Screen mesh number recommendation method based on screen printing and related device

Abstract

The invention provides a silk screen mesh number recommendation method based on silk screen printing and a related device. The method comprises the following steps: acquiring a to-be-analyzed ink sample as a target ink sample; collecting ink particle data in the target ink sample; collecting screen attributes in the candidate screen samples; the silk screen attributes at least comprise silk screen materials and silk screen mesh numbers of candidate silk screen samples; in combination with the ink particle data and the screen attributes, predicting a target screen mesh number of which the matching degree with the target ink sample reaches a preset condition through a mesh number matching model; and generating silk screen recommendation information based on the target silk screen mesh number, and sending the silk screen recommendation information to the user to assist the user in completing a silk screen selection process. According to the embodiment of the invention, accurate matching of the ink particles and the silk screen mesh number is realized, the printing quality and the production efficiency are improved, and meanwhile, the production debugging cost and resource waste are reduced.

Description

Technical Field

[0001] The present application relates to the technical field of curtain wall equipment, and in particular to a method for recommending screen mesh count based on screen printing and related devices. Background Art

[0002] Photovoltaic glass screen printing is a key process in the production of photovoltaic modules. The selection of screen mesh directly affects the ink permeability, pattern fineness and product quality. The application of photovoltaic glass in high-end buildings places higher requirements on aesthetics and functionality, and different ink particles and process conditions further increase the complexity of selection.

[0003] At present, in the actual production of photovoltaic glass screen printing, the selection of screen mesh count mainly depends on experience. This traditional method is difficult to meet the requirements of modern photovoltaic industry for efficient, accurate and stable production. In related technologies, the selection of appropriate screen mesh count usually relies on manual experience or multiple tests. The repeated tests in the debugging process waste ink, screen and other materials, increasing debugging time and production costs. In addition, human factors in related technologies may also cause the screen mesh count to not match the ink particle size, which will lead to unstable printing quality, affect production efficiency, and may delay production plans. Even more, mismatched screens may further accelerate the loss of equipment and further increase equipment maintenance costs.

[0004] Therefore, it is urgent to design a new technical solution to solve at least one of the above technical problems and improve the efficiency and precision of photovoltaic glass manufacturing. Summary of the invention

[0005] In order to solve or improve the above-mentioned problems, the embodiments of the present application provide a screen mesh recommendation method based on screen printing and related devices to achieve accurate matching of ink particles and screen mesh, improve printing quality and production efficiency, and reduce production debugging costs and resource waste.

[0006] In one embodiment of the present application, a method for recommending screen mesh count based on screen printing is provided, comprising: obtaining an ink sample to be analyzed as a target ink sample; Collecting ink particle data in the target ink sample; the ink particle data at least includes: particle distribution data, toner ratio, glass powder ratio, toner type; Collecting screen attributes from the candidate screen sample; the screen attributes at least include: screen material and screen mesh number of the candidate screen sample; Combining the ink particle data with the screen properties, predicting a target screen mesh whose matching degree with the target ink sample reaches a preset condition through a mesh matching model; Silk screen recommendation information is generated based on the target silk screen mesh count, and the silk screen recommendation information is sent to the user to assist the user in completing the silk screen selection process.

[0007] In another embodiment of the present application, a screen printing machine is provided, comprising at least the following units: An acquisition module, used for acquiring an ink sample to be analyzed as a target ink sample; A first acquisition module is used to acquire ink particle data in the target ink sample; the ink particle data at least includes: particle distribution data, toner ratio, glass powder ratio, and toner type; The second collection module is used to collect the screen properties of the candidate screen sample; the screen properties at least include: the screen material and screen mesh number of the candidate screen sample; A prediction module, for combining the ink particle data with the screen properties, and predicting a target screen mesh whose matching degree with the target ink sample meets a preset condition through a mesh matching model; The recommendation module is used to generate screen recommendation information based on the target screen mesh count and send the screen recommendation information to the user to assist the user in completing the screen selection process.

[0008] In one embodiment of the present application, an electronic device includes: at least one processor, memory, and input-output unit; The memory is used to store a computer program, and the processor is used to call the computer program stored in the memory to execute the screen mesh number recommendation method based on screen printing in the first embodiment.

[0009] In one embodiment of the present application, a computer-readable storage medium is provided, which includes instructions. When the instructions are executed on a computer, the computer executes the screen mesh recommendation method based on screen printing in the first embodiment described above.

[0010] In the technical solution provided in the embodiment of the present application, the ink sample to be analyzed is obtained as the target ink sample. Then, the ink particle data in the target ink sample is collected; the ink particle data at least includes: particle distribution data, toner ratio, glass powder ratio, toner type; the screen properties in the candidate screen sample are collected; the screen properties at least include: screen material and screen mesh number of the candidate screen sample. Then, the target screen mesh number whose matching degree with the target ink sample reaches the preset condition is predicted by the mesh number matching model in combination with the ink particle data and the screen properties. Finally, screen recommendation information is generated based on the target screen mesh number, and the screen recommendation information is sent to the user to assist the user in completing the screen selection process. In the embodiment of the present application, the screen mesh number is intelligently matched based on the ink sample and recommended to the user, so as to achieve the precise matching of ink particles and screen mesh number, solve the printing quality problem of improper matching between ink particle size and screen mesh number in the photovoltaic glass screen printing process in the related technology, and can significantly improve the printing quality. At the same time, it helps to reduce the production debugging cost and resource waste caused by repeated debugging, and greatly improves the production efficiency of the entire photovoltaic glass screen printing. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0012] Figure 1 A schematic diagram of a process flow of a method for recommending screen mesh count based on screen printing provided in one embodiment of the present application; Figure 2 A schematic diagram of the principle of a method for recommending screen mesh count based on screen printing provided in one embodiment of the present application; Figure 3 A schematic diagram of the structure of an electronic device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0013] The present application provides the following embodiments to solve or improve at least some of the above problems. In order to enable people in the technical field to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0014] In some features described in the specification, claims and the above-mentioned drawings of this application, descriptions such as "first" and "second" are used to distinguish different components, parts, modules, devices, etc., and do not represent the order of precedence, nor do they limit "first" and "second" to be different types. In addition, the embodiments described below are only some embodiments of the present application, not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of this application.

