LPBF grafting printing device and printing method based on CIS wide-width line scanning camera

Abstract

The invention belongs to the technical field of additive manufacturing, and provides an LPBF grafting printing device and method based on a CIS wide-width line scanning camera, and the LPBF grafting printing device based on the CIS wide-width line scanning camera comprises an LPBF printer main body, a CIS sensor assembly, a connection positioning device, a laser galvanometer system and a control system. The CIS sensor assembly is integrated on the powder spreading mechanism through the connecting and positioning device and is provided with an LED light source and a lens array, and an additional auxiliary light source is not needed. The control system adopts a polynomial model of three or more orders to realize coordinate mapping, and combines a centroid coarse registration algorithm and an iterative nearest point fine registration algorithm to complete path alignment; the problems that an existing area-array camera system is insufficient in large-size precision, light source interference exists, distortion correction is complex and coordinate mapping is inaccurate are solved in a targeted mode, the forming breadth is covered by a single CIS sensor assembly by 91.5 mm-2300 mm, the grafting precision is within 0.05 mm, and the large-size and high-precision grafting printing and online detection are suitable for large-size and high-precision grafting printing and online detection.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of additive manufacturing, and particularly relates to a LPBF graft printing device and method based on a CIS wide-line scanning camera. BACKGROUND

[0002] Laser powder bed fusion (LPBF) graft printing takes an existing solid member as a "base", and grows a complex structure by laser melting metal powder epitaxial, which can greatly save materials and time, and is widely used in batch mold manufacturing, heterogeneous material composite, and repair remanufacturing fields. With the upgrading of industrial demand, there is an increasingly urgent demand for large-scale grafting of more than one meter and high-precision grafting of less than 0.1 mm. However, the existing technology has four key defects: Large size and high precision are difficult to balance: the existing system relies on a face array camera for imaging, which can achieve 0.1 mm precision in a 400 mm x 400 mm small size device, but in a meter-level forming device, the field of view and precision of the face array camera cannot be balanced; multi-camera splicing will cause field of view registration error accumulation (error is superimposed once for each additional camera), and non-uniform gray scale, which seriously affects the accuracy of contour extraction and cannot meet the demand for large size and high precision; Multiple interferences caused by auxiliary light sources: the face array camera needs to additionally arrange 3-6 groups of ring / bar auxiliary light sources, which increases the complexity of the device, and the heat generated by the light sources (temperature rise of 3-5℃) will interfere with the stability of the laser galvanometer and focusing lens, causing the laser focal point to deviate; under a super large field of view, the light attenuation is obvious, and "bright center and dark edge" is easy to occur, further amplifying the contour extraction error; Image distortion and correction are complex: the industrial lens matched with the face array camera has a 0.5%-2% pillow / barrel distortion, and the depth of field fluctuation (±2mm) will make the full-width consistency decrease by 10%-15%; complex algorithms such as Zhang's calibration method are needed for compensation, but the operation is tedious and is easily disturbed by environmental light, and there is still an error of 0.03-0.05mm after correction, which is difficult to meet the high-precision grafting; Insufficient coordinate mapping and registration accuracy: the laser galvanometer has a 0.1mm / m nonlinear distortion due to lens error and installation gap, which cannot be corrected by the existing homography model (linear mapping), and the coordinate mapping error is 0.08-0.12mm; and the contour registration is only single translation / rotation correction, without considering the installation angle of the base, so that the step mark of the grafting part on the powder bed grafting printing focal plane is greater than 0.1mm, affecting the mechanical properties of the part; Therefore, a LPBF graft printing device and method based on a CIS wide-line scanning camera is needed to solve the above problems. SUMMARY

[0003] The application aims to provide a LPBF graft printing device and method based on a CIS wide-line scanning camera to solve the problems raised in the background.

[0004] To achieve the above object, the present application provides the following technical scheme: a LPBF grafting printing device based on a CIS wide-line scanning camera, comprising a LPBF printer main body, a CIS sensor assembly, a connecting positioning device, a laser galvanometer system and a control system; the connecting positioning device is installed on the powder laying mechanism of the LPBF printer main body, and comprises an integrated mounting support and an adapter; the integrated mounting support spans above the powder laying mechanism and is made of aluminum alloy or carbon fiber; one end of the adapter is fixed with the CIS sensor of the CIS sensor assembly through a screw, and the other end is positioned and screwed to the integrated mounting support through a pin hole, so that the perpendicularity deviation of the long axis of the CIS sensor assembly and the powder laying direction of the powder laying mechanism is less than 0.1°, and the distance between the light entrance surface of the CIS sensor assembly and the focal plane of the powder bed grafting printing is 10-20 mm; The CIS sensor assembly integrates a linear array sensor, an LED light source and a lens array, and the horizontal pixel number and the vertical line frequency are adapted to the forming width and the target grafting resolution of the printer; The laser galvanometer system is in communication connection with the control system, and the control system is configured with a coordinate mapping module and a path registration module; The device is based on the LPBF printer main body, and the core innovation is the integrated installation of the CIS sensor assembly and the powder laying mechanism: the integrated mounting support is made of aluminum alloy / carbon fiber, which takes into account rigidity and lightweight, and avoids increasing the load of the powder laying mechanism; the adapter is designed by "pin hole positioning + screw fixing", which ensures that the perpendicularity deviation of the long axis of the CIS sensor assembly and the powder laying direction of the powder laying mechanism is less than 0.1°, and eliminates the geometric distortion of the line scanning; the CIS sensor assembly integrates the CIS sensor, the LED light source and the lens array, without the need for additional light sources and lenses, and is adapted to the forming width and resolution requirements; the overall structure does not need to be spliced by multiple cameras, and a single CIS sensor assembly covers a width of 91.5mm-2300mm, thus solving the contradiction between large size and high precision from the root.

