A method and printing system for controlling overprint accuracy based on a digital printer

Abstract

The application relates to a control system technology and discloses a method for controlling overprint accuracy based on a digital printer and a printing system. By adopting a preset non-visible spectrum mark, the reference mark is deployed to provide accurate reference for subsequent compensation, and interference of the mark on the appearance of a finished product is avoided. Moreover, multi-dimensional overprint deviation parameters containing a deformation coefficient are calculated, and overprint compensation of geometric transformation and / or inkjet control is performed according to the parameters. The overprint accuracy and stability under complex and changeable working conditions are significantly improved, and the efficiency of overprint printing is improved. In addition, the application also discloses a control device and a computer readable storage medium.

Description

Technical Field

[0001] This application relates to the field of control system technology, and in particular to a method, control device, printing system and computer-readable storage medium for controlling the registration accuracy of a digital printer. Background Technology

[0002] With the increasing market demand for personalized, short-run, and fast-delivery printed materials, digital printing technology, with its advantages of no need for plate making, variable data printing, and rapid response, has been widely used in packaging, labeling, and commercial printing. In high-quality color printing, it is typically necessary to precisely overlay multiple color groups such as cyan, magenta, yellow, and black, as well as possible spot colors, onto the same printing substrate to form a final image with rich colors and clear details. This alignment accuracy between the image positions of different color groups is called "registration accuracy." Misregulation can lead to problems such as ghosting, color fringing, and color muddiness in printed materials, seriously affecting product quality.

[0003] In traditional printing (such as offset and gravure printing), registration marks (such as "crosshairs" or "markers") fixed to the edge of the printing plate are often used to achieve registration control. Operators use visual or photoelectric sensors to detect the positional deviation of the marks for each color group and manually or semi-automatically adjust the position of the printing cylinder or substrate. However, this method is inefficient and inaccurate in high-speed, multi-color printing, and relies on human experience, making it difficult to achieve fully automatic, real-time, closed-loop, and precise control.

[0004] The above content is only used to help understand the technical solution of this application and does not represent an admission that the above content is prior art. Summary of the Invention

[0005] The main objective of this application is to provide a method, control device, printing system, and computer-readable storage medium for controlling the registration accuracy of a digital printer, aiming to improve registration accuracy and efficiency while reducing the cost of manual intervention.

[0006] To achieve the above objectives, this application provides a method for controlling the registration accuracy of a digital printer, comprising the following steps:

[0007] At least one preset reference mark is generated for the current printing substrate, and the preset reference mark is fused with the printing graphic data; wherein, the preset reference mark is an identifier that can be detected under a preset non-visible spectrum;

[0008] Based on the fused printed graphic data, the digital printer is driven to execute the printing of the current color group;

[0009] After the current color group is printed, check whether there are any unprinted subsequent color groups;

[0010] If present, an image of the substrate containing the pre-printed pre-defined reference mark is obtained under the pre-defined non-visible spectrum;

[0011] The preset reference mark position in the image of the printing material is compared with the pre-stored target reference position to calculate the real-time overprinting deviation parameter; wherein, the overprinting deviation parameter includes at least one of offset, rotation angle and deformation coefficient;

[0012] Based on the aforementioned overprinting deviation parameters, overprinting compensation is performed on the printed image data of the next color group, and the adjusted printing instructions are fed back to the digital printer; wherein, the overprinting compensation includes geometric transformation compensation of the printed image data, and / or dynamic adjustment of the inkjet control corresponding to the printed image data; the inkjet control includes lateral offset of the inkjet head and / or ink droplet ejection timing calibration.

[0013] Drive the digital printer to perform printing of the next color group.

[0014] To achieve the above objectives, this application also provides a control device, comprising:

[0015] A mark generation module is used to generate at least one preset reference mark for the current printing substrate and to fuse the preset reference mark with the printing graphic data; wherein, the distribution position of the preset reference mark on the printing substrate is determined according to the deformation data of the pre-scanned printing substrate; the preset reference mark is an identifier that can be detected under a preset non-visible spectrum;

[0016] The printing execution module is used to drive the digital printer to execute the printing of the current color group based on the fused printing graphic data;

[0017] The judgment module is used to detect whether there are any unprinted subsequent color groups after the current color group has been printed;

[0018] A multispectral detection module is used to acquire, if present, an image of the substrate containing the pre-printed pre-defined reference mark under the pre-defined non-visible spectrum.

[0019] The deviation calculation module is used to compare the preset reference mark position in the image of the printing material with the pre-stored target reference position to calculate the real-time overprinting deviation parameter; wherein, the overprinting deviation parameter includes at least one of offset, rotation angle and deformation coefficient;

[0020] An adaptive compensation module is used to perform overprint compensation on the printed image data of the next color group based on the overprint deviation parameter, and to feed back the adjusted printing instruction to the digital printer; wherein, the overprint compensation includes geometric transformation compensation of the printed image data, and / or dynamic adjustment of the inkjet control corresponding to the printed image data; the inkjet control includes lateral offset of the inkjet head and / or ink droplet ejection timing calibration.

[0021] The printing execution module is also used to drive the digital printer to execute the printing of the next color group.

[0022] To achieve the above objectives, this application also provides a printing system, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements the steps of the above-described method for controlling the registration accuracy of a digital printer.

[0023] To achieve the above objectives, this application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described method for controlling the overprinting accuracy of a digital printer.

[0024] The method, control device, printing system, and computer-readable storage medium for controlling registration accuracy based on a digital printer provided in this application employ preset non-visible spectral markings. Deploying reference markings provides a precise reference for subsequent compensation while avoiding interference from the markings on the finished product's appearance. Furthermore, by calculating multi-dimensional registration deviation parameters, including deformation coefficients, and performing geometric transformations and / or inkjet control for registration compensation, it can not only correct overall offset and rotation but also accurately compensate for nonlinear local deformation of materials and mechanical micro-errors. The entire process forms a fully automated real-time closed-loop control, reducing the cost of manual intervention. In high-speed continuous printing, it can continuously adapt and adjust, significantly improving registration accuracy and stability under complex and variable working conditions, and increasing the efficiency of registration printing. Attached Figure Description

[0025] Figure 1 This is a schematic diagram of the steps of a digital printer-based method for controlling overprint accuracy in one embodiment of this application;

[0026] Figure 2 This is a schematic diagram of the control device in one embodiment of this application;

[0027] Figure 3 This is a schematic diagram of the internal architecture of the controller of a printing system according to an embodiment of this application.

[0028] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0029] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0030] Furthermore, descriptions using terms such as "first" and "second" in this application are for descriptive purposes only (e.g., to distinguish identical or similar features) and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, features defined with "first" or "second" may explicitly or implicitly include at least one of those features. Additionally, technical solutions from different embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. If a combination of technical solutions is contradictory or impossible to implement, such a combination should be considered nonexistent and not within the scope of protection claimed in this application.

[0031] Reference Figure 1 In one embodiment, the method for controlling the registration accuracy of a digital printer includes:

[0032] Step S10: Generate at least one preset reference mark for the current printing substrate, and fuse the preset reference mark with the printing graphic data; wherein, the preset reference mark is an identifier that can be detected under a preset non-visible spectrum;

[0033] Step S20: Based on the fused printed graphic data, drive the digital printer to perform printing of the current color group;

[0034] Step S30: After the current color group is printed, check whether there are any unprinted subsequent color groups;

[0035] Step S40: If present, obtain an image of the substrate containing the printed preset reference mark under the preset non-visible spectrum;

[0036] Step S50: Compare the preset reference mark position in the image of the printing material with the pre-stored target reference position, and calculate the real-time overprinting deviation parameter; wherein, the overprinting deviation parameter includes at least one of offset, rotation angle and deformation coefficient;

[0037] Step S60: Based on the registration deviation parameter, perform registration compensation on the printed image data of the next color group to obtain the adjusted printing instruction and feed it back to the digital printer; wherein, the registration compensation includes geometric transformation compensation of the printed image data and / or dynamic adjustment of the inkjet control corresponding to the printed image data; the inkjet control includes lateral offset of the inkjet head and / or ink droplet ejection timing calibration.

