Method and system for dynamic synchronization monitoring of molten pool temperature and three-dimensional topography

Abstract

The application provides a kind of dynamic synchronization monitoring method and system of molten pool temperature and three-dimensional topography, belongs to the field of additive manufacturing.The method comprises: obtaining double-band molten pool binocular image;The double-band molten pool binocular image is carried out perspective separation, and left perspective image of first wave band and right perspective image of second wave band are obtained;The right perspective image is carried out affine transformation, so that left perspective image of first wave band and right perspective image of second wave band realize spatial alignment;According to left perspective image of first wave band and right perspective image of second wave band, the temperature field reconstruction and three-dimensional topography reconstruction of molten pool are carried out.The method can realize the synchronous monitoring of temperature field and three-dimensional topography by obtaining a single double-band molten pool binocular image, decomposing double-band molten pool binocular image, and then reconstructing the temperature field and three-dimensional topography of molten pool.

Description

Technical Field

[0001] This application belongs to the field of additive manufacturing, and in particular relates to a method and system for dynamic synchronous monitoring of molten pool temperature and three-dimensional morphology. Background Technology

[0002] In recent years, metal additive manufacturing technology, as a core component of intelligent manufacturing, has been profoundly transforming the design and manufacturing paradigms of components in high-end equipment fields such as aerospace, biomedicine, and precision instruments. However, the process of this technology moving from "prototype manufacturing" to "industrial mass production" is facing severe challenges due to core bottlenecks such as quality consistency and process reliability caused by the lack of process monitoring. Currently, the metal additive manufacturing process is essentially a difficult-to-observe dynamic "black box," involving complex multi-physics field couplings under transient extreme conditions, making it difficult to predict and control metallurgical defects inside the molten pool, such as porosity, lack of fusion, and microcracks. The currently widely used offline trial-and-error and post-processing inspection methods are not only time-consuming and costly, but also cannot meet the stringent requirements for internal quality and service performance of key load-bearing components. This has become a major technical obstacle restricting industrial upgrading and large-scale application. Against this backdrop, developing online monitoring technology that can perceive and analyze key process signals in real time, and realizing the transformation of the manufacturing process from a "black box" to "transparency," has become an urgent and significant need for both academia and industry.

[0003] In molten pool temperature measurement, temperature detection methods include infrared thermal imaging technology, which mainly employs non-contact infrared thermometry. Its principle is based on the thermal radiation characteristics of an object, converting captured infrared radiation energy into a temperature reading. This method is extremely sensitive to the calibration of the object's surface emissivity, but the emissivity of the molten pool dynamically changes with temperature, wavelength, and surface condition, making it difficult to guarantee measurement accuracy. Another method is image colorimetric temperature measurement, which uses the dual-wavelength colorimetric principle to calculate the target temperature. It is widely used because it does not require absolute emissivity. However, existing colorimetric thermometry techniques all require images from the same viewpoint in both bands to ensure accurate matching of grayscale values ​​at the same spatial location in the two bands. This single-viewpoint constraint makes the system extremely sensitive to specular reflections on the molten pool surface: the surface of the liquid molten pool exhibits specular characteristics, and incident laser light or other strong light sources may form specular reflections on the molten pool surface before entering the camera. Since colorimetric thermometry assumes that all received radiation comes from the object's own thermal radiation, specular reflection components severely undermine this assumption, leading to false high temperatures or abnormal fluctuations in the calculated temperature value, reducing the reliability of the temperature measurement.

[0004] In the field of 3D imaging of molten pools, existing technologies mainly fall into three categories. The first is the laser structured light method based on optical triangulation. This method reconstructs the 3D shape by projecting laser stripes onto the molten pool surface and capturing the deformation from another angle using a camera. This method requires the projected laser stripes to be clearly imaged on the molten pool surface; however, the strong specular reflection of the molten pool surface and its high-temperature self-luminous radiation easily cause the laser stripes to be submerged or broken, resulting in the loss of structured light information and failure of 3D reconstruction. The second is phase measurement profilometry, which projects an coded grating pattern onto the molten pool surface and analyzes the deformation phase to obtain depth information. This method requires the projected pattern to maintain a clear phase distribution on the molten pool surface, but the rapid dynamic changes of the molten pool easily cause motion blur in the projected pattern, making stable 3D reconstruction difficult. The third is passive binocular stereo vision. This method does not require active projection of structured light; it only uses two cameras to simultaneously acquire images of the molten pool from different perspectives and reconstructs the 3D shape using the principle of parallax. However, traditional binocular vision technology requires two independent industrial cameras to work synchronously. During the high-speed dynamic changes of the molten pool, it is difficult to guarantee the trigger synchronization accuracy between the two cameras. Any slight trigger delay or temperature drift will cause the left and right view images to be misaligned in time, which will introduce parallax calculation errors and affect the accuracy of 3D reconstruction.

[0005] Under the existing optical path design, current single monitoring systems can only detect a single physical quantity, making it difficult to achieve simultaneous online monitoring of molten pool temperature and three-dimensional morphology. Although two independent devices can be used to perform monitoring tasks in parallel, this approach significantly increases equipment costs and reduces the convenience of the monitoring process. Furthermore, parallel operation of dual systems also brings problems such as an increased number of components, increased overall weight, and larger space requirements. More importantly, it is difficult to ensure strict synchronization of molten pool information in both time and space. Summary of the Invention

[0006] In view of this, this application provides a method and system for synchronous monitoring of molten pool temperature and three-dimensional morphology, aiming to achieve synchronous monitoring of molten pool temperature and three-dimensional morphology.

[0007] Firstly, this application provides a method for dynamic synchronous online monitoring of molten pool temperature and three-dimensional morphology, including: Acquire dual-band stereo images of the molten pool; Viewpoint separation is performed on the dual-band molten pool binocular image to obtain the left-view image of the first band and the right-view image of the second band. An affine transformation is performed on the right-view image to achieve spatial alignment between the left-view image of the first band and the right-view image of the second band. The temperature field and three-dimensional morphology of the molten pool were reconstructed based on the left-view image of the first band and the right-view image of the second band.