[0015] In order to solve the above technical problems, a screen printing machine provided in an embodiment of the present application includes the following units: An acquisition module, used for acquiring an ink sample to be analyzed as a target ink sample; A first acquisition module is used to acquire ink particle data in the target ink sample; the ink particle data at least includes: particle distribution data, toner ratio, glass powder ratio, and toner type; The second collection module is used to collect the screen properties of the candidate screen sample; the screen properties at least include: the screen material and screen mesh number of the candidate screen sample; A prediction module, for combining the ink particle data with the screen properties, and predicting a target screen mesh whose matching degree with the target ink sample meets a preset condition through a mesh matching model; The recommendation module is used to generate screen recommendation information based on the target screen mesh count and send the screen recommendation information to the user to assist the user in completing the screen selection process.

[0016] In the embodiment of the present application, the screen mesh number is intelligently matched based on the ink sample and recommended to the user, achieving accurate matching of ink particles and screen mesh number, solving the printing quality problem of improper matching of ink particle size and screen mesh number in the photovoltaic glass screen printing process in the related technology, and significantly improving the printing quality. At the same time, it helps to reduce the production debugging costs and resource waste caused by repeated debugging, and greatly improves the production efficiency of the entire photovoltaic glass screen printing.

[0017] It should be added here that the control logic of the control device will be introduced in detail in the method embodiment below.

[0018] An embodiment of the present application provides a method for recommending screen mesh count based on screen printing. The execution subject of the method described in this embodiment can be the screen mesh count recommendation device in the above embodiment, which is implemented on the basis of the structure of the above screen printing machine embodiment. Figure 1 As shown, the method may include the following steps: 101. Obtain an ink sample to be analyzed as a target ink sample; 102. Collect ink particle data in the target ink sample; the ink particle data at least includes: particle distribution data, toner ratio, glass powder ratio, and toner type; 103. Collecting screen attributes of the candidate screen sample; the screen attributes at least include: screen material and screen mesh number of the candidate screen sample; 104. Combining the ink particle data with the screen properties, predicting a target screen mesh whose matching degree with the target ink sample meets a preset condition through a mesh matching model; 105. Generate screen recommendation information based on the target screen mesh count, and send the screen recommendation information to the user to assist the user in completing the screen selection process.

[0019] The present application provides a method for recommending screen mesh based on screen printing, which obtains the ink sample to be analyzed as the target ink sample, which is the basis of the entire precise matching work. Only by clarifying the specific ink sample to be analyzed can the various particle data related to it be accurately collected subsequently, thereby taking the key first step to solve the problem of matching ink particles and screen mesh in photovoltaic glass screen printing, which helps to improve the printing quality in a targeted manner. The ink particle data in the target ink sample is collected, covering various information such as particle distribution data, toner ratio, glass powder ratio, toner type, etc. Comprehensively mastering these data makes the understanding of the characteristics of the ink more detailed and accurate, and can more accurately analyze the adaptation relationship between it and the screen properties, changing the previous situation of relying on manual experience and lacking accurate data support, and providing a detailed basis for achieving accurate matching of ink particles and screen mesh, thereby avoiding printing quality problems caused by improper matching. Collect screen properties in candidate screen samples, including key contents such as screen material and screen mesh. This helps to build a complete screen information library, allowing the system to fully consider the characteristics of different screens, conduct a comprehensive analysis with the ink particle data, and further improve the various conditions required for matching. It is no longer just a blind attempt like the traditional method, which improves the scientific nature of screen selection and helps improve production efficiency. Combining the ink particle data and screen properties collected previously, the mesh matching model is used to predict the target screen mesh that matches the target ink sample to meet the preset conditions. With the help of the model built with intelligent algorithms, it gets rid of the inefficient and unstable matching methods of traditional manual experience or repeated trials, and can quickly and accurately find the appropriate screen mesh, achieving accurate matching of ink particles and screen mesh, which can significantly improve printing quality, while reducing production debugging costs and resource waste caused by repeated debugging, greatly improving the production efficiency of the entire photovoltaic glass screen printing. Screen recommendation information is generated based on the target screen mesh and sent to the user to assist the user in completing the screen selection process. This provides users with a clear and accurate reference, allowing them to efficiently select the appropriate screen, further ensuring that operations can be performed in accordance with precise matching requirements in actual production, guaranteeing printing quality, and improving overall production efficiency. It avoids the blindness and uncertainty of manual screen selection in the past, and effectively facilitates the smooth development of photovoltaic glass screen printing.

[0020] In general, the above steps solve the printing quality problem of improper matching between ink particle size and screen mesh number in the photovoltaic glass screen printing process by constructing a screen mesh number recommendation system based on intelligent algorithms, thereby improving printing quality and production efficiency, while reducing production debugging costs and reducing resource waste.

[0021] As an optional embodiment, in 102, collecting ink particle data in the target ink sample includes: Randomly select some ink samples from the ink samples to be analyzed as the target ink samples; perform component analysis on the target ink samples to obtain the color powder ratio, glass powder ratio, and color powder type of the target ink samples; the analysis method at least includes spectrophotometry; based on the color powder ratio, glass powder ratio, and color powder type of the target ink samples, set the particle measurement parameters of the target ink samples; use the particle measurement parameters to obtain the scattered light intensity data of the ink particles in the target ink samples at different angles; the collection method at least includes multi-angle dynamic light scattering method; perform particle size analysis on the target ink samples based on the scattered light intensity data to obtain the particle distribution data.