[0038] Step S70: Drive the digital printer to print the next color group.

[0039] In this embodiment, the execution terminal can be a printing system or other equipment or device (such as a control device) that controls the printing system.

[0040] As described in step S10, the system generates at least one special mark (such as a crosshair, dot matrix, etc.), which can be detected under a preset non-visible spectrum (such as infrared or ultraviolet spectrum), but is invisible or does not interfere with the printed text under visible light.

[0041] Optionally, deformation data (such as elongation and local distortion) of the substrate can be obtained through pre-scanning (e.g., optical scanning) to calculate the optimal distribution position of the markings on the material, and deformation-sensitive areas or image edges can be selected. The graphic data of the preset reference markings is merged with the printing image data of the current color group to generate a final printing file containing hidden markings.

[0042] Optionally, before printing the current color group (e.g., cyan C) begins, a separate scanning unit is used to quickly scan the substrate that is about to be printed to obtain the global deformation (overall shrinkage ratio (e.g., paper “expansion and contraction” due to humidity)), local deformation (fluctuations at the material edges, wrinkling tendency in the middle area, uneven deformation caused by the texture or fiber orientation of the material itself), and position reference (the precise position and angle of the material on the transport platform (possible skew)).

[0043] Optionally, output a "deformation field map" covering the material region or a set of parameters describing the material's geometry (such as offset vectors of grid control points) as deformation data.

[0044] Optionally, based on the deformation data, an algorithm can be used to decide where the marker should be placed.

[0045] Optionally, the markers may be deliberately placed in areas of significant or representative deformation (e.g., where the deformation gradient is greatest, or where the material edges are prone to fluctuation). In this way, the markers can more sensitively capture key deformation information of the material.

[0046] The algorithm ensures that the marker is placed in the blank space or non-critical area of ​​the text and image content, so that it does not affect the main content even if it is faintly visible under visible light.

[0047] Optionally, generate multiple tags (≥3) and distribute them as dispersedly as possible (e.g., in a large triangle or quadrilateral distribution).

[0048] Finally, the system obtains a set of precise marker coordinates, which are the marker target positions tailored to this specific material, relative to the material's theoretical origin (or the previous coordinate system).

[0049] Optionally, the markers can be designed as graphics that are easy for machine vision to recognize, such as solid dots, crosshairs, concentric rings, etc.

[0050] Optionally, printers in the current color group use inks containing components sensitive to infrared (or near-infrared) or ultraviolet (UV) light to print these markings (such as in the 780nm~950nm, 365nm~395nm wavelength range, etc.). These inks are completely transparent or the same color as the substrate under visible light, and are imperceptible to the human eye.

[0051] The system prepares the standard printed graphic data (such as plate C in the CMYK color separation diagram) for the current color group in memory, and overlays it onto this graphic data according to the calculated preset reference marks and their determined position coordinates. This generates a new, composite rasterized image data (post-RIP data).

[0052] This involves creating a "virtual printing plate" at the data level. This plate contains both the patterns and text required by the customer and hidden positioning marks automatically added by the system for subsequent calibration.

[0053] As described in step S20, the digital printer completes the printing of the current color group (such as cyan C) based on the fused graphic data. At this time, the preset reference mark has been printed on the material along with the graphic, but since it is a non-visible spectral material, it is not visible to the naked eye.

[0054] Optionally, based on the fused printed image data, the printing press's control system (RIP server or printer controller) converts the raster data into inkjet control instructions that the printer can understand. These instructions are precise down to when and where each nozzle ejects which color of ink droplets.

[0055] Optionally, control commands are sent to the printhead drive module of the digital printer via a high-speed data interface (such as Ethernet or fiber optic). While the printing material moves precisely (via rollers, belts, or flatbeds), the printhead moves in coordination in the horizontal (scanning direction) and vertical (paper feed direction) directions according to the commands.

[0056] The preset reference mark and the customer's graphics are printed in one go using the same print head, the same print path, and in the exact same print stroke. This ensures that: the geometric center of the mark and the relative position of the surrounding customer graphics are determined instantly during printing, without any additional assembly or alignment errors; the mark and graphics undergo the same drying process and bear the same mechanical tension, and any subsequent deformation that may occur (such as localized paper deformation caused by ink penetration) is synchronized with the surrounding graphics, thus guaranteeing the accuracy of the mark as a representative of the local coordinate system.

[0057] In this way, visible customer graphics and invisible preset reference marks are left on the printing material, which together record the material state after the current color group printing is completed.

[0058] As described in step S30, this step is a logical decision node that determines whether the process enters the calibration loop or terminates the task.

[0059] Optionally, the system checks if there are any unfinished color groups in the printing task (such as magenta M, yellow Y, and black K). If there are no subsequent color groups, the process ends (waiting for the printing of the next substrate until the current batch of printing tasks is completed); if there are, it proceeds to step S40 to continue the calibration compensation process.

[0060] As described in step S40, when it is determined in step S30 that there is a subsequent color group, the image of the printing material containing the mark of the current color group is obtained.

[0061] Optionally, the printed substrate can be photographed using a dedicated sensor or camera under non-visible spectral conditions (such as infrared or ultraviolet light). The preset reference marks will be clearly displayed in the image, but the visible light text may be invisible or weak, facilitating mark extraction.

[0062] Optionally, a non-visible light source matching the characteristics of a preset reference marking ink can be activated. For example, if the marking uses infrared absorbing ink, an infrared LED array lamp with a specific wavelength (such as 850nm or 940nm) can be turned on; if ultraviolet fluorescent ink is used, a UV lamp can be turned on.

[0063] Alternatively, activate or move a dedicated non-visible light industrial camera. This camera is equipped with a filter corresponding to the wavelength of the excitation source to block most visible light and ambient stray light, receiving only the specific non-visible light signal reflected or emitted by the marker.

[0064] Optionally, under illumination by an excitation light source, the preset reference marks on the printing material, due to their special ink composition, appear as high-contrast bright spots or clear graphics in the camera's field of view. In contrast, ordinary customer graphic inks and printing material substrates respond weakly under this non-visible spectrum, appearing as a dull, uniform background in the image.

[0065] Optionally, the camera captures the entire material or a key area containing all the markings at high resolution (e.g., a few micrometers per pixel), generating a high signal-to-noise ratio binarized or grayscale image where the outlines and center points of the markings are extremely clear. The output is a corresponding digital image, the core information of which is the actual imaging position of one or more preset reference markings on the current printing substrate.

[0066] As described in step S50, the mark position in the captured image is compared with the pre-stored target reference position (i.e., the theoretically correct position). The deviation of the actual printing position is calculated by image processing algorithms (such as feature point matching, affine transformation analysis). The overprinting deviation parameters include at least one of the following: offset (translation error of the mark on the plane), rotation angle (tilt or rotation angle error of the mark), and deformation coefficient (such as scaling, twisting parameters; nonlinear deformation caused by material expansion or contraction or printer mechanical error).

[0067] Optionally, based on image processing algorithms (such as edge detection, binarization, morphological operations, and subpixel localization), the precise center point coordinates of all preset reference markers in the image can be extracted.

[0068] Assuming there are n markers, we obtain a set of two-dimensional coordinate points: P1 = { (x1′, y1′), (x2′, y2′),..., (xn′, yn′)}. These coordinates are based on the camera coordinate system and have been transformed to the physical coordinate system of the printing material plane using camera calibration parameters.

