Bimodal in-situ monitoring device and method for electron beam powder bed melting pool
By using a sandwich-type observation window and a dual-camera collaborative tracking system, combined with a multimodal data fusion method, the problem of molten pool monitoring in electron beam powder bed fusion additive manufacturing was solved, achieving high-precision real-time quality assessment and defect monitoring, and improving the forming quality of the parts.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- ZHEJIANG WANLI UNIV
- Filing Date
- 2026-01-07
- Publication Date
- 2026-04-17
AI Technical Summary
Existing technologies in electron beam powder bed fusion additive manufacturing struggle to achieve real-time, accurate, in-situ monitoring of the dynamic behavior and temperature field of the molten pool under extreme environments such as high vacuum, strong radiation, high temperature, and pervasive metal vapor. This results in high noise and significant information loss in the images and temperature data, affecting the reliability of quality judgment and process feedback control.
By employing a sandwich-type observation window component and a dual-camera collaborative tracking system, combined with a multimodal data fusion method, and through the collaborative work of a global camera and an infrared camera, precise tracking of the molten pool and high-precision restoration of the temperature field are achieved. Deep learning algorithms are used to restore image details and perform temperature compensation, ensuring the accuracy and continuity of monitoring data.
It enables comprehensive, high-precision in-situ monitoring of the electron beam additive manufacturing process, allowing for real-time online evaluation of part quality, timely identification and early warning of defects, and improvement of part forming quality consistency and reliability.
Smart Images

Figure CN121877795A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of additive manufacturing technology, and in particular to a dual-modal in-situ monitoring device and method for electron beam powder bed molten pool. Background Technology
[0002] In electron beam powder bed fusion additive manufacturing, real-time, accurate, in-situ monitoring of the dynamic behavior and temperature field of the molten pool is crucial for process optimization, online quality assessment, and defect control. However, the extreme environment of this process—high vacuum, strong radiation, high temperature, and pervasive metal vapor—poses significant challenges to the reliability and accuracy of the monitoring system. Existing technologies typically employ a single observation window and directly use radiation-shielding materials such as lead glass to shield X-rays. However, these materials exhibit significant absorption and interference in the infrared band, leading to severe attenuation and distortion of the infrared temperature measurement signal, making it impossible to obtain a true molten pool temperature field. Furthermore, to balance field of view and observation accuracy, the commonly used fixed single-camera system struggles to dynamically track a high-speed moving molten pool, failing to strike a balance between global monitoring and high-definition close-ups, resulting in untimely and inaccurate capture of transient changes in the molten pool. In addition, the metal vapor generated during the process easily contaminates the inside of the observation window, rapidly degrading image quality, while conventional protection or cleaning methods often fail to balance cleaning effectiveness with the continuity of observation. Due to the aforementioned factors, the acquired raw images and temperature data suffer from high noise and significant information loss. Even with conventional image processing and temperature correction methods, it is difficult to reconstruct a clear molten pool morphology and accurate temperature distribution, severely limiting the reliability of real-time quality assessment and process feedback control based on monitoring data. Therefore, developing an in-situ monitoring system capable of adapting to extreme process environments and achieving high-precision dynamic tracking and multimodal data fusion has become a pressing technical challenge for improving the quality control level of electron beam additive manufacturing. Summary of the Invention
[0003] To address the aforementioned shortcomings, this invention proposes a dual-modal in-situ monitoring device and method for electron beam powder bed molten pool. Through the systematic collaboration of a sandwich-type observation window component, dual-camera collaborative tracking, and multimodal data fusion, it achieves precise tracking of the molten pool and high-precision reconstruction of images and temperature fields, ultimately achieving comprehensive in-situ quality monitoring and significant improvement in the electron beam additive manufacturing process.