[0008] Optionally, the step of performing an affine transformation on the right-view image to achieve spatial alignment between the left-view image of the first band and the right-view image of the second band includes: A checkerboard calibration board was used and placed at multiple different spatial positions within the system's imaging field of view. Checkerboard images were acquired from both the left and right perspectives. The coordinates of all corner points of the checkerboard were extracted from each image. The corner point coordinates in the left perspective image were used as the target coordinates, and the corresponding corner point coordinates in the right perspective image were used as the source coordinates. A one-to-one coordinate pair was formed for each point. The above operation was repeated for all calibration images acquired at different spatial positions to obtain multiple sets of coordinate pairs, forming a transformation matrix between the source coordinate set and the target coordinate set. Based on the solved transformation matrix, the coordinates of the right-view image are remapped pixel by pixel, transforming each pixel in the right-view image to its corresponding position in the coordinate system of the left-view image, so that the right-view image and the left-view image achieve precise spatial correspondence at the pixel level, and obtain a well-matched binocular image pair.

[0009] Optionally, the process of reconstructing the temperature field of the molten pool is as follows: The grayscale ratio of corresponding pixel positions in the left-view image of the first band and the right-view image of the second band is calculated pixel by pixel, and it is determined whether specular reflection occurs at the corresponding pixel position. If it is determined that no specular reflection occurs in the molten pool at the corresponding pixel position, the temperature value of each point in the molten pool is calculated by substituting it into the dual-wavelength colorimetric thermometry formula, and a molten pool temperature field distribution map is generated. If it is determined that specular reflection occurs in the molten pool at the corresponding pixel position, light intensity correction is performed, and then the temperature value of each point in the molten pool is calculated by substituting it into the dual-wavelength colorimetric thermometry formula, and a molten pool temperature field distribution map is generated.

[0010] Optionally, the process of three-dimensional topography reconstruction is as follows: Calculate the grayscale ratio at corresponding pixel positions in the left-view image of the first band and the right-view image of the second band, and statistically obtain the average grayscale ratio of the entire image or a local area. Based on the average grayscale ratio, perform overall grayscale stretching on the images with relatively low grayscale values ​​in the left-view image of the first band and the right-view image of the second band, and align the brightness level of the images with relatively low grayscale values ​​in the left-view image of the first band and the right-view image of the second band with the images with relatively high grayscale values ​​in the left-view image of the first band and the right-view image of the second band, thereby compensating for the brightness loss caused by the difference in the optical path of the two bands. The left-view image of the first band and the right-view image of the second band are input into the residual network to extract the initial dense disparity map. A two-stage noise reduction method is achieved by performing two iterations of bilateral filtering on the initial dense disparity map. The initial depth map is calculated based on the initial dense disparity map; A two-dimensional noise reduction method is used to process the initial depth map through five iterations using bilateral filtering, thereby obtaining the depth map of the molten pool. The molten pool is reconstructed in three dimensions based on its depth map. The pixel coordinates and depth values ​​of each pixel in the depth map are converted into three-dimensional spatial points, and the set of all points constitutes the three-dimensional topography of the molten pool.

[0011] Optionally, it also includes: Spatiotemporal correlation analysis was performed on the temperature field and three-dimensional morphology of the reconstructed molten pool. By locating the coordinates of the highest temperature point in the molten pool and using the abscissa of the highest temperature point as the boundary, the molten pool was divided into the head and tail of the molten pool. By performing curve fitting of temperature and depth data on the head and tail of the molten pool respectively, a coupling relationship model between the temperature field and the three-dimensional morphology of the molten pool was obtained.

[0012] Secondly, this application provides a dynamic synchronous monitoring system for molten pool temperature and three-dimensional morphology, comprising: Additive manufacturing components used for additive manufacturing; The monitoring optical path is used to acquire binocular images of the molten pool in the dual-band process of additive manufacturing; The image processing unit is used to separate the viewing angles of the dual-band molten pool binocular images, so that the left-view image of the first band and the right-view image of the second band are spatially aligned. The right-view image is subjected to an affine transformation to achieve spatial alignment of the left-view image of the first band and the right-view image of the second band. The temperature field and three-dimensional morphology of the molten pool are reconstructed based on the left-view image of the first band and the right-view image of the second band.

[0013] Optionally, the monitoring optical path includes: The system includes a beam splitting component, a first-band adjustment component, a second-band adjustment component, a beam combining component, and an image acquisition unit. The beam splitter is used to split the acquired light from the molten pool into a first beam and a second beam; the first beam is transmitted to a first band adjustment component, and the first band adjustment component outputs a band of [missing information]. The first beam; the second beam is transmitted to the second band adjustment component, and the output band of the second band adjustment component is... The second beam; the beam combiner is used to combine the wavelength band of The first beam and wavelength are The second beam is combined into a dual-band molten pool signal light, and then the image acquisition unit obtains a dual-band molten pool binocular image based on the dual-band molten pool signal light.

[0014] Optionally, the beam splitting component includes: The first reflecting mirror and the first right-angle prism; The first reflector is used to split the acquired light into a first beam and a second beam, while the first right-angle prism transmits the first beam and the second beam to the first band adjustment component and the second band adjustment component, respectively.

[0015] Optionally, the first band adjustment component includes: The second reflector, the first narrowband filter, and the third reflector; The second reflector is used to transmit the received first beam to the first narrowband filter, and the beam is filtered by the first narrowband filter to obtain a wavelength band of [band value missing]. The first beam, followed by the third reflecting mirror, will have a wavelength of... The first beam of light is transmitted to the beam combining component.

[0016] Optionally, the second band adjustment component includes: The fourth reflector, the second narrowband filter, and the fifth reflector; The fourth reflecting mirror is used to transmit the received second beam to the second narrowband filter. After filtering by the second narrowband filter, the resulting wavelength is [band value missing]. The second beam, followed by the fifth reflecting mirror, will have a wavelength of... The second beam is transmitted to the beam combining component.

