In depth understanding of the formation of condensation rings in laser spot welding - machine learning and molecular dynamics simulation
Researchers from the Pacific Northwest National Laboratory and Johns Hopkins University have reported that machine learning and molecular dynamics simulations can help to gain a deeper understanding of the formation of condensation rings in laser spot welding. The related paper titled 'Machine learning and molecular dynamics simulations aided insights into conditioned ring formation in laser spot welding' was published in Scientific Reports.
The formation of a condensation ring can serve as a benchmark for evaluating the efficiency and quality of welding in welding processes. The formation of a condensation ring is crucial because the resulting condensate will precipitate into the powder, thereby altering the quality of the uncured powder. This study investigated the complex relationship between alloy composition, vapor pressure, and condensation ring thickness through two-dimensional micrographs. To investigate this process, laser spot welding was performed on 9 different alloys, and the formation of the inner diameter and condensation ring during spot welding was studied. By utilizing machine learning models, experimental observations, and molecular dynamics simulations, researchers have explored the fundamental factors that affect the formation of condensation rings. These models are good at predicting the diameter of solder joints and the thickness of condensation rings, determining that laser power is the main determining factor of solder joint diameter, followed by physical properties such as hardness and density. On the contrary, for the thickness of the condensation ring, vapor pressure and melting point descriptors are always the most important factors, which have been validated in all models. The molecular dynamics simulation of nickel chromium alloy elucidated the vaporization kinetics and confirmed the role of vapor pressure in controlling surface vaporization. The discovery by researchers emphasizes the crucial influence of vapor pressure and melting point descriptors in the formation of condensation rings. The fusion of machine learning prediction and simulation analysis elucidates the dominant role of these descriptors, providing important insights for alloy design strategies to minimize the formation of condensation rings during laser welding processes.
Figure 1 (a) Photo of laser platform through protective glass. (b) High speed static photos of molten pool and condensate luminescence, as well as weak ionized plasma above the workpiece.
Figure 2 Top view optical micrograph of laser welded joints in SS 316 cold working (a), Ni (b), SS 316 hot working (c), Ni70Cr30 (d), Hastelloy alloy (e), Ni80Cr20 (f), Mg5Al95 (g), Ni90Cr10 (h), Inconel (i), and Ni95Cr5 alloy (j). The scale bar in the figure is 415 μ m.
Figure 3 The effect of image processing algorithms on the welding points and condensation rings of 316 stainless steel cold working. (a) Convert the original image of the solder joint into (b) grayscale image, where the inner and outer edges are identified as the solder joint diameter and the condenser ring diameter, respectively. The scale bar in the figure is 415 μ m.
Figure 4 (a) Relationship between solder joint diameter of different alloys and (b) variation of condensing ring thickness with laser power.
Figure 5 SEM-SEI images at different magnifications show a significant increase in surface condensation formation in the 20Cr sample compared to the 5Cr sample. The energy spectrum confirms that the surface condensate (orange) is mainly chromium oxide.
Figure 6 shows the Pearson correlation coefficients between all descriptor pairs in the predicted dataset of solder joint diameter and condensation ring thickness. Blue and red represent strong negative correlation (-1) and positive correlation (1), respectively.
Figure 7: Distribution of elements on the dataset.
Figure 8 Ranking of descriptor importance for predicting solder joint diameter using the CatBoost model.
Figure 9: Importance ranking of descriptors for predicting condensation ring thickness using CatBoost model.
Figure 10: Molecular dynamics model of Ni70Cr30 alloy laser spot welding process (a) t=0ps, (b) t=20ps, (c) t=30ps, (d) t=40 ps, (e) t=50ps, describing the process of atoms vaporizing from the surface to the vacuum zone. The laser heated chromatographic column displays (f) t=0ps, (g) t=20ps, (h) t=30ps, (i) t=40ps, and (j) t=50ps. The melting process continues until all atoms in the chromatographic column escape beyond the boundaries of the cubic volume, leaving no particles in the column volume.
Figure 11 (a) Changes in the number of vaporized atoms on the surface of Ni70Cr30 (black), Ni80Cr20 (blue), and Ni100Cr00 (red) over time. (b) The functional relationship between the total system pressure and time in the simulation process of laser spot welding of the same alloy.
The basic factors causing the formation of condensation rings during laser spot welding were studied using machine learning and molecular dynamics simulations, and experimental observations. A machine learning model was trained on the collected experimental data using chemical composition descriptors, physical quantities of alloys, and process parameters to predict the solder joint diameter and condensation ring thickness of various alloys. The model shows that the formation of the condensation ring is mainly influenced by the vapor pressure of the metal in the alloy. The higher the vapor pressure of the constituent metal, the greater the evaporation of atoms in the molten pool, resulting in thicker condensation rings. Molecular dynamics simulations of Ni70Cr30, Ni80Cr20, and pure Ni confirmed this hypothesis, with Ni70Cr30 having the highest number of atoms vaporized from the surface and pure Ni having the lowest, due to the higher vapor pressure of chromium (approximately 458 Pa) and the lower vapor pressure of nickel (0.44 Pa). These insights greatly promote people's understanding of the intrinsic mechanism of condensation ring formation during welding processes, emphasizing the key role of steam pressure as a controlling factor in such phenomena.
Source: Yangtze River Delta Laser Alliance