[0005] In a further technical scheme, the adapter is provided with a positioning pin hole and a screw mounting hole; the positioning pin hole is used to limit the relative displacement of the adapter and the integrated mounting support in the X and Y directions; and the screw mounting hole realizes the detachable fixed connection of the adapter, the integrated mounting support and the CIS sensor of the CIS sensor assembly; The positioning pin hole of the adapter adopts H7 / g6 transition fit to limit the displacement deviation in the X and Y directions to be less than or equal to 0.01mm; the M3 internal hexagonal screw is pre-tightened at a torque of 5-8N·m to realize the rigid connection with the integrated mounting support and the CIS sensor; the screw hole adopts a countersunk design to avoid interference with the light entrance surface imaging; this structure ensures that the amplitude of the CIS sensor assembly is less than or equal to 0.005mm when the powder laying mechanism moves at a high speed of 200mm / s, guarantees the continuity of the line scanning image, and reduces the complexity of the subsequent correction algorithm.

[0006] Further technical solutions, the light entrance surface of the CIS sensor assembly is provided with a dust protection structure, and the dust protection structure is a quartz glass light window or an air curtain generating device; The light entrance surface protection scheme of the CIS sensor assembly: the quartz glass light window adopts 92% light transmittance borosilicate glass, is sealed and connected with the light entrance surface through a fluororubber O-shaped sealing ring, isolates 99% metal dust, and a surface anti-reflection film avoids glare; the air curtain generating device forms a 5mm thick uniform air curtain in front of the light entrance surface by using 0.3MPa filtered compressed air, and blows away the suspended dust in real time. The two schemes can be flexibly selected according to the printing material, ensuring that the dust coverage rate of the light entrance surface is less than 0.1%, and ensuring long-term imaging stability.

[0007] Further technical solutions, the number of horizontal pixels of the CIS sensor assembly adopts a 2-fold safety margin configuration, the vertical line frequency adopts a 2-fold safety margin configuration, and the scanning width is 91.5mm-2300mm; The 2-fold safety margin is optimized for actual working conditions: when the powder laying mechanism is reduced to 150mm / s due to powder flowability, or the forming width is deviated by +5mm due to substrate installation deviation, the 16K horizontal pixels of the CIS sensor assembly still maintain a resolution of 0.05mm / px under a width of 805mm, and the 4k vertical line frequency still maintains a resolution of 0.0375mm / px at a speed of 150mm / s, both of which are better than the target accuracy, and are suitable for the current maximum 2050mm forming size of the LPBF device.

[0008] Further technical solutions, the coordinate mapping module stores a third-order and above polynomial mapping model, and the path registration module integrates a centroid coarse registration algorithm and an iterative closest point fine registration algorithm; The third-order polynomial model of the coordinate mapping module collects 25 feature points covering the powder bed grafting printing focal plane, fits the coefficients by the least square method, and corrects the laser galvanometer distortion of more than 95%; in the path registration module, the centroid coarse registration is quickly positioned by the geometric moment, and the iterative closest point algorithm (≤20 times of convergence, registration error <0.005mm, realizing "high efficiency + high precision" matching.

[0009] A LPBF grafting printing method based on a CIS wide-width line-scan camera, applied to the LPBF grafting printing device based on the CIS wide-width line-scan camera described in any of the above, comprising the following steps: S1, determining the number of horizontal pixels and the vertical line frequency requirement of the CIS sensor assembly according to the forming width of the LPBF printer and the target grafting resolution, selecting and adapting the device, and then installing the CIS sensor assembly on the powder laying mechanism through the connection positioning device to complete the integration; S2, establishing the mapping relationship between the image pixel coordinates and the laser galvanometer coordinates by using the in-situ self-calibration method, verifying the accuracy by the mean Euclidean error (MRE), and ensuring that the MRE is less than or equal to 0.05mm; S3, install the base to be grafted on the lifting platform of the forming cylinder, adjust the height of the lifting platform, so that the upper surface of the base is in the best focal plane of the CIS sensor assembly; S4, control the CIS sensor assembly to move with the powder laying mechanism, perform single or multiple scans on the surface of the base, obtain the pixel coordinates of the profile of the base, reconstruct the physical profile through coordinate mapping, after registration with the slice profile of the model to be printed, the laser galvanometer system executes layer-by-layer powder laying and melting printing according to the corrected path; The selection needs to be combined with the equipment parameters: when the forming width is 1000mm and the target resolution is 0.1mm, the number of horizontal pixels of the CIS sensor assembly is ≥1000mm ÷ 0.1mm / px=10000, and a 20K pixel CIS sensor is selected with a 2 times margin; when the powder laying vehicle speed of the powder laying mechanism is 180mm / s, the vertical line frequency is ≥180mm / s ÷ 0.1mm / px=1800 lines / s, and a CIS sensor assembly with a line frequency of 4000 lines / s is selected with a 2 times margin. When installing, use a dial indicator to calibrate the distance between the light entrance surface and the powder bed grafting printing focal plane, ensure that the full-width gray scale deviation is <5%, and lay the accuracy foundation for the subsequent links; The base to be grafted is fixed to the lifting platform through a tool fixture, and the platform height is adjusted by a dial indicator, so that the upper surface of the base is parallel to the bottom surface of the powder laying mechanism scraper with a deviation <0.02mm, and is coplanar with the powder bed grafting printing focal plane, ensuring that it is in the best focal plane of the CIS sensor assembly, avoiding blurred profile edges caused by inaccurate focusing.