[0017] The technical solution provided in this application has at least one of the following beneficial effects: (1) This invention creatively achieves dual-band binocular imaging of the molten pool by adopting a dual-band binocular different-view imaging architecture, physically separating and spatially multiplexing the dual-band optical paths. By using a single optical path system to achieve coaxial synchronous online monitoring of the molten pool temperature and three-dimensional dynamic features in metal additive manufacturing, the system can significantly improve its integration and overall performance while ensuring data spatiotemporal consistency.

[0018] (2) This invention adopts a dual-band binocular anisotropic imaging architecture, which overcomes the defect of traditional single-view colorimetric thermometry being sensitive to the specular reflection of the molten pool in principle. The specular reflection of the molten pool has strong directional selectivity and can only be incident in one direction, and cannot simultaneously form equivalent interference to the left and right anisotropic views. Therefore, this application can suppress false high temperature and temperature fluctuations from the source, ensuring stable and reliable temperature measurement. In addition, this application uses a narrowband filter to achieve high-precision separation of the target characteristic band. Compared with a bandpass filter, the narrowband filter, with its extremely narrow half-width and high center wavelength transmittance, effectively suppresses background stray light and environmental radiation interference, further improving the signal-to-noise ratio and spectral purity of dual-band imaging, providing high-quality data for subsequent temperature field inversion and three-dimensional morphology reconstruction, and enhancing the overall anti-interference capability and measurement accuracy of the monitoring system.

[0019] (3) In the 3D reconstruction process of this application, the first-level denoising process uses bilateral filtering to perform two iterations on the disparity map extracted by the residual network to initially denoise and balance spatial continuity; the second-level denoising process uses bilateral filtering to perform five iterations on the depth map of geometric reconstruction to further remove noise, optimize details and improve reconstruction accuracy, and finally outputs the optimized depth map. Compared with the traditional method of using only Gaussian filtering to denoise the depth map, this method requires fewer filtering steps and the reconstruction result is more robust. The number of iterations is selected as the optimal combination obtained after combination optimization. Using this combination of iterations can effectively avoid the problem of losing edge details due to too many iterations, and the problem of poor reconstruction quality due to too few iterations.

[0020] (4) This application enables real-time monitoring and dynamic feature correlation analysis of molten pool temperature and molten pool scale. By performing coupled correlation analysis on the coaxial synchronous online monitoring results of the molten pool temperature field and three-dimensional morphology, the dynamic coupling law between molten pool temperature and keyhole features is revealed. Since the traditional reconstruction processes of temperature field and three-dimensional morphology are separate, it is difficult to synchronize the reconstruction results of temperature field and three-dimensional morphology in time, thus making it impossible to effectively determine the dynamic coupling law between molten pool temperature and keyhole features. However, this application obtains dual-band binocular images of the molten pool by designing a special optical path structure and combines it with a unique integrated reconstruction method to achieve synchronous reconstruction of temperature field and three-dimensional morphology. Based on the reconstruction structure, coupled correlation analysis is performed, thereby effectively determining the dynamic coupling law between molten pool temperature and keyhole features. This synchronous monitoring capability enables real-time diagnosis of processing instability, allowing the system to reflect anomalies in the coupling relationship of key parameters such as temperature and depth in real time. This allows for rapid identification of defects in the early stages of defect formation (such as determining process instability in the early stages of defects such as porosity and lack of fusion), thereby effectively preventing the formation of defective parts due to process instability. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0022] To address the aforementioned problems, this invention proposes a method and system for simultaneous online monitoring of molten pool dynamic temperature and three-dimensional morphology in metal additive manufacturing. Through an innovative optical path design, the system can acquire and process dual-band binocular images of the molten pool in real time, thereby achieving coaxial synchronous monitoring of molten pool temperature and three-dimensional morphology on a single device. This optical path structure not only effectively ensures the spatiotemporal consistency of temperature and morphology data, but also offers significant advantages over traditional dual-device parallel solutions, including compact structure, light weight, small size, and low cost.

[0023] Figure 1 A schematic diagram of the structure of a dynamic synchronous monitoring system for molten pool temperature and three-dimensional morphology provided in an embodiment of this application; Figure 2 This is a schematic diagram of the monitoring optical path provided in one embodiment of this application; Figure 3 A dual-band binocular image of a molten pool obtained by a dynamic synchronous monitoring system for molten pool temperature and three-dimensional morphology provided in an embodiment of this application; Figure 4 A flowchart illustrating the synchronous monitoring of molten pool temperature and three-dimensional morphology provided in an embodiment of this application; Figure 5 A flowchart illustrating the coupling relationship between molten pool temperature and depth provided in one embodiment of this application; Figure 6 A schematic diagram of a dynamic synchronous monitoring method for molten pool temperature and three-dimensional morphology provided in an embodiment of this application; Figure 7 A schematic diagram of a real-time monitoring curve of the dynamic behavior of the molten pool provided in an embodiment of this application; Figure 8 A schematic diagram of the temperature field and three-dimensional reconstruction results provided in an embodiment of this application; Figure 9 A temperature depth distribution variation diagram and fitting relationship diagram provided for an embodiment of this application.

[0024] The attached figures are labeled as follows: 1: Additive manufacturing component; 10: Cavity; 11: Laser; 12: Laser processing optical path; 121: Long-pass dichroic mirror; 122: Galvanometer; 123: Field mirror; 13: Substrate; 14: Molten pool; 2: Monitoring optical path; 21: Beam splitter assembly; 211: First reflector; 212: First right-angle prism; 22: First band adjustment assembly; 221: Second reflector; 222: First narrowband filter; 223: Third reflector; 23: Second band adjustment assembly; 231: Fourth reflector; 232: Second narrowband filter; 233: Fifth reflector; 24: Beam combiner assembly; 241: Second right-angle prism; 242: Lens group; 25: Image acquisition unit; 3: Image processing unit. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0026] Current methods involve separately monitoring the temperature field and the three-dimensional morphology of the molten pool. However, there is no method that can simultaneously monitor the temperature field and the three-dimensional morphology of the molten pool in both time and space. This results in temporal and spatial discrepancies between the separately monitored temperature field and the three-dimensional morphology, which is detrimental to the monitoring and analysis of molten pool characteristics.