[0022] Specifically, firstly, a part of the ink sample to be analyzed is randomly selected as the target ink sample, in order to represent the characteristics of the whole ink. Then, the toner ratio, glass powder ratio and toner type are determined by component analysis (spectrophotometry is at least one method). Spectrophotometry uses the different absorption characteristics of the material to different wavelengths of light to determine the material composition, so as to accurately obtain these component information of the ink. Then, the particle measurement parameters are set according to these component information. This is because factors such as different component ratios may affect the accuracy of particle measurement. Reasonable parameter settings can better adapt to the characteristics of the target ink sample. Then, the multi-angle dynamic light scattering method is used to obtain the scattered light intensity data of the ink particles at different angles. The principle of the multi-angle dynamic light scattering method is based on the light scattering theory. When light is irradiated on the particles, the particles will scatter the light. The scattered light intensity is related to the size of the particles and other properties. The scattered light intensity at different angles can provide more comprehensive particle information. Finally, the particle size analysis is performed based on these scattered light intensity data to obtain the particle distribution data. This is because there is a certain mathematical relationship between the scattered light intensity and the particle size. The particle size distribution of the particles can be inferred by analyzing the light intensity data.

[0023] In this way, the ink particle data in the target ink sample can be obtained very accurately. Through detailed component analysis, the composition details of the ink can be fully understood, providing a basis for the subsequent particle measurement parameter setting, thereby ensuring the scientificity and accuracy of particle data collection. The multi-angle dynamic light scattering method is used to obtain scattered light intensity data and use it for particle size analysis to obtain particle distribution data, which is more comprehensive and accurate than the traditional simple measurement method, thereby providing a high-precision data basis for the subsequent combination of ink particle data and screen properties, which is conducive to accurately matching the screen mesh count, improving the quality of photovoltaic glass screen printing, and avoiding the problem of improper screen mesh matching due to inaccurate particle data. At the same time, it also helps to improve the efficiency of the entire matching process and reduce repeated debugging due to data errors.

[0024] Further optionally, in the embodiment of the present application, when analyzing ink particles, a multi-angle dynamic light scattering method (DLS) is used. Specifically, the particle size range that can be measured by the multi-angle dynamic light scattering method is usually from nanometers to micrometers. In the analysis of ink particles, this range covers the sizes of most ink particles, whether fine toner particles or components containing slightly larger particles such as glass powder, which can be effectively measured. For example, for some nano-level high-performance toners, the particle size may be around tens of nanometers, and this method can accurately measure the size of these tiny particles, providing basic data for a comprehensive understanding of the distribution of ink particles. It obtains particle information by detecting the scattering signal of particles to light, and does not require complex pretreatment of samples, such as dyeing, labeling or chemical reactions. In the analysis of ink samples, changes that may be caused to ink particles due to pretreatment are avoided, and the size and distribution of ink particles in the original state can be truly reflected. This is crucial to maintaining the integrity and authenticity of the ink sample, ensuring that the measurement results can be accurately used for subsequent matching with the screen mesh count.

[0025] The multi-angle dynamic light scattering method can complete the measurement in a relatively short time. It can quickly obtain a large amount of particle scattered light intensity data, and usually obtain preliminary results of particle size distribution within a few minutes. For situations where rapid analysis of ink particle data is required, such as in the large-scale production of photovoltaic glass screen printing, the efficient measurement speed helps to adjust the printing parameters in time, reduce waiting time in the production process, and thus improve production efficiency.

[0026] The scattered light intensity at different angles contains information about particles in different dimensions, such as particle shape, size, and aggregation state. In ink particle analysis, multi-angle data can more comprehensively describe the characteristics of particles. For example, by changing the scattered light intensity at different angles, it is possible to determine whether ink particles exist alone or aggregated, and thus more accurately evaluate the particle distribution, providing more detailed particle information for accurate matching of screen mesh counts.

[0027] The multi-angle dynamic light scattering method can achieve high measurement accuracy under appropriate experimental conditions. Its measurement error can usually be controlled within a small range, and under the same measurement conditions, the repeatability of multiple measurements is good. This is very important for the accurate analysis of ink particle data, ensuring the reliability and comparability of the particle distribution data measured each time, and helping to establish a stable matching model between ink particles and screen mesh count, reducing the situation of improper screen mesh count matching caused by data errors.

[0028] For example, in the process of collecting and processing ink particle data, in order to ensure the high quality of the data and the accuracy of subsequent analysis, high-precision particle size analysis instruments, such as laser particle size analyzers or dynamic light scattering instruments, are used to carry out related work.

[0029] First of all, these advanced instruments play a key role. They have powerful detection capabilities and can perform very accurate detection of ink particles at the nanometer to micrometer level. For example, when analyzing inks containing a variety of different sized components, whether it is extremely fine nano-level toner particles or slightly larger micron-level particles of other components, the corresponding particle size information can be accurately captured. Subsequently, the instrument will generate a particle size distribution curve for the ink particles based on the test results. This curve intuitively shows the distribution of ink particles in different particle size ranges. Moreover, key parameters such as average particle size, median particle size and particle distribution range can be further calculated. The derivation of these parameters is of great significance for subsequent in-depth analysis and is an important input basis for the entire analysis process.

[0030] Specifically looking at the original data, the particle size measurement data is divided into several intervals (such as 0-5µm, 5-10µm, etc.), and in each interval, the corresponding number of particles or the frequency of occurrence will be recorded. Based on such detailed interval division and corresponding records, the average particle diameter can be calculated. The calculation is achieved through the corresponding formula. The center value of the interval is combined with the corresponding frequency for calculation, so as to obtain the average diameter, which is a key indicator that can represent the average size of the overall particles. As for the particle standard deviation, a specific formula is used to measure the degree of discreteness of the particle distribution. For example, when the standard deviation value is small, it means that the particle size is relatively concentrated and the distribution is more uniform. On the contrary, a larger value means that the particle size is highly discrete and the size difference is more obvious.

[0031] For the particle distribution range, the minimum diameter and the maximum diameter are extracted from the entire particle size distribution data, and then the distribution interval is determined by calculating the difference between the two. This interval can clearly describe the width of the particle distribution and is used to show the span of the ink particle size.

