Objective Laser powder bed fusion (LPBF) is an additive manufacturing (AM) process that has the advantages of forming complex-shaped parts and cutting costs. It is widely used in the aerospace, medical equipment, weapons manufacturing, and other industries. However, in the LPBF process, the material powder is repeatedly heated and melted under the effect of laser energy and then cooled and solidified, which facilitates the formation of a large thermal gradient and thermal stress in the parts, leading to warping deformation. This type of deformation significantly affects the dimensional accuracy and mechanical properties of parts. By combining sensor signal acquisition with data analysis, deformation defects can be detected during AM to reduce production costs and improve the quality of formed parts. The radiant light signal of the molten pool is sensitive to the thickness of the powder layer during the LPBF process, which may reflect the warping deformation that has already occurred. It is also correlated with the temperature of the molten pool, reflecting the peak temperature at that location, and is related to the temperature field of the sample. Therefore, it has the potential to monitor the thermal stress during warping deformation. To study the relationship between thermal stress-induced warping deformation and the radiant light signal of the molten pool, a method for monitoring warping deformation in the LPBF process by acquiring the radiant light signal of the molten pool is explored in this study. In this study, an overhanging sample is formed during the experiment, and the radiation signal of the molten pool is collected and analyzed. The results show that the radiant light signal can not only monitor warping deformation but also reflect formation process of warping deformation to a certain extent. Methods To collect and compare the radiation light signal of the molten pool during the forming process of the warped and normal samples, T-shaped overhanging structure samples are formed (Fig.2), and five samples with three different support structures and sizes are designed for the experiment (Table 1). In this process, three sensors collect the radiation intensity signals from the molten pool, and an upper computer records the coordinate data of the laser spots (Fig.1). After data alignment, each light intensity value corresponds to the coordinates of the laser spot during scanning. To further explain the variation trend of the light intensity signal along the long side ( Y -direction) of the sample, the scanning section of the sample is divided into regions, and the average light intensity of each region is calculated. Three measurement points are selected on the sample, and the heights of the measurement points relative to the substrate plane are measured using a coordinate apparatus. Results and Discussions No evident warping deformation is observed in the forming process of samples S80-1 and S80-2, whereas the warping deformations of samples S25-1, S25-2, and S20 are larger (Fig.6). This result indicates that samples with smaller support areas are prone to warping deformation; however, no noticeable linear correlation is observed. The normal samples S80-1 and S80-2 produce a larger average light intensity at both ends, with a minimum value of 0.93 V, while warped samples S25 -1, S25-2 and S20-1 produce lower light intensity at the same area (Fig.7). This phenomenon indicates that sample warping can be distinguished from the light signal of the molten pool. The light intensity distribution of the first overhanging layer is different between the warped and normal samples. The light intensity of the warped sample in the region where the corresponding lower layer is solid is significantly higher than that in other regions, forming a "wave peak" in the curve (Fig.8) . The above phenomena indicate a correlation between the radiant intensity distribution and peak temperature at the corresponding position and reveal that the evolution trend of the light intensity between the layers of the samples with the same geometric structure. The light intensity of the normal sample fluctuates more between layers, whereas that of the warped sample fluctuates less (Fig.9). Conclusions In this study, three types of overhanging samples with different structures are formed, and the radiation light signal of the molten pool is collected. Combined with sample deformation measurements and statistical methods, the data are analyzed, and the following conclusions are obtained: 1) In the layer after warping deformation, the light intensity of the warped specimen decreases significantly in the warped region, while the distribution of the light intensity of the normal specimen is uniform without a notable gradient. 2) For the warped specimen, when the overhanging layer has just been formed, and the deformation has not yet occurred, the light intensity "crest" corresponding to the central solid region of the specimen is quite different from the light intensity in other regions of the layer. 3) The interlayer evolution trends of the light-intensity values of the warped and normal samples are different. With an increase in the number of formed layers, the influence of the overhanging structure on the light intensity signal gradually decreases, and the light intensity tends to stabilize after the fifth layer. 4) A sample with a smaller support area is more likely to produce warping deformation, but no notable linear correlation exists between these two factors.