[0027] Therefore, the purpose of this application is to achieve simultaneous monitoring of the temperature field and three-dimensional morphology in both time and space, thereby facilitating the monitoring and analysis of molten pool characteristics. To achieve this objective, the core concept of this application includes: acquiring a dual-band binocular image of the molten pool through a monitoring optical path. This dual-band binocular image can simultaneously contain information from two bands. Subsequently, by decomposing the dual-band binocular image, the temperature field and three-dimensional morphology can be reconstructed separately. In this way, the reconstructed temperature field and three-dimensional morphology are synchronized in both space and time.

[0028] Figure 1 This is a schematic diagram of a dynamic synchronous monitoring system for molten pool temperature and three-dimensional morphology provided in an embodiment of this application. See also... Figure 1 ,include: Additive manufacturing component 1, used for additive manufacturing; Monitoring optical path 2 is used to acquire binocular images of the dual-band molten pool during additive manufacturing; Image processing unit 3 is used to separate the viewing angles of the dual-band molten pool binocular images, so that the left-view image of the first band and the right-view image of the second band are spatially aligned, and to perform an affine transformation on the right-view image so that the left-view image of the first band and the right-view image of the second band are spatially aligned; and to reconstruct the temperature field and three-dimensional morphology of the molten pool based on the left-view image of the first band and the right-view image of the second band.

[0029] In some examples, additive manufacturing component 1 includes: Cavity 10, laser 11, laser processing optical path 12, and substrate 13; The substrate 13 is disposed inside the cavity 10, the laser 11 is used to generate laser light, and the laser processing optical path 12 is used to transmit the laser light generated by the laser 11 to the substrate 13 inside the cavity 10, thereby forming a molten pool 14 on the substrate 13.

[0030] The additive manufacturing component 1 described above only illustrates the process of molten pool generation. In reality, the complete additive manufacturing process is as follows: First, powder is spread onto the substrate using a powder spreading mechanism. Once the powder is fully spread on the substrate, the laser is activated. The laser beam is transmitted to the substrate surface via a laser processing optical path, thereby generating a molten pool on the substrate surface. The molten pool processes the powder on the substrate surface, thus achieving the printing of the current layer. Generally, components manufactured through additive manufacturing typically consist of several layers; by repeating the powder spreading and printing operations described above, the component can be fabricated.

[0031] In some examples, the laser processing optical path 12 includes: Long-wavelength dichroic mirror 121, galvanometer 122, field mirror 123.

[0032] See Figure 2 In some examples, the monitoring optical path 2 includes: The beam splitting component 21, the first band adjustment component 22, the second band adjustment component 23, the beam combining component 24, and the image acquisition unit 25; The beam splitter 21 is used to split the acquired light from the molten pool into a first beam and a second beam; the first beam is transmitted to the first band adjustment component 22, and the first band adjustment component 22 outputs a band of [missing information]. The first beam; the second beam is transmitted to the second band adjustment component 23, and the output band of the second band adjustment component 23 is... The second beam; the beam combining component 24 is used to combine the wavelength band of The first beam and wavelength are The second beam is combined into a dual-band molten pool signal light, and then the image acquisition unit 25 obtains a dual-band molten pool binocular image based on the dual-band molten pool signal light.

[0033] In some examples, the beam-splitting component 21 includes: First reflecting mirror 211 and first right-angle prism 212; The first reflector 211 is used to divide the acquired beam into a first beam and a second beam, while the first right-angle prism 212 transmits the first beam and the second beam to the first band adjustment component 22 and the second band adjustment component 23, respectively.

[0034] In some examples, the first band adjustment component 22 includes: Second reflector 221, first narrowband filter 222 and third reflector 223; The second reflector 221 is used to transmit the received first beam to the first narrowband filter 222, and the beam is filtered by the first narrowband filter 222 to obtain a wavelength band of [missing information]. The first beam, followed by the third reflecting mirror 223, will have a wavelength of... The first beam of light is transmitted to the light combining component 24.

[0035] In some examples, the second band adjustment component 23 includes: Fourth reflector 231, second narrowband filter 232 and fifth reflector 233; The fourth reflector 231 is used to transmit the received second beam to the second narrowband filter 232. After filtering by the second narrowband filter 232, the resulting wavelength is [band value missing]. The second beam, followed by the fifth reflector 233, will then reflect the wavelength of... The second beam is transmitted to the beam combining component 24.

[0036] In some examples, the light combining component 24 includes: The second right-angle prism 241 and the lens group 242.

[0037] This invention employs an integrated optical path design with dual optical paths and dual center band filters. Through the symmetrical arrangement of a first and second right-angle prism, combined with a first, second, third, and fourth reflecting mirror, it effectively filters... and The dual-band optical path is physically separated and spatially multiplexed to achieve dual-band binocular imaging of the molten pool. The acquired binocular images of the molten pool have... and With two wavelengths, a single device can simultaneously monitor the dynamic temperature and three-dimensional morphology of the molten pool during metal additive manufacturing.