[0004] To achieve the above objectives, the present invention provides the following technical solution: a dual-mode in-situ monitoring device for electron beam powder bed molten pool, comprising: The equipment casing contains a vacuum chamber. A sandwich-type observation window assembly is installed on the wall of a vacuum chamber. The observation window assembly includes an installation frame that forms a sealed cavity, a radiation-proof glass layer and a high-temperature resistant glass layer that are installed parallel to each other at predetermined intervals on the optical path channel of the sealed cavity to separate the cavity, forming a sandwich structure that has the functions of sealing, high temperature resistance and radiation protection. A dual-modal vision module is positioned between the radiation-proof glass layer and the high-temperature resistant glass layer, and is oriented towards the high-temperature resistant glass layer. The dual-modal vision module includes a global camera and an infrared camera. The drive module, connected to the infrared camera, is used to drive the infrared camera to perform at least two degrees of freedom of motion in order to adjust its optical axis orientation; The global camera has a field of view covering the entire powder bed area and is used to detect the position of the electron beam. The infrared camera is configured to track and capture images of the molten pool based on the position of the electron beam detected by the global camera and is driven by the drive module.
[0005] As an improvement, the global camera is positioned above the infrared camera, and the dual-modal vision module also includes a global camera bracket fixedly installed in the upper part of the sealed cavity for mounting the global camera.
[0006] As an improvement, the drive module includes a turntable bracket mounted on a mounting frame, a turntable motor disposed on one side of the turntable bracket, a turntable mounted between the turntable brackets, a turntable at least partially rotatable within the turntable, a turntable motor disposed at the lower end of the turntable, and an infrared camera bracket mounted on the turntable. The infrared camera is mounted at one end of the infrared camera bracket to achieve monitoring through a high-temperature resistant glass layer. The output shaft of the turntable motor drives the turntable to rotate relative to the turntable bracket to adjust the rotation angle of the infrared camera in the A-axis direction. The output shaft of the turntable motor drives the turntable to rotate relative to the turntable to adjust the rotation angle of the infrared camera in the B-axis direction through the infrared camera bracket.
[0007] As an improvement, a first glass mounting frame is vertically installed at the end of the mounting frame closest to the vacuum chamber wall, corresponding to the high-temperature resistant glass layer, and a second glass mounting frame is vertically installed at the end of the mounting frame furthest from the vacuum chamber wall, corresponding to the radiation-proof glass layer. The high-temperature resistant glass layer is embedded in the first glass mounting frame, and the radiation-proof glass layer is embedded in the second glass mounting frame.
[0008] As an improvement, the mounting frame is provided with an outwardly extending mounting edge, which is removably and sealed to the wall of the vacuum chamber by fasteners.
[0009] As an improvement, several elastic seals are provided at the connection between the high-temperature resistant glass layer and the first glass mounting frame, the connection between the radiation-proof glass layer and the second glass mounting frame, the connection between the mounting edge and the side wall of the equipment housing, and the connection between the global camera and the global camera bracket.
[0010] The monitoring device also includes an adsorption module, which includes an adsorption motor assembly set in a sealed cavity and an adsorption wheel set in a vacuum chamber and adjacent to a high-temperature resistant glass layer and driven to rotate by the motor assembly, for resisting or adsorbing metal vapor inside the high-temperature resistant glass layer. The adsorption wheel is made of a porous high-temperature resistant adsorption material with a partially circular cross-section, so that it forms a periodic field of view window when rotating.
[0011] A dual-modal in-situ monitoring method for electron beam powder bed molten pool, employing any of the electron beam additive manufacturing in-situ monitoring and quality control devices described above, comprising: Adaptive visual tracking steps: The powder bed is monitored in a grid pattern by a fixed global camera, and when the electron beam is detected to enter a specific grid, the camera tracking module is driven to adjust the optical axis of the infrared camera to achieve real-time and automatic tracking of the moving molten pool. Multimodal data fusion and compensation steps: Simultaneously acquire infrared thermal images and RGB images of the molten pool from an infrared camera and a global camera. Utilize multiple deep learning algorithms to dehaze and reconstruct the RGB images, which are degraded due to metal vapor and adsorption wheel obstruction. Based on the restored RGB image information, correct the infrared temperature field data using a dual-modal temperature compensation algorithm to obtain an accurate compensated temperature field T. c ; Quality assessment steps: Based on the accurate compensated temperature field T c The restored molten pool image enables real-time quality assessment and defect monitoring of the additive manufacturing process.
