A method for preparing a PCB board with high carbon ink

Abstract

This application provides a method for preparing a PCB board with high-carbon ink, comprising: acquiring a conductivity distribution dataset including region identifiers, conductivity values, and boundary coordinates; generating a motion control instruction set corresponding to timestamps, nozzle positions, and deposition coordinates using a micrometer-level precision control algorithm; acquiring an ink deposition image dataset during the deposition process and calculating the deviation between the actual deposition amount and the target deposition amount; controlling the deposition process according to an updated spray frequency configuration file and detecting the transition width of the boundary region; if the transition width exceeds a micrometer-level threshold, generating an ink deposition dataset including optimized boundary coordinates; generating partitioned heating control parameters corresponding to region identifiers and curing temperatures based on the ink deposition dataset; measuring the final conductivity parameters and generating a verification dataset corresponding to the region identifiers and measured conductivity. This invention achieves high-precision and controllable deposition of high-carbon ink on PCB boards, effectively improving its conductivity and consistency.

Description

Technical Field

[0001] This invention relates to the field of PCB board manufacturing technology, and in particular to a method for preparing a PCB board using high-carbon ink. Background Technology

[0002] Printed circuit board (PCB) manufacturing is a core area of ​​the electronics industry, directly impacting the performance, reliability, and miniaturization of electronic products. With the rapid development of smart devices and flexible electronics, PCBs need to achieve complex functions within limited space, making the regionalized design of differentiated conductivity properties a key technological direction. This technology can improve the integration and functional diversity of circuit boards, which is of great significance for promoting the development of 5G communications, the Internet of Things, and wearable devices. However, traditional PCB manufacturing methods have significant limitations in achieving regionalized conductivity properties.

[0003] Currently, conventional PCB manufacturing primarily relies on etching and electroplating processes, which struggle to achieve differentiated conductivity across different areas of the same board. Traditional processes typically use a single conductive material, making it difficult to meet the precise partitioning requirements for high and low conductivity areas in complex circuit designs. Furthermore, existing printing technologies suffer from insufficient nozzle switching precision and ink deposition control during multi-material deposition processes, leading to blurred boundaries between conductive areas and impacting circuit performance stability. During curing, a single heating method also struggles to achieve optimal curing conditions for different conductive materials, further limiting manufacturing precision and reliability. In the fabrication of PCBs containing high-carbon inks, the core challenge lies in achieving high-precision regional ink deposition and clear boundaries for differentiated conductivity. Key technical factors include: the ability of multi-nozzle printing equipment to control nozzle switching with micron-level precision, the ability to monitor and adjust ink deposition in real time, and the adaptability of the partitioned heating curing process to different conductive inks. Unresolved issues hinder precise control of regional conductivity, limiting the manufacture of high-performance PCBs.

[0004] Therefore, the key issue of this research is how to achieve micron-level precise deposition, clear demarcation, and differentiated curing of high and low conductivity areas using a regional printing process with high-carbon ink on the same PCB board. Summary of the Invention

[0005] This invention provides a method for preparing a PCB board with high-carbon ink, mainly comprising:

[0006] Obtain a dataset of conductivity distribution that includes region identifiers, conductivity values, and boundary coordinates;

[0007] Based on the conductivity distribution dataset, a mapping relationship between printhead number, ink type, and target area coordinates is generated;

[0008] A micron-level precision control algorithm is used to generate a set of motion control commands corresponding to timestamps, nozzle positions, and deposition coordinates.

[0009] Acquire an image dataset of ink deposition during the deposition process and calculate the deviation between the actual deposition amount and the target deposition amount;

[0010] If the deviation value exceeds the preset threshold, the injection frequency configuration file containing proportional-integral-derivative parameters is updated.

[0011] The deposition process is controlled according to the updated jet frequency profile, and the transition width of the boundary region is detected.

[0012] If the transition width exceeds a micrometer-level threshold, an ink deposition dataset containing optimized boundary coordinates is generated.

[0013] Based on the ink deposition dataset, generate zone heating control parameters corresponding to the region identifier and curing temperature;

[0014] Acquire actual temperature distribution data during the curing process; if the temperature deviation exceeds the threshold, generate a new power control command.

[0015] Measure the final conductivity parameters and generate a verification dataset corresponding to the region identifiers and measured conductivity.

[0016] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0017] This invention discloses a method for the precise deposition of high-carbon ink on PCB boards. The method first acquires conductivity distribution data, generates a mapping relationship between the nozzle and the target area, and uses a micron-level precision control algorithm to generate motion commands. During deposition, deviations are calculated through real-time image analysis, and the spraying frequency is dynamically adjusted. For boundary areas, the transition width is detected and deposition data is optimized. During curing, heating is performed in zones according to regional characteristics, and the temperature distribution is monitored and power adjusted in real time. Finally, conductivity is measured to generate a verification dataset. This invention achieves high-precision and controllable deposition of high-carbon ink on PCB boards, effectively improving the conductivity and consistency of PCB boards, and providing a new technical solution for high-performance electronic circuit manufacturing. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating a high-carbon ink PCB board and its preparation method according to the present invention.

[0019] Figure 2 This is a schematic diagram of a PCB board with high carbon ink and its preparation method according to the present invention;

[0020] Figure 3 This is another schematic diagram of a PCB board with high carbon ink and its preparation method according to the present invention. Detailed Implementation

[0021] 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.

[0022] like Figure 1-3 The specific method for preparing a PCB board with high-carbon ink in this embodiment may include:

[0023] Step S101: Obtain a conductivity distribution dataset containing region identifiers, conductivity values, and boundary coordinates.

[0024] An initial dataset is acquired from a sensor network, containing region identifiers, conductivity values, and boundary coordinates. Noise detection is performed on the initial dataset; if noise is detected, a median filtering algorithm is applied to obtain a denoised dataset. Based on the region identifiers and boundary coordinates in the denoised dataset, a K-means clustering algorithm is used to divide the dataset into regions, determining the number of clusters to obtain the region division results. The conductivity values ​​of each region in the region division results are extracted from the denoised dataset to generate a conductivity distribution feature. The variance of the conductivity distribution feature is calculated; if the variance exceeds a preset threshold, a linear interpolation algorithm is used to smooth the conductivity values, obtaining a smoothed conductivity distribution. Based on the smoothed conductivity distribution and the region division results, a conductivity distribution dataset containing region identifiers, conductivity values, and boundary coordinates is generated. A boundary coordinate consistency verification algorithm is used to compare the boundary coordinates of the conductivity distribution dataset with those of the initial dataset to determine the final conductivity distribution dataset.

[0025] For example, the initial dataset collected by the sensor network includes region identifiers, conductivity values, and boundary coordinates. Taking a soil monitoring area as an example, assuming the area is divided into three sub-regions A, B, and C, the initial dataset records: the conductivity values ​​of region A are 50, 52, and 48, and the boundary coordinates are (0,0)-

[0026] (10,10); the conductivity values ​​of region B are 60, 65, and 70, with boundary coordinates of (10,0)-(20,10); the conductivity values ​​of region C are 45, 50, and 55, with boundary coordinates of (0,10)-(10,20). These data may contain noise due to sensor jitter or environmental interference.

[0027] In one possible implementation, noise detection is achieved by examining abnormal fluctuations in conductivity values. If the difference between values ​​within a certain region exceeds a preset threshold (e.g., 10), it is determined to be noise.

[0028] For example, the difference between 65 and 70 in region B is 5, which does not exceed the threshold and is therefore considered noise; however, if a value suddenly changes to 90, it is considered noise. After noise is detected, a median filtering algorithm is used to process it, selecting the median value of adjacent data to replace the outlier.

[0029] For example, if 48 in region A becomes 90, then the median of 50, 52, and 90, 52, is used to replace it, resulting in a denoised dataset where region A has 50, 52, and 52. This method effectively smooths outliers, preserves data trends, and improves the reliability of subsequent analysis.

[0030] Specifically, based on the region identifiers and boundary coordinates of the denoised dataset, the K-means clustering algorithm is used to divide the regions. Assuming the initial number of clusters K = 3, the algorithm assigns data points to the nearest cluster according to the geometric center of the boundary coordinates.

