Online monitoring and penetration recognition in all-position TIG welding of nuclear power pipeline

Abstract

In the construction of a nuclear island, a nuclear power pipeline is the “aorta” of the nuclear power plant, which performs the critical role of a nuclear safety protection barrier. By automating welding, nuclear power pipeline can maintain high quality and reduce the need for skilled welders. We developed an all-position TIG welding quality monitoring system to address the low internal forming quality assessment accuracy for nuclear power pipeline welding. An algorithm for data processing extracted the features of the pipeline groove, back molten pool, and solidified weld via sensors mounted on the internal crawler mechanism. Based on the process parameters collected during welding, a penetration state recognition model was constructed using a random forest. The results show that, compared with the decision tree and support vector machine, the random forest model has an accurate penetration recognition performance and can realize automatic assessment of pipeline welding quality.