[0069] Furthermore, the system acquires pre-stored target reference positions, which are generated during step S10. When the system generates preset reference marks for the current material to be printed, it not only determines their positions but also stores this "design position" in memory or a database. This set of coordinates is relative to the material's theoretical origin or ideal layout, denoted as: P2 = { (X1,Y1), (X2, Y2), ..., (Xn, Yn)}. This is the position where the system expects these marks to appear, already incorporating dynamic adjustments to accommodate material pre-scan deformation.

[0070] Optionally, the overall deviation of the entire layout can be described by fitting an optimal coordinate transformation model (such as an affine transformation or a higher-order model). The system takes P2 as input and P1 as output, and uses a mathematical algorithm (such as the least squares method) to solve for the transformation parameters A and b that minimize the overall error of the two sets of points; where matrix A contains rotation, scaling (isotropic / anisotropic) and shear information; vector b represents the overall translation (offset) in the X and Y directions.

[0071] From the solved transformation matrix A and vector b, the specific overprinting deviation parameters mentioned in the steps can be analyzed.

[0072] The offset comes directly from the translation vector b = [ΔX, ΔY]. T This represents the overall translation of the entire printed page in the X (horizontal) and Y (vertical) directions. For example, ΔX = +0.15mm means that the entire page has shifted to the right by 0.15mm.

[0073] It should be noted that this offset is generally caused by material slippage during transport, paper feed deviation, or mechanical positioning error.

[0074] The rotation angle can be derived from the transformation matrix A. For a pure rotation, A = [[cosθ, -sinθ],[sinθ, cosθ]]. It represents the planar rotation angle of the entire printed page around a central point (usually the page center or a reference point). It is typically expressed in radians or degrees, for example, θ = +0.2° (counterclockwise rotation).

[0075] It should be noted that this rotation angle is generally caused by the material being tilted when it is loaded onto the machine, or by the scanning direction of the print head not being absolutely perpendicular to the material's forward direction.

[0076] The deformation coefficients can refer to local or global geometric changes other than overall translation and rotation. They can be interpreted from more elements of matrix A, mainly including:

[0077] Scale Factor: If A is a diagonal matrix [[Sx, 0], [0, Sy]], then Sx and Sy represent the scaling ratios in the X and Y directions, respectively. Sx = 1.001 indicates that the material is stretched by 0.1% in the X direction (e.g., uniform expansion and contraction of the material during printing due to moisture absorption (paper) or heat (film).

[0078] Anisotropic scaling: When Sx ≠ Sy, the page is scaled differently in the horizontal and vertical directions.

[0079] Skew: If the off-diagonal elements of A are not zero, it indicates that the layout has undergone parallelogram deformation, which will turn the originally rectangular image into a rhombus (generally caused by uneven tension on both sides of the material or unstable speed of the print head during scanning).

[0080] Higher-order deformation: For more complex local deformations (such as wavy edges or central bulges), the system may use a polynomial transformation model or a thin-plate spline interpolation model for fitting. In this case, the deformation coefficients correspond to the polynomial coefficients or the displacement vectors of the mesh control points. This can describe more refined, nonlinear material deformation.

[0081] As described in step S60, based on the registration deviation parameter, the printing data or control command for the next color group is adjusted in advance, in reverse, and precisely, so that the ink droplets to be printed can land exactly on the correct position of the dried previous color group. The inkjet control adjustment methods include:

[0082] Lateral offset: Physically adjusts the lateral position of the inkjet head to compensate for mechanical deviation;

[0083] Ink droplet ejection timing calibration: Adjust the trigger time of the inkjet to match the material movement speed or deformation.

[0084] For example, if displacement, rotation, or deformation of the printing material is detected relative to the ideal state, then the printing data or printhead action of the next color group will be adjusted to make an equal and opposite adjustment.

[0085] Optionally, the system performs a mathematical geometric transformation on the full-page printed image data to be sent to the next color group in a raster image processor or a dedicated image processing module. This transformation is the inverse transformation of the deviation transformation calculated in step S50. If step S50 calculates the transformation from the target to the actual as P1 = A × P2 + b, then for compensation, its inverse transformation needs to be applied to the image data I1 of the next color group:

[0086] I2=A -1 ×(I1-b);

[0087] In effect, this is equivalent to "pre-" offsetting, rotating, and deforming" the image data of the next color group in the opposite direction to the detected material deviation.

[0088] Example: If the material is detected to have shifted 0.2mm to the right, then shift the next color group image data to the left by 0.2mm; if the material is detected to have rotated 0.1° counterclockwise, then rotate the next color group image data clockwise by 0.1°; if the material is detected to have been stretched by 0.05% in the Y direction, then compress the next color group image data by 0.05% in the Y direction.

[0089] Alternatively, inkjet control can be dynamically adjusted to perform overprint compensation.

[0090] Optional, lateral offset of the printhead: In single-pass scanning digital printers, the printhead reciprocates laterally (typically along the X-axis). The system can fine-tune the physical position of each printhead assembly in the lateral direction at the moment of printhead scanning using a precision servo motor or piezoelectric actuator. This is primarily used to compensate for overall lateral offset or as part of rotational compensation (achieving small-angle rotation by combining it with changes in the longitudinal material feed rate).

[0091] Example: X-axis offset +0.15mm detected. Before the next scan, instruct the nozzle assembly's starting position to pre-adjust 0.15mm to the left.

[0092] Optional, droplet ejection timing calibration: This compensates for minute velocity variations or deformations of the material in its direction of motion (typically the Y-axis) by dynamically adjusting the ejection trigger time of each nozzle (or nozzle array). The printer control system precisely knows the theoretical feed rate of the material and the relative position of the printhead to the material. When stretching, compression, or local velocity fluctuations in the material's direction of motion are detected, the system calculates in real time the precise timing when ink should be ejected for each pixel row (or each marking cycle). If a part of the material moves faster, droplets are ejected earlier; if it slows down, ejection is delayed. This essentially "stretches" or "compresses" the image in the time dimension.

[0093] Example: The system detects that the material is locally stretched in the Y direction due to uneven tension in a certain area. During the corresponding printing time period in that area, the system slightly slows down the ink droplet ejection frequency to match the pixel pitch of the printed image with the stretched material, preventing local image compression. This compensates for dynamic, non-uniform deformation, especially the slight speed fluctuations that are unavoidable during material transport.

[0094] In practical systems, geometric transformation compensation and inkjet control can be used in combination. Geometric transformation compensation handles global, static deviations (such as overall offset, rotation, and uniform scaling); inkjet control adjusts and handles local, dynamic deviations (such as instantaneous jitter during material movement and local tension deformation), especially time-sensitive compensation.

[0095] After the above compensation calculations, the system generates a new set of "pre-distorted" printing instructions. This set of instructions may include at least one of the following: compensated dot matrix image data (if geometric transformations were used), updated printhead motion trajectory parameters (such as lateral offset), and a finely calibrated droplet ejection timing table.

[0096] This set of instructions is the adjusted printing instruction, which can be fed back to the digital printer in real time.

[0097] Step S70: After receiving the printing instructions after the overprint compensation processing in step S60, the main controller of the digital printer decomposes these instructions into synchronous and precise hardware actions: driving the material feeding system to move the printing material at a set speed; and / or driving the printhead motion system to perform lateral scanning, and possibly in conjunction with the lateral offset parameters in the instructions; and / or, at a precise spatiotemporal point, triggering the corresponding nozzle to eject ink droplets according to the compensated image data and timing table.

[0098] Because the instruction has been reversed and compensated, when the ink droplet falls on the printing material where the deviation has actually occurred, it is perfectly aligned with the image and text already printed in the previous color group.

[0099] Once completed, steps S30-S70 can be repeated until all color groups are printed, achieving real-time calibration for each color.