[0032] Finally, the ink particle data after the above standardization is stored in the matching algorithm module, making it the input content for subsequent related calculations. When analyzing the matching problem between ink particles and screen mesh count in photovoltaic glass screen printing, these accurate and standardized data can be combined with information such as screen properties, and the corresponding matching algorithm can be used to accurately predict the appropriate screen mesh count, thereby avoiding improper matching problems caused by inaccurate and non-standard data, and effectively improving printing quality and production efficiency.

[0033] In the embodiment of the present application, when the particle standard deviation is small, it means that the particle size distribution of the ink particles is relatively concentrated. In the photovoltaic glass screen printing process, the quality of this ink is relatively high. Because the uniform particle size can make the ink have better fluidity and transferability during screen printing. For example, when printing fine photovoltaic glass patterns, ink with uniform particle size can pass through the screen holes more smoothly and adhere evenly to the glass surface. This helps to form high-quality printed patterns and reduce printing defects caused by differences in particle size, such as unclear patterns, uneven colors, and other problems. Moreover, a smaller standard deviation also indicates that the ink has better stability, and it is less likely to aggregate or stratify particles during storage and use, thereby ensuring the consistency of ink performance.

[0034] On the contrary, when the particle standard deviation is large, the particle size distribution of the ink particles is relatively dispersed. This type of ink may cause various problems during the printing process. Larger particles may clog the screen holes, affecting the normal transfer of the ink, resulting in breakpoints and ink shortages in the printed pattern. At the same time, due to the large difference in particle size, particles of different sizes may settle unevenly during the drying process, which in turn affects the flatness and glossiness of the printed product. Moreover, during the storage process of this type of ink, due to the complex interaction between particles, it is more likely to agglomerate, change the original properties of the ink, reduce the content of the effective ingredients of the ink, and ultimately affect the printing quality and the service life of the ink.

[0035] As an optional embodiment, in 104, setting the particle measurement parameters of the target ink sample based on the toner ratio, glass powder ratio, and toner type of the target ink sample includes: selecting a measurement mode that matches the characteristics of the toner ratio, glass powder ratio, and toner type of the target ink sample, and configuring the particle measurement parameters used by the measuring device in the selected measurement mode; wherein the measurement mode is a static light scattering mode or a dynamic light scattering mode. Furthermore, in the above steps, performing a particle size analysis on the target ink sample based on the scattered light intensity data to obtain the particle distribution data includes: performing a particle size analysis on the target ink sample based on the scattered light intensity data to obtain a particle size distribution curve of ink particles in the target ink sample; and obtaining the particle distribution data based on the particle size distribution curve.

[0036] When measuring ink particle data, the particle measurement parameters must first be set according to the toner ratio, glass powder ratio, and toner type of the target ink sample. This is because different combinations of components (toner ratio, glass powder ratio, and toner type) will cause ink particles to have different optical properties. For example, inks with a high toner ratio and inks with a low toner ratio may scatter and absorb light differently, and the presence or absence of glass powder and its ratio will also affect the interaction between light and particles. Therefore, it is necessary to select a measurement mode that matches these characteristics, namely, static light scattering mode or dynamic light scattering mode. The static light scattering mode mainly obtains information based on the scattering of light by particles in a static state. When light is irradiated on a particle, according to the Mie scattering theory, the intensity of the scattered light is related to factors such as the size, shape, and refractive index of the particle. By detecting the intensity of scattered light at different angles, information such as the size of the particle can be inferred. The dynamic light scattering mode uses the Brownian motion of particles in a solution for measurement. Because the relative positions of particles are constantly changing due to Brownian motion, the scattered light produces a Doppler shift. By analyzing this frequency shift information, the diffusion coefficient of the particles can be obtained, and then the particle size can be calculated according to the Stokes-Einstein equation.

[0037] After obtaining the scattered light intensity data, the target ink sample is subjected to particle size analysis based on these data. Since there is a certain mathematical relationship between scattered light intensity and particle size, a particle size distribution curve can be constructed for different scattering angles and intensities. This curve reflects the distribution probability or relative content of ink particles at different particle sizes. From this particle size distribution curve, particle distribution data can be further extracted, such as the percentage of particles in different particle size ranges, average particle size, particle size distribution range and other parameters.

[0038] This way of setting the particle measurement parameters can improve the accuracy of the measurement. By selecting the appropriate measurement mode (static light scattering or dynamic light scattering) according to the specific composition characteristics of the ink, you can better adapt to the optical properties of the ink particles and avoid data deviation caused by inappropriate measurement modes. For example, if the Brownian motion of the ink particles is more obvious, the dynamic light scattering mode may obtain particle size information more accurately.

[0039] In terms of particle size analysis, the particle size distribution curve and particle distribution data can provide detailed and accurate information for subsequent matching with screen properties. These data can more comprehensively describe the size and distribution of ink particles, and help to accurately determine which screen mesh number is the best match for the ink. For example, when it is known that the particle size distribution of ink particles is very narrow and the average particle size is small, a screen with a relatively high mesh number can be selected to ensure that the ink can pass through the screen smoothly for printing, thereby improving the printing quality and reducing printing defects caused by particle-screen mesh mismatch. At the same time, detailed and accurate data also helps to establish a more reliable matching model. In the long-term production process, the matching algorithm can be continuously optimized based on these data to improve production efficiency and product quality stability.

[0040] In the embodiments of the present application, the normal distribution curve is the most common type of particle size distribution curve, which presents a symmetrical bell shape. This means that the probability of the particle size appearing near the average value is the highest, and it gradually decreases to both sides. In this distribution, the average particle size is the particle size value corresponding to the position of the symmetry axis of the curve. For example, in the ink of some monodisperse systems under ideal conditions, the particle size is relatively uniform, and its particle size distribution curve is close to the normal distribution. The particle size of most particles is concentrated near the average value, and particles with small and large particle sizes account for a small proportion.

[0041] For inks that present a normal distribution, their printing performance is relatively stable. In the screen printing process, this type of ink can pass through the screen relatively evenly, because the particle size is relatively concentrated, and there will be no problems such as some large particles blocking the screen or too many small particles causing the ink to be too fluid. Therefore, the printed pattern is of relatively high quality, with uniform colors and clear lines.