[0031] For example, if the coordinate centers of regions A, B, and C are (5,5), (15,5), and (5,15) respectively, clustering preserves the original divisions, resulting in the region partitioning. The advantage of K-means clustering lies in its ability to automatically optimize region boundaries, adapt to complex terrain, and improve partitioning accuracy.

[0032] In one embodiment, the conductivity performance value of each region is extracted from the denoised dataset to generate a performance distribution feature.

[0033] For example, the performance distribution characteristics of region A are an average of 51.3, region B is 65, and region C is 50. This characteristic reflects the spatial differences in soil conductivity and can be used to assess soil fertility or salinity distribution. The variance of the performance distribution characteristics is calculated; assuming a variance of 50, exceeding a preset threshold of 30 indicates significant fluctuations in performance values, which may affect the stability of the analysis.

[0034] Preferably, a linear interpolation algorithm is used to smooth the performance values.

[0035] For example, the conductivity values ​​of region B, 60, 65, and 70, fluctuate significantly. By interpolating to generate intermediate values ​​of 62.5 and 67.5, a smoothed sequence of 60, 62.5, 67.5, and 70 is obtained. Smoothing reduces the impact of abrupt changes, making the performance distribution more continuous and facilitating subsequent modeling and visualization.

[0036] For example, based on the smoothing performance distribution and region segmentation results, a conductivity performance distribution dataset is generated, containing region identifiers, performance values, and boundary coordinates. For instance, region A: performance value 51.3, coordinates (0,0)-

[0037] (10,10). This dataset integrates spatial and performance information and can be used for soil management decisions.

[0038] Understandably, the boundary coordinates of the conductivity performance distribution dataset and the initial dataset are compared using a boundary coordinate consistency verification algorithm to ensure data consistency.

[0039] For example, verify whether the coordinates (0,0)-(10,10) of region A are consistent. If a deviation is found, adjust the coordinates to match the initial dataset. The final conductivity distribution dataset has high accuracy and consistency, which can support precision agriculture or environmental monitoring.

[0040] It should be noted that each step of the above method revolves around improving data quality and analytical accuracy. Noise detection and filtering ensure data reliability, clustering optimizes region partitioning, smoothing enhances distribution continuity, and boundary verification guarantees spatial consistency. These technical effects collectively support the accurate analysis of soil conductivity, providing a reliable basis for scientific decision-making.

[0041] It is understood that this embodiment uses soil and farmland as examples to describe the method used in this application and the specific application of high carbon ink on PCB boards. Therefore, the above content related to soil and farmland is only illustrative and is used to verify and explain the methods and means used in this application, and does not deviate from the subject matter of this application.

[0042] Step S102: Generate a mapping relationship between printhead number, ink type and target area coordinates based on the conductivity distribution dataset.

[0043] The system obtains region identifiers, conductivity values, and boundary coordinates from a conductivity distribution dataset. A data extraction algorithm is used to generate a data table containing these identifiers, values, and coordinates, resulting in an initial data table. Based on the region identifiers and coordinates in the initial data table, a K-means clustering algorithm is used to divide the regions, yielding region partitioning results. If the boundary coordinates in the partitioned regions differ from those in the initial data table, a coordinate consistency verification algorithm is used to adjust the partitioning results, resulting in consistent region partitioning. From the consistent partitioning, the system extracts the region identifier and target region coordinates for each region, generating a region coordinate table containing these identifiers and coordinates. Based on the region identifiers in the coordinate table, a preset printhead allocation rule is used to match printhead numbers and ink type names, generating a printhead allocation table containing printhead numbers, ink type names, and target region coordinates. Finally, the printhead allocation table extracts the printhead numbers, ink type names, and target region coordinates, generating a mapping table containing these relationships. Based on the target region coordinates in the mapping table and the boundary coordinates in the consistent region division, the mapping table is verified using a coordinate consistency verification algorithm to obtain the final mapping table.

[0044] For example, the conductivity distribution dataset includes region identifiers, conductivity values, and boundary coordinates to describe the spatial distribution of soil conductivity. Taking a farmland as an example, the dataset records three regions: region X has conductivity values ​​of 40, 42, and 45, with boundary coordinates of (0,0)-(8,8); region Y has conductivity values ​​of 50, 55, and 60, with boundary coordinates of (8,0)-(16,8); and region Z has conductivity values ​​of 35, 38, and 40, with boundary coordinates of (0,8)-(8,16). The data extraction algorithm generates an initial data table from this dataset, containing region identifiers such as X, Y, and Z, a sequence of conductivity values, and corresponding boundary coordinates. The structure is clear and facilitates subsequent processing.

[0045] In one possible implementation, the K-means clustering algorithm divides the regions based on the region identifiers and boundary coordinates of the initial data table. Assuming K=3, the algorithm uses the geometric centers of the boundary coordinates, such as (4,4) for region X, (12,4) for region Y, and (4,12) for region Z, as the initial cluster centers, assigning data points to the nearest cluster. If the boundary coordinates of region X become (0,0)-(9,9) after clustering, which is inconsistent with the initial data table, a coordinate consistency verification algorithm is used to compare the boundary point deviations and adjust them to (0,0)-(8,8) to obtain a consistent region division, ensuring accurate spatial partitioning.

[0046] Specifically, region identifiers and target region coordinates are extracted from the consistent region division to generate a region coordinate table.

[0047] For example, region X is identified by the identifier X, and the target region coordinates are (0,0)-(8,8). This table provides the basis for subsequent printhead assignment. The printhead assignment rules match the printhead number value and ink type name based on the region identifier.

[0048] For example, region X is matched with printhead S1, and the ink type is nutrient solution A; region Y is matched with printhead S2, and the ink type is nutrient solution B. A printhead assignment table is generated, containing printhead S1, nutrient solution A, and coordinates (0,0)-

[0049] Information such as (8,8) clarifies the correspondence between the nozzles and the areas.

[0050] Preferably, the printhead number, ink type name, and target area coordinates are extracted from the printhead allocation table to generate a mapping table.

[0051] For example, the mapping table records the nutrient solution A corresponding to sprinkler S1 and the coordinates (0,0)-(8,8). To ensure accuracy, a coordinate consistency verification algorithm is used to compare the target area coordinates in the mapping table with the boundary coordinates of the consistent area division. If the coordinate deviation of sprinkler S2 is found to be (8,0)-(17,8), it is adjusted to (8,0)-(16,8), resulting in the final mapping table. This table supports precise scheduling of spraying equipment.

[0052] Understandably, the above process, through data extraction, clustering, coordinate verification, and mapping relationship generation, progresses step by step to ensure high consistency and usability of the data table from initial to final.

[0053] For example, the initial data table provides complete information, the area coordinate table focuses on spatial division, and the nozzle allocation table and mapping relationship table directly serve the spraying operation. The logic is rigorous and the steps are closely linked.

[0054] Step S103: A motion control command set corresponding to the timestamp, nozzle position, and deposition coordinates is generated using a micron-level precision control algorithm.

[0055] Timestamps and nozzle motion trajectory data are obtained from time-series data. Threshold segmentation is used to separate the position coordinates in the nozzle motion trajectory data, resulting in a set of nozzle position coordinates. Target deposition coordinates are acquired using a vision sensor, yielding an initial coordinate dataset. Based on the timestamps and nozzle position coordinates in the initial coordinate dataset, Kalman filtering is used to calculate the motion acceleration of adjacent coordinate points. If the motion acceleration exceeds a preset acceleration threshold, a discontinuity in the nozzle motion trajectory is identified. Cubic spline interpolation is used between discontinuities to generate new coordinate points, resulting in a smooth motion trajectory where the rate of change of velocity is lower than a preset smoothness threshold. The interpolated nozzle position coordinates are extracted from the smooth motion trajectory, and a mapping relationship is established between the timestamps and the nozzle position coordinates and the target deposition coordinates through time window matching, resulting in a mapping table of timestamps, nozzle position coordinates, and target deposition coordinates. Based on the target deposition coordinates corresponding to each timestamp in the mapping table, a preset G-code instruction template is matched to generate a preliminary instruction set. The actual motion path during the execution of the preliminary instruction set is measured using a laser interferometer. If the deviation between the actual motion path and the target deposition coordinates exceeds a preset tolerance threshold, the parameters of the preliminary instruction set are adjusted to obtain the final motion control instruction set.

[0056] For example, time series data typically includes timestamps and nozzle movement trajectory data to describe changes in the position of the nozzles during precision spraying of farmland.