[0012] As an improvement, an intelligent pre-setting step is included before the adaptive visual tracking step: before the device starts, the infrared camera returns to the center of the powder bed and uses this point as the origin. , ), initial angle , And load the parameters focal length f and infrared weighting coefficients. , The value range is 0.9-0.94.
[0013] As an improvement, the adaptive visual tracking steps specifically include: The powder bed is divided into an M×N grid; Preset a corresponding pitch angle for each grid center With horizontal angle parameters ; When the global camera detects that the electron beam has entered a certain grid, it sends the grid number to the control system of the infrared camera. The infrared camera calls the corresponding pitch and horizontal angle parameters according to the number, and adjusts its optical axis direction through the turntable motor and turntable motor to achieve precise tracking of the molten pool.
[0014] As an improvement, a deep learning algorithm based on an atmospheric scattering model is used to restore details and enhance contrast of RGB images that are attenuated by metal vapor interference, and to restore the morphological details of the molten pool area. In addition, a deep learning network based on a deformable alignment mechanism is used to perform multi-frame fusion and view reconstruction of RGB images that are missing information due to periodic occlusion.
[0015] As an improvement, the dual-modal temperature compensation algorithm is implemented using the following formula: Among them, T c For the compensated temperature, T r R represents the original temperature measured by the infrared camera, R, G, and B represent the three-channel pixel values of the restored RGB image, K1 and K2 are weighting coefficients, and a, b, and c are the contribution coefficients of the RGB channels.
[0016] Compared with the prior art, the advantages of the present invention are as follows: Through an innovative sandwich-type observation window component design, a square window is opened at the rear of the device, with a high-temperature resistant glass layer and a radiation-proof glass layer arranged sequentially from the inside out. The inner high-temperature resistant glass layer has extremely low absorption of infrared light waves, effectively ensuring the monitoring performance of the infrared camera; the outer lead glass layer reliably blocks harmful radiation. Together, they form a sealed cavity that is physically and functionally doubly isolated. This ensures the vacuum level of the device and the structural safety of the observation window, while eliminating the interference and attenuation caused by the lead glass directly serving as the observation window to infrared radiation transmission, thus guaranteeing the accuracy of the infrared temperature monitoring signal. Simultaneously, the dual-modal vision module integrated within this sandwich structure achieves intelligent, real-time, and precise tracking of the dynamic molten pool through the collaborative work of the global camera and the infrared camera. The global camera divides the powder bed into a grid matrix and pre-stores the pitch and horizontal angle parameters at the center of each grid. When an electron beam is detected entering a specific grid, the grid number is broadcast immediately. The infrared camera then instantly calls up the corresponding angle parameters and adjusts the optical axis direction through the AB axis mechanism driven by the turntable motor and the turntable motor. This ensures continuous and accurate imaging of the molten pool area, effectively solving the shortcomings of traditional single-camera monitoring in capturing dynamic changes in the molten pool. It achieves precise zonal monitoring of the target area and provides a stable and reliable data source for process quality control.
[0017] Meanwhile, addressing the issue of periodic obstruction by metal vapor and the adsorption wheel, the adsorption module achieves a balance between maintaining a clean observation window and preserving a periodically complete field of view through the intermittent rotation of partially circular adsorption wheels made of high-temperature resistant adsorption material. Furthermore, the system utilizes a complementary multimodal data fusion and compensation mechanism to achieve high-precision data restoration. Using a deep learning algorithm based on an atmospheric scattering model and a deformable alignment mechanism, dehazing, detail enhancement, multi-frame fusion, and view reconstruction are performed on RGB images with fog attenuation and partial obstruction, effectively restoring the morphological details of the molten pool. Building upon this, a pixel-weighted temperature compensation algorithm is employed, using the three-channel values of the restored clear RGB image to compensate for errors in the infrared temperature measurement system. Through a linear combination of specific weighting coefficients and channel coefficients, the final output is a more accurate temperature field reflecting the actual thermal state of the molten pool, significantly improving the accuracy of temperature monitoring.