[0042] The lognormal distribution curve is also a common distribution type. Its shape is similar to the normal distribution curve, but the horizontal axis is usually the logarithm of the particle size. This distribution curve is suitable for describing particle systems with a wide range of particle sizes and a large particle size span. In ink particles, when there are multiple particle components of different sizes and the size difference is large, a lognormal distribution often occurs. For example, when there are both nano-scale fine toner particles and micron-scale glass powder particles in the ink, the particle size distribution curve may show a lognormal distribution.

[0043] For log-normally distributed inks, the synergistic effect of particles of different sizes needs to be considered during the printing process. Due to the large span of particle sizes, it may be necessary to select an appropriate screen mesh count based on the particle size of the main component and its proportion in the distribution. If large particles account for a large proportion, the screen mesh count cannot be too high, otherwise the large particles will not be able to pass through the screen; conversely, if small particles dominate, it is necessary to note that too many small particles may cause the ink's adhesion and other properties to decrease, and the screen mesh count and printing process parameters need to be comprehensively considered to ensure printing quality.

[0044] The bimodal distribution curve has two obvious peaks, which indicates that there are two main particle size ranges in the ink particle system. This may be due to the fact that there are two different types of particle components in the ink formulation, their particle sizes are quite different, and each occupies a certain proportion in the system. For example, an ink may contain both fine particle toner for color development and coarse particle additives for increasing the wear resistance of the ink. The particle sizes of these two components are different, and two peaks will appear on the particle size distribution curve.

[0045] For inks with bimodal distribution, the screen mesh and printing parameters need to be adjusted more finely during the printing process. Because the two different particle size ranges have different requirements for the screen mesh, a balance point must be found so that both particles can be printed well through the screen. If the screen mesh is selected based on the particle size of only one particle, the other particle may not be printed properly, thus affecting the printing quality, such as pattern defects or insufficient wear resistance.

[0046] In practical applications, in log-normally distributed inks, the average particle size plays a key role in the fluidity of the ink. If the average particle size is smaller, the ink usually has better fluidity. This is because the friction between smaller particles is relatively small, and during the printing process, the ink can pass through the screen more smoothly, just like fine sand flows more easily than coarse sand. For example, in photovoltaic glass screen printing, ink with a small average particle size can be more evenly distributed on the screen, which is conducive to forming a flat and uniform printing layer on the glass surface.

[0047] Smaller average particle size helps improve printing accuracy. When the average particle size of ink particles is smaller, they can better fill the small holes of the screen, so that the pattern details can be reproduced more accurately during printing. For some fine circuit patterns or tiny logos printed on photovoltaic glass, ink with a small average particle size can ensure the clarity and accuracy of the pattern and reduce the blurring of the pattern or uneven line thickness.

[0048] Generally speaking, inks with a small average particle size may dry faster. This is because small particles have a larger specific surface area, a larger contact area with air or other drying media, and water or solvents are more likely to evaporate. In the production process, this helps improve printing efficiency and reduce the time waiting for drying.

[0049] The geometric standard deviation is an important indicator of the dispersion of particle sizes in a lognormal distribution. A smaller geometric standard deviation means that the particle size distribution is relatively concentrated and the composition of the ink is relatively uniform. This ink is more stable during the printing process because the consistency of the particle size makes the behavior of the ink more predictable when passing through the screen. For example, during the screen printing process, ink with uniform particle distribution is less likely to have some large particles blocking the screen holes and small particles over-penetrating, thus ensuring the stability of printing quality.

[0050] Inks with small geometric standard deviations may perform better when mixed or used in conjunction with other materials. In photovoltaic glass screen printing, inks may need to work with some additives or curing agents. If the ink itself has a concentrated particle size distribution, after mixing with other materials, the chemical reaction or physical interaction between them will be more uniform and stable, which is conducive to the formation of high-quality printed coatings.

[0051] When the geometric standard deviation is large, the dispersion of ink particle size is high, which increases the difficulty of quality control. During the production process, due to the large difference in particle size, different batches of ink may have different printing performance due to differences in particle distribution. For example, large particles may cause screen clogging or defects in the printed pattern during screen printing, while too many small particles may affect the hiding power or adhesion of the ink. Therefore, inks with large geometric standard deviations require stricter quality control measures during the production process to ensure printing quality.

[0052] As an optional embodiment, in the above steps, before using particle measurement parameters to obtain the scattered light intensity data of ink particles in the target ink sample at different angles, a calibration ink sample with a similar composition to the target ink sample can be selected as a reference sample; the measurement equipment is precision calibrated based on the reference sample, and the reference sample deviation data obtained by the precision calibration is set as the compensation parameter in the particle measurement parameters.

[0053] It is worth noting that, first, a calibration ink sample with similar composition to the target ink sample is selected as a reference sample. This is because similar composition means a certain correlation in light scattering characteristics. During the measurement process, the measuring device itself may have certain systematic errors, which may come from various factors such as the accuracy of the optical components of the device and the signal processing algorithm. By placing a known reference sample into the measuring device, the accuracy correction is performed based on the difference between the theoretical value of the reference sample (these theoretical values ​​are obtained through precise experimental methods or standard data) and the value actually measured by the measuring device. For example, for the particle size in the reference sample, there is theoretically an accurate numerical range, but the measuring device may obtain a deviated value due to its own reasons. This deviation can be quantified, and setting it as a compensation parameter in the particle measurement parameter is equivalent to compensating for the inherent deviation of the measuring device when the target ink sample is measured later.

[0054] Thus, in terms of precision calibration, the accuracy of the measuring equipment can be significantly improved by calibrating it with a reference sample. This makes the subsequent acquisition of scattered light intensity data of ink particles at different angles in the target ink sample more accurate. For example, when measuring the particle size of ink particles, the calibrated equipment can more closely reflect the actual size of the particles, reducing the particle size measurement deviation caused by equipment errors.