[0100] In one embodiment, a preset non-visible spectral marker is used to dynamically deploy a reference marker. This allows the marker's positioning to anticipate and adapt to the initial deformation of the material, providing a precise reference for subsequent compensation while avoiding interference with the finished product's appearance. Furthermore, by calculating multi-dimensional registration deviation parameters, including the deformation coefficient, and performing geometric transformations and / or inkjet control registration compensation accordingly, not only can overall offset and rotation be corrected, but also nonlinear local deformation and mechanical micro-errors of the material can be accurately compensated. The entire process forms a fully automated real-time closed-loop control, reducing the cost of manual intervention. In high-speed continuous printing, it can continuously adapt and adjust, significantly improving registration accuracy and stability under complex and variable working conditions, and increasing the efficiency of registration printing.

[0101] Among them, the reference mark position is pre-adjusted according to the material deformation to avoid the amplification of errors caused by fixed mark position; the non-visible spectrum mark does not interfere with the appearance of the finished product and is suitable for high-quality color printing; each color group is detected and compensated immediately after printing to effectively offset the influence of factors such as material deformation and mechanical drift; through the dual compensation of geometric transformation and inkjet control, the multi-color alignment accuracy is significantly improved (up to sub-pixel level).

[0102] In one embodiment, based on the above embodiments, before the step of generating at least one preset reference mark for the current printing substrate and fusing the preset reference mark with the printing graphic data, the method further includes:

[0103] The printing substrate is pre-scanned to obtain a feature map of surface deformation characteristics;

[0104] Based on the feature map, the region with the smallest expected deformation during the printing process is selected as the distribution position of the preset reference mark on the printing substrate.

[0105] In this embodiment, the preset reference mark is placed in the "stable area" where deformation is least likely to occur during the printing process, rather than being evenly or randomly distributed. This makes the subsequent detected registration deviation more accurately reflect the misalignment of the image and text, rather than being disturbed by the displacement of the reference mark itself.

[0106] Optionally, before printing begins, a comprehensive pre-scan of the substrate is performed using a dedicated sensor (such as a high-resolution optical scanner, a 3D profilometer, or a specific spectral imaging device), and combined with digital image correlation (DIC) to generate a feature map of the material surface deformation characteristics.

[0107] Optionally, deformation characteristics may include uneven material thickness, pre-existing stress areas, and areas of change in elastic modulus. These factors can all cause the material to stretch, shrink, or twist to varying degrees under printing tension and changes in temperature and humidity.

[0108] In addition, for textured materials (such as fabric, leather texture, and certain types of paper), the direction and density of the texture can also affect ink adhesion and local deformation.

[0109] Optionally, the system software performs intelligent analysis on the generated feature map to identify the regions with the lowest expected deformation during subsequent printing as stable regions. These stable regions may possess the following characteristics:

[0110] (1) The material has uniform thickness and a dense structure;

[0111] (2) No pre-existing creases, wrinkles, or stress concentration points;

[0112] (3) The texture direction is consistent and there are no obvious interference features;

[0113] (4) It is far from the edge of the material or the contact point of the clamp, and is less affected by mechanical tension.

[0114] On the feature map, these stable regions will be marked (e.g., blue indicates low deformation regions and red indicates high deformation regions).

[0115] Optionally, based on the above analysis results, preset reference markers can be actively and precisely placed within one or more identified stable regions.

[0116] Optionally, the number and location of the markers can be dynamically adjusted according to the material size and deformation complexity, but the core principle is to ensure that the preset reference markers are located on the most stable anchor points.

[0117] In this way, even if the material undergoes overall or partial deformation during the printing process, the positional change is relatively small because the reference point is set in the stable region. This allows the compensation system to more clearly distinguish whether the overall displacement is caused by material deformation or registration error caused by the printing system itself, thus making a more accurate compensation decision.

[0118] Moreover, stable regions usually mean smooth surfaces and clear features, which is beneficial for rapid and accurate image recognition and localization in the non-visible spectrum, reducing false detection or missed detection rates.

[0119] In one embodiment, a feature map is obtained by pre-scanning the printing material, and a preset reference mark is arranged in the area with the least expected deformation. This improves the stability of the reference mark from the source, so that it is less affected by the material deformation in the subsequent printing process. This provides a more reliable and accurate reference for calculating the registration deviation, and further improves the accuracy and robustness of the overall control system.

[0120] In one embodiment, based on the above embodiments, the method for controlling the registration accuracy of a digital printer further includes:

[0121] Initial parameters, registration deviation parameters, and final compensation effects from historical printing tasks are collected in advance to generate training samples; wherein, the initial parameters include printing parameters, environmental parameters, and material parameters;

[0122] A model for predicting overprinting deviations is trained based on training samples.

[0123] The registration deviation prediction model is used to predict the initial registration deviation based on the input initial parameters before a new printing task begins; and during the registration compensation process, it performs composite compensation on the relevant printed graphic data based on the initial registration deviation and the real-time detected registration deviation parameters.

[0124] In this embodiment, multi-dimensional data, including initial parameters and process and result data, are collected from historically successful printing tasks.

[0125] The initial parameters include:

[0126] (1) Printing parameters: These can include printing resolution, inkjet frequency, printhead temperature, ink type / viscosity, printing speed, number of passes, etc.

[0127] (2) Environmental parameters: These can include workshop temperature, humidity, and environmental cleanliness, etc.

[0128] (3) Material parameters: These can be the type of substrate (such as PET, PP, cotton cloth), weight / thickness, surface tension, initial deformation data obtained from pre-scanning, roll tension settings, etc.

[0129] The process and result data may include registration deviation parameters (sequence data such as offset, rotation angle, and deformation coefficient calculated in real time after printing each color group) and final compensation effect (the registration accuracy evaluation result of the final product (such as the deviation value measured by professional measuring instruments), or the parameters when the system finally reaches a stable compensation state).

[0130] Optionally, the {initial parameters, color group overprinting deviation sequence, and final result} of each historical task can be linked together to form a complete training sample.

[0131] The core logic of the model training based on the above sample patterns is to learn the mapping relationship from initial conditions to the evolution of the bias and the result. It is evident that the model's learning process is a supervised learning problem, with the goal of predicting continuous bias values. Therefore, the model type can be a time-series prediction model or a regression model combined with time-series features (because overprinting bias evolves with the color group sequence (time step)). The model's operating settings are as follows:

[0132] Model input: initial parameters (as conditional features);

[0133] Model output: the predicted initial overprinting error, or even the predicted sequence of the first few color errors;

[0134] Training objective: To enable the model to learn to identify the key factors and their combinations that affect the accuracy of printing from a variety of complex initial conditions, and to predict the type and magnitude of the deviation that is most likely to occur in the printing process under specific initial conditions (i.e., initial registration deviation).

[0135] Example: The model may learn that "when using a thin PET material in a low-tension, high-humidity environment, the initial lateral offset of the system is large and accompanied by a slight clockwise rotation" is a common pattern.

[0136] The specific process of model training is illustrated below:

[0137] The goal is to transform raw, multi-source, heterogeneous industrial data into clean, consistent, and model-learnable numerical feature vectors.

[0138] Optionally, historical data can be aggregated from systems such as the printing press PLC, MES system, environmental sensors, material database, and machine vision inspection unit, indexed by printing task ID and color group serial number. Since the timestamps of different systems may not be completely synchronized, the moment when the visual inspection is triggered after the color group printing is completed can be used as the reference point to align the printing parameters, environmental parameters, material parameters and inspection results (registration deviation parameters) obtained at the moment before that moment.

[0139] For the collected data, non-numerical data are numericalized (e.g., material type (e.g., PET, cotton) is one-hot encoded), and all numerical features (e.g., temperature, humidity, speed) are scaled to a uniform dimension (e.g., 0-1 range) to prevent the model from being dominated by large numerical features.