[0010] In a further technical solution, in step S2, the in-situ self-calibration method comprises: Laying a photosensitive color developing medium or metal powder on the working plane corresponding to the powder bed grafting printing focal plane, and scanning and marking the laser galvanometer according to the MxN feature point array; Control the CIS sensor assembly to scan the calibration medium with the powder laying mechanism, and obtain a full-width calibration image containing laser marking traces; Preprocess the calibration image by denoising and binarization, extract the center coordinates of the feature points in the calibration image using a sub-pixel edge detection algorithm, obtain the pixel coordinates (u, v) of each feature point in the image coordinate system, and record the pixel coordinate set of all feature points.

[0011] In a further technical solution, in step S2, the polynomial mapping model is: ; Wherein, n≥3 (polynomial order), x, y are the physical coordinates of X, Y directions in the laser galvanometer coordinate system, u, v are the image pixel coordinates collected by the CIS sensor assembly, 、 are the fitting coefficients of x, y directions, and the coefficients are obtained by least squares fitting of the pixel coordinates of the feature points and the physical coordinates of the galvanometer.

[0012] Further technical solutions, in step S4, the profile extraction adopts Canny operator, Sobel operator or gradient-based algorithm, and the profile reconstruction is realized by curve fitting or polygon approximation; in step S2, the formula for calculating the longitudinal equivalent resolution of the CIS sensor assembly is: ; Wherein, Δy is the longitudinal equivalent resolution, is the movement speed of the CIS sensor assembly with the powder laying mechanism, is the longitudinal line frequency of the CIS sensor assembly; The CIS sensor assembly scans at a speed of 180mm / s, and a single acquisition of 1000mm×800mm, 20K×4K resolution image, JPEG2000 compression transmission time <0.5s; Canny operator separates the base profile from the powder bed grafting printing focal plane background, B-spline curve (3 order) reconstructs the profile, and the fitting error is <0.005mm; after registration, the path alignment error with the base is <0.01mm, the laser galvanometer is exposed on the powder bed grafting printing focal plane at a power of 200W and a speed of 1000mm / s, the metallurgical bonding strength of the metal powder with the base is ≥80% of the substrate strength, and the step mark of the grafting site is <0.01mm.

[0013] Further technical solutions, in step S4, the registration process includes: first, obtaining the initial rotation / translation offset through the centroid coarse registration, and then solving the optimal rotation matrix and translation vector through the iterative closest point algorithm, and performing rigid body transformation on the printing path coordinates, and the transformation formula is: ; In step S2, the MRE verification formula is: ; Wherein, in the transformation formula: , is the scanning path coordinate of the corrected laser galvanometer on the powder bed grafting printing focal plane, is the final rotation offset, , is the path coordinate before correction, , is the final translation offset; in the MRE formula: MRE is the average positioning accuracy error of the coordinate mapping model, M is the total number of test feature points, , is the true value of the physical coordinates of the kth test point on the powder bed grafting printing focal plane, , is the pixel coordinate of the kth test point in the CIS sensor assembly scanning image, , a mapping function from pixels to galvanometer coordinates; The calibration medium is selected according to scenes: 0.1mm thick thermal paper is used for normal temperature scenes, and 20-53um metal powder is used for high temperature scenes; a 6*6 chessboard is selected as a feature point array, and covers the full width. After preprocessing, the image signal-to-noise ratio is greater than or equal to 30dB, and the sub-pixel edge detection positioning accuracy is less than 0.1px; after fitting by a third-order polynomial model, 10 test points are qualified when the MRE is less than or equal to 0.05mm, the calibration time is less than 5min, and the efficiency is improved by 5 times compared with traditional offline calibration.

[0014] Compared with the prior art, the present application has the following advantages: The present application solves the contradiction between large size and high precision: the CIS sensor assembly is designed to cover a forming width of 91.5mm-2300mm, without the need for multi-camera splicing, avoiding registration errors and gray level unevenness; the 2 times safety margin parameter configuration ensures that the resolution is still maintained at the level of 0.1mm under extreme working conditions, and the accurate installation of the integrated mounting support ensures that the grafting accuracy is within 0.05mm, meeting the requirements of large size and high precision such as meter-level mold repair; The present application eliminates the interference of auxiliary light sources: the CIS sensor assembly integrates an LED strip light source, eliminating the need for additional light sources and simplifying the device structure; the light source temperature rise is less than 1℃, avoiding laser galvanometer focal point deviation; the full-width illumination is uniform, and the profile extraction error is reduced from 0.05mm to 0.01mm; The present application simplifies distortion correction and improves consistency: the distortion rate of the integrated lens array of the CIS sensor assembly is less than 0.01%, and the depth of field fluctuation has less than 1% impact on the full-width consistency, without the need for complex correction algorithms; the dust protection structure on the light entrance surface ensures that the long-term imaging consistency deviation is less than 3%, solving the problem of unstable precision of existing systems; The present application improves the accuracy of coordinate mapping and registration: the third-order polynomial model corrects more than 95% of the galvanometer distortion, and the MRE quantitative verification ensures the mapping accuracy; the double registration algorithm realizes sub-pixel level matching, the rigid body transformation corrects the base inclination, and the step mark of the grafting part on the powder bed is less than 0.01mm, and the tensile strength of the part is improved by 10%-15%.

[0015] In order to more clearly illustrate the structural features and effects of the present application, the present application will be described in detail below in combination with the drawings and specific embodiments. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 It is a structure schematic diagram of the present application in perspective view; Figure 2 It is a structure schematic diagram of the present application in perspective view; Figure 3 It is a structure schematic diagram of the present application in perspective view; Figure 4The flowchart of the whole application.