[0055] The application of compensation parameters can effectively compensate for the systematic error of the measuring equipment. When obtaining the particle measurement data of the target ink sample, the deviation of the equipment itself can be taken into account, making the final data more reliable. In this way, the particle size analysis and particle distribution evaluation based on these more accurate data will be more accurate, providing a more accurate basis for the matching of ink particles and screen mesh count, thereby improving the quality control and production efficiency of the entire photovoltaic glass screen printing process.

[0056] As an optional embodiment, in the above steps, obtaining the particle distribution data based on the particle size distribution curve includes: Extracting the original particle size measurement data of multiple measurements from the particle size distribution curve; marking each curve distribution interval with the corresponding number of ink particles or ink particle hit frequency in the current distribution interval; and calculating the corresponding particle distribution data in each distribution interval based on the original particle size measurement data.

[0057] The particle distribution data corresponding to each distribution interval at least include: the average particle diameter, the particle standard deviation, and the particle diameter distribution range corresponding to each distribution interval.

[0058] Specifically, first of all, extracting the original particle size measurement data from the particle size distribution curve is the basis. Because the particle size distribution curve is an intuitive presentation of the particle size distribution of ink particles, it contains the particle size information obtained by multiple measurements. The corresponding number of ink particles or the hit frequency is marked in each distribution interval of the curve, which reflects the occurrence of particles in this specific particle size interval. Based on these original data, the particle distribution data in each distribution interval is calculated. For the calculation of the average particle diameter, all particle sizes in the interval and their corresponding numbers (or frequencies) are comprehensively considered. Through a suitable mathematical formula, the particle size and frequency in the interval are combined to obtain a value that can represent the average size of the particles in this interval. The calculation of the particle standard deviation is to measure the degree of discreteness of the particle size in the interval. Using statistical methods, the standard deviation is calculated based on the original particle size data in the interval and their distribution. It can reflect whether the particle size is relatively concentrated or dispersed in the interval. The particle diameter distribution range directly obtains the minimum and maximum particle sizes in the interval from the original particle size data. The difference between these two values ​​is the distribution range, which is used to describe the span of the particle size in the interval.

[0059] This method can obtain detailed and accurate particle distribution data. By extracting information from the particle size distribution curve and performing calculations, rich particle characteristic data in each distribution interval can be obtained. Determining the average particle diameter helps to understand the typical size of particles in each particle size interval, which is very important for subsequent matching with the screen mesh. For example, when it is known that the average particle diameter in a certain interval is large, it can be considered that the screen mesh should not be too fine, otherwise these particles may not pass through the screen smoothly.

[0060] The particle standard deviation can reflect the uniformity of particle size. If the standard deviation is small, it means that the particle size within this range is relatively uniform, which is beneficial to the quality of the ink and printing performance, because uniform particles are easier to control and transfer during the screen printing process.

[0061] The particle diameter distribution range can intuitively show the width and narrowness of the particle diameter in each interval. A wider distribution range may mean that there are both large particles that may clog the screen and small particles that are prone to excessive penetration in this interval, which requires more caution in screen selection and printing process adjustment, thereby effectively improving printing quality and stability of the printing process.

[0062] It can be understood that in the above steps, different types of toners will directly affect the final color rendering effect of the ink. For example, in photovoltaic glass screen printing, if organic toners and inorganic toners are used, the color vividness, saturation and light resistance will be different. Organic toners may have brighter colors, but poor light resistance. Although inorganic toners are relatively dark in color, they have better durability and can maintain color stability under long-term light exposure.

[0063] The type of toner also affects the compatibility of the ink with the photovoltaic glass substrate. Some toners may interact better with the chemical properties of the glass surface, forming a stronger adhesion layer, thereby improving the adhesion of the ink on the glass. Inappropriate toners may cause the ink to fall off easily, affecting the printing quality.

[0064] Glass powder plays a major role in increasing the hardness and wear resistance of ink. A higher proportion of glass powder will make the pattern formed after the ink is printed more wear-resistant. In the application scenarios of photovoltaic glass, such as surface printing of solar panels, ink with a high proportion of glass powder can resist the wear of the printed pattern by natural factors such as wind, sand, rain, etc., and extend the service life.

[0065] The proportion of glass powder also affects the transparency of the ink. The right amount of glass powder can improve the physical properties of the ink without affecting the color and transparency of the ink. However, if the proportion of glass powder is too high, it may cause the transparency of the ink to decrease, affecting the lighting efficiency of the photovoltaic glass, especially for solar panels that need to convert photoelectricity through glass, which is a factor that needs to be considered.

[0066] The viscosity of ink is a reflection of its internal friction. Ink with moderate viscosity has good fluidity and transferability during screen printing. If the viscosity is too high, the ink will have poor fluidity on the screen and will have difficulty passing through the screen holes evenly, which may result in incomplete printed patterns or uneven thickness. On the contrary, if the viscosity is too low, the ink may flow excessively, causing problems such as blurred patterns and ink bleeding.

[0067] Viscosity also affects the drying speed of ink. Generally speaking, inks with higher viscosity may dry more slowly because it is more difficult for the solvent to evaporate. At the same time, the stability of viscosity is also important. During the printing process, if the viscosity of the ink changes, for example due to environmental factors such as temperature and humidity, it may cause fluctuations in printing quality.

[0068] Solvents are an important component of inks, and different types of solvents have different volatilities. For example, water-based solvents and organic solvents differ greatly in drying speed and drying method. Water-based solvents usually evaporate more slowly but are more environmentally friendly; organic solvents may evaporate faster, but may pose safety risks and environmental pollution problems. The solvent content will also affect the drying speed of the ink, and solvents with a high content will prolong the drying time of the ink. The type of solvent determines its solubility and dispersibility for components such as toner and glass powder. A suitable solvent can ensure that toner and glass powder are evenly dispersed in the ink system to form a stable ink. If the solvent cannot dissolve or disperse these components well, it may cause particle agglomeration, affecting the performance of the ink and the printing quality.

[0069] Further optionally, it is assumed that the original particle size measurement data includes: the interval particle distribution center value and the particle hit frequency in each distribution interval; the interval particle distribution center value is the position of the geometric center of the particle distribution shape.

