Laser cleaning and repairing integrated method for nozzle micropores of ink box nozzle

Abstract

The invention relates to a laser cleaning and repairing integrated method for a nozzle micropore of an ink box nozzle in the technical field of new-generation information, which comprises the following steps of: scanning the nozzle fixed on a platform through image acquisition equipment to obtain first micropore coordinate data and a first surface feature image; the first surface feature image captures microstructure details of the surface of the nozzle; an image analysis algorithm is adopted to process the first surface feature image sequence, a first micropore blockage distribution pattern is obtained, and the first micropore blockage distribution pattern represents the overall layout of the blockage position and degree; and uniformity verification is conducted on the first repair layer thickness distribution diagram, if the fluctuation range exceeds a threshold value, the first dynamic repair parameter set is iteratively adjusted, the uniformity verification ensures the layer thickness consistency, and final regeneration nozzle quality data are obtained through verification circulation.

Description

Technical Field

[0001] This invention relates to the field of next-generation information technology, and in particular to an integrated method for laser cleaning and repair of the nozzle micropores of an ink cartridge nozzle. Background Technology

[0002] The recycling and reuse of waste ink cartridge nozzles is a key area of ​​the circular economy for printing consumables, directly related to resource conservation and environmental protection. Current recycling methods primarily involve soaking the nozzles in chemical cleaning agents combined with mechanical needle unclogging to remove ink residue. However, this approach is inconsistent in its effectiveness when dealing with sintered and carbonized blockages, and solvent residues often remain on the nozzle surface after cleaning, leading to subsequent soldering failures or inkjet deviations, severely limiting the grade of recyclable materials.

[0003] At the microscopic level, nozzle recovery involves the simultaneous treatment of multiple micrometer-level channels and welded areas. Existing methods struggle to achieve high-precision centralized positioning, leading to uneven laser energy distribution. Some areas suffer from excessive ablation, while others retain contaminants, creating a contradiction between thorough cleaning and substrate integrity. Further complicating matters, if the welded area contains an oxide layer or minor dents after cleaning, laser cladding repair requires extremely rapid powder fusion and solidification control; otherwise, analytical challenges arise. Current recovery processes rely on chemical cleaning and mechanical unclogging, but these methods introduce environmental issues such as wastewater treatment and solvent evaporation. Porosity or insufficient bonding strength means that while the repair layer covers defects, it fails to restore weldability.

[0004] In production, once a batch of nozzles is fixed on the platform, the system needs to quickly identify the coordinates of dozens of micro-holes and plan the laser path. If the positioning deviation exceeds two micrometers, the cleaning will deviate from the center of the blockage. If the powder feeding amount during the repair stage is mismatched with the laser power, the thickness of the cladding layer will fluctuate excessively, resulting in the scrapping of the entire batch of nozzles.

[0005] Therefore, how to achieve precise positioning of micropores and dynamic matching of repair parameters within the same workstation has become a key issue restricting the consistency of nozzle regeneration quality. Summary of the Invention

[0006] This invention provides an integrated laser cleaning and repair method for the micropores of an ink cartridge nozzle, mainly comprising:

[0007] The nozzle fixed to the platform is scanned using an image acquisition device to obtain first micropore coordinate data and a first surface feature image. The first surface feature image captures the microstructural details of the nozzle surface. An image analysis algorithm is used to process the first surface feature image sequence to obtain a first micropore blockage distribution pattern, which represents the overall layout of blockage location and degree. A first laser path is planned based on the first micropore blockage distribution pattern, and an optimization algorithm is used to adjust the path order and energy distribution parameters. The energy distribution parameters relate to cleaning efficiency and energy uniformity, determining the first laser cleaning path and its corresponding first power value. The coordinates of a first target area are extracted from the first laser cleaning path and the corresponding first power value. If the degree of blockage in the first micropore blockage distribution pattern exceeds a preset threshold, the power increase is adjusted. The preset threshold is set based on historical cleaning data. The cleaning priority is determined by threshold comparison to obtain a first optimized cleaning parameter set. The laser device is driven by the first optimized cleaning parameter set to perform point-by-point scanning and obtain real-time feedback of the first surface residue data. The core theme of the first surface residue analysis is to review the provided nozzle repair technology description, focusing on the process of laser cleaning and repairing nozzle micropore blockage and defects. The data includes residue density and distribution information, which is integrated through data fusion technology to obtain a nozzle surface state image after the first cleaning. The first oxide layer and the first impact defect are detected from the nozzle surface state image after the first cleaning. The first oxide layer and the first impact defect affect the nozzle's functional durability. An edge detection algorithm is used to identify the defect boundary and determine the coordinates and first depth value of the first repair area. The first powder feeding rate and the second laser power are matched according to the first repair area coordinates and the first depth value. The matching considers material compatibility and repair efficiency. If the first depth value is greater than the average value, the powder feeding rate is increased. The first dynamic repair parameter set is obtained through proportional calculation. A first cladding control sequence is generated from the first dynamic repair parameter set. The first cladding control sequence controls the laser cladding process and obtains the first solidification time window data. The first solidification time window data reflects the material cooling dynamics. The fusion quality is judged through time series analysis to obtain the thickness distribution map of the first repair layer. The uniformity of the first repair layer thickness distribution map is checked. If the fluctuation range exceeds the threshold, the first dynamic repair parameter group is iteratively adjusted. The uniformity check ensures the consistency of the layer thickness. The final regeneration nozzle quality data is obtained through the check cycle.

[0008] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0009] This invention discloses an integrated laser cleaning and repair method for micropores in ink cartridge nozzles. The method involves scanning and acquiring micropore coordinates and surface feature images using an image acquisition device. Algorithm analysis forms a blockage distribution pattern, and the laser path and power allocation are calculated and optimized. If the blockage exceeds a threshold, the power is dynamically increased to form a first optimized cleaning parameter set. The laser is used to clean point-by-point, and residual data is fused in real time to generate a post-cleaning state image. Oxide layer and impact defects are detected to determine the repair area and depth. The toner feed rate and laser power are matched to generate a dynamic repair parameter set and cladding control sequence. Solidification time is monitored to determine fusion quality and verify the uniformity of the repair layer thickness. Parameters are iteratively adjusted until the target is met, ultimately obtaining a high-quality regenerated nozzle. This invention achieves precise blockage cleaning and intelligent defect analysis. The description of the provided nozzle repair technology is currently under review. The core theme focuses on the process of laser cleaning and repairing micropore blockage and defects in nozzles. Closed-loop control of the repair process improves nozzle functionality, durability, and repair efficiency. Attached Figure Description

[0010] Figure 1 This is a flowchart of an integrated laser cleaning and repair method for the nozzle micropores of an ink cartridge nozzle according to the present invention.