[0017] In the figure: 1, powder laying mechanism; 2, CIS sensor assembly; 201, CIS sensor; 202, adapter; 203, light entrance surface; 3, integrated mounting support; 4, powder bed grafting printing focal plane. DETAILED DESCRIPTION

[0018] The application will be further described below in combination with examples.

[0019] The following examples are used to illustrate the application, but cannot be used to limit the protection scope of the application. The conditions in the examples can be further adjusted according to specific conditions, and simple improvements to the method of the application under the concept of the application also belong to the scope of protection of the application.

[0020] Please refer to Figures 1-4 The application provides a LPBF grafting printing device based on a CIS wide-line scanning camera, which comprises a LPBF printer main body, a CIS sensor assembly 2, a connecting and positioning device, a laser galvanometer system and a control system; the connecting and positioning device is installed on a powder laying mechanism 1 of the LPBF printer main body and comprises an integrated mounting support 3 and an adapter 202; the integrated mounting support 3 spans above the powder laying mechanism 1 and is made of aluminum alloy or carbon fiber; one end of the adapter 202 is fixed with the CIS sensor 201 of the CIS sensor assembly 2 through a screw, and the other end is positioned and screw-connected with the integrated mounting support 3 through a hole, so that the perpendicularity deviation between the long axis of the CIS sensor assembly 2 and the powder laying movement direction of the powder laying mechanism 1 is less than 0.1°, and the distance between the light entrance surface 203 of the CIS sensor assembly 2 and the powder bed grafting printing focal plane 4 is 10-20 mm. The CIS sensor assembly 2 integrates a linear array sensor (i.e. the CIS sensor 201), an LED light source and a lens array, and the lateral pixel number and the longitudinal line frequency are adapted to the forming width of the printer and the target grafting resolution. The laser galvanometer system is in communication connection with the control system, and the control system is configured with a coordinate mapping module and a path registration module.

[0021] In this embodiment, the contradiction between large size and high precision is structurally broken through. The single CIS sensor assembly 2 can cover a forming width of 91.5 mm-2300 mm, directly adapting to the current largest LPBF equipment with a size of 2050 mm, without multi-camera partition splicing, completely avoiding the accumulated field of view registration error (from 0.08 mm to 0) and the gray level non-uniformity problem (the full-width gray level deviation is reduced from 15% to 5%) caused by traditional area array camera splicing. The integrated design eliminates the interference of light source. The relative position of the LED strip light source of the CIS sensor assembly 2 and the photosensitive array (CIS sensor 201) is fixed, and the illumination uniformity is >90%. Without additional arrangement of ring / strip auxiliary light source, the internal structure of the equipment is simplified (3-6 groups of light sources and supports are reduced, and the space occupation is reduced by 40%), and the influence of the light source thermal effect (temperature rise is reduced from 3-5°C to <1°C) on the stability of the laser galvanometer is avoided, and the laser focal point offset is reduced from 0.05 mm to 0.01 mm. The precision closed loop is established. The "pin hole positioning + screw fixing" connected to the positioning device ensures that the CIS sensor assembly 2 has no vibration offset (amplitude ≤0.005 mm), and the fixed distance (10-20 mm) between the light entrance surface 203 and the printing focal plane 4 of the powder bed guarantees the consistency of the full-width imaging definition, providing high-quality original data for subsequent coordinate mapping and contour extraction. The final grafting precision can reach within 0.05 mm.

[0022] Specifically, the adapter 202 is provided with a positioning pin hole and a screw mounting hole. The positioning pin hole is used to limit the relative displacement of the adapter 202 and the integrated mounting support 3 in the X and Y directions, and the screw mounting hole realizes the detachable fixed connection of the adapter 202, the integrated mounting support 3 and the CIS sensor 201 of the CIS sensor assembly 2.

[0023] In this embodiment, the installation deviation is accurately controlled. The positioning pin hole adopts H7 / g6 transition fit, which can limit the relative displacement deviation of the adapter 202 and the integrated mounting support 3 to ≤0.01 mm. Combined with the rigid connection of the M3 internal hexagonal screw, the perpendicularity deviation of the long axis of the CIS sensor assembly 2 and the powder laying motion direction of the powder laying mechanism 1 is ensured to be <0.1°, avoiding geometric stretching or compression of the line scan image. The dynamic stability is ensured. When the powder laying mechanism 1 moves at a high speed (200 mm / s), the rigid fixation of the screw can suppress the vibration of the CIS sensor assembly 2, avoid the line scan image from being broken or blurred due to vibration, and reduce the edge positioning error to 0.01 mm from 0.03 mm during contour extraction. The detachable screw connection design shortens the disassembly and assembly time of the CIS sensor assembly 2 from 1 hour to 15 minutes, reduces the equipment maintenance cost, and avoids the interference of the screw head with the imaging light path of the light entrance surface 203.

[0024] Specifically, the light entrance surface 203 of the CIS sensor assembly 2 is provided with a dust protection structure, which is a quartz glass light window or an air curtain generating device.