[0057] For example, a set of time series data records the position coordinates of a nozzle at timestamp t1 = 10:00:00 as (2,3) and at t2 = 10:00:01 as (2.5,3.2). Using a threshold segmentation method, a position change threshold of 0.3 can be set to separate the continuously moving coordinate points, resulting in a set of nozzle position coordinates, such as {(2,3),}

[0058] (2.5,3.2)}. This process ensures the removal of outliers in the data, providing a reliable foundation for subsequent analysis.

[0059] In one possible implementation, a vision sensor is used to acquire the target deposition coordinates.

[0060] For example, a sensor scans a farmland area and identifies the target deposition coordinates as (3,4), generating an initial coordinate dataset containing a timestamp t1 and the coordinates (3,4). This dataset provides a spatial basis for matching nozzle movement with the target area, reducing the possibility of indiscriminate spraying.

[0061] Specifically, Kalman filtering is used to calculate the motion acceleration of adjacent coordinate points of the nozzle.

[0062] For example, based on coordinates (2,3) and (2.5,3.2), combined with a 1-second timestamp interval, the filter estimates acceleration. If the preset acceleration threshold is 0.5 m / s²... 2 The calculated result is 0.7 m / s. 2 If a discontinuity is found, it is determined that a discontinuity exists. This method improves the robustness of trajectory analysis through dynamic prediction and correction.

[0063] Preferably, cubic spline interpolation is used to generate new coordinate points between discontinuities.

[0064] For example, inserting the point (2.3, 3.1) between (2, 3) and (2.5, 3.2) generates a smooth motion trajectory with a velocity change rate below the smoothness threshold of 0.2 m / s. 2 The smooth trajectory reduces vibration during nozzle movement and improves spray uniformity.

[0065] For example, the interpolated coordinates, such as (2.3, 3.1), are extracted from the smooth motion trajectory. A mapping relationship is established with the target deposition coordinates (3, 4) through time window matching, generating a mapping relationship table that records t1 corresponding to (2.3, 3.1) and (3, 4). This table provides accurate time-space correlation for subsequent instruction generation.

[0066] In one embodiment, a preliminary instruction set is generated by matching G-code instruction templates according to a mapping table.

[0067] For example, for (3,4), the command "G1X3Y4F100" is generated, indicating that the nozzle moves to (3,4) at a speed of 100 mm / s. This process transforms abstract coordinates into commands that the device can execute.

[0068] Understandably, laser interferometers measure the actual motion path.

[0069] For example, after executing the command, the actual path is (3.1, 4.2), which deviates from the target (3, 4) by 0.22, exceeding the tolerance threshold of 0.1. At this point, the command parameters are adjusted, such as reducing the speed to 80 mm / s, to generate the final motion control command set. This verification and adjustment mechanism ensures high precision in spraying operations.

[0070] It should be noted that the above method, through time series analysis, trajectory smoothing, coordinate mapping, and command optimization, provides systematic support for precision spraying in a progressive manner. The implementation of each step revolves around the core needs of farmland spraying, with rigorous logic and mutual support.

[0071] Step S104: Obtain the ink deposition image dataset during the deposition process and calculate the deviation between the actual deposition amount and the target deposition amount.

[0072] Images of the ink deposition process are captured using an industrial camera at a preset frame rate, and images are cropped at preset time intervals to generate an original image set. An image segmentation algorithm is used to process the original image set, with a preset number of iterations, to extract the contours of the ink deposition areas, generating a segmented image set. The pixel area of ​​each contour region in the segmented image set is calculated using a contour area calculation function and multiplied by a preset conversion coefficient to generate an actual deposition amount dataset. The target deposition amount value is read from a configuration file to generate a target deposition amount dataset. If the absolute difference between the actual deposition amount and the target deposition amount exceeds a preset threshold, a mean filtering function is used to filter the difference sequence with a preset window size to generate a deposition deviation dataset. The deposition deviation dataset is matched with a pre-established compensation parameter lookup table to generate an adjustment parameter set; the compensation parameter lookup table contains the correspondence between deviation ranges and compensation coefficients. Based on the compensation coefficients in the adjustment parameter set, a contour detection function is used to re-detect the deposition region boundaries in the segmented image set, generating an optimized deposition amount dataset.

[0073] For example, an industrial camera captures images of the ink deposition process at a preset frame rate, such as 30 frames per second, to ensure the capture of dynamic deposition details. The camera is aimed at the inkjet area and captures a frame every 0.5 seconds to generate the original image set.

[0074] For example, 20 frames of images can be generated within 10 seconds, recording the continuous changes in ink deposition.

[0075] It should be noted that the choice of frame rate and interval needs to balance image quality and processing efficiency. High frame rate is suitable for fast deposition scenarios, while low frame rate is suitable for stable processes.

[0076] In one possible implementation, the image segmentation algorithm processes the original image set, often employing a threshold-based segmentation method. The number of iterations is set to 5, and the ink regions are separated from the background through multiple optimizations.

[0077] For example, by setting a grayscale threshold of 150, the outline of the ink region can be extracted to generate a segmented image set. If the background is complex, edge detection algorithms can be used to enhance the clarity of the outline. The segmented image set provides a reliable foundation for subsequent area calculations.

[0078] Specifically, the contour area calculation function statistically calculates the pixel area of ​​each contour in the segmented image set.

[0079] For example, if the outline pixel area of ​​a frame is 10,000 pixels, multiplying it by a preset conversion factor of 0.01 square millimeters per pixel yields an actual deposition amount of 100 square millimeters. Repeating this process generates a dataset of actual deposition amounts containing deposition amounts from multiple frames.

[0080] Preferably, the conversion factor needs to be calibrated according to the camera resolution and actual size to ensure data accuracy.

[0081] For example, the target deposition amount value is read from the configuration file, such as 120 square millimeters per frame, and a target deposition amount dataset is generated. If the absolute value of the difference between the actual deposition amount of 100 square millimeters and the target of 120 square millimeters is 20 square millimeters, exceeding the preset threshold of 10 square millimeters, further processing is required. The configuration file can store multiple target values ​​to adapt to different inkjet requirements.

[0082] In one embodiment, the mean filtering function processes the difference sequence, with a window size of 3 frames.

[0083] For example, with difference sequences of 20, 18, and 22 square millimeters, filtering generates a smoothed depositional deviation dataset, such as 19.33 square millimeters. Filtering reduces the impact of noise and improves the stability of the deviation data. The window size can be adjusted according to the deposition rate; smaller windows are needed for rapidly changing scenarios.

[0084] Understandably, the sedimentation deviation dataset is matched with the compensation parameter lookup table, which records the correspondence between the deviation range and the compensation coefficient.

[0085] For example, a deviation of 10-20 square millimeters corresponds to a compensation coefficient of 1.1. After matching, an adjustment parameter set is generated, which includes coefficient 1.1.

[0086] It should be noted that the comparison table needs to be pre-constructed based on experimental data to ensure the appropriateness of the compensation.

[0087] In one embodiment, the contour detection function re-detects the deposition region boundary of the segmented image set according to the compensation coefficient of 1.1 of the adjusted parameter set.

[0088] For example, by adjusting the detection threshold and re-extracting the contours, an optimized deposition dataset can be generated, such as adjusting the deposition amount to 118 square millimeters, which is close to the target value.

[0089] Preferably, morphological operations can be combined to smooth the boundaries, improving detection accuracy. The optimized dataset provides a precise basis for subsequent inkjet control.

[0090] Step S105: If the deviation value exceeds a preset threshold, update the injection frequency configuration file containing proportional-integral-derivative parameters.

[0091] The deposition deviation value sequence is obtained. If the deposition deviation value sequence exceeds a preset threshold, the time series features of the deposition deviation value sequence are extracted using an ARIMA model to obtain the deviation change trend. Based on the deviation change trend, Kalman filtering is used to smooth the deposition deviation value sequence to obtain a smoothed deviation sequence. If the fluctuation amplitude of the smoothed deviation sequence exceeds a preset amplitude threshold, a normal distribution is used to fit the smoothed deviation sequence to determine the deviation distribution parameters. Based on the deviation distribution parameters, the proportional, integral, and derivative parameters in the configuration file are updated using the gradient descent method to obtain the updated control parameter set. The injection frequency value is calculated using a PID controller based on the updated control parameter set to obtain the adjusted frequency parameter set. If the difference between the adjusted frequency parameter set and the historical frequency parameter set exceeds a preset difference threshold, the root locus method is used to perform stability analysis on the adjusted frequency parameter set to obtain the verified frequency parameter set. Based on the verified frequency parameter set, the injection frequency configuration file is updated using JSON format to obtain the final frequency configuration file.