[0070] Based on the above assumptions, the average particle diameter corresponding to the i-th distribution interval is calculated using the following formula: ;in, is the center value of the interval particle distribution in the i-th distribution interval, is the particle hit frequency corresponding to the i-th distribution interval.

[0071] The particle standard deviation corresponding to the i-th distribution interval is calculated using the following formula: ;in, is the average diameter of particles in the i-th distribution interval, is the center value of the interval particle distribution in the i-th distribution interval, is the corresponding particle hit frequency in the i-th distribution interval.

[0072] For the calculation of the average particle diameter, the formula is based on the center value of the interval particle distribution and the particle hit frequency within the distribution interval. The center value of the interval particle distribution represents the geometric center position of the particle distribution shape in the interval, which can be regarded as a typical value of the particle size in this interval. The particle hit frequency reflects how often the particles appear in this interval during the measurement. By multiplying the center value of each interval with the hit frequency and summing them, and then normalizing them (dividing by the sum of the frequencies), an average diameter that takes into account the probability of particles appearing in different intervals can be obtained. This calculation method can integrate the information of each interval and more accurately reflect the average particle size of the entire particle system. When calculating the particle standard deviation, the formula is based on the calculated average particle diameter. First, the square of the difference between the center value and the average diameter in each distribution interval is calculated, multiplied by the hit frequency, and then the sum of this result for all intervals is taken, and finally the square root is taken to obtain the standard deviation. This process is based on statistical principles, and by measuring the degree of deviation of the particle size in each interval from the average particle size and considering the frequency of particles appearing in each interval, a value that can quantify the degree of particle size dispersion is obtained.

[0073] This calculation method can more accurately describe the overall size of ink particles in the calculation of the average particle diameter. Unlike the simple arithmetic mean, it fully considers the frequency of particles in different particle size ranges, and can more accurately give an average diameter representing the overall particle size for ink particle systems with uneven particle size distribution. This is very important for subsequent operations such as matching with the screen mesh number, because the selection of the screen mesh number requires a precise grasp of the ink particle size. In terms of the calculation of the particle standard deviation, the result obtained by the above formula can effectively reflect the degree of discreteness of the particle size. It provides a quantitative indicator for evaluating the uniformity of ink particles. A smaller standard deviation indicates that the particle size is more concentrated, and the ink quality is more stable in terms of particle size, which is conducive to ensuring uniform ink transfer and good printing quality during the screen printing process. On the contrary, a larger standard deviation indicates that the particle size is highly dispersed, and problems caused by particle size differences may occur during the printing process, such as clogging the screen or uneven printing patterns. These accurate calculation results provide more reliable data support for ink performance evaluation and printing process optimization.

[0074] Further optionally, assuming that the original particle size measurement data also includes: the minimum particle diameter and the maximum particle diameter in each distribution interval. Based on this, the particle diameter distribution range corresponding to the i-th distribution interval is calculated using the following formula: ;in, is the maximum particle diameter in the i-th distribution interval, is the minimum particle diameter in the i-th distribution interval.

[0075] In this calculation process, the particle diameter distribution range is calculated by the maximum particle diameter and the minimum particle diameter in each distribution interval in the original particle size measurement data. This is a very intuitive way, because the essence of the distribution range is to reflect the span of the particle diameter in this interval. Subtracting the minimum particle diameter from the maximum particle diameter directly gives the difference in particle diameters in the interval, and this difference is the particle diameter distribution range. It is based on such a simple mathematical logic that the particle diameter span in the interval is defined by the maximum and minimum particle diameters.

[0076] This calculation method can clearly show the distribution width of the ink particle size in each distribution interval. It is very useful for evaluating the uniformity of ink particles. If the calculated particle diameter distribution range is small, it means that the particle size of the ink particles in this interval is relatively concentrated and the particle size is relatively uniform. This is beneficial to the quality and printing performance of the ink, because in the screen printing process, uniform particles can better control the transfer and deposition of the ink, making the printed pattern clearer and more uniform. On the contrary, if the particle diameter distribution range is large, it means that there is a large difference in particle size in this interval, and there may be problems such as large particles blocking the screen holes or small particles over-penetrating. This reminds us that we need to consider this particle size non-uniformity when selecting the screen and adjusting the printing process, so as to effectively improve the printing quality and the stability of the printing process. At the same time, by calculating the particle diameter distribution range of each distribution interval, we can fully understand the uniformity of the ink particles on the entire particle size distribution curve, and provide detailed data support for the comprehensive evaluation of ink performance and the optimization of printing process.

[0077] Further optionally, after collecting the screen properties in the candidate screen sample in 103, the ink particle data and the screen properties can also be stored in a dynamic data space, and corresponding update parameters can be configured to achieve dynamic update; through the dynamic data space, the historical measurement data and historical analysis data of multiple candidate screen samples and ink samples are dynamically saved for the mesh matching model to use for screen mesh prediction analysis.

[0078] After collecting the ink particle data and screen properties, they are stored in the dynamic data space. This dynamic data space is like a constantly updated information warehouse. The configured update parameters enable this warehouse to update data according to certain rules. It records the historical measurement data and historical analysis data of multiple candidate screen samples and ink samples. For ink particle data, it includes information such as particle distribution data and toner ratio, and screen properties include screen material, screen mesh number, etc. When new data is generated or old data needs to be modified, the update parameters will drive the dynamic change of data. The mesh matching model extracts these rich data from this dynamic data space when performing screen mesh prediction analysis. The model uses these historical data to learn the complex relationship between ink particles and screen mesh number, because as the data accumulates, it can better understand how to best match different ink characteristics with different screen properties, and continuously optimize the prediction by analyzing a large amount of historical data.

[0079] From the perspective of data storage, the dynamic data space provides a flexible storage environment. It can continuously accumulate data so that all ink and screen sample information can be properly preserved. This is very beneficial for long-term production processes and quality control, and it is convenient to trace the characteristic data of different batches of ink and screen. In terms of model prediction, dynamically updated data provides richer materials for the mesh matching model. By using historical measurement and analysis data of multiple candidate screen samples and ink samples, the mesh matching model can continuously learn and adapt to new situations. This helps to improve the accuracy of model predictions because it can judge the best matching relationship between ink particles and screen mesh based on more actual cases. Moreover, dynamic updates can ensure that the model always uses the latest and most comprehensive data for predictions, reducing matching errors caused by outdated data, thereby better assisting the screen printing process and improving printing quality and production efficiency.