[0018] Ultimately, the synergistic effect of all the aforementioned technical aspects enables comprehensive, high-precision, in-situ monitoring of the electron beam additive manufacturing process. Based on the restored, clear image of the molten pool and the precisely compensated temperature field data, the system can perform real-time online quality assessment and defect monitoring of the manufacturing process, thereby promptly identifying and issuing warnings for defects such as poor fusion, warping, and cracks, fundamentally improving the consistency and reliability of the formed parts. Attached Figure Description
[0019] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments: Figure 1 This is a schematic diagram of a dual-mode in-situ monitoring device for an electron beam powder bed molten pool. Figure 2 This is a schematic cross-sectional view of a dual-mode in-situ monitoring device for electron beam powder bed molten pool. Figure 3 This is a schematic diagram of the adsorption module structure; Figure 4 This is a schematic diagram of the dual-modal vision module structure; Figure 5 This is a flowchart of a dual-modal in-situ monitoring method for electron beam powder bed molten pool.
[0020] The markings in the above figures are as follows: 1. Shell; 1.1. Vacuum chamber; 2. Sandwiched observation window assembly; 2.1. Radiation-proof glass layer; 2.2. High-temperature resistant glass layer; 2.3. Mounting frame; 2.3.1. First glass mounting frame; 2.3.2. Second glass mounting frame; 2.3.3. Mounting edge; 3. Dual-modal vision module; 3.1. Infrared camera; 3.2. Global camera bracket; 3.3. Global camera; 4. Drive module; 4.1. Turntable bracket; 4.2. Turntable motor; 4.3. Turntable; 4.4. Turntable; 4.5. Turntable motor; 4.6. Infrared camera bracket; 5. Elastic seal; 6. Powder bed; 7. Electron gun; 8. Adsorption module; 8.1. Adsorption motor assembly; 8.2. Adsorption wheel. Detailed Implementation
[0021] In this invention, it should be understood that the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise", "axial", "planar direction", "circumferential", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing this invention and simplifying the description, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this invention.
[0022] like Figures 1 to 3As shown, a dual-modal in-situ monitoring device for electron beam powder bed molten pool includes a housing 1, a sandwich-type observation window assembly 2, a dual-modal vision module 3, and a drive module 4. The housing 1 contains a vacuum chamber 1.1, within which is a powder bed 6 for carrying powder and an electron gun 7 for emitting an electron beam to the powder bed 6. The powder bed 6 is embedded in a mounting groove within the vacuum chamber 1.1 and can move precisely along the Z-direction via a lifting mechanism. The electron gun 7 is fixedly mounted on the top of the vacuum chamber 1.1. The sandwich-type observation window assembly 2 is disposed on the chamber wall of the vacuum chamber 1.1 and includes a mounting frame 2.3 forming a sealed cavity and radiation shielding components installed parallel to the optical path channel of the sealed cavity from the outside to the inside at predetermined intervals. Glass layer 2.1 and high-temperature resistant glass layer 2.2 separate the cavity, forming a sandwich structure that combines sealing, high-temperature resistance, and radiation protection functions. The dual-modal vision module 3 is disposed between the radiation-proof glass layer 2.1 and the high-temperature resistant glass layer 2.2 and faces the high-temperature resistant glass layer 2.2. The dual-modal vision module 3 includes a global camera 3.3 and an infrared camera 3.1. The drive module 4 is connected to the infrared camera 3.1 and is used to drive the infrared camera 3.1 to perform at least two degrees of freedom of movement to adjust its optical axis orientation. The field of view of the global camera 3.3 covers the entire powder bed 6 area and is used to detect the position of the electron beam. The infrared camera 3.1 is configured to be driven by the drive module 4 based on the position of the electron beam detected by the global camera 3.3 to track and capture images of the molten pool.