[0025] In this embodiment, the harsh printing environment is adapted to the high metal dust suspension rate in the LPBF printing process. The quartz glass light window can physically isolate more than 99% of the dust, and the air curtain generating device can blow away the suspended dust in front of the light entrance surface 203 in real time, ensuring that the dust coverage rate of the light entrance surface 203 of the CIS sensor assembly 2 is less than 0.1%; the long-term imaging accuracy is ensured; without protection, the light entrance surface 203 of the CIS sensor assembly 2 is prone to dust accumulation, and the imaging clarity decreases by 30% within 1 month. After adding the protection structure, the clarity decreases by less than 5% within 6 months, avoiding contour extraction errors caused by dust pollution; the equipment life is extended; dust entering the CIS sensor assembly 2 will wear out the CIS sensor 201, and the protection structure can extend the service life of the CIS sensor assembly 2 from 1 year to 2.5 years, reducing equipment replacement costs, and the O-ring sealing design of the quartz glass light window can also prevent the powder from being damp and clumped to pollute the light entrance surface 203.

[0026] Specifically, the number of horizontal pixels of the CIS sensor assembly 2 is configured with a 2-fold safety margin (i.e. the actual number of horizontal pixels selected is greater than or equal to 2 times the target grafting resolution divided by the forming width of the printer), the vertical line frequency is configured with a 2-fold safety margin (i.e. the actual vertical line frequency selected is greater than or equal to 2 times the powder laying mechanism 1 powder laying vehicle movement speed divided by the target grafting resolution), and the scanning width is 91.5mm-2300mm.

[0027] In this embodiment, the actual working condition fluctuations are addressed; the powder laying vehicle speed of the powder laying mechanism 1 may decrease from 200mm / s to 150mm / s due to powder flowability (such as poor flowability of stainless steel powder), and the forming width may increase by 5mm due to substrate installation deviation. The 2-fold safety margin can ensure that when the number of horizontal pixels is 16K, the resolution remains 0.05mm / px on a 805mm wide surface; when the vertical line frequency is 4k lines / s, the resolution remains 0.0375mm / px at a speed of 150mm / s, both of which are better than the target resolution of 0.1mm; the full-scene device is adapted; the scanning width of 91.5mm to 2300mm can cover more than 95% of the forming width of mainstream LPBF devices, and users do not need to purchase CIS sensor assemblies 2 separately for different size devices, reducing device adaptation costs by 30%; defects caused by multi-camera splicing are avoided; traditional multi-surface array camera splicing requires 2-4 hours of calibration, and there is a gray scale inconsistency. The single CIS sensor assembly 2 in this configuration can complete full-width scanning, the calibration time is shortened to 0.5 hours, and the full-width imaging consistency deviation is less than 3%, providing uniform data for subsequent contour reconstruction.

[0028] Specifically, the coordinate mapping module stores a third-order or higher polynomial mapping model, and the path registration module integrates a centroid coarse registration algorithm and an iterative closest point fine registration algorithm.

[0029] In this embodiment, the galvanometer nonlinear distortion is corrected. The laser galvanometer has a 0.1 mm / m pillow-shaped / bucket-shaped distortion due to lens processing errors and installation gaps. The traditional homography model (linear mapping) cannot be corrected. The third-order polynomial model can be fitted by least squares method with 25 full-width feature points, and more than 95% of the nonlinear distortion can be corrected, with a residual distortion of less than 0.01 mm, and a coordinate mapping error reduced from 0.08 mm to 0.02 mm. High-efficiency and high-precision registration is achieved. The centroid coarse registration calculates the geometric moment of the contour, and can reduce the initial deviation of the theoretical contour and the actual contour from 5 mm to 0.5 mm within 0.1 seconds, avoiding the iterative closest point algorithm falling into local optimum. The iterative closest point algorithm (point-to-surface distance measurement) can converge within 20 iterations, with a registration error of less than 0.005 mm, ensuring that the theoretical model and the actual base minimize the mean square error, laying a foundation for error-free grafting.

[0030] A LPBF grafting printing method based on a CIS wide-line scanning camera, applied to the LPBF grafting printing device based on the CIS wide-line scanning camera in the above embodiment, comprising the following steps: S1, according to the forming width of the LPBF printer and the target grafting resolution, determine the number of horizontal pixels and the demand for vertical line frequency of the CIS sensor assembly 2, select and adapt the equipment, and then install it on the powder laying mechanism 1 through the connection positioning device to complete the integration; S2, establish the mapping relationship between the image pixel coordinates and the laser galvanometer coordinates by using the in-situ self-calibration method, verify the accuracy by the mean Euclidean error (MRE), and ensure that the MRE is less than or equal to 0.05 mm; S3, install the base to be grafted on the lifting platform of the forming cylinder, adjust the height of the lifting platform, so that the upper surface (grafting joint surface) of the base is in the best focal plane of the CIS sensor assembly 2 (i.e. the distance between the light entrance surface 203 and the grafting joint surface is 10-20 mm, and the powder bed grafting printing focal plane 4 is coplanar); S4, control the CIS sensor assembly 2 to move with the powder laying mechanism 1, and perform single or multiple scanning on the surface of the base to obtain the pixel coordinates of the base contour, reconstruct the physical contour through coordinate mapping, and after registration with the model slice contour to be printed, the laser galvanometer system performs layer-by-layer powder laying and fusion printing according to the corrected path.

[0031] In this embodiment, a full-process precision closed loop is formed - S1 ensures that the CIS sensor assembly 2 parameters are fully matched with the device working condition, S2 avoids offline installation error through in-situ calibration, S3 guarantees accurate focusing of the light entrance surface 203, S4 realizes adaptive adjustment of profile-path-printing, and systematically solves the four problems of traditional methods, i.e. insufficient precision of large size, light source interference, complex distortion correction, and inaccurate coordinate mapping; improves batch efficiency - after the first integrated calibration, subsequent batches only need to perform S3-S4, and the single batch printing preparation time is shortened from 2 hours to 30 minutes, while the function reuse rate is improved by 40% for the powder laying mechanism 1 powder laying uniformity and online detection of the fused section.