[0011] Figure 2 This is a schematic diagram of an integrated laser cleaning and repair method for the nozzle micropores of an ink cartridge nozzle according to the present invention. Detailed Implementation

[0012] The technical solutions of the embodiments of the present invention will be clearly and thoroughly described below with reference to the accompanying drawings. The described embodiments are merely some embodiments of the present invention.

[0013] like Figure 1-2 This embodiment of the integrated laser cleaning and repair method for the nozzle micropores of an ink cartridge nozzle may specifically include:

[0014] S101. The nozzle fixed on the platform is scanned by an image acquisition device to obtain the first micro-hole coordinate data and the first surface feature image. The first surface feature image captures the microstructural details of the nozzle surface.

[0015] A high-resolution scan of a nozzle fixed to a platform is performed using an image acquisition device to obtain first micro-orifice coordinate data and a first surface feature image. An image preprocessing algorithm is used to denoise and enhance the first surface feature image to obtain a second surface feature image. If the clarity of the second surface feature image reaches a preset threshold, an edge detection algorithm is used to extract the microstructure contour of the nozzle surface to obtain first contour data; if the clarity does not reach the preset threshold, the image preprocessing algorithm is used again to enhance the first surface feature image to obtain an updated second surface feature image. Based on the first micro-orifice coordinate data and the first contour data, the spatial correspondence between the micro-orifice and the surface microstructure is calculated to determine first spatial mapping data. A template matching algorithm is used to analyze the first spatial mapping data to identify the deviation between the micro-orifice and the microstructure, obtaining first deviation data. Based on the first deviation data, a clustering algorithm is used to classify the micro-orifice position and surface features to generate a first classification result. The first classification result is compared with a preset nozzle structure standard to determine whether the micro-orifice position and surface features meet the requirements, obtaining the final analysis result.

[0016] Specifically, a high-resolution industrial camera (2048×1536 pixels, 60fps) was used to perform a panoramic scan of the nozzle fixed on the platform. The camera was equipped with a 50mm focal length lens, a working distance of 100mm, and used an LED ring light source (5500K color temperature) to ensure uniform illumination, acquiring image data of the nozzle surface. During the scanning process, the camera moved the platform along the XY axis (step accuracy 0.01mm) to perform a grid scan, moving 0.5mm per step, covering a 10mm×10mm area on the nozzle surface, acquiring a total of 400 frames. For each frame, an edge detection algorithm (Canny algorithm, threshold range 50-150) was used to extract the nozzle micro-orifice contour, calculate the micro-orifice center coordinates, and obtain the first micro-orifice coordinate data (e.g., center point coordinates (x1, y1) = (2.35, 3.47)mm, accuracy ±0.005mm). The coordinate data was fitted with a circular contour using the least squares method to remove noise points (points more than 3σ from the center) to ensure coordinate accuracy. Next, surface features are extracted from the acquired images. The Gray-Level Co-occurrence Matrix (GLCM) algorithm is used to analyze the microstructure of the nozzle surface. The GLCM distance is set to 1 pixel, and the angles are 0°, 45°, 90°, and 135°. Texture feature parameters (contrast 0.85, correlation 0.92, entropy 6.73) are calculated to generate the first surface feature image, capturing the nozzle surface roughness (Ra=0.2μm) and micro-orifice edge defects (crack width <0.01mm). To ensure logical rigor, coordinate data is associated with the feature image. The micro-orifice positions are matched with corresponding texture features through coordinate mapping to generate a nozzle surface quality distribution map for subsequent defect detection. The entire process is implemented using automated image processing software (such as OpenCV), and the data is stored in a database for real-time retrieval and analysis.

[0017] S102. The first surface feature image sequence is processed using an image analysis algorithm to obtain a first micropore blockage distribution pattern, which represents the overall layout of the blockage location and degree.

[0018] A first surface feature image sequence is obtained, which contains multiple surface feature images. A preset image preprocessing algorithm is used to denoise and enhance the first surface feature image sequence to obtain a second surface feature image sequence. A convolutional neural network algorithm is used to extract features from the second surface feature image sequence to obtain a micropore feature set. If the feature values ​​in the micropore feature set exceed a preset threshold, a clustering algorithm is used to classify the micropore feature set to determine the micropore blockage location set. Based on the micropore blockage location set, spatial analysis techniques are used to calculate the spatial distribution characteristics of the blockage locations to obtain the blockage location layout. Statistical analysis techniques are used to quantify the feature values ​​in the micropore blockage location set to obtain a blockage degree representation. Visualization techniques are used to fuse the blockage location layout and the blockage degree representation to generate a first micropore blockage distribution pattern. If the spatial distribution characteristics of the first micropore blockage distribution pattern are inconsistent with a preset reference pattern, the second surface feature image sequence is re-analyzed to obtain an updated first micropore blockage distribution pattern.