[0038] After the metal additive manufacturing equipment is started, the laser emitted from the laser is reflected by a long-pass dichroic mirror and a galvanometer mirror before acting on the substrate to form a molten pool. During processing, the molten pool radiation light passes through the field mirror and enters the galvanometer mirror, then through the galvanometer mirror and the long-pass dichroic mirror to the monitoring module. The molten pool radiation light is reflected by the first right-angle prism to obtain two beams, namely optical path one and optical path two. Optical path one is reflected by the first reflecting mirror and reaches the first narrowband filter. After being filtered by the first narrowband filter, a wavelength of [wavelength band missing] is obtained. The light beam continues to propagate to the second reflecting mirror. After being reflected by the second reflecting mirror, the light path is guided by the second right-angle prism to the lens group and the industrial camera. The industrial camera acquires the image and transmits it synchronously to the image processing unit to obtain a dual-band fused pool binocular image. The second light path is reflected by the third reflecting mirror and reaches the second narrowband filter. After being filtered by the second narrowband filter, a wavelength of [band unspecified] is obtained. The light beam continues to propagate to the fourth reflecting mirror. After being reflected by the fourth reflecting mirror, the light path is guided by the second right-angle prism to the lens group and the industrial camera. The industrial camera acquires dual-band binocular images of the molten pool and sends them to the image processing unit to reconstruct the temperature and depth distribution of the molten pool online.

[0039] Figure 3 The image shows a dual-band binocular image of a molten pool captured by a dynamic synchronous monitoring system for molten pool temperature and three-dimensional morphology provided in an embodiment of this application. See also... Figure 3 This includes left-view images of the first band and right-view images of the second band. The left-view image (left image) has the following bands: The bands of the right-view image (right image) are .

[0040] Figure 4 A flowchart illustrating the simultaneous monitoring of molten pool temperature and three-dimensional morphology according to an embodiment of this application. See also... Figure 4 ,include: S101. Obtain a dual-band molten pool stereo image.

[0041] In some examples, by means of Figure 1 The system shown obtains as follows Figure 3 The image shown is a dual-band binocular image of a molten pool.

[0042] S102. Perform viewpoint separation on the dual-band molten pool binocular image to obtain the left-view image of the first band and the right-view image of the second band.

[0043] S103. Perform an affine transformation on the right-view image to achieve spatial alignment between the left-view image of the first band and the right-view image of the second band.

[0044] In some examples, affine transformation is a two-dimensional coordinate transformation used to describe geometric distortions such as translation, rotation, scaling, and shearing between images. Its function is to map pixel coordinates in the right-view image to the coordinate system of the left-view image, achieving spatial alignment between the two images. This transformation can be represented by a transformation matrix, which is obtained through least-squares fitting based on the established mapping relationship between the source and target coordinate sets, minimizing the error between the transformed source and target coordinates. The specific steps include: A checkerboard calibration board is used and placed at multiple different spatial positions within the system's imaging field of view. Checkerboard images are acquired from both left and right perspectives. The coordinates of all corner points of the checkerboard are extracted from each image. The corner point coordinates in the left perspective image are used as the target coordinates, and the corresponding corner point coordinates in the right perspective image are used as the source coordinates. A one-to-one coordinate pair is formed point by point. The above operation is repeated for all calibration images acquired at different spatial positions to obtain multiple sets of coordinate pairs, forming a transformation matrix between the source coordinate set and the target coordinate set.

[0045] Based on the solved transformation matrix, the coordinates of the right-view image are remapped pixel by pixel, transforming each pixel in the right-view image to its corresponding position in the coordinate system of the left-view image, so that the right-view image and the left-view image achieve precise spatial correspondence at the pixel level, and obtain a well-matched binocular image pair.

[0046] S104. Based on the left-view image of the first band and the right-view image of the second band, reconstruct the temperature field and three-dimensional morphology of the molten pool.

[0047] In some examples, the process of reconstructing the temperature field of the molten pool is as follows: The grayscale ratio of corresponding pixel positions in the left-view image of the first band and the right-view image of the second band is calculated pixel by pixel, and it is determined whether specular reflection occurs at the corresponding pixel position. If it is determined that no specular reflection occurs in the molten pool at the corresponding pixel position, the temperature value of each point in the molten pool is calculated by substituting it into the dual-wavelength colorimetric thermometry formula, and a molten pool temperature field distribution map is generated. If it is determined that specular reflection occurs in the molten pool at the corresponding pixel position, light intensity correction is performed, and then the temperature value of each point in the molten pool is calculated by substituting it into the dual-wavelength colorimetric thermometry formula, and a molten pool temperature field distribution map is generated.

[0048] The process of determining specular reflection and correcting light intensity includes: The specular reflection determination process involves comparing the grayscale values ​​of corresponding pixels in the left-view image of the first band with those in the right-view image of the second band, pixel by pixel. The difference between the current pixel and the average grayscale value of its neighborhood in the left-view image, and the difference between the corresponding pixel and its neighborhood in the right-view image, are calculated respectively. The neighborhood is defined as the 24 pixels within a 5×5 pixel window centered on the current pixel, excluding the pixel itself, and is taken from the same viewpoint image as the current pixel. When the difference between one viewpoint and the other is significantly greater, the pixel is determined to be a specular reflection interference point, and specular reflection occurs at that location in the molten pool at the current moment.

[0049] Light intensity correction steps: Under stable conditions without specular reflection, acquire multiple sets of synchronized images of the first band left-view image and the second band right-view image. Calculate and statistically analyze the standard grayscale ratio of the corresponding positions of the two views under pure thermal radiation conditions pixel by pixel. For pixels identified as subject to specular reflection interference, use the actual grayscale of the unaffected view as a benchmark, combined with the pre-statistically calculated standard grayscale ratio, to calculate the theoretical grayscale value of the affected view under non-reflection conditions. Replace the abnormal grayscale of the affected view with this theoretical grayscale value, eliminate the extra light intensity component caused by specular reflection, restore the true grayscale ratio between the two bands determined by the wavelength difference, and complete the light intensity correction.

[0050] The formula for the dual-wavelength colorimetric thermometry principle used is as follows:

[0051] in, and These are the center wavelengths of the first and second bands, respectively. Absolute temperature , It is the ratio of the grayscale values ​​of an image under two coaxial wavelengths; R ( ) / R ( To monitor the spectral response coefficient of the system, the dual-wavelength temperature measurement system and the blackbody furnace were placed in a constant temperature and light-proof environment. Multiple stable temperature points were selected, and dual-wavelength images were acquired simultaneously, with the grayscale ratio extracted. Multiple groups (ln ,1 / T Substitute the data into the dual-wavelength temperature measurement formula, use the least squares method for linear fitting to solve the intercept, and invert to obtain the spectral response coefficient.