[0032] Specifically, in step S2, the in-situ self-calibration method includes: A photosensitive color medium (thermal paper / photographic paper) or metal powder is laid on the working plane (substrate) corresponding to the powder bed grafting printing focal plane 4, and the laser galvanometer scans and marks according to the MxN feature point array (M and N are positive integers, and the feature point is a cross / circle / chessboard); The CIS sensor assembly 2 is controlled to scan the calibration medium with the powder laying mechanism 1, and a full-width calibration image containing laser marking traces is obtained; The calibration image is preprocessed by denoising and binarization, the center coordinates of the feature points in the calibration image are extracted by using a sub-pixel edge detection algorithm (Canny operator combined with sub-pixel interpolation), the pixel coordinates (u, v) of each feature point in the image coordinate system are obtained (u is the pixel coordinate in the scanning row direction, and v is the pixel coordinate in the sensor array direction), and the pixel coordinate set of all feature points is recorded.

[0033] In this embodiment, offline calibration error is avoided - traditional offline calibration requires disassembly of the CIS sensor assembly 2 to install the calibration plate, and after reinstallation, the error increases by 0.05 mm, in-situ self-calibration is directly operated on the working plane corresponding to the powder bed grafting printing focal plane 4, without disassembly, and the installation error is reduced to <0.01 mm; adapt to multiple scene printing - 0.1 mm thick thermal paper is selected for normal temperature printing, and 20-53 μm metal powder is selected for high temperature printing, the adaptability of the calibration medium is improved by 90%; provide high-quality mapping data - the MxN feature point array covers the full-width surface, avoids insufficient sampling in the edge area, the image signal-to-noise ratio after preprocessing is ≥30 dB, the sub-pixel edge detection positioning accuracy is <0.1 px, provides sufficient and accurate samples for polynomial model fitting, and the mapping model stability is improved by 60%.

[0034] Specifically, in step S2, the polynomial mapping model is: ; wherein n≥3 (polynomial order), x, y are X, Y direction physical coordinates (unit: mm, corresponding to the physical coordinates of the powder bed grafting printing focal plane 4) in the laser galvanometer coordinate system, u, v are image pixel coordinates collected by the CIS sensor assembly 2, 、 are x, y direction fitting coefficients (j, k are non-negative integers, j+k≤n), and the coefficients are obtained by fitting the pixel coordinates of the characteristic points and the physical coordinates of the galvanometer by the least square method.

[0035] In this embodiment, the non-linear relationship is accurately described - the distortion (pillow-shaped, barrel-shaped) of the laser galvanometer and the installation angle of the CIS sensor assembly 2 are both non-linear errors, and the n=3 polynomial model can accurately fit these errors through cross terms. Compared with the n=2 model, the distortion correction rate is increased from 60% to 95%, and the coordinate mapping error of the powder bed grafting printing focal plane 4 in the edge area is reduced from 0.07 mm to 0.02 mm; the fitting stability is ensured - the least square method can use the data of all characteristic points for global fitting, avoiding the influence of individual abnormal points on the model, and the standard deviation of the fitting coefficient is <0.001, ensuring that the consistency deviation of the mapping model is <3% when printing in different batches; it is compatible with higher precision requirements - when the forming width of the LPBF device is >2000 mm, a polynomial model with n=4 can be selected to further improve the distortion correction precision and meet the grafting needs of super large parts.

[0036] Specifically, in step S4, the contour extraction adopts Canny operator, Sobel operator or gradient-based algorithm, and the contour reconstruction is realized by curve fitting (Bezier / B-spline curve) or polygon approximation (Douglas-Peucker algorithm); in step S2, the formula for calculating the longitudinal equivalent resolution of the CIS sensor assembly 2 is: ; wherein Δy is the longitudinal equivalent resolution (unit: mm / px), is the movement speed of the CIS sensor assembly 2 with the powder laying mechanism 1 (i.e. the powder laying vehicle speed of the powder laying mechanism 1, unit: mm / s), is the longitudinal line frequency of the CIS sensor assembly 2 (unit: lines / s).

[0037] In this embodiment, the high-efficiency separation contour and background - the contour extraction accuracy of Canny operator in the pepper noise environment reaches 95%, and the anti-interference ability is stronger than that of the traditional threshold segmentation; Sobel operator is suitable for the scene with obvious base edge gradient, and the extraction speed is improved by 20%; restoring complex contour morphology - the base contour may have complex structures such as circular arc and inclined surface, B-spline curve (3 order) can fit the smooth contour through control points, with a deviation of <0.005mm, avoiding the contour corner distortion caused by polygon approximation; Quantitative guidance for CIS selection - the longitudinal equivalent resolution formula clearly quantifies the relationship between Δy and v, f, such as the powder laying vehicle speed of powder laying mechanism 1 being 200mm / s and the target Δy being 0.1mm, f≥2000 lines / s can be directly calculated, combined with a 2 times safety margin to select a CIS sensor component 2 of 4000 lines / s, to ensure that the longitudinal resolution is consistent with the transverse resolution (Δx), avoiding the contour stretching caused by the mismatch between the longitudinal and transverse resolutions.