[0140] Then perform the derived feature creation:

[0141] Physical interaction characteristics: Calculate combinations of characteristics that may reflect physical mechanisms, such as "ambient humidity × material moisture absorption coefficient" and "printing speed ÷ ink viscosity";

[0142] Statistical characteristics: Calculate the moving average and standard deviation of parameters for consecutive tasks on the same roll of material to capture slow drift in equipment status;

[0143] Time series characteristics: For a deviation sequence, its rate of change (first-order difference) and acceleration (second-order difference) can be calculated to describe the dynamic trend of the deviation.

[0144] Optionally, for occasional data loss from sensors, interpolation (such as the average of the preceding and following color groups) or task-type-based filling can be used; statistical methods (such as the 3σ principle) or business rules (such as offset exceeding physical limits) can be used to identify anomalous samples; "harmful anomalies" (data caused by equipment failure or operational errors) and "critical anomalies" (rare but reflecting important operating conditions) can be distinguished; "harmful anomaly" samples can be removed or corrected; "critical anomaly" samples can be marked, and different weights can be assigned during training or used for training a dedicated anomaly detection model.

[0145] Optionally, a suitable model architecture for time series forecasting can be selected and trained using preprocessed data. Based on the prediction objective, the model architecture includes two modules:

[0146] Optionally, Module 1 is a static model that directly predicts the initial deviation. It is primarily used for initial prediction before the task begins. It can be constructed using a high-level regression model or a shallow neural network (preferably a gradient boosting decision tree, such as XGBoost or LightGBM; these models excel at handling tabular data, can automatically select features, have strong fitting capabilities for nonlinear relationships, and offer relatively good model interpretability). The input to Module 1 is a feature vector containing the initial parameters; the output is the predicted deviation of the first color group.

[0147] Optionally, a dynamic model for predicting multi-color deviation sequences in two-dimensional modules is used to predict deviation trends throughout the printing process, enabling more refined composite compensation. A time-series model is employed (preferably a recurrent neural network, such as LSTM (Long Short-Term Memory) or GRU (Gated Recurrent Unit); specifically designed to process sequential data and capture long-term dependencies in deviation evolution over time). The input unit is a time series (the first time step is the "initial parameters," and subsequent time steps can incorporate measured deviations of already printed color groups (using true values ​​during training and the model's own previous predictions or measured values ​​for rolling predictions during deployment)); the output is a deviation sequence for one or more future color groups.

[0148] Optionally, the training dataset can be divided into training, validation, and test sets (e.g., 7:2:1) according to the actual situation, and then the model can be iteratively trained based on this until the model converges.

[0149] Optionally, mean squared error or mean absolute error can be used as the loss function to minimize the difference between the prediction bias and the actual bias.

[0150] Optionally, gradient descent can be performed using optimizers such as Adam or SGD.

[0151] The integrated application of the trained model in the control flow is divided into two stages, incorporating it into the original closed-loop control:

[0152] Pre-New Task Prediction: After a new printing task is loaded but before actual printing begins, the system collects the initial parameters (printing, environment, and material parameters) for this task. These parameters are then input into a pre-trained registration error prediction model. The model outputs the predicted initial registration error (e.g., the predicted offset and rotation that may be detected after the first color group). This prediction provides the system with an expected error, eliminating the need to wait for the first color to finish printing and detect the actual error.

[0153] Composite compensation during execution: When driving the digital printer to print the first color group, the predicted initial registration deviation can be directly used to pre-compensate the printed image data or inkjet control of the first color group. Starting from the second color group, the preset reference mark of the previous color is detected, and the real-time registration deviation parameter is calculated. At this time, the compensation amount calculation is no longer based solely on the real-time deviation, but comprehensively considers both the "predicted initial deviation" and the "real-time detected deviation".

[0154] Optionally, the predicted initial overprinting bias (prior experience) can be organically combined with the real-time detected overprinting bias parameters (posterior facts) to form a more robust and adaptive compensation quantity. The weights of the two can be dynamically allocated based on the historical accuracy of the prediction model and the noise level of the real-time data.

[0155] Optionally, the deviation can be decomposed into systematic error (predictable) and random error (requiring real-time correction) and compensated separately.

[0156] The following uses a specific deviation parameter—lateral offset (ΔX)—as an example to illustrate the entire composite compensation process (the processing logic for other parameters (ΔY, rotation angle, deformation coefficient) is the same).

[0157] The prediction bias (ΔXp) is given by the prediction model based on initial parameters before the task begins. For example, ΔXp = +0.15 mm (the predicted image will shift 0.15 mm to the right). The real-time detection bias (ΔXm1) is obtained by actual measurement after the first color group is printed, using a preset reference mark. For example, ΔXm1 = +0.10 mm.

[0158] Then, based on the predicted and detected deviations, the deviation residuals and the confidence level are calculated.

[0159] Optional, the residual error: E = |ΔXm1 - ΔXp| = 0.05 mm. This value reflects the accuracy of the prediction.

[0160] Optionally, the detection weight (Wp) can be calculated based on the historical performance of the prediction model (such as the average residual and the accuracy of the most recent N tasks) and the similarity between the current task parameters and historical samples. For example, if the model has historically been accurate and the current material is a commonly used type, Wp can be set to 0.7.

[0161] Optionally, Wp = (1 / (1+k×e))×s;

[0162] Where e is the historical average residual of the model (the mean of multiple E values), k is the adjustment coefficient, and s is the similarity between the current task and the historical task set (0~1).

[0163] Optionally, the detection weight Wm = 1 - Wp.

[0164] Optionally, the formula for calculating the composite compensation amount (ΔXc) is as follows:

[0165] ΔXc = (Wp×ΔXp+Wm×ΔXm1) ×α;

[0166] Where α is a global compensation gain coefficient (slightly greater than 1, such as 1.05), used to perform moderate overcompensation to offset some of the hysteresis effects.

[0167] Substituting the example value: ΔXc = (0.7×0.15+0.3×0.10)×1.05 = 0.14175 mm.

[0168] This means that the system will decide to apply a lateral compensation of approximately -0.142 mm (i.e., shift to the left) to the next color group image data.

[0169] Then, the calculated compensation amount is converted into specific execution instructions:

[0170] Optionally, for geometric transformation compensation, the entire layout or a portion of the digital layout file (such as PDF or TIFF) of the next color group can be translated in reverse. This can be done using an image processing library (such as OpenCV) or the API of a RIP (raster image processor) to apply an affine transformation (-0.142 mm, corresponding to ΔYc, corresponding to rotation angle c).

[0171] Optionally, for dynamic adjustment of inkjet control (lower-level, faster), the printer's paper feed mechanism or printhead bracket can be fine-tuned for physical alignment; alternatively, the ejection timing of specific nozzles can be adjusted to compensate for minute errors. The compensation time Δt = ΔXc / paper speed. If the paper speed is 1 m / s, then Δt = 142 μs. In this case, the control system commands the printhead to eject ink droplets 142 microseconds earlier, so that when the ink droplets land on the paper, their position is corrected to the left by 0.142 mm.

[0172] After completing the second color group printing, the dynamic mark is detected again to obtain a new ΔXm2. ΔXm2 is compared with the expected value (e.g., 0) to obtain the residual error of this round of compensation. This residual error will be used as feedback data to fine-tune subsequent compensations for the current task. The historical database is updated to optimize the prediction model.

[0173] Optionally, after the current task is completed, its complete initial parameters, actual deviation sequence, and final results will be used as new training samples and fed back into the historical database. The overprinting deviation prediction model can be fine-tuned or retrained periodically or online with new data to adapt to changes brought about by equipment aging, new materials, and new processes, thus achieving continuous model evolution.