[0019] Specifically, the first step is to acquire a sequence of surface feature images. For example, 50 frames are continuously acquired at a resolution of 1000×1000 pixels using a high-resolution optical microscope, one frame every 9 seconds, for a total duration of 450 seconds. The image format is 16-bit grayscale TIFF to ensure the dynamic changes of the micropore region are captured. Next, each frame is preprocessed, including using a median filter (kernel size 5×5) to remove noise, and then using histogram equalization to enhance contrast, raising the grayscale value of the micropore edges to the range of 150-200, while normalizing the background to grayscale 50. Subsequently, the Canny edge detection algorithm (low threshold 50, high threshold 150) is applied to extract the micropore contours, and Hough circle transform is used to detect the diameter of individual micropores, setting a diameter threshold range of 2.0-5.0 micrometers to filter out non-target areas. Further, the average grayscale value inside each micropore in the sequence is calculated. If it is lower than 70% (84) of the average grayscale value of 120 in the normal open state, it is marked as a potential blockage point, and the blockage degree index is calculated as (120 - actual grayscale) / 120 × 100%. For example, when the grayscale value is 60, the blockage degree is 50%. Then, a spatiotemporal blockage matrix is ​​constructed with a size of 50 frames × N micropores (assuming 200 micropores are detected), and the matrix elements are the blockage degree of each frame. The K-means clustering algorithm (K=4) is used to cluster the row vectors of the matrix to identify the blockage pattern, such as rapid blockage (degree > 80% in the first 10 frames) and progressive blockage (linearly increasing to 60%). Finally, principal component analysis (PCA) was used to reduce the dimensionality to 2D, calculate the coordinates of each micropore in the feature space, and draw a heatmap of the blockage distribution. The heatmap color ranges from blue (0%) to red (100%). The coordinates of the center of the blockage location (x=320.5 pixels, y=410.3 pixels) and the overall blockage rate of 28.7% were statistically analyzed to form the first micropore blockage distribution pattern, which will be used for subsequent process optimization.

[0020] S103. Calculate the first laser path planning based on the first micropore blockage distribution pattern, and use an optimization algorithm to adjust the path sequence and energy distribution parameters. The energy distribution parameters involve cleaning efficiency and energy uniformity. Determine the first laser cleaning path and the corresponding first power value.

[0021] First, micropore blockage distribution data is acquired, and image processing techniques are used to analyze the micropore distribution pattern to obtain the spatial characteristics of the micropore blockage. Based on these spatial characteristics, a genetic algorithm is used to calculate an initial laser path plan, resulting in a first set of candidate laser paths. For this first set, a simulated annealing algorithm is used to adjust the path order, resulting in a second laser path. Based on the second laser path, cleaning efficiency and energy uniformity requirements are extracted from energy allocation parameters, and an initial power allocation scheme is calculated, resulting in a first power allocation set. If the energy uniformity of the first power allocation set is lower than a preset threshold, a gradient descent algorithm is used to optimize the power allocation, resulting in a second power allocation set. Using the second laser path and the second power allocation set, a final laser cleaning path and a corresponding first power value are generated, and a laser cleaning control command is output to determine the cleaning execution scheme.

[0022] Specifically, the first laser path planning is calculated based on the first micropore blockage distribution pattern. An optimization algorithm is used to adjust the path sequence and energy distribution parameters, involving cleaning efficiency and energy uniformity, to determine the first laser cleaning path and its corresponding first power value. First, micropore blockage distribution data is acquired using a high-resolution imaging system. Assuming the micropore array is a 10×10 grid, and the blockage degree is represented by grayscale values ​​ranging from 0 to 255, a blockage distribution matrix is ​​randomly generated, where one micropore has a blockage value of 180 and another has a value of 50. The K-means clustering algorithm (K=3) is used to classify the blockage degree, resulting in lightly blocked (0-85), moderately blocked (86-170), and severely blocked (171-255) areas. Analysis shows that severely blocked areas account for 20%, moderately blocked areas account for 50%, and lightly blocked areas account for 30%. Subsequently, based on this distribution, laser paths were planned, and a genetic algorithm was used to optimize the path order. The population size was set to 50, and 100 iterations were performed. The fitness function was a weighted sum of path length and cleaning time (weights 0.4 and 0.6). The shortest path was calculated to cover all heavily clogged micropores, reducing the path length by 15% to 8.5 cm. Energy distribution parameters considered cleaning efficiency and energy uniformity, setting the laser power range to 100-500W. 450W was allocated for heavily clogged pores, 300W for moderately clogged pores, and 150W for lightly clogged pores. A simulated annealing algorithm was used to optimize the energy distribution. The initial temperature was 1000°C, the cooling coefficient was 0.95, and 200 iterations were performed, reducing the energy uniformity (standard deviation) from the initial 25W to 10W and increasing the cleaning efficiency to 95%. Finally, the first laser cleaning path was determined to be a polygonal path from grid (1,1) to (10,10), with the power value dynamically adjusted according to the degree of clog. A control command file containing path coordinates and power values ​​was generated for the laser cleaning system to execute. This process forms a closed loop through data analysis, algorithm optimization, and parameter adjustment to ensure efficient cleaning.

[0023] S104. Extract the coordinates of the first target area from the first laser cleaning path and the corresponding first power value. If the degree of blockage in the first micropore blockage distribution pattern exceeds a preset threshold, adjust the power increase. The preset threshold is set based on historical cleaning data. The cleaning priority is determined by threshold comparison to obtain the first optimized cleaning parameter set.

[0024] The coordinates of the target area are extracted from the first laser cleaning path and the first power value to obtain a first target area coordinate set. The micropore blockage distribution pattern is mapped using this first target area coordinate set, and image processing techniques are used to analyze the pixel density of the blockage area to obtain a quantified value of the blockage degree. If the quantified value of the blockage degree exceeds a preset threshold, a power increase adjustment value is predicted using a linear regression algorithm to obtain an adjusted power increase set. The correspondence between cleaning effect and power value is extracted from historical cleaning data, and a support vector machine algorithm is used to classify cleaning priorities to obtain a priority ranking list. Based on the priority ranking list and the adjusted power increase set, the matching relationship between the first power value and the cleaning path is optimized to obtain a first optimized cleaning parameter set. The control commands of the laser cleaning equipment are adjusted using the first optimized cleaning parameter set, and the applicability of the cleaning parameters is verified based on threshold comparison to obtain the final cleaning execution parameters. The laser cleaning equipment is driven by the final cleaning execution parameters, and the cleaning path and power value are iteratively updated based on real-time feedback data to obtain a dynamically optimized cleaning scheme.

[0025] Specifically, the coordinates of the first target area are extracted from the first laser cleaning path and the corresponding first power value. It is assumed that the laser cleaning path data is stored in the form of a two-dimensional array, with the path point coordinates being (x, y) and the corresponding power value being P.