[0023] The global camera 3.3 is positioned above the infrared camera 3.1. The dual-modal vision module 3 also includes a global camera bracket 3.2, which is fixedly installed in the upper part of the sealed cavity for mounting the global camera 3.3.
[0024] The first glass mounting frame 2.3.1 is vertically installed at the end of the mounting frame 2.3 closest to the wall of the vacuum chamber 1.1, corresponding to the high-temperature resistant glass layer 2.2. The second glass mounting frame 2.3.2 is vertically installed at the end of the mounting frame 2.3 furthest from the wall of the vacuum chamber 1.1, corresponding to the radiation-proof glass layer 2.1. The high-temperature resistant glass layer 2.2 is embedded in the first glass mounting frame 2.3.1, and the radiation-proof glass layer 2.1 is embedded in the second glass mounting frame 2.3.2.
[0025] The mounting frame 2.3 has an outwardly extending mounting edge 2.3.3, which is removably and sealed to the chamber wall of the vacuum chamber 1.1 by fasteners.
[0026] Several elastic sealing elements 5 are provided at the connection between the high-temperature resistant glass layer 2.2 and the first glass mounting frame 2.3.1, the connection between the radiation-proof glass layer 2.1 and the second glass mounting frame 2.3.2, between the mounting edge 2.3.3 and the side wall of the equipment housing 1, and between the global camera 3.3 and the global camera bracket 3.2.
[0027] The monitoring device also includes an adsorption module 8, which includes an adsorption motor assembly 8.1 disposed in a sealed cavity and an adsorption wheel 8.2 disposed in a vacuum chamber 1.1 adjacent to a high-temperature resistant glass layer 2.2 and driven to rotate by the motor assembly. The adsorption wheel 8.2 is used to resist or adsorb metal vapor inside the high-temperature resistant glass layer 2.2. The adsorption wheel 8.2 is made of a porous high-temperature resistant adsorption material with a partially circular cross-section, so that it forms a periodic field of view window when rotating. Preferably, the adsorption wheel 8.2 is a clay mesh with a 3 / 4 circular cross-section.
[0028] like Figure 4 As shown, the drive module 4 includes a turntable bracket 4.1 mounted on the mounting frame 2.3, a turntable motor 4.2 disposed on one side of the turntable bracket 4.1, a turntable 4.3 mounted between the turntable brackets 4.1, a turntable 4.4 at least partially rotatably disposed within the turntable 4.3, a turntable motor 4.5 disposed at the lower end of the turntable 4.3, and an infrared camera bracket 4.6 mounted on the turntable 4.4. An infrared camera 3.1 is mounted on one end of the infrared camera bracket 4.6 to achieve monitoring through the high-temperature resistant glass layer 2.2. The output shaft of the turntable motor 4.2 drives the turntable 4.3 to rotate relative to the turntable bracket 4.1 to adjust the rotation angle of the infrared camera 3.1 in the A-axis direction. The output shaft of the turntable motor 4.4 drives the turntable 4.4 to rotate relative to the turntable 4.3 to adjust the rotation angle of the infrared camera 3.1 in the B-axis direction through the infrared camera 3.1 bracket.