[0174] In one embodiment, a two-layer intelligent control architecture was constructed, which not only optimized the performance of the original real-time feedback loop but also added a feedforward prediction loop to the loop. This enables the entire registration accuracy control system to learn, predict, and evolve. Through this composite compensation mechanism, the system utilizes the foresight of historical experience while maintaining the accuracy of real-time feedback, thereby achieving high-precision registration control in complex and ever-changing printing environments.

[0175] In one embodiment, based on the above embodiments, the registration deviation prediction model is further used to adjust the initial printing parameters according to the first generated initial registration deviation, perform the printing task based on the adjusted initial printing parameters, and update the initial registration deviation based on the adjusted initial printing parameters.

[0176] The updated initial registration deviation is used to perform composite compensation on the relevant printed graphic data by combining the real-time detected registration deviation parameters.

[0177] In this embodiment, this stage occurs before the first color group is printed, but before the task is loaded, and is an iterative decision-making process.

[0178] Optionally, the system receives an initial combination of parameters for a new printing task, such as material type, ambient temperature and humidity, and printing parameters. Based on these initial parameters, the prediction model calculates the expected registration error, i.e., the initial prediction error ΔP1.

[0179] The system determines the magnitude of the initial prediction deviation ΔP1. If the absolute value of ΔP1 is less than the system's preset process tolerance threshold, the current parameters are considered acceptable, and the printing task is executed directly using the currently set parameters.

[0180] If ΔP1 exceeds the threshold, the parameter optimization process is triggered. At this point, the system enters an optimization loop, where the parameter optimization module aims to find a new combination of parameters that reduces ΔP1. This module works backwards to calculate the direction and magnitude of the initial parameter adjustments in order to minimize prediction bias. The tuning process can involve accessing the prediction model's own gradient information, querying a rule-based database of historical successes, or running a physical simulation model.

[0181] After calculation, a set of adjusted initial printing parameters is obtained, called the optimized parameters P1. These parameters are typically variables that can be quickly adjusted and are sensitive to registration, such as printhead temperature, media preheating temperature, or printing speed. The optimized parameters P1 are re-inputted into the prediction model to obtain a new, updated prediction bias value ΔP2. ΔP2 is compared with the original ΔP1. If ΔP2 is significantly better (i.e., its absolute value is smaller) and the convergence condition is met, then P1 and ΔP2 are adopted. Otherwise, multiple iterations can be performed until a satisfactory result is found or the iteration limit is reached.

[0182] Ultimately, the system will determine a set of initial printing parameters for this task, along with a corresponding, more precise baseline prediction deviation. The system will then begin the printing task using these locked initial printing parameters.

[0183] The first color group is printed using parameter P1. After printing, an online inspection system (such as a vision camera) immediately captures a preset reference mark and measures the actual, real-time registration deviation parameter ΔM. Then, the composite compensation amount ΔC is calculated using the composite compensation amount calculation formula (composite compensation amount = (predicted deviation weight × ΔP2 + detection deviation weight × ΔM) × compensation gain coefficient). The generated composite compensation amount ΔC will be converted into specific execution instructions and applied to the printing of subsequent color groups.

[0184] Optionally, after the second color group printing is completed, the system performs another test to verify the compensation effect and feeds back the residual error to the system for fine-tuning the compensation of subsequent color groups. Finally, it is used as learning data to update the entire prediction model and optimize the rule base.

[0185] Thus, the initial predicted ΔP1 is updated to ΔP2 based on the optimized parameter P1. ΔP2 represents the residual deviation that the system is still expected to have after optimizing the process parameters. This value is usually smaller than ΔP1, making subsequent composite compensation start-up more optimal and less burdensome.

[0186] In one embodiment, the initial printing parameters are first adjusted based on model predictions, and the task is executed and the initial deviation prediction is updated based on the adjusted parameters. This is equivalent to a pre-correction before printing begins, which can reduce system errors at the physical level in advance, reduce the pressure of subsequent real-time compensation, and make the entire control process more forward-looking and efficient.

[0187] That is, before printing begins, the system will proactively use predictive models to optimize printing process parameters in reverse, striving to reduce errors from the source, and then combine real-time detection for dynamic compensation.

[0188] In one embodiment, based on the above embodiments, the method for controlling the registration accuracy of a digital printer further includes:

[0189] The change in the registration deviation parameter is calculated based on the registration deviation parameter detected multiple times.

[0190] The analysis of the weight of the initial overprinting deviation in the composite compensation and its impact on the amount of change is as follows;

[0191] Based on the impact results, with the goal of reducing the overprinting deviation parameter, the weight of the initial overprinting deviation in the composite compensation is adjusted.

[0192] In this embodiment, after performing composite compensation, the system continuously performs real-time detection on subsequent color groups (such as the Nth color group, the (N+1)th color group, etc.). Each detection yields an actual, compensated overprinting deviation parameter, denoted as ΔMn. Here, n represents the number of detections.

[0193] At this point, the system does not view each ΔMn in isolation, but focuses on its sequence changes, calculating the amount of change (or trend) between two or more adjacent detection results.

[0194] Adjacent changes: δn = ΔMn - ΔM{n-1}, which reflects whether the error after compensation converges, diverges, or oscillates.

[0195] Then calculate the mean, variance, and slope of ΔMn within a window (e.g., the last 5 times), which reflects the long-term stability of the error.

[0196] Based on the initially set detection weight Wp, a causal relationship model is established between the "setting of weight Wp" and the "observed change δn or trend". The change pattern of the compensated error is analyzed under the current weight Wp (correspondingly, Wm = 1 - Wp). For example:

[0197] If Wp is set too high (over-reliance on prediction), and the actual printing conditions differ from the prediction model assumptions, it is possible to observe ΔMn continuously biased in one direction (systematic deviation), and δn changing slowly. If Wm is set too high (over-reliance on real-time detection), and the detection itself has noise or slight jitter, it is possible to observe ΔMn oscillating at high frequency near zero, and δn fluctuating between positive and negative.

[0198] This indicates that the system needs to determine the extent to which the current undesirable error change pattern (such as divergence or oscillation) is caused by improper weight allocation, rather than other uncontrollable factors (such as sudden material defects or sudden environmental changes).

[0199] Based on the impact results obtained from the above analysis, the system adjusts the weight Wp with the goal of reducing the final overprinting deviation and making it converge to zero quickly and stably.

[0200] Optionally, if the error system deviates systematically, it indicates that the prediction baseline ΔP2 may be inaccurate, or the environment may have drifted. In this case, Wp should be reduced and the weight Wm of the real-time detection ΔM should be increased to make the system more confident in "what it sees".

[0201] Optionally, if the error oscillates at high frequencies, it indicates that there may be an over-reliance on noisy detection signals. In this case, Wp should be increased, and a relatively smooth and stable prediction value ΔP2 should be used for filtering to suppress noise.

[0202] Optionally, since the objective function is to minimize the mean of the most recent |ΔMn|, the decision variable is Wp. Therefore, a simple gradient trial method or a more complex reinforcement learning approach can be used to find the optimal weights online.

[0203] The adjusted weights are immediately applied to the next (n+1) composite compensation calculation. The system then continues to observe the new ΔM{n+1} and its changes, entering the next round of analysis-adjustment loop. This process can continue throughout the entire printing task.

[0204] In this way, the system no longer relies on fixed fusion rules that may not be suitable for the current task, but can adjust its decision logic in real time based on the actual feedback during the printing process. This enables it to better cope with the uncertainty of the prediction model and the changes in the noise level of the detection signal, improving the control accuracy and stability of the system under different operating conditions.

[0205] Optionally, rapid optimization can be performed within this task through weight adaptation. After the task is completed, the history of successful / failed weight adjustments, along with the corresponding working conditions (materials, environment), can be recorded and used to train a more advanced meta-model. The task of this meta-model is: given the initial state of the current task, recommend an initial weight Wp, thereby making the adjustment process start better and converge faster.

