Surface Feature Extraction And Comparative Analysis of The Molten Pool in Laser Intelligent Manufacturing Process Monitoring

01 Introduction
Laser melting-based manufacturing technologies (including laser welding, additive manufacturing, and remelting) are key development directions in advanced manufacturing. Their processes involve complex multi-phase coupling behaviors among solid, liquid, gas, and plasma states. Under high-power laser irradiation, when the power density reaches the order of 10⁶ W/cm², intense material evaporation occurs, forming a typical keyhole structure. Although this deep penetration mode improves processing efficiency, it can also lead to spatter, collapse, and unstable molten pool behavior, severely affecting forming quality.
In the context of intelligent manufacturing, real-time monitoring of laser processing has become a critical technology. The molten pool, as the most intuitive physical carrier, directly reflects process stability through its surface morphology. Current studies mainly focus on single parameters such as molten pool width, area, and grayscale features, but these indicators often fail to accurately describe the complex dynamic behavior inside the keyhole. Therefore, extracting key features that truly reflect molten pool dynamic evolution and establishing their relationship with laser–material interaction have become core challenges for intelligent monitoring in laser manufacturing.

02 Overview of the Study
This paper develops an in-situ monitoring method for molten pool surfaces based on the principle of specular reflection. By introducing auxiliary illumination beams and a high-speed camera, high-resolution observation of the molten pool surface (MPS) is achieved. First, image preprocessing techniques (such as median filtering and unsharp masking) are applied to denoise and enhance molten pool images. Then, threshold segmentation and morphological operations are used to extract three key feature regions: the molten pool surface region (MPS), the keyhole opening region, and the specular micro-region.
Based on this approach, the study systematically analyzes the area and fluctuation behavior of these regions under different laser power conditions. The results show that as the laser power increases from 2 kW to 5 kW, the molten pool area, keyhole opening area, and specular micro-region area all increase significantly. Among them, the molten pool area exhibits the most pronounced growth, while the specular micro-region shows the largest relative fluctuation amplitude. Further analysis reveals that the temporal variation of the specular micro-region exhibits clear periodicity (approximately 0.5–0.6 ms), effectively reflecting the periodic interaction of the laser with the keyhole front wall, thereby providing a new method to study internal energy transfer mechanisms.

03 Results and Discussion
Figure 1 illustrates the in-situ molten pool surface monitoring system constructed in this study. A high-power fiber laser is vertically applied to the material surface to form the molten pool. Auxiliary illumination lasers and high-speed cameras are symmetrically arranged on both sides, and optical filters are used to suppress plasma interference signals. The illumination light forms specular reflections on the molten pool surface, and the high-brightness regions are captured and recorded as specular micro-regions. This figure clearly demonstrates the core concept of indirectly characterizing molten pool states through reflected light, forming the basis for subsequent feature extraction and analysis.

Figure 2 presents the complete workflow for molten pool image feature extraction. Through image enhancement, logical masking, threshold segmentation, and region filtering, noise and interference are progressively removed, enabling effective extraction of specular micro-regions and keyhole opening areas. This figure highlights the crucial role of image processing in complex molten pool identification and provides reliable data for quantitative analysis.

Figure 3 shows typical molten pool morphologies under different laser power conditions. The molten pool exhibits a “tadpole-like” shape, with the front corresponding to the keyhole region and the rear forming the molten pool tail. As laser power increases, the molten pool size expands, the keyhole becomes more pronounced, and both spatter and surface fluctuations increase significantly. This indicates that laser power directly affects molten pool stability and is a key parameter for process quality control.

Figure 4 presents the temporal evolution of the specular micro-region, keyhole opening, and molten pool area. The results show that the specular micro-region exhibits clear periodic fluctuations, while the keyhole area fluctuates irregularly and the molten pool area remains relatively stable. Combined with the schematic of laser transmission inside the keyhole, this periodic behavior originates from the periodic interaction of the laser with the keyhole front wall. Therefore, the specular micro-region is identified as the most representative parameter for monitoring molten pool dynamics.

04 Conclusion
This study proposes a molten pool surface monitoring method based on specular reflection features and successfully achieves the extraction and comparative analysis of the molten pool surface region, keyhole opening, and specular micro-region. The results indicate that all three features increase with laser power, but the specular micro-region is the most sensitive for dynamic characterization, exhibiting both the largest fluctuation amplitude and stable periodic behavior. Further analysis confirms that this periodicity is closely related to the periodic interaction of the laser with the keyhole front wall, reflecting the internal energy transfer process within the molten pool. Therefore, the specular micro-region can serve as a key parameter for characterizing molten pool stability and keyhole behavior, providing important theoretical support and technical guidance for real-time monitoring and intelligent control in laser manufacturing.

References: https://doi.org/10.1016/j.optlastec.2025.114628

**--Cite the article published by 高能束加工技术 on April 21th, 2026, in the WeChat public account "High-Energy Beam Processing Technology and Applications."