[0026] For example, the path data is [(10,20,50W), (15,25,55W), (20,30,60W)]. The coordinates (x,y) are extracted from the parsed array, resulting in [(10,20), (15,25), (20,30)], and stored as the target region coordinate set. Next, the distribution pattern of the first micropore blockage is analyzed. It is assumed that the degree of blockage is calculated using image processing algorithms, with grayscale values ​​ranging from 0 to 255, and a preset threshold of 180. Image segmentation algorithms (such as the Otsu thresholding method) are used to analyze the micropore region, calculating the average grayscale value of the blockage degree at each coordinate point. For example, the grayscale value at (10,20) is 200, exceeding the threshold of 180. For points where the blockage exceeds a threshold, the power increase is adjusted using the following algorithm: If the grayscale value G > 180, the new power P' = P × (1 + (G - 180) / 100). For point (10, 20), P' = 50 × (1 + (200 - 180) / 100) = 60W. Based on historical cleaning data, assuming that historical data shows a grayscale value of 180 corresponds to a standard cleaning power of 50W, and that an increase of 5W is required for every 10 increase in grayscale value, linear regression analysis confirms that the threshold of 180 is reasonable. The cleaning priority is determined by comparing the blockage degree at each point using the following algorithm: Priority = Grayscale value / Maximum grayscale value. For point (10, 20), the priority = 200 / 255 ≈ 0.784. The coordinates are sorted from highest to lowest priority to generate the first optimized cleaning parameter set, in the format [(x,y,P',priority)], such as [(10,20,60W,0.784), (15,25,55W,0.706), (20,30,60W,0.667)]. The entire process is automated and logically rigorous, ensuring both cleaning efficiency and accuracy.

[0027] S105. The laser device is driven to perform point-by-point scanning based on the first optimized cleaning parameter set to obtain real-time feedback of the first surface residue data. The core theme of the first surface residue analysis is reviewing the provided nozzle repair technology description, focusing on the process of laser cleaning and repairing nozzle micropore blockage and defects. The data includes residue density and distribution information, which are integrated using data fusion technology to obtain an image of the nozzle surface state after the first cleaning.

[0028] First, surface residue data is acquired, and the laser device is driven by point-by-point scanning to obtain a first surface residue distribution image. Data fusion technology is used to integrate and process the first surface residue distribution image to obtain a first post-cleaning nozzle surface state image. It is determined whether the residue density in the first post-cleaning nozzle surface state image exceeds a preset threshold. If the residue density exceeds the preset threshold, the cleaning parameters are adjusted, and the laser device is driven again to perform point-by-point scanning to acquire second surface residue data, obtaining a second surface residue distribution image. A convolutional neural network algorithm is used to extract features from the second surface residue distribution image, identifying micropore blockage locations and defect areas, obtaining micropore blockage and defect features. Based on the micropore blockage and defect features, a targeted laser cleaning path plan is generated, and the laser device is driven to perform a repair scan to acquire post-repair surface state data, obtaining a post-repair surface state image. An image segmentation algorithm is used to process the post-repair surface state image, extracting residue distribution and micropore state information, and determining whether the repair effect meets a preset standard. If the repair effect does not meet the preset standard, a support vector machine algorithm is used to classify the residue distribution and micropore state information, identifying unrepaired areas, generating a new cleaning parameter set, and re-executing the point-by-point scanning and repair process to obtain the final nozzle surface state image.

[0029] Specifically, the laser device performs point-by-point scanning based on the first optimized cleaning parameter set. The laser device uses a 1064nm nanosecond pulsed laser with a pulse width of 10ns, an energy density of 2.5J / cm², and a scanning speed of 500mm / s. The point-by-point scanning employs a gridded path with a grid spacing of 0.1mm to ensure coverage of all micropores on the nozzle surface. Real-time feedback of the first surface residue data is acquired using a high-speed CCD camera with a resolution of 1280×1024 pixels and a frame rate of 1000fps. Laser-induced fluorescence (LIF) technology is used to detect the fluorescence signal of the residue. The residue density is expressed as particles per square millimeter, with an initial residue density of 150 particles / mm². Data processing employs a Kalman filter algorithm to denoise the fluorescence signal, with filtering parameters Q=0.01 and R=0.1, generating a residue distribution matrix with a resolution of 0.05mm. Data fusion technology integrates information from multiple sensors, combining CCD images and LIF signals. A weighted average algorithm (weight ratio 0.6:0.4) generates a post-cleaning nozzle surface state image. Pixel grayscale values ​​range from 0-255, with values ​​below 50 indicating clean areas and values ​​above 150 indicating areas with severe residue. The analysis process uses a K-means clustering algorithm (K=3) to categorize surface states into clean, slightly residual, and heavily residual areas, with cluster centers at grayscale values ​​of 30, 100, and 180, respectively. Analysis results show that the residual density is reduced to 20 particles / mm² after cleaning, with clean areas accounting for 85%, slightly residual areas for 12%, and heavily residual areas for 3%. For heavily residual areas, the system automatically adjusts the laser energy density to 3.0 J / cm², rescans, generates secondary cleaning data, and feeds it back to the control system to form a closed-loop optimization. In the logic chain, initial parameters drive laser cleaning, real-time data acquisition and fusion ensure accurate state images, and cluster analysis guides parameter adjustments, achieving adaptive cleaning, reducing residual density, and improving nozzle micro-orifice patency. If the nozzle micropore blockage involves metal oxides, the system can call an infrared spectrometer for auxiliary analysis to detect the characteristic peaks of the oxides (such as Fe2O3 at 550 cm⁻¹) and further optimize the cleaning parameters.

[0030] S106. Detect the first oxide layer and the first impact defect from the first cleaned nozzle surface state image. The first oxide layer and the first impact defect affect the nozzle's functional durability. Use an edge detection algorithm to identify the defect boundary and determine the coordinates of the first repair area and the first depth value.