[0092] For example, the deposition deviation value sequence can be obtained by combining image processing and data analysis. In an inkjet deposition scenario, an industrial camera captures images of the inkjet area, which are then segmented and their areas calculated to obtain the actual deposition amount. Assuming the target deposition amount is 100 square millimeters and the actual amount is 95 square millimeters, the deviation is 5 square millimeters. Multiple frames are continuously acquired to generate a deviation sequence, such as 5, 6, 4, and 7 square millimeters.

[0093] It should be noted that the deviation sequence needs to be updated regularly to reflect the dynamic changes in the inkjet process.

[0094] In one possible implementation, if the deviation sequence exceeds a preset threshold, such as 3 square millimeters, the ARIMA model is used to analyze the time series characteristics. The ARIMA model extracts trend and periodic features by analyzing the autoregressive, differencing, and moving average components of the sequence.

[0095] For example, if the model prediction bias shows an upward trend over the next 5 frames, it suggests that there may be persistent bias in the inkjet system.

[0096] Preferably, the parameters of the ARIMA model need to be optimized based on historical data to adapt to different deposition rates.

[0097] Specifically, Kalman filtering is used to smooth biased sequences and reduce the impact of noise. Assuming the biased sequence is 5, 6, 4, 7 square millimeters, the filtered sequence will be smoothed to 5.2, 5.8, 4.3, 6.5 square millimeters.

[0098] Understandably, Kalman filtering dynamically adjusts the bias value through state estimation and measurement updates to ensure a more stable sequence. The filter parameters need to be calibrated according to the noise characteristics of the inkjet environment.

[0099] For example, if the fluctuation range of the smoothed deviation sequence exceeds a preset threshold, such as 2 square millimeters, a normal distribution is used for fitting to determine the mean and standard deviation. Assuming the fitting result shows a mean of 5.5 square millimeters and a standard deviation of 0.8 square millimeters, it indicates that the deviation distribution is relatively concentrated.

[0100] In one embodiment, the distributed parameters are used to assess system stability and guide subsequent parameter adjustments.

[0101] In one embodiment, the gradient descent method updates the proportional, integral, and derivative parameters of the PID controller based on the deviation distribution parameters. Assuming an initial proportional parameter of 0.5, it is iteratively adjusted to 0.55 based on the mean and standard deviation of the deviation.

[0102] Preferably, the learning rate should be moderate to avoid parameter oscillations. The updated control parameter set improves inkjet accuracy.

[0103] For example, the PID controller calculates the injection frequency based on the updated parameters. Assuming the original frequency is 100Hz and it is adjusted to 105Hz, a set of frequency parameters is generated.

[0104] It should be noted that the nozzle response time must be taken into account when adjusting the frequency to ensure a smooth transition.

[0105] Specifically, if the difference between the adjusted frequency parameter set and the historical parameter set exceeds a threshold, such as 5 Hz, the root locus method is used to analyze stability. The root locus method confirms that the frequency adjustment will not cause oscillations by checking the location of the system poles.

[0106] In one embodiment, the analysis results show that the poles are located in the stable region, verifying the validity of the frequency parameter set.

[0107] Understandably, the final frequency parameter set updates the configuration file in JSON format.

[0108] For example, the configuration file records the frequency of 105Hz, timestamp, and inkjet mode for easy system reading and execution.

[0109] Preferably, the JSON format should have a clear structure to ensure compatibility.

[0110] Step S106: Control the deposition process according to the updated jet frequency configuration file and detect the transition width of the boundary region.

[0111] Image data of the boundary region during the deposition process is acquired, and the boundary region is extracted using an image segmentation algorithm to obtain a segmented boundary image. If the pixel distribution entropy value of the segmented boundary image exceeds a preset entropy threshold, median filtering is used to denoise the segmented boundary image to obtain a denoised boundary image. Based on the denoised boundary image, the Canny edge detection algorithm is used to extract the boundary transition region to obtain the boundary transition contour. If the continuity parameter of the boundary transition contour is lower than a preset continuity threshold, morphological dilation is used to repair contour breaks to obtain a repaired boundary contour. Based on the repaired boundary contour, the pixel width of the boundary transition region is calculated to obtain a boundary transition width sequence. If the variance of the boundary transition width sequence exceeds a preset variance threshold, the K-means clustering algorithm is used to classify the width sequence to obtain a classified width sequence. Based on the classified width sequence, the sliding window method is used to calculate the local mean to obtain a smoothed width sequence. If the fluctuation amplitude of the smoothed width sequence exceeds a preset amplitude threshold, the smoothed width sequence is optimized using linear interpolation to obtain an optimized width sequence. Based on the optimized width sequence, a boundary transition width configuration file is generated in JSON format to obtain a width configuration file. Based on the width configuration file, the control parameters of the deposition process are adjusted to obtain an adjusted control parameter set.

[0112] Specifically, in inkjet deposition scenarios, acquiring image data of the boundary regions during the deposition process is a crucial step.

[0113] For example, an industrial camera captures images of the inkjet area at a frequency of 30 frames per second, generating raw image data including the boundary areas. The image resolution is 1280×720 pixels, and the grayscale value is 0 to 255.

[0114] It should be noted that the boundary region typically refers to the interface between the inkjet-deposited material and the substrate, and its image characteristics are characterized by drastic changes in grayscale values. An image segmentation algorithm is used to extract the boundary region, resulting in a segmented boundary image.

[0115] Preferably, based on the Otsu thresholding method, the threshold is automatically determined by analyzing the image grayscale histogram to separate the boundary region from the background.

[0116] For example, assuming a threshold of 150, pixels with gray values ​​higher than 150 are classified as boundary regions, generating a binarized segmented image.

[0117] Understandably, the Otsu thresholding method can adapt to image characteristics under different lighting conditions. If the pixel distribution entropy value of the segmentation boundary image exceeds a preset entropy threshold, denoising processing is performed. The pixel distribution entropy value reflects the degree of disorder in the gray-level distribution of image pixels.

[0118] In one embodiment, the entropy value is calculated using the Shannon entropy formula. Assuming a result of 5.2 and a preset threshold of 5.0, this indicates significant image noise. In this case, a 3×3 window median filter is used to replace each pixel with the median of its neighboring pixels, resulting in a denoised boundary image. After denoising, the image edges are smoother, and noise points are reduced. Based on the denoised boundary image, the Canny edge detection algorithm is used to extract the boundary transition region. The Canny algorithm generates the boundary transition contour through Gaussian filtering, gradient calculation, and double thresholding.

[0119] For example, a low threshold of 50 and a high threshold of 150 are set to detect continuous boundary lines. If the continuity parameter of the boundary transition contour is lower than a preset threshold, such as 80%, the break is repaired through morphological dilation. In one possible implementation, a 3×3 structuring element is used for dilation to fill small gaps in the contour, resulting in a repaired boundary contour. Based on the repaired boundary contour, the pixel width of the boundary transition region is calculated to obtain a boundary transition width sequence.

[0120] Specifically, the pixel width of the boundary region is measured along the contour normal, assuming a sequence of 10, 12, 9, and 11 pixels. If the sequence variance exceeds a preset threshold, such as 2.0, the K-means clustering algorithm is used to classify the width sequence.

[0121] For example, setting K=2 divides the widths into two categories: narrower widths of approximately 9 pixels and wider widths of approximately 12 pixels, generating a categorized width sequence. A sliding window method is then used to calculate the local mean, resulting in a smooth width sequence.

[0122] In one embodiment, the window size is 3, and the calculated results are 10.3, 10.7, and 10.3 pixels. If the fluctuation amplitude of the smooth width sequence exceeds a preset threshold, such as 1.0 pixel, the sequence is optimized using linear interpolation.

[0123] For example, intermediate values ​​are inserted at points with large fluctuations to generate an optimized width sequence, such as 10.4, 10.6, 10.4 pixels. Based on the optimized width sequence, a boundary transition width configuration file in JSON format is generated.

