Classification of molten pool modes in variable polarity plasma arc welding based on acoustic signature

摘要

The relationships between the acoustic signal and the modes of the welding pool such as no-keyhole (melt-in), keyhole and cutting in variable polarity plasma arc welding (VPPAW) are investigated. Welch power spectral density (PSD) estimate is used for preprocessing the sound data. A neural network (NN) is used to distinguish the keyhole mode from the cutting mode. The results show that the keyhole mode can be distinguished from the cutting mode under the experiment conditions in this paper.