[0052] In some examples, the process of 3D topography reconstruction is as follows: The first step is to calculate the grayscale ratio at corresponding pixel positions in the left-view image of the first band and the right-view image of the second band, and to obtain the average grayscale ratio of the entire image or a local area. Based on the average grayscale ratio, the images with relatively low grayscale values ​​in the left-view image of the first band and the right-view image of the second band are subjected to overall grayscale stretching. The brightness level of the images with relatively low grayscale values ​​in the left-view image of the first band and the right-view image of the second band is aligned with the images with relatively high grayscale values ​​in the left-view image of the first band and the right-view image of the second band, thereby compensating for the brightness loss caused by the difference in the optical path of the two bands.

[0053] The second step involves inputting the left-view image of the first band and the right-view image of the second band into the residual network to extract the initial dense disparity map.

[0054] The third step involves using bilateral filtering to perform two iterations on the initial dense disparity map, thereby achieving first-level noise reduction of the initial dense disparity map.

[0055] Step 4: Calculate the initial depth map based on the initial dense disparity map.

[0056] In some examples, the initial depth map is calculated based on the initial dense disparity map using the principle of binocular vision.

[0057] The depth calculation formula based on the binocular vision principle used is as follows:

[0058] in, This is the distance from the molten pool to the camera. The focal length of the main lens. As the baseline, The parallax is the difference between two viewpoint images.

[0059] Step 5: Use bilateral filtering to perform five iterations on the initial depth map to achieve two-stage noise reduction of the initial depth map and obtain the depth map of the melt pool.

[0060] Step 6: Reconstruct the molten pool in three dimensions based on the depth map of the molten pool. Convert the pixel coordinates and depth values ​​of each pixel in the depth map into three-dimensional spatial points. The set of all points constitutes the three-dimensional topography map of the molten pool.

[0061] In this embodiment, the bilateral filtering is a nonlinear filtering method that can smooth noise while preserving edge information. In the first-stage denoising of the disparity map, two bilateral filtering iterations are used to suppress local disparity anomalies and isolated noise points in the residual network output, while maintaining the disparity continuity of the melt pool edge region, providing accurate input for subsequent geometric reconstruction. In the second-stage denoising of the depth map, five bilateral filtering iterations are used to successively eliminate outliers using the statistical characteristics of the depth data, while avoiding the loss of fine geometric features such as keyholes due to over-smoothing. Experiments have verified that the cascaded filtering strategy using this combination of iterations, compared to the traditional Gaussian filtering method, improves depth reconstruction accuracy by approximately 15% while maintaining the sharpness of the melt pool edges, and exhibits stronger robustness to interference from splashes, smoke, and other contaminants.

[0062] Figure 5 A flowchart illustrating the coupling relationship between molten pool temperature and depth, provided for one embodiment of this application. See also... Figure 5 ,include: S201. Obtain a dual-band molten pool stereo image.

[0063] See step S101.

[0064] S202. Perform viewpoint separation on the dual-band molten pool binocular image to obtain the left-view image of the first band and the right-view image of the second band.

[0065] See step S102.

[0066] S203. Perform an affine transformation on the right-view image to achieve spatial alignment between the left-view image of the first band and the right-view image of the second band.

[0067] See step S103.

[0068] S204. Based on the left-view image of the first band and the right-view image of the second band, reconstruct the temperature field and three-dimensional morphology of the molten pool.

[0069] See step S104.

[0070] S205. Spatiotemporal correlation analysis is performed on the temperature field and three-dimensional morphology of the reconstructed molten pool. By locating the coordinates of the highest temperature point of the molten pool, and using the abscissa of the coordinates of the highest temperature point of the molten pool as the boundary, the molten pool is divided into the head and tail of the molten pool. By performing curve fitting of temperature and depth data on the head and tail of the molten pool respectively, the coupling relationship model of the temperature field and three-dimensional morphology of the molten pool is obtained.

[0071] In some examples, step S205 includes: The temperature field of the molten pool obtained synchronously was analyzed in a spatiotemporal correlation with the three-dimensional morphology. The molten pool was divided into a head region and a tail region with the highest temperature point as the boundary. Coupled relationship models of temperature change with depth were established for each region. The temperature-depth relationship of the head region of the molten pool conforms to a quadratic function distribution, while the temperature-depth relationship of the tail region of the molten pool conforms to a cubic function distribution.

[0072] Based on the fitting relationship, the dynamic coupling relationship between molten pool temperature and depth is obtained as follows:

[0073] in The temperature of the molten pool. The temperature of the molten pool. This represents the peak temperature. , , , , , , These parameters are related to material properties and processing parameters. The temperature-depth relationship at the head of the molten pool follows a quadratic function distribution, indicating a positive correlation between temperature and depth in the energy input-dominated region. The temperature-depth relationship at the tail of the molten pool follows a cubic function distribution, indicating a nonlinear temperature decay with depth in the heat conduction and convection-dominated region, with obvious abrupt temperature gradient points.

[0074] Figure 6 The flowchart of a method for dynamic synchronous online monitoring of molten pool temperature and three-dimensional morphology provided in an embodiment of this application illustrates the method framework. After acquiring dual-band binocular images of the molten pool, the image acquisition unit sets up two parallel branches: three-dimensional morphology reconstruction and temperature field reconstruction. In the three-dimensional morphology reconstruction branch, after matching a precise binocular image pair according to step S103, the image is input to a residual network to extract a disparity map. The extracted disparity map is then iterated twice by bilateral filtering for first-level noise reduction. The initial depth distribution is then calculated based on the triangulation principle, followed by five iterations of bilateral filtering to complete second-level noise reduction. Finally, online back-projection yields the three-dimensional depth distribution map of the molten pool at different times. In the temperature field reconstruction branch, the grayscale ratio of corresponding pixel positions in the left-view image of the first band and the right-view image of the second band is calculated pixel-by-pixel. This ratio is then substituted into the dual-wavelength colorimetric thermometry formula to calculate the temperature value at each point in the molten pool, generating a molten pool temperature field distribution map. The two branches are executed synchronously, ultimately achieving dynamic synchronous online monitoring of the molten pool temperature and three-dimensional morphology.