[0124] For example, the configuration file includes a width sequence, timestamp, and inkjet pattern, such as

[0125] {"widths":[10.4,10.6,10.4],"timestamp":"2025-04-19T10:00:00",

[0126] "mode":"standard"}. Adjust the control parameters of the deposition process according to the width configuration file.

[0127] For example, based on the width sequence, the nozzle movement speed is adjusted from 50 mm / s to 52 mm / s to generate an adjustment control parameter set, thereby improving deposition uniformity.

[0128] Step S107: If the transition width exceeds a micrometer-level threshold, an ink deposition dataset containing optimized boundary coordinates is generated.

[0129] A boundary region image is acquired, including pixel distribution characteristics. If the pixel distribution characteristics exceed a preset distribution threshold, the boundary region image is segmented using the Otsu algorithm to obtain a segmented boundary image. Based on the segmented boundary image, the Canny edge detection algorithm is used to extract the boundary contour, obtaining an initial boundary contour. If the continuity of the initial boundary contour is lower than a preset continuity threshold, the initial boundary contour is repaired using a closing operation to obtain a repaired boundary contour. Based on the repaired boundary contour, a boundary tracking algorithm is used to calculate the coordinate point set of the boundary transition region, obtaining a boundary coordinate dataset. If the coordinate point distribution density of the boundary coordinate dataset exceeds a preset density threshold, the boundary coordinate dataset is classified using a K-means clustering algorithm to obtain a classified coordinate dataset. Based on the classified coordinate dataset, cubic spline interpolation is used to optimize the coordinate points in the classified coordinate dataset to obtain optimized boundary coordinates. The optimized boundary coordinates are integrated with ink deposition parameters using JSON format to obtain an ink deposition dataset.

[0130] Specifically, in inkjet deposition scenarios, acquiring boundary region images is a crucial step in analyzing the interface between the deposited material and the substrate. Boundary region images typically contain pixel distribution characteristics, reflecting the spatial variations in grayscale values.

[0131] For example, pixel distribution characteristics can be characterized by the standard deviation of the image's grayscale histogram. Assuming a 1280×720 pixel grayscale image with a calculated standard deviation of 20 and a preset distribution threshold of 15, this indicates high pixel distribution complexity, requiring further processing. The Otsu algorithm automatically determines the segmentation threshold to separate boundary regions from the background by analyzing the grayscale histogram.

[0132] For example, with a threshold of 140, pixels with gray values ​​higher than 140 are classified as boundary regions, generating a binarized segmented image.

[0133] Understandably, the Otsu algorithm can adapt to different lighting conditions, improving segmentation robustness. Based on the segmentation boundary image, the Canny edge detection algorithm is used to extract boundary contours. The Canny algorithm smooths the image using Gaussian filtering and then calculates the gradient to detect edge points.

[0134] Preferably, a low threshold of 60 and a high threshold of 160 are set to generate an initial boundary profile. If the continuity of the initial boundary profile is lower than a preset threshold, such as 85%, it needs to be repaired.

[0135] In one embodiment, the closing operation uses a 3×3 structuring element to first expand and then erode, filling small breaks in the contour to obtain a repaired boundary contour. The closing operation effectively connects discontinuous edges, enhancing contour integrity. Based on the repaired boundary contour, a boundary tracing algorithm calculates the coordinate point set of the boundary transition region.

[0136] It should be noted that the boundary tracing algorithm traverses the contour pixel by pixel, recording the coordinates of each boundary point.

[0137] For example, a boundary coordinate dataset containing 5000 coordinate points is generated. If the density of coordinate points exceeds a preset threshold, such as more than 5 coordinate points per 10 pixels, then classification processing is required.

[0138] Specifically, the K-means clustering algorithm divides coordinate points into two categories: dense regions and sparse regions. Assuming the clustering result is 3000 dense points and 2000 sparse points, a categorized coordinate dataset is generated. Clustering helps distinguish the complexity and regularity of boundary regions. Cubic spline interpolation is used to optimize the coordinate points in the categorized coordinate dataset. Cubic spline interpolation uses a piecewise polynomial to fit the coordinate points, generating a smooth curve. In one possible implementation, the spacing between coordinate points is uniform after interpolation, and the optimized boundary coordinates better match the actual shape of the boundary curve. A smooth coordinate point set improves the accuracy of subsequent analysis. The optimized boundary coordinates and ink deposition parameters are integrated using JSON format to generate an ink deposition dataset.

[0139] For example, the dataset contains a sequence of boundary coordinates, inkjet flow rate such as 0.5 ml / s, and timestamps such as 2025-04-19T12:00:00, in the format {"coordinates":"[(x1,y1),

[0140] (x2,y2), ...]", "flow":"0.5", "timestamp":"2025-04-19T12:00:00"}. The JSON structure facilitates parameter passing and storage, which helps in the precise control of the deposition process.

[0141] In one embodiment, the boundary coordinate dataset can be further combined with nozzle position parameters to generate a dynamic adjustment scheme.

[0142] For example, the nozzle path can be adjusted based on the coordinates of dense areas to reduce deposition overlap.

[0143] Preferably, the inkjet frequency can be increased in sparse areas to ensure uniform coverage. These adjustments improve deposition quality and consistency through synergistic optimization of multi-dimensional parameters.

[0144] Step S108: Generate zone heating control parameters corresponding to the region identifier and curing temperature based on the ink deposition dataset.

[0145] An ink deposition dataset is obtained, containing boundary coordinates and ink deposition parameters. Gaussian filtering is used to denoise the ink deposition parameters, resulting in a preprocessed deposition parameter set. Based on the gradient distribution of the preprocessed deposition parameter set, region identifiers are extracted using threshold segmentation. A quadtree spatial partitioning algorithm is used to partition the region identifiers, setting a minimum partition area as a preset value, resulting in a partition identifier set. The coefficient of variation (COP) of the boundary points of the partition identifier set is calculated; if the COP is greater than a preset threshold, region merging is performed. Adjacent partitions are detected using Delaunay triangulation; if the standard deviation of the area of ​​adjacent partitions is greater than a preset threshold, they are merged into the same partition, resulting in an adjusted partition identifier set. For the adjusted partition identifier set, the mean pixel coordinates of each partition are calculated, resulting in a geometric center coordinate set. Based on the geometric center coordinate set and a preset curing temperature lookup table, a two-dimensional bilinear interpolation algorithm is used to generate partition heating control parameters, resulting in a partition heating control parameter set. The heating parameter values ​​of each partition are extracted from the partition heating control parameter set, and a sliding window mean filter is used to obtain a smoothed heating parameter set. The range of the smoothed heating parameter set is calculated. If the range is greater than a preset threshold, an inverse distance weighting method is used to weight the parameters of adjacent zones, with the weighting coefficient being the reciprocal of the zone center distance, to obtain the optimized heating parameter set. Based on the optimized heating parameter set, a linear proportional converter is used to generate PWM duty cycle instructions, resulting in the zone heating control instruction set.

[0146] It should be noted that the Delaunay triangulation mentioned above refers to constructing a triangular network that satisfies certain specific properties given a set of points. This network divides all points into triangles, aiming to maximize the minimum angle of each triangle and ensure that the circumcircle of each triangle does not contain any other points. Therefore, this embodiment uses Delaunay triangulation to detect adjacent partitions, which can improve computational stability and optimize computational efficiency; it can also simplify adjacent partition detection and adapt to dynamic changes.

[0147] Specifically, obtaining ink deposition datasets is a core step in the inkjet printing process. Datasets typically include boundary coordinates and ink deposition parameters, such as nozzle flow rate or deposition time.

[0148] For example, boundary coordinates can be obtained through image processing, while ink deposition parameters are extracted from inkjet equipment logs, such as a nozzle flow rate of 0.3 mL / s and a deposition time of 2 seconds. This dataset provides the basis for subsequent optimization. Gaussian filtering is used to denoise the ink deposition parameters.

[0149] It should be noted that Gaussian filtering reduces noise interference by using a weighted average to smooth the parameters.

[0150] In one embodiment, for the nozzle flow rate data, the standard deviation of the Gaussian kernel is set to 1.5, and the flow rate fluctuation is reduced from ±0.05 mL / s to ±0.01 mL / s after filtering, ensuring parameter stability. Region identifiers are extracted by threshold segmentation based on the gradient distribution of the preprocessed deposition parameter set.