[0029] like Figure 5 As shown, a dual-mode in-situ monitoring method for electron beam powder bed molten pool includes: Intelligent preset steps: Before the equipment starts, the infrared camera returns to the center of the powder bed and uses this point as the origin. , ), initial angle , And load the parameters focal length f and infrared weighting coefficients. , The value range is 0.9-0.94, preferably... The value is 0.92; Adaptive vision tracking steps: Divide the powder bed into an M×N grid, preferably a 6×6 grid; preset the pitch angle for the center of each grid. With horizontal angle parameters When the global camera detects that the electron beam has entered a certain grid, it sends the grid number to the control system of the infrared camera. The infrared camera calls the corresponding pitch and horizontal angle parameters according to the number, and adjusts its optical axis pointing through the turntable motor and turntable motor to achieve precise tracking of the molten pool. Multimodal data fusion and compensation steps: Simultaneously acquire infrared thermal images and RGB images of the molten pool from an infrared camera and a global camera. Utilize a deep learning algorithm based on an atmospheric scattering model to restore details and enhance contrast in the RGB images, which are attenuated due to metal vapor and adsorption wheel interference, thus restoring the morphological details of the molten pool region. Then, employ a deep learning network based on a deformable alignment mechanism to perform multi-frame fusion and view reconstruction on the RGB images, which suffer from information loss due to periodic occlusion. Based on the restored RGB image information, correct the infrared temperature field data using a dual-modal temperature compensation algorithm to obtain an accurate compensated temperature field T. c The dual-mode temperature compensation algorithm is implemented using the following formula: Among them, T c For the compensated temperature, T r The original temperature is measured by the infrared camera. R, G, and B are the three-channel pixel values of the restored RGB image. K1 and K2 are weighting coefficients, with K2 set to 0.08. a, b, and c are the contribution coefficients of the RGB channels, with a set to 0.6, b set to 0.3, and c set to 0.1. Quality assessment steps: Based on the accurate compensated temperature field T c The restored molten pool image enables real-time quality assessment and defect monitoring of the additive manufacturing process.
[0030] The above description only illustrates the preferred embodiments of the present invention and should not be construed as limiting the scope of the claims. The present invention is not limited to the above embodiments, and variations in its specific structure are permitted. All modifications made within the scope of the independent claims of this invention are also within the scope of protection of this invention.
Claims
1. A dual-mode in-situ monitoring device for electron beam powder bed molten pool, characterized in that, include: The equipment casing contains a vacuum chamber. A sandwich-type observation window assembly is disposed on the wall of the vacuum chamber. The observation window assembly includes an installation frame constituting a sealed cavity, a radiation-proof glass layer and a high-temperature resistant glass layer installed parallel to each other at predetermined intervals on the optical path channel of the sealed cavity to separate the cavity, forming a sandwich structure that has the functions of sealing, high temperature resistance and radiation protection. A dual-modal vision module is disposed between a radiation-proof glass layer and a high-temperature resistant glass layer and faces the high-temperature resistant glass layer. The dual-modal vision module includes a global camera and an infrared camera. A drive module, connected to the infrared camera, is used to drive the infrared camera to perform at least two degrees of freedom of motion in order to adjust its optical axis orientation; The global camera has a field of view covering the entire powder bed area and is used to detect the position of the electron beam. The infrared camera is configured to track and capture images of the molten pool based on the position of the electron beam detected by the global camera and is driven by the driving module.
2. The dual-mode in-situ monitoring device for electron beam powder bed molten pool according to claim 1, characterized in that, The global camera is positioned above the infrared camera, and the dual-modal vision module also includes a global camera bracket fixedly installed in the upper part of the sealed cavity for mounting the global camera.
3. The dual-mode in-situ monitoring device for electron beam powder bed molten pool according to claim 1, characterized in that, The drive module includes a turntable bracket mounted on a mounting frame, a turntable motor disposed on one side of the turntable bracket, a turntable mounted between the turntable brackets, a turntable at least partially rotatable within the turntable, a turntable motor disposed at the lower end of the turntable, and an infrared camera bracket mounted on the turntable. The infrared camera is mounted at one end of the infrared camera bracket to achieve monitoring through a high-temperature resistant glass layer. The output shaft of the turntable motor drives the turntable to rotate relative to the turntable bracket to adjust the rotation angle of the infrared camera in the A-axis direction. The output shaft of the turntable motor drives the turntable to rotate relative to the turntable to adjust the rotation angle of the infrared camera in the B-axis direction via the infrared camera bracket.
