| Abstract Scope |
Welding spatter is most likely to occur during GMAW. The welding spatters increase the cost because of waste of filler metal and clean time. The issue is caused by inadequate welding parameters such as welding current, voltage, shielding gas, and wire feeding speed. It can be used as an indicator of poor welding quality.
In this study, algorithms were developed to automatically count spatter based on the weld bead images for MIG welding. The image can be obtained with a personal mobile phone camera instead of a commercial one. The software pipeline to quantify the numbers and area of spatters was proposed in this work. Algorithms that integrate conventional image operations, and convolutional neural network (CNN) classification models were developed to recognize the location and size of the spatters.
The algorithm was tested in the lab environment. The quantity and density distribution of spatter was reported based on the calculation. Finally, the calculated result from software was compared with the weight of spatters. The software can be further applied to evaluate the quality of the weld. |