[0151] Specifically, the gradient reflects the spatial rate of change of parameters, and high gradient regions usually correspond to uneven deposition.

[0152] Preferably, a gradient threshold of 0.1 is set to segment high-variability regions and generate labeled images, with a label value of 1 indicating regions requiring optimization. A quadtree spatial partitioning algorithm is used to partition the region labels.

[0153] Understandably, a quadtree divides a region into sub-regions of varying sizes through recursive partitioning. One possible implementation sets the minimum partition area to 100 pixels, generating a partition identifier set containing 50 sub-regions for finer-grained analysis. The coefficient of variation of the boundary points of the partition identifier set is calculated; if it exceeds a preset threshold, region merging is performed.

[0154] For example, a coefficient of variation of 0.4 exceeds the threshold of 0.3, indicating that the distribution of boundary points is uneven.

[0155] In one embodiment, adjacent partitions with fewer than 10 boundary points are merged to reduce fragmented partitions and improve processing efficiency. Adjacent partitions are detected and merged based on Delaunay triangulation.

[0156] For example, triangulation constructs the topological relationships between partitions. If the standard deviation of the area of ​​adjacent partitions is greater than 200 pixels, they are merged into a single partition. After adjustment, the number of partitions is reduced from 50 to 30, generating an optimized partition identifier set and reducing computational complexity. For the adjusted partition identifier set, the mean pixel coordinates of each partition are calculated to obtain the geometric center coordinate set.

[0157] Specifically, a certain partition contains 1000 pixels with a mean coordinate of (500, 300), forming a set of 30 center points, representing the spatial distribution of the partition. Based on the geometric center coordinate set and the curing temperature lookup table, two-dimensional bilinear interpolation is used to generate the partition heating control parameters.

[0158] Preferably, the lookup table defines a temperature range of 100-150 degrees Celsius. After interpolation, the heating temperature of a certain zone is 120 degrees Celsius, generating a parameter set containing 30 temperature values ​​to ensure accurate zone heating. Heating parameter values ​​are extracted from the zone heating control parameter set and smoothed using a sliding window mean filter.

[0159] In one embodiment, the window size is 3, and the filtered temperature fluctuation is reduced from ±5 degrees Celsius to ±2 degrees Celsius, generating a smooth heating parameter set and improving control stability. The range of the smooth heating parameter set is calculated, and if it exceeds a preset threshold, an inverse distance weighting method is used for optimization.

[0160] For example, the range is 20 degrees Celsius, exceeding the threshold of 15 degrees Celsius. In one possible implementation, the temperature difference is reduced to 10 degrees Celsius after weighted averaging using the reciprocal of the center distance between partitions, generating an optimized heating parameter set. Based on the optimized heating parameter set, a linear proportional converter is used to generate PWM duty cycle instructions.

[0161] For example, a temperature of 120 degrees Celsius corresponds to a duty cycle of 60%, generating a partitioned heating control instruction set containing 30 instructions, which is directly used for equipment control to ensure heating uniformity.

[0162] Step S109: Obtain actual temperature distribution data during the curing process. If the temperature deviation exceeds the threshold, generate a new power control command.

[0163] Temperature distribution data during the curing process is acquired from a temperature sensor to obtain a real-time temperature distribution. The real-time temperature distribution is compared with a preset standard temperature distribution to calculate the temperature deviation, which is obtained by subtracting the preset standard temperature distribution from the real-time temperature distribution. If the temperature deviation exceeds a preset threshold, the temperature deviation is used in conjunction with a PID controller to calculate an adjustment amount, generating a power control command. The control system executes the power control command to adjust the power output of the curing equipment, resulting in an updated temperature distribution. The updated temperature distribution data is acquired from the temperature sensor and compared with the preset standard temperature distribution to determine if the temperature deviation is within the preset threshold. If the temperature deviation is still outside the preset threshold, the gradient descent method is used to adjust the power control parameters, generating a new power control command.

[0164] Specifically, obtaining temperature distribution data during the curing process from temperature sensors is a key step in the inkjet printing curing process.

[0165] For example, a temperature sensor array is arranged above the curing area to collect the temperature value of each pixel in real time, forming a two-dimensional temperature distribution map.

[0166] It should be noted that the sensor typically uses infrared temperature measurement technology, with an accuracy of ±1 degree Celsius.

[0167] For example, in a solidified area measuring 1000 × 1000 mm, a sensor collects temperature data at 10,000 points at 10 mm intervals, generating a temperature distribution map and recording a maximum temperature of 150 degrees Celsius and a minimum temperature of 100 degrees Celsius. This high-resolution data provides the basis for subsequent deviation analysis. The temperature deviation is calculated by comparing the real-time temperature distribution with a preset standard temperature distribution.

[0168] Specifically, the preset standard temperature distribution is a temperature template designed based on the ideal curing effect. For example, the target temperature is 120 degrees Celsius, and fluctuations of ±5 degrees Celsius are allowed.

[0169] In one embodiment, the real-time temperature distribution shows that the temperature of a certain area is 130 degrees Celsius, with a deviation of 130-120=10 degrees Celsius.

[0170] Preferably, the deviation calculation covers the entire curing area, generating a deviation distribution map that visually reflects areas of uneven temperature. This deviation analysis helps to accurately locate areas requiring adjustment. If the temperature deviation exceeds a preset threshold, a PID controller is used to calculate the adjustment amount and generate a power control command.

[0171] Understandably, a PID controller adjusts the output power through three parts: proportional, integral, and derivative, to quickly respond to deviations.

[0172] For example, the preset threshold is ±5 degrees Celsius, and a certain area deviates by 10 degrees Celsius, exceeding the threshold.

[0173] In one possible implementation, the PID controller adjusts the heater power based on the magnitude of the deviation, increasing the output power control command from 50% to 60% to reduce the temperature deviation. This method ensures precise power adjustment and rapid response. By executing the power control command through the control system, the power output of the curing equipment is adjusted to obtain the updated temperature distribution.

[0174] Specifically, the control system translates commands into actual power signals for the heater, for example, translating a 60% power command into a 300-watt output.

[0175] In one embodiment, the temperature in a certain area was adjusted from 130 degrees Celsius to 122 degrees Celsius, reducing the deviation to 2 degrees Celsius.

[0176] It should be noted that the updated temperature distribution is collected again by the sensor to ensure real-time data accuracy. This closed-loop control method improves the stability of temperature regulation. The updated temperature distribution data is obtained from the temperature sensor and compared with the preset standard temperature distribution to determine whether the temperature deviation is within the preset threshold.

[0177] For example, if the deviation in a certain area is 2 degrees Celsius after the update, it falls within the ±5 degree Celsius threshold range, indicating that the adjustment is effective.

[0178] In one possible implementation, if all regional deviations meet the threshold, the system maintains the current power output; otherwise, it proceeds to the next optimization step. This deviation judgment mechanism ensures the temperature consistency of the curing process. If the temperature deviation is still outside the preset threshold, the power control parameters are adjusted using the gradient descent method to generate a new power control command.

[0179] Preferably, the gradient descent method gradually reduces the deviation by iteratively optimizing the controller parameters.

[0180] For example, the deviation in a certain area is still 8 degrees Celsius.

[0181] In one embodiment, the gradient descent method adjusts the proportional gain in the PID parameters from 1.0 to 0.8, generating a new power command that fine-tunes the power from 60% to 58%. This method avoids over-adjustment through small-step optimization, ensuring that the temperature distribution gradually approaches the target.

[0182] Step S1010: Measure the final conductivity parameters and generate a verification dataset corresponding to the region identifier and the measured conductivity.

[0183] Obtain the original dataset, which includes multiple records, each containing a region identifier and measured conductivity data. For missing conductivity values ​​in the original dataset, calculate the mean conductivity of each group by region identifier, and impute the missing values ​​using the within-group mean to obtain a complete conductivity dataset. Perform Z-score standardization on the conductivity column of the complete conductivity dataset to obtain a standardized conductivity dataset. Analyze the standardized conductivity dataset using the elbow method to determine the optimal number of clusters. Use the K-means algorithm to cluster the standardized conductivity dataset to obtain a conductivity dataset with category labels. Based on the conductivity dataset with category labels, calculate the mean conductivity and standard deviation of each cluster by category label to obtain the statistical characteristics of each cluster. If the standard deviation of a cluster is greater than a preset threshold, the cluster is marked as an anomaly category, obtaining an anomaly category set. Extract the region identifiers corresponding to the anomaly category set from the complete conductivity dataset to obtain an anomaly region set. Based on the region identifiers in the abnormal region set, the measured conductivity records are indexed and matched from the original dataset to obtain the abnormal conductivity verification set.