[0038] Specifically, in step S4, the registration process includes: first obtaining an initial rotation / translation offset through centroid coarse registration, and then solving the optimal rotation matrix and translation vector through the iterative closest point algorithm, and performing rigid transformation on the printing path coordinates, and the transformation formula is: ; In step S2, the MRE verification formula is: ; In the transformation formula: 、 is the scanning path coordinate of the modified laser galvanometer in the powder bed grafting printing focal plane 4 (unit: mm), is the final rotation offset (unit: rad), 、 is the path coordinate before correction (unit: mm), 、 is the final translation offset (unit: mm, corresponding to the coordinate offset of the powder bed grafting printing focal plane 4); in the MRE formula: MRE is the average positioning accuracy error of the coordinate mapping model (unit: mm), M is the total number of test feature points, 、 is the true value of the kth test point in the laser galvanometer physical coordinate of the powder bed grafting printing focal plane 4 (unit: mm), 、 is the pixel coordinate of the kth test point in the CIS sensor component 2 scanning image, 、 is the mapping function from pixel to galvanometer coordinate.

[0039] In this embodiment, sub-pixel-level registration is achieved—after rapid reduction of positional deviation through coarse centroid registration, the iterative nearest point algorithm solves for the optimal rotation matrix and translation vector by minimizing the mean square error of the point set. Compared with traditional single translation registration, the registration error is reduced from 0.05mm to 0.005mm; the base mounting tilt angle is corrected—when there is a slight tilt angle in the base mounting, the rotation term in the rigid body transformation formula can accurately correct the path direction, avoiding step marks at the grafting part on the powder bed grafting printing focal plane 4, thus improving the mechanical properties of the part; the mapping accuracy is quantitatively verified—the MRE formula uses 10 test points not involved in the fitting to quantitatively evaluate the reliability of the coordinate mapping model. MRE ≤ 0.05mm ensures that the mapping error is within a controllable range, avoiding batch printing scrap due to inaccurate mapping. At the same time, the MRE data can be used as a basis for equipment calibration.

[0040] Working principle and usage process of this invention: This invention addresses existing shortcomings through a closed-loop process of "CIS high-precision imaging → precise coordinate mapping → adaptive contour registration → laser precision printing," as detailed below: Equipment preparation and CIS integration: Based on the LPBF printer's forming width, target resolution, and the speed of the powder spreading mechanism 1, determine the horizontal pixel count and vertical line frequency (2 times safety margin) of the CIS sensor component 2. After selection, install it through the connection positioning device: the integrated mounting bracket 3 spans the powder spreading mechanism 1, the adapter 202 is positioned by the pin hole, and the CIS sensor component 2 is fixed by screws. The distance between the light-incident surface 203 and the powder bed grafting printing focal plane 4 is calibrated (10-20mm), and the dust protection structure is installed. In-situ self-calibration and coordinate mapping establishment: Thermal paper / metal powder is laid on the substrate corresponding to the focal plane 4 of the powder bed printing. The laser galvanometer scans and marks the feature points in a 6×6 checkerboard pattern. The CIS sensor assembly 2 scans across the calibration medium with the powder laying mechanism 1 to acquire a full-frame image. After denoising and binarization, the Canny operator is used in conjunction with sub-pixel interpolation to extract the pixel coordinates of the feature points. A third-order polynomial model is used, and the least squares method is used to fit the coefficients to establish the mapping between the pixel and the galvanometer coordinates. The MRE is calculated for 10 test points, and ≤0.05mm indicates that the calibration is qualified. Positioning of the grafting base: The base is fixed to the lifting platform by tooling fixtures. The height of the platform is adjusted by dial indicator so that the parallelism deviation between the upper surface of the base and the bottom surface of the scraper of the powder spreading mechanism 1 is <0.02mm, and it is coplanar with the printing focal plane 4 of the powder bed grafting, ensuring that it is at the optimal focal plane of the CIS sensor assembly 2. Vacuum adsorption fixes the base to prevent displacement. Contour acquisition, registration, and printing: The CIS sensor assembly 2 scans the base with the powder laying mechanism 1 to obtain high-resolution images; the Canny operator extracts the contour pixel coordinates, calls the polynomial model to convert into the physical coordinates of the laser galvanometer in the powder bed grafting printing focal plane 4, and the B-spline curve reconstructs the contour; the centroid coarse registration calculates the initial offset, the iterative closest point algorithm solves the optimal rotation matrix and translation vector, and the rigid body transformation corrects the printing path; the powder laying mechanism 1 lays metal powder on the powder bed grafting printing focal plane 4, the laser galvanometer exposes according to the corrected path, and repeats layer by layer until the grafting is completed. Subsequent batches do not need to repeat integration and calibration, and the laser galvanometer can be re-executed after calibration every half year.

[0041] Although embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, replacements and changes can be made to these embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A CIS wide-line-scan camera-based LPBF graft printing device, characterized in that, The application relates to a laser powder bed fusion (LPBF) printer, which comprises an LPBF printer body, a CIS sensor assembly (2), a connecting and positioning device, a laser galvanometer system and a control system; the connecting and positioning device is installed on a powder laying mechanism (1) of the LPBF printer body and comprises an integrated mounting support (3) and an adapter (202); the integrated mounting support (3) is arranged above the powder laying mechanism (1) and is made of aluminum alloy or carbon fiber; one end of the adapter (202) is fixed with the CIS sensor (201) of the CIS sensor assembly (2) through screws; the other end of the adapter (202) is positioned and screwed with the integrated mounting support (3) through a screw hole, so that the perpendicularity deviation between the long axis of the CIS sensor assembly (2) and the powder laying movement direction of the powder laying mechanism (1) is less than 0.1 degrees, and the distance between the light entrance surface (203) of the CIS sensor assembly (2) and the powder bed grafting printing focal plane (4) is 10-20 mm; The CIS sensor assembly (2) is integrated with a linear array sensor, an LED light source and a lens array, the lateral pixel number and the longitudinal line frequency are adapted to the forming width of the printer and the target grafting resolution; The laser galvanometer system is in communication connection with the control system, and the control system is provided with a coordinate mapping module and a path registration module.

