thermograph, performs color to gray level conversion, edge detection by Sobel filter mask and morphological processing such as dilation, erosion and region filling to detect and quantify the defect. The feature vectors used for quantifying the defect are area, major axis length and minor axis length.