[0184] Specifically, acquiring area identification and measured conductivity data through sensors and storing them as raw datasets is a crucial step in monitoring material performance during the inkjet printing curing process.

[0185] For example, a sensor array is arranged within a solidified area to collect the conductivity value and corresponding identifier of each area.

[0186] For example, the curing area is 500×500 mm in size, and the sensor collects data from 10,000 points at 5 mm intervals. Each point records the area identifier such as A1, B2, etc., and the conductivity value such as 10 mS / cm.

[0187] It should be noted that electrical conductivity reflects the material's electrical conductivity and directly affects the curing quality. The sensor accuracy is typically ±0.1 mS / cm. The collected data is stored in CSV format, containing a region identifier column and an electrical conductivity column, forming the original dataset. For missing electrical conductivity values ​​in the original data, the mean of each group is calculated based on the region identifier for imputation.

[0188] In one possible implementation, a region A1 contains 100 points, of which 10 points have missing conductivity values. The average conductivity of the 90 valid points in region A1 is calculated to be 12 mS / cm, and this average value is used to fill in the missing points.

[0189] Preferably, the data distribution is checked before interpolation to ensure the representativeness of the mean. After interpolation, a complete conductivity dataset is generated, retaining the region identifiers and conductivity column to ensure data integrity. The conductivity column of the complete conductivity dataset is then Z-score standardized to eliminate the influence of dimensions.

[0190] Understandably, the Z-score converts conductivity values ​​into standard values ​​with a mean of 0 and a standard deviation of 1, which facilitates subsequent clustering.

[0191] For example, if the conductivity at a certain point is 15 mS / cm, the dataset mean is 12 mS / cm, the standard deviation is 2 mS / cm, and the standardized value is 1.5, then the standardized dataset is more suitable for machine learning algorithms. The elbow method is used to determine the optimal number of clusters k, and the K-means algorithm is used for clustering.

[0192] In one embodiment, the elbow method finds the inflection point k=3 by plotting the relationship between the number of clusters and the sum of squared errors within each cluster. The K-means algorithm divides the standardized data into 3 clusters and outputs a dataset with class labels, with each row recording the region identifier, conductivity, and cluster label such as C1, C2.

[0193] It should be noted that the K-means algorithm iteratively optimizes cluster centers to classify regions with similar electrical conductivity. Based on the clustering results, the mean and standard deviation of conductivity for each cluster are calculated according to the category label.

[0194] Specifically, cluster C1 contains 5000 points with a mean of 11 mS / cm and a standard deviation of 0.3; cluster C2 has a mean of 13 mS / cm and a standard deviation of 0.6. The preset standard deviation threshold is 0.5. If the standard deviation of cluster C2 exceeds the threshold, it is marked as an anomaly.

[0195] Preferably, an excessively high standard deviation reflects large fluctuations in conductivity, which may indicate uneven curing. Region identifiers corresponding to the anomaly categories are extracted from the complete conductivity dataset to form a set of anomaly regions.

[0196] For example, cluster C2 contains 1000 identifiers such as regions B1 and B2, forming an anomalous region set. Based on this set, measured conductivity records are indexed and matched from the original dataset to generate an anomalous conductivity verification set.

[0197] In one embodiment, the validation set includes conductivity records for the B1 region, such as 13.2 mS / cm and 14.1 mS / cm, for subsequent analysis of the cause of the anomaly.

[0198] Understandably, validation sets provide accurate data support for process optimization.

[0199] It will be apparent to those skilled in the art that this application is not limited to the details of the exemplary embodiments described above, and that this application can be implemented in other specific forms without departing from the spirit or essential characteristics of this application. Therefore, the embodiments should be considered illustrative and non-limiting in all respects, and the scope of this application is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within this application. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A method for preparing a PCB board with high-carbon ink, characterized in that, The method includes: Obtain a dataset of conductivity distribution that includes region identifiers, conductivity values, and boundary coordinates; Based on the conductivity distribution dataset, a mapping relationship between printhead number, ink type, and target area coordinates is generated; A micron-level precision control algorithm is used to generate a set of motion control commands corresponding to timestamps, nozzle positions, and deposition coordinates. Acquire an image dataset of ink deposition during the deposition process and calculate the deviation between the actual deposition amount and the target deposition amount; If the deviation value exceeds the preset threshold, the injection frequency configuration file containing proportional-integral-derivative parameters is updated. The deposition process is controlled according to the updated jet frequency profile, and the transition width of the boundary region is detected. If the transition width exceeds a micrometer-level threshold, an ink deposition dataset containing optimized boundary coordinates is generated. Based on the ink deposition dataset, generate zone heating control parameters corresponding to the region identifier and curing temperature; Acquire actual temperature distribution data during the curing process; if the temperature deviation exceeds the threshold, generate a new power control command. Measure the final conductivity parameters and generate a verification dataset corresponding to the region identifiers and measured conductivity. The step of generating a motion control command set corresponding to timestamps, nozzle positions, and deposition coordinates using a micrometer-level precision control algorithm includes: Timestamps and nozzle motion trajectory data are obtained from time series data. The position coordinates in the nozzle motion trajectory data are separated by a threshold segmentation method to obtain a set of nozzle position coordinates. The target deposition coordinates are obtained using a visual sensor to obtain an initial coordinate dataset; Based on the timestamps and nozzle position coordinates in the initial coordinate dataset, Kalman filtering is used to calculate the motion acceleration of adjacent coordinate points. If the motion acceleration exceeds a preset acceleration threshold, it is determined that there are discontinuities in the nozzle motion trajectory. New coordinate points are generated between the discontinuities using cubic spline interpolation to obtain a smooth motion trajectory, wherein the rate of change of the velocity of the smooth motion trajectory is lower than a preset smoothness threshold. The interpolated nozzle position coordinates are extracted from the smooth motion trajectory, and a mapping relationship is established with the target deposition coordinates through time window matching to obtain a mapping relationship table between timestamps and nozzle position coordinates and target deposition coordinates; Based on the target deposition coordinates corresponding to each timestamp in the mapping table, a preset instruction template is matched to generate a preliminary instruction set; The actual motion path during the execution of the preliminary instruction set is measured using a laser interferometer. If the deviation between the actual motion path and the target deposition coordinates exceeds a preset tolerance threshold, the parameters of the preliminary instruction set are adjusted to obtain the final motion control instruction set.

2. The method according to claim 1, characterized in that, The process of obtaining a conductivity distribution dataset containing region identifiers, conductivity values, and boundary coordinates includes: An initial dataset is obtained from the sensor network, the initial dataset containing region identifiers, conductivity values, and boundary coordinates; Noise detection is performed on the initial dataset to obtain a denoised dataset; Based on the denoised dataset, the K-means clustering algorithm was used to divide the regions, and the region division results were obtained. Extract the conductivity performance value of each region from the denoised dataset to generate a performance distribution feature; Calculate the variance of the performance distribution characteristics. If the variance is greater than a preset threshold, use a linear interpolation algorithm to smooth the conductivity values ​​to obtain a smooth performance distribution. Based on the smoothed performance distribution and the region division results, a conductivity performance distribution dataset containing region identifiers, performance values, and boundary coordinates is generated. The final conductivity distribution dataset is determined by comparing the boundary coordinates of the conductivity distribution dataset with those of the initial dataset using a boundary coordinate consistency verification algorithm.