4. The dual-mode in-situ monitoring device for electron beam powder bed molten pool according to claim 2, characterized in that, The mounting frame has a first glass mounting frame vertically arranged at the end near the vacuum chamber wall corresponding to the high-temperature resistant glass layer, and a second glass mounting frame vertically arranged at the end away from the vacuum chamber wall corresponding to the radiation-proof glass layer. The high-temperature resistant glass layer is embedded in the first glass mounting frame, and the radiation-proof glass layer is embedded in the second glass mounting frame.
5. The dual-mode in-situ monitoring device for electron beam powder bed molten pool according to claim 4, characterized in that, The mounting frame has an outwardly extending mounting edge, which is detachably and sealed to the wall of the vacuum chamber by fasteners.
6. The dual-mode in-situ monitoring device for electron beam powder bed molten pool according to claim 5, characterized in that, Several elastic seals are provided at the connection between the high-temperature resistant glass layer and the first glass mounting frame, the connection between the radiation-proof glass layer and the second glass mounting frame, between the mounting edge and the side wall of the equipment housing, and between the global camera and the global camera bracket.
7. The dual-mode in-situ monitoring device for electron beam powder bed molten pool according to claim 1, characterized in that, The monitoring device also includes an adsorption module, which includes an adsorption motor assembly disposed in a sealed cavity and an adsorption wheel disposed in a vacuum chamber and adjacent to a high-temperature resistant glass layer and driven to rotate by the motor assembly, for resisting or adsorbing metal vapor inside the high-temperature resistant glass layer. The adsorption wheel is made of a porous high-temperature resistant adsorption material with a partially circular cross-section, so that it forms a periodic field of view window when rotating.
8. A method for in-situ monitoring of a dual-mode molten pool in an electron beam powder bed, comprising using an in-situ monitoring device for a dual-mode molten pool in an electron beam powder bed as described in any one of claims 1-7, characterized in that, include: Adaptive visual tracking steps: The powder bed is monitored in a grid pattern by a fixed global camera, and when the electron beam is detected to enter a specific grid, the camera tracking module is driven to adjust the optical axis of the infrared camera to achieve real-time and automatic tracking of the moving molten pool. Multimodal data fusion and compensation steps: Simultaneously acquire infrared thermal images and RGB images of the molten pool from the infrared camera and the global camera. Utilize multiple deep learning algorithms to dehaze and reconstruct the RGB images degraded by metal vapor and adsorption wheel obstruction. Based on the restored RGB image information, correct the infrared temperature field data using a dual-modal temperature compensation algorithm to obtain an accurate compensated temperature field T. c ; Quality assessment steps: Based on the precise compensated temperature field T c The restored molten pool image enables real-time quality assessment and defect monitoring of the additive manufacturing process.
9. The method for dual-mode in-situ monitoring of an electron beam powder bed molten pool according to claim 8, characterized in that: Before the adaptive visual tracking step, an intelligent preset step is also included: before the device starts, the infrared camera returns to the center position of the powder bed and uses this point as the origin. , ), initial angle , And load the parameters focal length f and infrared weighting coefficients. , The value range is 0.9-0.
94.
10. The method for in-situ monitoring of a dual-mode molten pool in an electron beam powder bed according to claim 8, characterized in that: The adaptive visual tracking steps specifically include: The powder bed is divided into an M×N grid; Preset a corresponding pitch angle for each grid center With horizontal angle parameters ; When the global camera detects that an electron beam has entered a certain grid, it sends the grid number to the control system of the infrared camera. The infrared camera calls the corresponding pitch and horizontal angle parameters according to the number, and adjusts its optical axis direction through the turntable motor and turntable motor to achieve precise tracking of the molten pool.
11. The method for dual-mode in-situ monitoring of an electron beam powder bed molten pool according to claim 8, characterized in that: The dual-modal temperature compensation algorithm is implemented using the following formula: Among them, T c For the compensated temperature, T r R represents the original temperature measured by the infrared camera, R, G, and B represent the three-channel pixel values of the restored RGB image, K1 and K2 are weighting coefficients, and a, b, and c are the contribution coefficients of the RGB channels.