[0075] like Figure 7 As shown, the real-time variation curves of key parameters of the molten pool obtained by online monitoring using the method of this invention are presented. These curves, based on the synchronous processing results of the temperature algorithm module and the three-dimensional topography algorithm module on the dual-band molten pool binocular image, reflect the dynamic changes of the maximum temperature, maximum height, and maximum keyhole depth of the molten pool over time, providing an intuitive quantitative basis for molten pool stability analysis and process quality assessment.

[0076] like Figure 8 As shown, the method of using the coaxial integration of the present invention into the LPBF online monitoring is illustrated. t 0, t 0+30ms t 0+60ms tThe temperature field and 3D reconstruction results were obtained at four time points: 0, 90, and 100ms. The first row shows the original molten pool images directly captured by the sensor at these four time points, clearly recording the initial morphology of the molten pool. The second row shows the temperature distribution map after system processing at the corresponding time points, dynamically displaying the specific temperature distribution of the molten pool from the high-temperature core to the surrounding area using color gradients. The third row shows the synchronously reconstructed molten pool depth distribution map, revealing in detail the geometric morphology and undulating contours of the molten pool in three dimensions. This series of results fully demonstrates that the present invention successfully achieves synchronous, coaxial, and online monitoring of molten pool temperature and 3D morphology during metal additive manufacturing.

[0077] like Figure 9 As shown, the temperature-depth fitting relationship of the cross-section at the peak temperature of the molten pool is presented for two substrates, aluminum alloy (6061T6) and stainless steel (304L), at a certain moment. First, the highest temperature point of the molten pool is located, and the molten pool is divided into a head and a tail section based on this point. By fitting the temperature and depth data curves of the left and right sections respectively, the coupling relationship between temperature and depth within the two sections is obtained. Based on the fitting relationship, the dynamic coupling relationship between molten pool temperature and depth is as follows:

[0078] in The temperature of the molten pool. The temperature of the molten pool. This represents the peak temperature. , , , , , , These parameters are related to material properties and processing parameters. The temperature-depth relationship at the head of the molten pool follows a quadratic function distribution, indicating a positive correlation between temperature and depth in the energy input-dominated region. The temperature-depth relationship at the tail of the molten pool follows a cubic function distribution, indicating a nonlinear temperature decay with depth in the heat conduction and convection-dominated region, with obvious abrupt temperature gradient points.

[0079] Specifically, such as Figure 9 In the embodiment shown, the temperature-depth fitting relationship obtained at the temperature peak point of the aluminum alloy (6061T6) substrate at this moment is as follows:

[0080] In this embodiment, the coefficient of determination (R²) is used. 2 As an evaluation index of the goodness of fit of a regression model, the closer its value is to 1, the stronger the model's explanatory power of the data. The coefficient of determination R0 for the head region of the molten pool is particularly important. 2 =0.9673, coefficient of determination R for the tail region of the molten pool2 =0.8526, Specifically, such as Figure 9 In the embodiment shown, the temperature-depth fitting relationship of the stainless steel (304L) substrate at the temperature peak point at this moment is as follows:

[0081] Among them, the coefficient of determination R of the head region of the molten pool 2 =0.9752, coefficient of determination R for the tail region of the molten pool 2 =0.9508.

[0082] The quantitative analysis above reveals the following dynamic coupling between the molten pool temperature field and its three-dimensional morphology: Molten pool temperature and depth exhibit a significant positive correlation, with temperature increases directly leading to increased molten pool depth. The head region of the molten pool shows a significantly larger temperature gradient than the tail region, reflecting different thermophysical mechanisms—one dominated by energy input and the other by convection—indicating different thermophysical mechanisms. The curvature variation of the temperature-depth curve is directly related to molten pool stability; when the fitting parameters deviate from the baseline range, this characteristic can serve as an important basis for determining process instability and defect generation. These results clearly reveal the differences in temperature and three-dimensional morphological behavior across different regions of the molten pool, providing precise quantitative evidence for a deeper understanding of keyhole dynamic evolution and for optimizing process parameters to suppress defects.

[0083] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for dynamic synchronous online monitoring of molten pool temperature and three-dimensional morphology, characterized in that, include: Acquire dual-band stereo images of the molten pool; Viewpoint separation is performed on the dual-band molten pool binocular image to obtain the left-view image of the first band and the right-view image of the second band. An affine transformation is performed on the right-view image to achieve spatial alignment between the left-view image of the first band and the right-view image of the second band. Based on the left-view image of the first band and the right-view image of the second band, the temperature field and three-dimensional morphology of the molten pool are reconstructed.

2. The method for dynamic synchronous online monitoring of molten pool temperature and three-dimensional morphology according to claim 1, characterized in that, The steps for performing an affine transformation on the right-view image to achieve spatial alignment between the left-view image of the first band and the right-view image of the second band include: A checkerboard calibration board was used and placed at multiple different spatial positions within the system's imaging field of view. Checkerboard images were acquired from both the left and right perspectives. The coordinates of all corner points of the checkerboard were extracted from each image. The corner point coordinates in the left perspective image were used as the target coordinates, and the corresponding corner point coordinates in the right perspective image were used as the source coordinates. A one-to-one coordinate pair was formed for each point. The above operation was repeated for all calibration images acquired at different spatial positions to obtain multiple sets of coordinate pairs, forming a transformation matrix between the source coordinate set and the target coordinate set. Based on the solved transformation matrix, the coordinates of the right-view image are remapped pixel by pixel, transforming each pixel in the right-view image to its corresponding position in the coordinate system of the left-view image, so that the right-view image and the left-view image achieve precise spatial correspondence at the pixel level, and obtain a well-matched binocular image pair.