3. The method according to claim 1, characterized in that, The step of generating the mapping relationship between printhead number, ink type, and target area coordinates based on the conductivity distribution dataset includes: Based on the conductivity performance distribution dataset, an initial data table was generated using a data extraction algorithm. Based on the region identifiers and boundary coordinates in the initial data table, the K-means clustering algorithm is used to divide the regions, and the region division results are obtained. If the boundary coordinates in the region division result are inconsistent with the boundary coordinates in the initial data table, the region division result is adjusted using a coordinate consistency verification algorithm to obtain a consistent region division. Extract the region identifier and target region coordinates of each region from the consistent region division, and generate a region coordinate table containing the region identifier and the target region coordinates; Based on the region identifier in the region coordinate table, the printhead number value and ink type name are matched using a preset printhead allocation rule to obtain the printhead allocation table; Extract the printhead number, ink type name, and target area coordinates from the printhead allocation table to obtain a mapping table; Based on the target region coordinates in the mapping table and the boundary coordinates in the consistent region division, the mapping table is verified using a coordinate consistency verification algorithm to obtain the final mapping table.

4. The method according to claim 1, characterized in that, The process of acquiring an image dataset of ink deposition during the deposition process and calculating the deviation between the actual deposition amount and the target deposition amount includes: Images of the ink deposition process are acquired and cropped at preset time intervals to generate an original image set; The original image set is processed using an image segmentation algorithm to extract the contours of the ink deposition areas to generate a segmented image set; The pixel area of ​​each contour region in the segmented image set is calculated using a contour area calculation function to generate an actual deposition amount dataset. Read the target deposition value from the configuration file and generate the target deposition dataset; If the absolute difference between the actual deposition amount and the target deposition amount exceeds a preset threshold, the mean filtering function is used to filter the difference sequence to generate a deposition deviation dataset. The set of adjustment parameters is generated by matching the sedimentation deviation dataset with a pre-established compensation parameter lookup table; the compensation parameter lookup table contains the correspondence between deviation ranges and compensation coefficients. Based on the compensation coefficients in the set of adjustment parameters, the boundary of the deposition region in the segmented image set is re-detected using a contour detection function to generate an optimized deposition amount dataset.

5. The method according to claim 1, characterized in that, If the deviation value exceeds a preset threshold, the injection frequency configuration file containing proportional-integral-derivative parameters is updated, including: Obtain the deposition deviation value sequence. If the deposition deviation value sequence exceeds a preset threshold, extract the time series features of the deposition deviation value sequence using the ARIMA model to obtain the deviation change trend. Based on the deviation change trend, the deposition deviation value sequence is smoothed using Kalman filtering to obtain a smoothed deviation sequence. If the fluctuation amplitude of the smoothed deviation sequence exceeds a preset amplitude threshold, then a normal distribution is used to fit the smoothed deviation sequence to determine the deviation distribution parameters; Based on the deviation distribution parameters, the proportional parameters, integral parameters, and derivative parameters in the configuration file are updated using the gradient descent method to obtain the updated control parameter set. The injection frequency value is calculated based on the updated control parameter set using a PID controller, thus obtaining the adjusted frequency parameter set. If the difference between the adjusted frequency parameter set and the historical frequency parameter set exceeds a preset difference threshold, the root locus method is used to perform stability analysis on the adjusted frequency parameter set to obtain the verified frequency parameter set. The injection frequency profile is updated based on the verified set of frequency parameters to obtain the final frequency profile.

6. The method according to claim 1, characterized in that, The step of controlling the deposition process according to the updated jet frequency profile and detecting the transition width of the boundary region includes: Image data of the boundary region during the deposition process are acquired, and the boundary region is extracted using an image segmentation algorithm to obtain a segmented boundary image; If the pixel distribution entropy value of the segmentation boundary image exceeds a preset entropy threshold, then median filtering is used to denoise the segmentation boundary image to obtain a denoised boundary image. Based on the denoised boundary image, the Canny edge detection algorithm is used to extract the boundary transition region to obtain the boundary transition contour. If the continuity parameter of the boundary transition contour is lower than the preset continuity threshold, the contour break is repaired by morphological dilation operation to obtain the repaired boundary contour. Calculate the pixel width of the boundary transition region based on the repaired boundary contour to obtain the boundary transition width sequence; If the variance of the boundary transition width sequence exceeds a preset variance threshold, the K-means clustering algorithm is used to classify the width sequence to obtain a classified width sequence. Based on the classification width sequence, the local mean is calculated using the sliding window method to obtain a smooth width sequence; If the fluctuation amplitude of the smooth width sequence exceeds a preset amplitude threshold, the smooth width sequence is optimized by linear interpolation to obtain an optimized width sequence. Generate a boundary transition width configuration file based on the optimized width sequence to obtain the width configuration file; The control parameters of the deposition process are adjusted according to the width configuration file to obtain the set of adjustment control parameters.

7. The method according to claim 1, characterized in that, If the transition width exceeds a micrometer-level threshold, an ink deposition dataset containing optimized boundary coordinates is generated, including: Acquire a boundary region image, the boundary region image including pixel distribution characteristics; If the pixel distribution characteristics exceed a preset distribution threshold, the boundary region image is segmented using the Otsu algorithm to obtain a segmented boundary image. Based on the segmented boundary image, the Canny edge detection algorithm is used to extract the boundary contour to obtain the initial boundary contour. If the continuity of the initial boundary contour is lower than a preset continuity threshold, the initial boundary contour is repaired by closing operation to obtain a repaired boundary contour. Based on the repaired boundary contour, the coordinate point set of the boundary transition region is calculated using a boundary tracing algorithm to obtain the boundary coordinate dataset. If the distribution density of coordinate points in the boundary coordinate dataset exceeds a preset density threshold, the boundary coordinate dataset is classified using the K-means clustering algorithm to obtain a classified coordinate dataset. Based on the classification coordinate dataset, cubic spline interpolation is used to optimize the coordinate points in the classification coordinate dataset to obtain optimized boundary coordinates; The optimized boundary coordinates and ink deposition parameters are integrated to obtain the ink deposition dataset.

8. The method according to claim 1, characterized in that, The step of generating zone heating control parameters corresponding to region identifiers and curing temperatures based on the ink deposition dataset includes: Obtain an ink deposition dataset, which includes boundary coordinates and ink deposition parameters; Gaussian filtering was used to reduce noise in the ink deposition parameters to obtain a preprocessed deposition parameter set. Based on the gradient distribution of the preprocessed deposition parameter set, region identifiers are extracted through threshold segmentation; The region identifiers are partitioned using a quadtree space partitioning algorithm, and the minimum partition area is set as a preset value to obtain a partition identifier set. Calculate the coefficient of variation of the boundary points of the partition identifier set. If the coefficient of variation is greater than a preset threshold, then perform region merging. Based on Delaunay triangulation to detect adjacent partitions, if the standard deviation of the area of ​​adjacent partitions is greater than a preset threshold, they are merged into the same partition to obtain an adjusted partition identifier set. For the adjusted partition identifier set, the mean pixel coordinates of each partition are calculated to obtain the geometric center coordinate set; Based on the geometric center coordinate set and the preset curing temperature lookup table, a two-dimensional bilinear interpolation algorithm is used to generate the partitioned heating control parameters, resulting in the partitioned heating control parameter set. The heating parameter values ​​of each partition are extracted from the partition heating control parameter set, and a smooth heating parameter set is obtained by using a sliding window mean filter. Calculate the range of the smoothed heating parameter set. If the range is greater than a preset threshold, use the inverse distance weighting method to perform a weighted average of the parameters of adjacent partitions. The weighting coefficient is the reciprocal of the partition center distance to obtain the optimized heating parameter set. Based on the optimized heating parameter set, a PWM duty cycle instruction is generated using linear proportional conversion to obtain a zoned heating control instruction set.

9. The method according to claim 1, characterized in that, The process of acquiring actual temperature distribution data during the curing process, and generating a new power control command if the temperature deviation exceeds a threshold, includes: Temperature distribution data during the curing process is obtained from temperature sensors to obtain real-time temperature distribution; The temperature deviation is calculated by comparing the real-time temperature distribution with the preset standard temperature distribution, and the temperature deviation is obtained by subtracting the preset standard temperature distribution from the real-time temperature distribution. If the temperature deviation exceeds the preset threshold, the temperature deviation and the PID controller are used to calculate the adjustment amount and generate a power control command. By executing the power control command through the control system, the power output of the curing equipment is adjusted to obtain the updated temperature distribution; The updated temperature distribution data is obtained from the temperature sensor and compared with the preset standard temperature distribution to determine whether the temperature deviation is within the preset threshold. If the temperature deviation is still outside the preset threshold, the power control parameters are adjusted using the gradient descent method to generate a new power control command.

Images