A study on the dynamic behavior of plasma plume in laser deep penetration welding of A304 stainless steel based on three-probe detection

In laser deep penetration welding process, the generation of laser-induced plasma exhibits strong randomness and time-varying characteristic. The electrical signal acquisition devices can obtain large samples of thermodynamic information with high acquisition frequency, which is suitable for an integrated analysis of plasma behavior by electrical signals. A new type of passive three-channel acquisition device was designed to acquire plasma electrical signals containing information about the thermodynamics of the plasma. After processing and analyzing the electrical signals, the temperature change rate of plasma eruption was primarily distributed in the range of 600-3000 K/ms, and the temperature change rate of plasma sway was mainly distributed in the range of 60-700 K/ms. Based on the significant difference in the temperature change rate of plasma eruption and sway, the dynamic behavior of the plasma was accurately clustered by Gaussian Mixture Modeling (GMM), a method of clustering analysis, which showed the best clustering performance compared to K-means and spectral clustering. The results showed that the frequency of plasma eruption was mainly concentrated in the range of 640-1420 Hz, and the frequency of plasma sway was mainly in the range of 620-1040 Hz. The ranges in the frequency of plasma sway were smaller than plasma eruption.