[0031] A nozzle surface condition image is acquired, and image preprocessing techniques are used to denoise the image to obtain a first image. From the first image, the Canny edge detection algorithm is used to identify the boundaries of the oxide layer and impact defects, resulting in a defect boundary set. For this defect boundary set, region segmentation techniques are used to determine the repair areas for the oxide layer and impact defects, resulting in a first repair area coordinate set. If the boundary point density of the first repair area coordinate set is greater than a preset threshold, a depth estimation algorithm is used to calculate the depth values ​​of the first repair area coordinate set, resulting in a first depth value set. Based on the first depth value set, cluster analysis is used to distinguish the depth features of the oxide layer and impact defects, resulting in a defect classification result. Using the defect classification result, coordinate mapping techniques are used to associate the first repair area coordinate set with the first depth value set, resulting in a repair area depth mapping. If the depth value in the repair area depth mapping exceeds a preset durability threshold, it is marked as a high-priority repair area, resulting in a final repair area list.

[0032] Specifically, the first oxide layer and the first impact defect are detected from the nozzle surface condition image after the first cleaning. An image processing method based on the Canny edge detection algorithm, combined with deep learning and geometric analysis, is employed to ensure detection accuracy and the precision of the repair area. First, an RGB image of the nozzle surface with a resolution of 1920×1080 pixels is acquired. The image is converted to grayscale using the grayscale conversion formula Y=0.299R+0.587G+0.114B to reduce computational complexity. Next, a Gaussian filter (kernel size 5×5, standard deviation σ=1.5) is applied to smooth the image and reduce noise interference. Then, the Canny edge detection algorithm is used, with a low threshold of 50 and a high threshold of 150, to extract the boundaries of the oxide layer and the impact defect, generating a binarized edge map. For oxide layers, based on color features (RGB values ​​leaning towards green or yellow, e.g., R: 180-220, G: 150-200, B: 50-100), suspected oxide regions are extracted through threshold segmentation, and the region area is calculated in conjunction with the edge map. An oxide layer area greater than 1000 pixels is considered a significant defect. For impact defects, irregular closed contours in the edge map are analyzed, and the perimeter-to-area ratio (e.g., perimeter / area > 0.1) is calculated to filter out impact regions with a depth greater than 0.5 mm. A deep learning model (such as U-Net, pre-trained on 10,000 nozzle defect images) is used for semantic segmentation of the edge map to further confirm the categories of oxide layers and impact regions, outputting a probability map (confidence > 0.9 is considered valid). Subsequently, the coordinates of the first repair area are determined using a contour fitting algorithm (minimum bounding rectangle). For example, the center coordinates of the oxide layer area are (x1, y1) = (500, 300), with an area of ​​2000 pixels; the center coordinates of the impact area are (x2, y2) = (700, 450), with an area of ​​1500 pixels. Finally, combined with laser ranging data (accuracy 0.01 mm), the defect depth is calculated, for example, the oxide layer depth is 0.3 mm, and the impact depth is 0.7 mm. The three-dimensional coordinates and depth values ​​of the repair area are generated and stored in JSON format: {"oxidation":{"center":[500,300],"depth":0.3},"dent":{"center":[700,450],"depth":0.7}}. Through the above steps, from image detection to defect classification, boundary recognition, and repair area localization, a closed logical chain is formed, ensuring the automation and accuracy of nozzle functional durability analysis.

[0033] S107. Match the first powder feeding amount and the second laser power according to the coordinates of the first repair area and the first depth value. The matching takes into account material compatibility and repair efficiency. If the first depth value is greater than the average value, the powder feeding rate is increased. The first dynamic repair parameter group is obtained by proportional calculation.

[0034] Based on the aforementioned business content and extracted relevant attributes, the following business solution is generated: The coordinates of the repair area and the first depth value are obtained. Material compatibility data is extracted from a preset database. A matching algorithm model is used to calculate the first powder feeding rate and the second laser power, resulting in a preliminary parameter set. If the first depth value is greater than a preset threshold, the powder feeding rate is adjusted according to a proportional calculation method, resulting in an adjusted powder feeding rate value. Based on the adjusted powder feeding rate value and the preliminary parameter set, the first powder feeding rate is updated, generating a first dynamic repair parameter set. The second laser power is extracted from the first dynamic repair parameter set. Combined with the material compatibility data, it is determined whether the repair efficiency meets a preset standard, resulting in an efficiency evaluation result. If the efficiency evaluation result is lower than the preset standard, the second laser power is optimized according to parameter adjustment rules, resulting in an optimized dynamic repair parameter set. Using the optimized dynamic repair parameter set and the coordinates of the repair area, a final repair control command is generated to determine the repair execution plan. Through the final repair control command, the equipment operating parameters are adjusted to complete the processing of the repair area, resulting in a repair completion status.

[0035] Specifically, the coordinates of the first repair area (e.g., X=150.5mm, Y=200.3mm, Z=50.0mm) and the first depth value (measured as D1=3.2mm) are first obtained and compared with the preset average depth value D_avg=2.0mm. The depth difference ΔD=D1-D_avg=1.2mm is calculated. If ΔD>0, it is determined that the powder feed rate needs to be increased to match the material stacking requirements. At the same time, Analyzingpower is used to analyze the material compatibility (Inconel 625 powder, particle size 45-90μm is selected). m, melting point 1300℃), its thermal conductivity k=15W / m·K, laser absorptivity α=0.35, according to the repair efficiency formula η=(α·P) / (ρ·Cp·V·ΔT), where ρ=8440kg / m³, Cp=410J / kg·K, V is the scanning speed 2m / s, ΔT=1600K, the calculated baseline efficiency η0≈0.62, for ΔD=1.2mm, an efficiency improvement of 15.3% is required, therefore the first powder feeding amount Q1 is calculated using the proportional algorithm Q1=Q0·(1+ΔD / D_avg·k_eff), Given Q0 = 15 g / min and k_eff = 0.85, calculate Q1 = 15 × (1 + 1.2 / 2.0 × 0.85) = 24.66 g / min. The second laser power P2 is calculated as P2 = P1·(D1 / D_avg)^0.5·(1 + α·ΔD / 10), where P1 = 1200 W. Substituting these values, we get P2 = 1200 × (3.2 / 2.0)^0.5 × (1 + 0.35 × 1.2 / 10) ≈ 1697 W. The algorithm incorporates the material thermal diffusion time t_diff = L^2 / D_therm. Given al, L=0.5mm, D_thermal=3.65×10^-6m² / s, t_diff≈68ms, and ensuring that the power increase does not exceed the material overheating threshold of 1800W, the first dynamic repair parameter set (Q1=24.66g / min, P2=1697W, V=2m / s) is finally formed. The powder feeding rate error is corrected in real time to ≤0.5g / min through a closed-loop feedback PID controller (Kp=0.12, Ki=0.08, Kd=0.05), achieving fully automated parameter matching and efficient repair.