2. The CIS wide-line-scan camera based LPBF graft printing apparatus of claim 1, wherein, The adapter (202) is provided with a positioning pin hole and a screw mounting hole; the positioning pin hole is used for limiting the relative displacement of the adapter (202) and the integrated mounting support (3) in the X and Y directions; and the screw mounting hole is used for detachably fixedly connecting the adapter (202) with the integrated mounting support (3) and the CIS sensor (201) of the CIS sensor assembly (2).

3. The CIS wide-line-scan camera based LPBF graft printing apparatus of claim 1, wherein, The light entrance surface (203) of the CIS sensor assembly (2) is provided with a dust protection structure, and the dust protection structure is a quartz glass light window or an air curtain generating device.

4. The CIS wide-line-scan camera-based LPBF graft printing apparatus of claim 1, wherein, The lateral pixel number of the CIS sensor assembly (2) is configured with a 2-fold safety margin, the longitudinal line frequency is configured with a 2-fold safety margin, and the scanning width is 91.5mm-2300mm.

5. The CIS wide-line-scan camera-based LPBF graft printing apparatus of claim 1, wherein, The coordinate mapping module stores a third-order and above polynomial mapping model, and the path registration module integrates a centroid coarse registration algorithm and an iterative closest point fine registration algorithm.

6. A CIS wide-line-scan camera-based LPBF grafting printing method applied to the CIS wide-line-scan camera-based LPBF grafting printing device of any one of claims 1-5, characterized in that, The application further discloses a method for grafting a model on a base, which comprises the following steps: S1, determining the lateral pixel number and the longitudinal line frequency requirement of the CIS sensor assembly (2) according to the forming width of the LPBF printer and the target grafting resolution, selecting an adaptive device, installing the device on the powder laying mechanism (1) through the connecting and positioning device, and completing integration; S2, establishing a mapping relationship between image pixel coordinates and laser galvanometer coordinates by using an in-situ self-calibration method, verifying the accuracy by an average Euclidean error, and ensuring that MRE is less than or equal to 0.05mm; S3, installing the base to be grafted on a lifting platform of a forming cylinder, adjusting the height of the lifting platform, and making the upper surface of the base be located on the optimal focal plane of the CIS sensor assembly (2); S4, controlling the CIS sensor assembly (2) to move with the powder laying mechanism (1), performing single or multiple scanning on the surface of the base, obtaining the pixel coordinates of the base profile, reconstructing the physical profile through coordinate mapping, registering the physical profile with the model slice profile to be printed, and enabling the laser galvanometer system to perform layer-by-layer powder laying and melting printing according to the corrected path.

7. The CIS wide-line-scan camera-based LPBF graft printing method according to claim 6, characterized in that, In step S2, the in-situ self-calibration method comprises: A photosensitive color developing medium or metal powder is laid on a working plane corresponding to the focal plane (4) of the powder bed grafting printing, and a laser galvanometer scans and marks according to an M×N characteristic point array; The CIS sensor assembly (2) is controlled to scan the calibration medium with the powder laying mechanism (1) to obtain a full-width calibration image containing laser marking traces; The calibration image is preprocessed by denoising and binarization, the center coordinates of the characteristic points in the calibration image are extracted by using a sub-pixel edge detection algorithm, the pixel coordinates (u, v) of each characteristic point in the image coordinate system are obtained, and the pixel coordinate set of all characteristic points is recorded.

8. The CIS wide-line-scan camera based LPBF graft printing method according to claim 6, wherein, In step S2, the polynomial mapping model is: ； Wherein, n≥3, x, y are physical coordinates of X, Y direction in laser galvanometer coordinate system, u, v are image pixel coordinates collected by CIS sensor assembly (2), , are fitting coefficients of x, y direction, which are obtained by fitting pixel coordinates of feature points and physical coordinates of galvanometer through least square method.

9. The CIS wide-line-scan camera based LPBF graft printing method according to claim 6, wherein, In step S4, the contour extraction adopts a Canny operator, a Sobel operator or a gradient-based algorithm, and the contour reconstruction is realized by curve fitting or polygon approximation; in step S2, the formula for calculating the longitudinal equivalent resolution of the CIS sensor assembly (2) is: ； where Δy is the longitudinal equivalent resolution, is the movement speed of the CIS sensor assembly (2) with the powder spreading mechanism (1), is the longitudinal line frequency of the CIS sensor assembly (2).

10. The CIS wide-line-scan camera based LPBF graft printing method according to claim 6, wherein, In step S4, the registration process includes: first, obtaining an initial rotation / translation offset through centroid coarse registration, and then solving an optimal rotation matrix and a translation vector through an iterative nearest point algorithm, and performing a rigid body transformation on the printing path coordinates, and the transformation formula is: ； In step S2, the MRE verification formula is: ； Wherein, in the transformation formula: , is the scanning path coordinate of the laser galvanometer after correction in the powder bed grafting printing focal plane (4), is the final rotation offset, , is the path coordinate before correction, , is the final translation offset; In the MRE formula: MRE is the average positioning accuracy error of the coordinate mapping model, M is the total number of test feature points, , is the true value of the kth test point in the laser galvanometer physical coordinate of the powder bed grafting printing focal plane (4), , is the pixel coordinate of the kth test point in the CIS sensor assembly (2) scanning image, , is the mapping function from pixel to galvanometer coordinate.

Images