[0206] In one embodiment, by analyzing the impact of the weight of the predicted deviation (initial overprinting deviation) in the composite compensation on the actual deviation change, and dynamically adjusting this weight, the control system can self-evaluate and optimize the effectiveness of the compensation strategy. This enhances the system's adaptability under different operating conditions, ensuring that the predictive model and the real-time feedback system can work together in the optimal ratio to achieve optimal accuracy under long-term stable operation.

[0207] In one embodiment, based on the above embodiments, before the step of generating at least one preset reference mark for the current printing substrate and fusing the preset reference mark with the printing graphic data, the method further includes:

[0208] A digital twin corresponding to the digital printer is pre-constructed; wherein the digital twin shares the predictive capability of the overprinting deviation prediction model;

[0209] Before executing a new printing task, the initial parameters and printing graphic data of the new printing task are injected into the digital twin, and the entire multi-color group overprinting process is pre-run through the simulation engine. During the simulation, the digital twin predicts the cumulative deformation and overprinting deviation that may occur after each color group is printed based on the material deformation model and historical data, and generates prediction results.

[0210] Based on the prediction results, adjust the overprinting related parameters when performing new printing tasks; wherein, the overprinting related parameters include at least one of the following: initial printing parameters, baseline settings of inkjet control parameters, initial distribution strategy of preset reference marks, and compensation amplitude of overprinting compensation.

[0211] In this embodiment, a high-fidelity virtual image—a digital twin—is constructed for the physical digital printer. This twin not only includes the printer's geometry, motion, and control models, but also fully shares the registration error prediction model and material deformation model used by the physical printer.

[0212] Optionally, the twin continuously learns from historical missions, accumulating a knowledge base about "what kind of deformation and error is caused by different materials, parameters, and environments".

[0213] When a new printing job arrives, the operator injects the initial parameters of the job (such as material type, thickness, ambient temperature and humidity, preset speed, etc.) and complete multi-color printing graphic data into the digital twin. Then, the simulation engine is started, commanding the digital twin to completely and sequentially simulate the entire multi-color printing process.

[0214] During the simulation, the digital twin performs the following key calculations:

[0215] (1) Color-by-color simulation: Simulate the printing of the first color. Based on the material deformation model, calculate the local or global deformation of the medium (such as expansion, contraction, warping) caused by the deposition of the ink of this color.

[0216] (2) Cumulative transfer: The deformation state caused by the first color is used as a new initial state and transferred to the simulation of the second color. When simulating the printing of the second color, not only the deformation caused by the second color itself should be considered, but also the deformation effect left by the first color should be superimposed. And so on, until the last color.

[0217] (3) Prediction Generation: Finally, the digital twin outputs a detailed prediction result. This result not only includes the final total registration error, but also reveals how deformation and registration error accumulate and evolve step by step throughout the printing process. It predicts the intermediate state of each color group after printing.

[0218] Based on the above predictions, the system can proactively and comprehensively adjust the entire strategy employed in physical printing, rather than just individual parameters. Adjustable registration-related parameters include:

[0219] Initial printing parameters: Adjust printhead temperature, media preheating temperature, printing speed, etc., to suppress excessive deformation detected in the prediction from the source;

[0220] Baseline settings for inkjet control parameters: pre-adjusting droplet size, jet waveform, etc., to adapt to predicted material deformation characteristics;

[0221] The initial distribution strategy for the preset benchmark markers is as follows: increase the marker density in the predicted deformation-sensitive areas or near key alignment points; select the marker shape most suitable for detecting the deformation (e.g., change the crosshairs into a fan-shaped array) based on the predicted deformation pattern (e.g., anisotropic shrinkage); and preset an initial compensation value close to the predicted deviation for subsequent color groups, so that the online compensation system can start working from a starting point closer to the optimal solution and accelerate convergence.

[0222] A set of virtual-verified and optimized overprint-related parameters, which may include optimized initial parameters, inkjet baseline, marking strategy, and compensation presets.

[0223] The physical printing press begins its task using registration-related parameters obtained from the digital twin. The optimized initial parameters provided by the digital twin minimize registration deviations between subsequent color groups. Furthermore, the system can perform a final round of fine-tuning based on more precise environmental sensor readings to ensure a perfect match with the current physical environment.

[0224] In one embodiment, introducing a digital twin and pre-running the entire printing process in a simulation environment allows for the early detection of potential problems (such as cumulative deformation and deviation trends) without consuming actual materials or production time. Based on the simulation predictions, a series of key parameters (initial parameters, inkjet control baseline, marking distribution strategy, and compensation amplitude) can then be adjusted in advance. This significantly reduces trial-and-error costs and scrap rates in actual production, making it particularly suitable for high-value, small-batch customized printing tasks, achieving virtual machine setup and risk prevention before production.

[0225] In one embodiment, based on the above embodiments, the step of adjusting the overprinting related parameters when executing a new printing task based on the prediction results includes:

[0226] The operating environment of the genetic algorithm is set based on the prediction results, and the genetic algorithm is configured to improve the overprinting accuracy and material utilization rate, and to optimize the combination of overprinting related parameters.

[0227] In this embodiment, all the overprinting related parameters to be optimized are encoded into a "chromosome". For example: chromosome = [printer temperature, printing speed, marker density, compensation preset value_color2, compensation preset value_color3, ...].

[0228] Optionally, reasonable upper and lower limits can be set for each parameter variable to form a multi-dimensional space that the algorithm can explore.

[0229] Set the objective function (fitness function) to explicitly tell the genetic algorithm what constitutes a "good" solution. The objective function can be set as a multi-objective optimization, for example:

[0230] Adaptability = A × (reciprocal of final registration accuracy error) + B × (material utilization rate)

[0231] Here, the smaller the overprinting accuracy error, the better (hence the reciprocal), and the higher the material utilization rate, the better. Coefficients A and B are used to balance the weights of the two indicators. The accuracy error comes directly from the simulation prediction results of the digital twin; the material utilization rate can be estimated through simulated inkjet volume, scrap rate, etc.

[0232] Then, set the genetic algorithm parameters, such as population size, number of iterations, crossover rate, mutation rate, etc.

[0233] The process involves iterative evolution and simulation. Initially, a population with multiple different parameter combinations (chromosomes) is randomly generated. For each individual in the population (i.e., a set of parameter combinations), a complete, end-to-end multi-color printing simulation is run in a digital twin. After each simulation, the fitness value of that individual is calculated based on the simulation results (predicted final registration error, estimated material consumption, etc.).

[0234] Optionally, individuals with superior fitness values ​​are selected to enter the next generation (survival of the fittest). The selected individuals are then subjected to crossover (exchanging some parameters) and mutation (randomly fine-tuning some parameters) to generate a new population. This simulates gene recombination and mutation in biological evolution.

[0235] Repeat the "evaluation-selection-crossover mutation" cycle until the set number of iterations is reached, or the fitness value no longer increases significantly.

[0236] After the genetic algorithm finishes, it outputs the parameter combination represented by the individual with the highest fitness. This combination is a solution that achieves the best balance between objectives such as "registration accuracy" and "material utilization" after hundreds or thousands of "trial prints" in the virtual world.

[0237] In one embodiment, the system can incorporate multiple, sometimes conflicting, objectives such as accuracy, cost (materials), and efficiency (speed) into the optimization framework. The genetic algorithm automatically seeks solutions on the Pareto optimal front for decision-makers to choose from.

[0238] In this context, digital twins offer cost-free, risk-free, and rapid "virtual experimentation" capabilities. Feedback data from physical printing continuously optimizes the digital twin model, making it more realistic. A more realistic model makes the optimization results of the genetic algorithm more reliable. More reliable optimization results, in turn, improve the performance of physical printing. This forms a positive feedback loop of intelligent enhancement.