[0036] S108. Generate a first cladding control sequence from the first dynamic repair parameter group. The first cladding control sequence controls the laser cladding process. Obtain first solidification time window data. The first solidification time window data reflects the material cooling dynamics. Determine the fusion quality through time series analysis and obtain the first repair layer thickness distribution map.

[0037] Extract key parameters from the first dynamic repair parameter group, and generate the first cladding control sequence through a preset mapping rule. Use the first cladding control sequence to drive the laser cladding equipment to obtain real-time solidification time window data. Process the solidification time window data through time series analysis to extract the dynamic characteristics of material cooling. If the cooling dynamic characteristics meet the preset cooling rate threshold, it is determined that the fusion quality is qualified, and a quality analysis result is obtained. According to the quality analysis result, calculate the thickness distribution of the first repair layer. Through the thickness distribution data, adjust the first dynamic repair parameter group to generate the second cladding control sequence. Use the second cladding control sequence to drive the laser cladding equipment to obtain an optimized thickness distribution of the repair layer.

[0038] Specifically, first extract parameters such as a laser power of 350 W, a scanning speed of 800 mm / min, a powder feeding rate of 15 g / min, and a spot diameter of 2.5 mm from the first dynamic repair parameter group, and use a parameter interpolation algorithm to generate the first cladding control sequence. For example, use cubic spline interpolation to smoothly transition the power from 300 W to 400 W and the speed from 700 mm / min to 900 mm / min within a time step of 0.1 s, forming a control instruction array with a sequence length of 500 frames. This array is output to the laser cladding equipment in real time through a PLC interface to control the dynamic state of the molten pool. Subsequently, during the cladding process, infrared thermal imager data is collected in real time at a sampling frequency of 100 Hz, the temperature decay curve of each pixel point is calculated, and the finite difference method is applied to solve the one-dimensional heat conduction equation ∂T / ∂t = α∂²T / ∂x² (where the thermal diffusivity α = 1.2×10⁻ 5 m² / s), and the time interval between the liquidus temperature of 1450 °C and the solidus temperature of 1350 °C is extracted as the first solidification time window data, obtaining a distribution data set with an average window width of 4.2 s and a standard deviation of 0.8 s. Then, perform autoregressive moving average model ARMA(2,1) fitting on this time series data, and calculate that when the residual sum of squares RSS < 0.05, it is determined as high-quality fusion. Otherwise, if RSS > 0.12, mark the defect area. Decompose the signal at scale 4 through the db4 mother wavelet function of wavelet transform to extract that the proportion of high-frequency noise energy is less than 15% to confirm good fusion quality. Finally, integrate the solidification window and position coordinates to generate the thickness distribution map of the first repair layer, and use the Kriging interpolation algorithm to predict the thickness value at a grid resolution of 0.1 mm × 0.1 mm, with an average thickness of 0.85 mm and a maximum deviation of 0.12 mm, forming a visualized thermal map for subsequent process optimization. All processes are automatically executed in a chain through a Python script to ensure a logical closed loop.

[0039] S109. Perform uniformity verification on the thickness distribution map of the first repair layer. If the fluctuation range exceeds the threshold, iteratively adjust the first dynamic repair parameter group. The uniformity verification ensures the consistency of the layer thickness. The final regeneration nozzle quality data is obtained through the verification cycle.

[0040] A first repair layer thickness distribution map is obtained, which includes thickness data at multiple locations. The thickness fluctuation range of the repair layer is calculated based on the thickness distribution map. The relationship between the thickness fluctuation range and a preset threshold is determined. If the thickness fluctuation range exceeds the preset threshold, the adjustment direction of the first dynamic repair parameter group is determined. The first dynamic repair parameter group is modified according to the adjustment direction to obtain a second dynamic repair parameter group. The thickness distribution map of the repair layer is regenerated based on the second dynamic repair parameter group. A uniformity check is performed by analyzing the regenerated thickness distribution map to obtain a layer thickness consistency index. It is determined whether the layer thickness consistency index meets a preset condition. If it does, the final regeneration nozzle quality data is output.

[0041] Specifically, firstly, the thickness distribution data of the first repair layer is loaded, such as a 100×100 pixel grid obtained from a laser scanner, with each point's thickness ranging from 50.0 μm to 150.0 μm. The global uniformity index is calculated using Analyzing. The standard deviation algorithm σ = √[Σ(xi-μ)² / N], where μ is the average thickness of 102.3 μm and N = 10000, yielding σ = 18.7 μm. Next, a threshold T = 10.0 μm is set. If σ > 10.0 μm, iterative adjustment is triggered, entering a dynamic parameter optimization loop. The initial first dynamic repair parameter set includes spraying speed v = 200.0 mm / s, gas flow rate q = 15.0 L / min, and powder supply rate p = At a speed of 8.0 g / min, 500 sets of parameter perturbation samples were generated through Monte Carlo simulation: Δv∈[-20.0,20.0] mm / s, Δq∈[-2.0,2.0] L / min, and Δp∈[-1.0,1.0] g / min. A response surface model (RSM) was constructed with the thickness standard deviation σ as the objective function: f(v,q,p)=a0+a1v+a2q+a3p+a12vq+a13vp+a23qp. The least squares method was used to fit the coefficients a0=25.3, a1=-0.12, a2=-1.8, a3=-2.4, a12=0.005, a13=0.008, a23=0.1. The parameter set was iteratively updated using gradient descent with a step size α=0.05, updating the formula v_{k+1}=v_k-α∂f / ∂v until σ≤10.0μm or convergence after 50 iterations. The change in σ was recorded for each iteration, such as σ= The thickness was measured at 18.7 μm, σ=12.4 μm for the 10th test, and σ=9.8 μm for the 25th test. After the final verification was passed, the quality data of the regenerated nozzle was output, including the optimized parameter set v=195.6 mm / s, q=14.3 L / min, p=7.7 g / min, as well as the average thickness of 101.8 μm, standard deviation of 9.6 μm, maximum deviation of +8.2 μm, and minimum deviation of -7.9 μm, forming a closed-loop automated uniformity control process.