3. The method for dynamic synchronous online monitoring of molten pool temperature and three-dimensional morphology according to claim 1, characterized in that, The process of reconstructing the temperature field of the molten pool is as follows: The grayscale ratio of corresponding pixel positions in the left-view image of the first band and the right-view image of the second band is calculated pixel by pixel, and it is determined whether specular reflection occurs at the corresponding pixel position. If it is determined that no specular reflection occurs in the molten pool at the corresponding pixel position, the temperature value of each point in the molten pool is calculated by substituting it into the dual-wavelength colorimetric thermometry formula, and a molten pool temperature field distribution map is generated. If it is determined that specular reflection occurs in the molten pool at the corresponding pixel position, light intensity correction is performed, and then the temperature value of each point in the molten pool is calculated by substituting it into the dual-wavelength colorimetric thermometry formula, and a molten pool temperature field distribution map is generated.

4. The method for dynamic synchronous online monitoring of molten pool temperature and three-dimensional morphology according to claim 1, characterized in that, The process of 3D topography reconstruction is as follows: Calculate the grayscale ratio at corresponding pixel positions in the left-view image of the first band and the right-view image of the second band, and statistically obtain the average grayscale ratio of the entire image or a local area. Based on the average grayscale ratio, perform overall grayscale stretching on the images with relatively low grayscale values ​​in the left-view image of the first band and the right-view image of the second band, and align the brightness level of the images with relatively low grayscale values ​​in the left-view image of the first band and the right-view image of the second band with the images with relatively high grayscale values ​​in the left-view image of the first band and the right-view image of the second band, thereby compensating for the brightness loss caused by the difference in the optical path of the two bands. The left-view image of the first band and the right-view image of the second band are input into the residual network to extract the initial dense disparity map. A two-stage noise reduction method is achieved by performing two iterations of bilateral filtering on the initial dense disparity map. The initial depth map is calculated based on the initial dense disparity map; A two-dimensional noise reduction method is used to process the initial depth map through five iterations using bilateral filtering, thereby obtaining the depth map of the molten pool. The molten pool is reconstructed in three dimensions based on its depth map. The pixel coordinates and depth values ​​of each pixel in the depth map are converted into three-dimensional spatial points, and the set of all points constitutes the three-dimensional topography of the molten pool.

5. The method for dynamic synchronous online monitoring of molten pool temperature and three-dimensional morphology according to any one of claims 1 to 4, characterized in that, Also includes: Spatiotemporal correlation analysis was performed on the temperature field and three-dimensional morphology of the reconstructed molten pool. By locating the coordinates of the highest temperature point in the molten pool and using the abscissa of the highest temperature point as the boundary, the molten pool was divided into the head and tail of the molten pool. By performing curve fitting of temperature and depth data on the head and tail of the molten pool respectively, a coupling relationship model between the temperature field and the three-dimensional morphology of the molten pool was obtained.

6. A dynamic synchronous monitoring system for molten pool temperature and three-dimensional morphology, characterized in that, include: Additive manufacturing components used for additive manufacturing; The monitoring optical path is used to acquire binocular images of the molten pool in the dual-band process of additive manufacturing; The image processing unit is used to separate the viewing angles of the dual-band molten pool binocular images, so that the left-view image of the first band and the right-view image of the second band are spatially aligned. The right-view image is subjected to an affine transformation to achieve spatial alignment of the left-view image of the first band and the right-view image of the second band. The temperature field and three-dimensional morphology of the molten pool are reconstructed based on the left-view image of the first band and the right-view image of the second band.

7. The dynamic synchronous monitoring system for molten pool temperature and three-dimensional morphology according to claim 6, characterized in that, The monitoring optical path includes: The system includes a beam splitting component, a first-band adjustment component, a second-band adjustment component, a beam combining component, and an image acquisition unit. The beam splitter is used to split the acquired light from the molten pool into a first beam and a second beam; the first beam is transmitted to a first band adjustment component, and the first band adjustment component outputs a band of [missing information]. The first beam; the second beam is transmitted to the second band adjustment component, and the output band of the second band adjustment component is... The second beam; the beam combiner is used to combine the wavelength band of The first beam and wavelength are The second beam is combined into a dual-band molten pool signal light, and then the image acquisition unit obtains a dual-band molten pool binocular image based on the dual-band molten pool signal light.

8. The dynamic synchronous monitoring system for molten pool temperature and three-dimensional morphology according to claim 6, characterized in that, The beam-splitting component includes: The first reflecting mirror and the first right-angle prism; The first reflector is used to split the acquired light into a first beam and a second beam, while the first right-angle prism transmits the first beam and the second beam to the first band adjustment component and the second band adjustment component, respectively.

9. The dynamic synchronous monitoring system for molten pool temperature and three-dimensional morphology according to claim 7, characterized in that, The first band modulation component includes: The second reflector, the first narrowband filter, and the third reflector; The second reflector is used to transmit the received first beam to the first narrowband filter, and the beam is filtered by the first narrowband filter to obtain a wavelength band of [band value missing]. The first beam, followed by the third reflecting mirror, will have a wavelength of... The first beam of light is transmitted to the beam combining component.

10. The dynamic synchronous monitoring system for molten pool temperature and three-dimensional morphology according to claim 6, characterized in that, The second band adjustment component includes: The fourth reflector, the second narrowband filter, and the fifth reflector; The fourth reflecting mirror is used to transmit the received second beam to the second narrowband filter. After filtering by the second narrowband filter, the resulting wavelength is [band value missing]. The second beam, followed by the fifth reflecting mirror, will have a wavelength of... The second beam is transmitted to the beam combining component.

Images