[0239] In one embodiment, genetic algorithms are combined with digital twin simulation to automatically search and optimize combinations of overprinting-related parameters with multiple objectives such as overprinting accuracy and material utilization. This can handle complex optimization problems with multiple variables and nonlinearity, find solutions close to the global optimum, and far exceed the effect of manual experience adjustment, so that the control strategy of the printing process reaches a highly intelligent and optimized level.

[0240] In addition, refer to Figure 2 This application also provides a control device Z10, comprising:

[0241] The mark generation module Z11 is used to generate at least one preset reference mark for the current printing substrate and fuse the preset reference mark with the printing graphic data; wherein, the preset reference mark is an identifier that can be detected under a preset non-visible spectrum;

[0242] The printing execution module Z12 is used to drive the digital printer to execute the printing of the current color group based on the fused printing graphic data;

[0243] The judgment module Z13 is used to detect whether there are any unprinted subsequent color groups after the current color group has been printed;

[0244] The multispectral detection module Z14 is used to acquire an image of the substrate containing the printed preset reference mark under the preset non-visible spectrum, if present.

[0245] The deviation calculation module Z15 is used to compare the preset reference mark position in the image of the printing material with the pre-stored target reference position to calculate the real-time overprinting deviation parameter; wherein, the overprinting deviation parameter includes at least one of offset, rotation angle and deformation coefficient;

[0246] The adaptive compensation module Z16 is used to perform overprint compensation on the printed image data of the next color group based on the overprint deviation parameter, and to feed back the adjusted printing instruction to the digital printer; wherein, the overprint compensation includes geometric transformation compensation of the printed image data, and / or dynamic adjustment of the inkjet control corresponding to the printed image data; the inkjet control includes lateral offset of the inkjet head and / or ink droplet ejection timing calibration.

[0247] The printing execution module Z12 is also used to drive the digital printer to execute the printing of the next color group.

[0248] Optionally, the control device Z10 can be a virtual control device (such as a virtual machine) or a physical device (such as a physical device other than a printing system that can execute the corresponding method).

[0249] Furthermore, this application also provides a printing system, the internal architecture of which can be as follows: Figure 3 As shown, the system includes a processor, memory, communication interface, and input interface connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores data called by the computer programs. The communication interface is used for data communication with external terminals. The input interface is used to receive signals input from external devices. When the computer program is executed by the processor, it implements a digital printer-based registration accuracy control method as described in the above embodiment.

[0250] Those skilled in the art will understand that Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the printing system to which the present application is applied.

[0251] Furthermore, this application also proposes a computer-readable storage medium comprising a computer program that, when executed by a processor, implements the steps of the digital printer-based registration accuracy control method described in the above embodiments. It is understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0252] In summary, the digital printer-based registration accuracy control method, control device, printing system, and computer-readable storage medium provided in this application employ preset non-visible spectral markers and dynamically deploy reference markers. This allows the marker positioning to anticipate and adapt to the initial deformation of the material, providing a precise reference for subsequent compensation while avoiding interference with the finished product's appearance. Furthermore, by calculating multi-dimensional registration deviation parameters, including deformation coefficients, and performing geometric transformations and / or inkjet control registration compensation accordingly, not only can overall offset and rotation be corrected, but also precise compensation can be made for nonlinear local deformation and mechanical micro-errors of the material. The entire process forms a fully automatic real-time closed-loop control, reducing the cost of manual intervention. In high-speed continuous printing, it can continuously adapt and adjust, significantly improving registration accuracy and stability under complex and variable working conditions, and increasing the efficiency of registration printing.

[0253] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media provided in this application and in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual-speed SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0254] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, apparatus, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0255] The above description is only a preferred embodiment of this application and does not limit the patent scope of this application. Any equivalent structural or procedural changes made based on the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A method for controlling the registration accuracy of a digital printer, characterized in that, include: Initial parameters, registration deviation parameters, and final compensation effects from historical printing tasks are collected in advance to generate training samples; wherein, the initial parameters include initial printing parameters, environmental parameters, and material parameters; a registration deviation prediction model is trained based on the training samples; A digital twin corresponding to the digital printer is pre-constructed; wherein the digital twin shares the predictive capability of the overprinting deviation prediction model; Before executing a new printing task, the initial parameters and printing graphic data of the new printing task are injected into the digital twin, and the entire multi-color group overprinting process is pre-run through the simulation engine. During the simulation, the digital twin predicts the cumulative deformation and overprinting deviation that may occur after each color group is printed based on the material deformation model and historical data, and generates prediction results. Based on the prediction results, the overprinting-related parameters are adjusted when executing a new printing task. These overprinting-related parameters include initial printing parameters, as well as at least one of the following: baseline settings for inkjet control parameters, initial distribution strategy for preset reference marks, and compensation amplitude for overprinting compensation. The overprinting deviation prediction model is used to predict the initial overprinting deviation based on the input initial parameters before the start of a new printing task. Furthermore, during the overprinting compensation process, composite compensation is performed on the relevant printed graphic data based on the initial overprinting deviation and the real-time detected overprinting deviation parameters. At least one preset reference mark is generated for the current printing substrate, and the preset reference mark is fused with the printing graphic data; wherein, the preset reference mark is an identifier that can be detected under a preset non-visible spectrum; Based on the fused printed graphic data, the digital printer is driven to execute the printing of the current color group; After the current color group is printed, check whether there are any unprinted subsequent color groups; If present, an image of the substrate containing the pre-printed pre-defined reference mark is obtained under the pre-defined non-visible spectrum; The preset reference mark position in the image of the printing material is compared with the pre-stored target reference position to calculate the real-time overprinting deviation parameter; wherein, the overprinting deviation parameter includes at least one of offset, rotation angle and deformation coefficient; Based on the aforementioned overprinting deviation parameters, overprinting compensation is performed on the printed image data of the next color group, and the adjusted printing instructions are fed back to the digital printer; wherein, the overprinting compensation includes geometric transformation compensation of the printed image data, and / or dynamic adjustment of the inkjet control corresponding to the printed image data; the inkjet control includes lateral offset of the inkjet head and / or ink droplet ejection timing calibration. Drive the digital printer to perform printing of the next color group.

2. The method for controlling printing accuracy based on a digital printer as described in claim 1, characterized in that, Before the step of generating at least one preset reference mark for the current printing substrate and fusing the preset reference mark with the printing graphic data, the method further includes: The printing substrate is pre-scanned to obtain a feature map of surface deformation characteristics; Based on the feature map, the region with the smallest expected deformation during the printing process is selected as the distribution position of the preset reference mark on the printing substrate.

3. The method for controlling printing accuracy based on a digital printer as described in claim 1, characterized in that, The method for controlling the overprinting accuracy based on a digital printer also includes: The change in the registration deviation parameter is calculated based on the registration deviation parameter detected multiple times. The analysis of the weight of the initial overprinting deviation in the composite compensation and its impact on the amount of change is as follows; Based on the impact results, with the goal of reducing the overprinting deviation parameter, the weight of the initial overprinting deviation in the composite compensation is adjusted.

4. The method for controlling printing accuracy based on a digital printer as described in claim 1, characterized in that, The steps for adjusting the overprinting parameters when executing a new printing task based on the prediction results include: The operating environment of the genetic algorithm is set based on the prediction results, and the genetic algorithm is configured to improve the overprinting accuracy and material utilization rate, and to optimize the combination of overprinting related parameters.

5. A printing system, characterized in that, The controller of the printing system includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements the steps of the digital printer-based registration accuracy control method as described in any one of claims 1 to 4.

6. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the digital printer-based overprinting accuracy control method as described in any one of claims 1 to 4.

Images