[0042] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for laser cleaning and repairing the micropores of an ink cartridge nozzle, characterized in that, The method includes: S101. Scan the nozzle fixed to the platform using an image acquisition device to obtain first micropore coordinate data and a first surface feature image. The first surface feature image captures the microstructural details of the nozzle surface. S102. Process the first surface feature image sequence using an image analysis algorithm to obtain a first micropore blockage distribution pattern. The first micropore blockage distribution pattern represents the overall layout of blockage location and degree. S103. Calculate a first laser path plan based on the first micropore blockage distribution pattern. Use an optimization algorithm to adjust the path order and energy distribution parameters. The energy distribution parameters involve cleaning efficiency and energy uniformity. Determine the first laser cleaning path and the corresponding first power value. S104. Extract the coordinates of a first target area from the first laser cleaning path and the corresponding first power value. If the blockage degree in the first micropore blockage distribution pattern exceeds a preset threshold, adjust the power increase. The preset threshold is set based on historical cleaning data. Determine the cleaning priority by comparing the thresholds to obtain a first optimized cleaning parameter set. S105. Drive the laser device to perform point-by-point scanning based on the first optimized cleaning parameter set to obtain real-time feedback of the first surface residue data. The core theme of the first surface residue analysis is to review the provided nozzle repair technology description. The core theme focuses on the process of laser cleaning and repairing nozzle micropore blockage and defects. The data includes residue density and distribution information. The data is integrated through data fusion technology to obtain the first cleaning nozzle surface state image. S106. Detect the first oxide layer and the first impact defect from the first cleaning nozzle surface state image. The first oxide layer and the first impact defect affect the nozzle's functional durability. Use an edge detection algorithm to identify the defect boundary and determine the coordinates and first depth value of the first repair area. S107. Match the first powder feeding rate and the second laser power according to the first repair area coordinates and the first depth value. The matching considers material compatibility and repair efficiency. If the first depth value is greater than the average value, increase the powder feeding rate. Obtain the first dynamic repair parameter set through proportional calculation. S108. Generate a first cladding control sequence from the first dynamic repair parameter group. The first cladding control sequence controls the laser cladding process and obtains first solidification time window data. The first solidification time window data reflects the material cooling dynamics. The fusion quality is judged through time series analysis, and a first repair layer thickness distribution map is obtained. S109. Perform uniformity verification on the first repair layer thickness distribution map. If the fluctuation range exceeds the threshold, iteratively adjust the first dynamic repair parameter group. The uniformity verification ensures the consistency of the layer thickness. The final regeneration nozzle quality data is obtained through verification loop.

2. The integrated laser cleaning and repair method for the nozzle micropores of an ink cartridge nozzle according to claim 1, characterized in that, S101 includes: The nozzle fixed to the platform is scanned at high resolution using an image acquisition device to obtain the coordinate data of the first micropore and the image of the first surface feature. The first surface feature image is denoised and enhanced using an image preprocessing algorithm to obtain the second surface feature image; If the clarity of the second surface feature image reaches a preset threshold, the microstructure contour of the nozzle surface is extracted by an edge detection algorithm to obtain the first contour data. If the clarity does not reach the preset threshold, the image preprocessing algorithm is used again to enhance the first surface feature image to obtain an updated second surface feature image. Based on the first micropore coordinate data and the first contour data, calculate the spatial correspondence between the micropores and the surface microstructure, and determine the first spatial mapping data; The first spatial mapping data is analyzed by template matching algorithm to identify the deviation between micropores and microstructures, and the first deviation data is obtained. Based on the first deviation data, a clustering algorithm is used to classify the micropore locations and surface features to generate a first classification result; By comparing the first classification result with the preset nozzle structure standard, it is determined whether the micropore position and surface features meet the requirements, and the final analysis result is obtained.

3. The integrated laser cleaning and repair method for the nozzle micropores of an ink cartridge nozzle according to claim 1, characterized in that, S102 includes: A first surface feature image sequence is obtained, wherein the first surface feature image sequence contains multiple surface feature images; The first surface feature image sequence is denoised and enhanced using a preset image preprocessing algorithm to obtain the second surface feature image sequence. The second surface feature image sequence is used to extract features using a convolutional neural network algorithm to obtain a micropore feature set; If the feature values ​​in the micropore feature set exceed a preset threshold, a clustering algorithm is used to classify the micropore feature set and determine the set of micropore blockage locations. Based on the set of micropore blockage locations, spatial analysis techniques are used to calculate the spatial distribution characteristics of the blockage locations, thereby obtaining the blockage location layout. The characteristic values ​​in the set of micropore blockage locations are quantified using statistical analysis techniques to obtain a representation of the degree of blockage; Visualization technology is used to fuse the layout of the blockage locations and the representation of the blockage degree to generate a first micropore blockage distribution pattern; If the spatial distribution characteristics of the first micropore blockage distribution pattern are inconsistent with the preset reference pattern, the second surface feature image sequence is re-analyzed to obtain the updated first micropore blockage distribution pattern.

4. The integrated laser cleaning and repair method for the nozzle micropores of an ink cartridge nozzle according to claim 1, characterized in that, S103 includes: First micropore blockage distribution data is obtained, and image processing technology is used to analyze the micropore distribution pattern of the first micropore blockage distribution data to obtain the spatial characteristics of micropore blockage. Based on the spatial characteristics of the micropore blockage, a genetic algorithm is used to calculate the initial laser path planning and obtain the first laser path candidate set; For the first set of laser path candidates, the simulated annealing algorithm is used to adjust the path order to obtain the second laser path; Based on the second laser path, the cleaning efficiency and energy uniformity requirements are extracted from the energy allocation parameters, and the initial power allocation scheme is calculated to obtain the first power allocation set; If the energy uniformity of the first power allocation set is lower than a preset threshold, then the gradient descent algorithm is used to optimize the power allocation to obtain the second power allocation set. The final laser cleaning path and the corresponding first power value are generated using the second laser path and the second power allocation set. Laser cleaning control commands are then output to determine the cleaning execution scheme.

5. The integrated laser cleaning and repair method for the nozzle micropores of an ink cartridge nozzle according to claim 1, characterized in that, S104 includes: The coordinates of the target area are extracted from the first laser cleaning path and the first power value to obtain the coordinate set of the first target area; By mapping the micropore blockage distribution pattern through the coordinate set of the first target region, and using image processing technology to analyze the pixel density of the blockage region, a quantitative value of the blockage degree is obtained. If the quantified value of the congestion level exceeds the preset threshold, the power increase adjustment value is predicted by the linear regression algorithm to obtain the adjusted power increase set. The correlation between cleaning effect and power value is extracted from historical cleaning data. The cleaning priority is classified by support vector machine algorithm to obtain a priority ranking list. Based on the priority sorting list and the adjusted power increase set, the matching relationship between the first power value and the cleaning path is optimized to obtain the first optimized cleaning parameter set; The control commands of the laser cleaning equipment are adjusted by adjusting the first optimized cleaning parameter set, and the applicability of the cleaning parameters is verified by threshold comparison to obtain the final cleaning execution parameters. The laser cleaning equipment is driven by the final cleaning execution parameters, and the cleaning path and power value are iteratively updated based on real-time feedback data to obtain a dynamically optimized cleaning scheme.

6. The integrated laser cleaning and repair method for the nozzle micropores of an ink cartridge nozzle according to claim 1, characterized in that, S105 includes: First surface residue data is acquired, and the laser device is driven by point-by-point scanning to obtain a first surface residue distribution image. Data fusion technology is used to integrate and process the first surface residue distribution image to obtain a first post-cleaning nozzle surface state image. It is then determined whether the residue density in the first post-cleaning nozzle surface state image exceeds a preset threshold. If the residue density exceeds the preset threshold, the cleaning parameters are adjusted, and the laser device is driven again to perform point-by-point scanning to acquire second surface residue data, obtaining a second surface residue distribution image. A convolutional neural network algorithm is used to extract features from the second surface residue distribution image to identify micropore blockage locations and defect areas. Based on the micropore blockage and defect characteristics, a targeted laser cleaning path plan is generated, driving the laser equipment to perform a repair scan, acquiring the surface state data after repair, and obtaining a post-repair surface state image. The post-repair surface state image is processed using an image segmentation algorithm to extract the residue distribution and micropore state information, and to determine whether the repair effect meets the preset standard. If the repair effect does not meet the preset standard, the residue distribution and micropore state information are classified using a support vector machine algorithm to determine the unrepaired areas, generate a new cleaning parameter set, and re-execute the point-by-point scanning and repair process to obtain the final nozzle surface state image.

7. The integrated laser cleaning and repair method for the nozzle micropores of an ink cartridge nozzle according to claim 1, characterized in that, S106 includes: A nozzle surface condition image is acquired, and image preprocessing techniques are used to denoise the nozzle surface condition image to obtain a first image. The Canny edge detection algorithm is used to identify the boundaries of the oxide layer and the impact defects in the first image to obtain a set of defect boundaries; For the set of defect boundaries, region segmentation technology is used to determine the repair areas of the oxide layer and the impact defects, and the coordinate set of the first repair area is obtained; If the density of boundary points in the first repair area coordinate set is greater than a preset threshold, then a depth estimation algorithm is used to calculate the depth value of the first repair area coordinate set to obtain a first depth value set. Based on the first set of depth values, cluster analysis is used to distinguish the depth features of oxide layers and impact defects, and the defect classification results are obtained. Based on the defect classification results, coordinate mapping technology is used to associate the coordinate set of the first repair area with the first depth value set to obtain the depth mapping of the repair area; If the depth value in the depth mapping of the repair area exceeds the preset durability threshold, it is marked as a high-priority repair area, and the final repair area list is obtained.

8. The integrated laser cleaning and repair method for the nozzle micropores of an ink cartridge nozzle according to claim 1, characterized in that, S107 includes: Based on the above business content and the extracted relevant attributes, the following business solution is generated: obtain the coordinates of the repair area and the first depth value, extract material compatibility data from the preset database, and use a matching algorithm model to calculate the first powder feeding amount and the second laser power to obtain a preliminary parameter set. If the first depth value is greater than the preset threshold, the powder feeding rate is adjusted according to the proportional calculation method to obtain the adjusted powder feeding rate value. Based on the adjusted powder delivery rate value and the preliminary parameter set, update the first powder delivery amount and generate the first dynamic repair parameter set. The second laser power is extracted from the first dynamic repair parameter group, and combined with material compatibility data, it is determined whether the repair efficiency meets the preset standard, and the efficiency evaluation result is obtained. If the efficiency evaluation result is lower than the preset standard, the second laser power is optimized according to the parameter adjustment rules to obtain the optimized dynamic repair parameter set; The optimized dynamic repair parameter set is used in conjunction with the coordinates of the repair area to generate the final repair control command and determine the repair execution plan. By using the final repair control command to adjust the equipment operating parameters, the repair area is processed, and the repair is completed.

9. The integrated laser cleaning and repair method for the nozzle micropores of an ink cartridge nozzle according to claim 1, characterized in that, S108 includes: Key parameters are extracted from the first dynamic repair parameter group, and the first cladding control sequence is generated through preset mapping rules; The laser cladding equipment is driven by the first cladding control sequence to obtain real-time solidification time window data; The solidification time window data is processed by time series analysis to extract the dynamic characteristics of material cooling. If the cooling dynamic characteristics meet the preset cooling rate threshold, the fusion quality is determined to be qualified, and the quality analysis result is obtained. Based on the quality analysis results, the thickness distribution of the first repair layer is calculated; By adjusting the first dynamic repair parameter group using the thickness distribution data, a second cladding control sequence is generated; The laser cladding equipment is driven by the second cladding control sequence to obtain an optimized repair layer thickness distribution.

Images