[0080] In practical applications, a large number of actual screen printing cases can be collected. When using dynamically updated data, the model is used to predict the screen mesh count, and the predicted value is compared with the screen mesh count (actual value) that works best in actual applications. For example, the difference between the predicted mesh count and the actual mesh count is calculated. The smaller the difference, the higher the prediction accuracy of the model. If the screen mesh count predicted by the model is very close to the mesh count used to achieve the best printing quality in actual production, it means that the dynamic update data has played a positive role in improving the prediction accuracy of the model. Count the proportion of correct predictions (the predicted mesh count and the actual best mesh count are consistent within a certain error range) in a certain number of prediction cases. This proportion can intuitively reflect the performance of the model. With the continuous accumulation of dynamically updated data, the proportion of correct predictions should show an upward trend, which means that the model can more accurately judge the matching relationship between ink particles and screen mesh count by learning new data.

[0081] Analyze whether the prediction results of the model are stable in different batches. If dynamically updating data helps optimize the model, then when faced with similar ink particle data and screen property inputs, the model's prediction results should be relatively stable and will not fluctuate greatly. For example, for the same type of ink (similar particle distribution, toner ratio, etc.) and screen (same material and approximate mesh range), the mesh count predicted by the model multiple times should vary within a small range.

[0082] Observe the performance of the model when there are certain interference factors in the data (such as measurement errors, a small amount of abnormal data, etc.). If the dynamic update of data can enable the model to better learn the inherent laws of the data, then the model should have a certain anti-interference ability and will not produce unreasonable prediction results due to these small interferences. You can artificially add some noise or small outliers to the input data to observe whether the prediction results of the model still remain within a reasonable range.

[0083] When introducing new ink formulas (such as new toners, different particle composition ratios) or new screen materials and mesh specifications, evaluate whether the model can adapt quickly and give reasonable predictions. Dynamically updated data should contain information on various types of inks and screens so that the model can learn a wider range of matching patterns. If the model can make relatively accurate predictions on new ink and screen combinations in a shorter period of time based on new data, it means that dynamically updated data has enhanced the adaptability of the model.

[0084] Consider the impact of production environment factors (such as changes in temperature and humidity) on ink and screen performance. If the dynamically updated data is comprehensive enough, it should be able to include ink and screen data under different environments. The model can use this data to learn the relationship between environmental factors and ink-screen matching, so that it can make appropriate screen mesh predictions under different production environments, reflecting good adaptability to changing environments.

[0085] Taking the database as an example, in the photovoltaic glass screen mesh recommendation system, a screen database based on a SQL relational database can be established, which plays a core data support role. It aims to reasonably store the technical parameters of various screens and ink-related parameters, and with its efficient data storage, retrieval and dynamic update functions, it provides a strong data foundation for the subsequent matching algorithm module to perform accurate calculations and reasonable recommendations, and is the key link to achieve accurate matching of particle data and screen mesh.

[0086] The database records in detail the key parameters of various aspects of ink, which will be used as input data for subsequent matching work. Among them, the particle distribution data can directly reflect the size and uniformity of ink particles, and is an important indicator for measuring the physical properties of ink; the toner ratio and glass powder ratio are related to the composition of the ink, and different ratios will affect the color presentation, hardness, wear resistance and other properties of the ink; the record of toner type is also very important, because different types of toner have differences in color effects and adhesion to the substrate. Combining these ink-related parameters can fully describe the properties of the ink and provide basic information for finding a screen mesh that matches it.

[0087] As the core content of screen-related parameters in the database, screen mesh count is the direct target of the entire matching work. It represents the density of the screen and plays a decisive role in whether the ink can pass through the screen smoothly for uniform printing. Through matching analysis with ink-related parameters, the most suitable screen mesh count can be determined, thereby ensuring the quality of photovoltaic glass screen printing.

[0088] Considering that new situations will constantly emerge in actual production, such as the introduction of new suppliers' products or the emergence of new ink formulas, users can use the data addition function to enter new types of screens or new ink parameters into the database. This function ensures that the database can keep pace with the times and always covers all kinds of screens and inks involved in current actual production, so that the matching algorithm module can perform calculations based on the latest and most complete data, improving the accuracy and practicality of recommendations.

[0089] During the course of business cooperation, the supplier's product specifications may change, such as some technical parameters of the screen being adjusted. At this time, the data modification function comes into play, and users can easily update the relevant data that has been entered into the database to ensure the accuracy of the information in the database, avoid matching errors caused by outdated data, and maintain the consistency between the database and actual production conditions.

[0090] As time goes by and production processes improve, some data may become obsolete or no longer valid, such as parameters corresponding to some obsolete ink formulas or no longer used screen models. Through the data deletion function, users can remove these redundant data from the database, keep the database clean and efficient, optimize the use of data storage space, and also help improve the efficiency of data retrieval and matching calculations.

[0091] Before inputting the ink parameters and screen mesh count in the database into the convolutional neural network (CNN) matching algorithm module, feature normalization is an important preprocessing step. Using the normalization method, all feature values ​​such as ink parameters (such as particle size, color powder ratio, glass powder ratio, etc.) and screen mesh count are mapped to a specific standardized range, such as 0 to 1 or with zero mean and unit variance. The advantage of doing this is that it can eliminate the differences between different features caused by factors such as dimensions, so that each feature is equally important in subsequent model operations, which helps to improve the accuracy and stability of CNN model training and prediction.

[0092] For classification features such as toner type and screen material, they cannot be directly used in numerical calculations, so they need to be converted into numerical vectors using methods such as One-Hot Encoding or Embedding Vector. For example, if there are several different types of toner, One-Hot Encoding can be used to convert them into a vector consisting of only 0 and 1. Each category corresponds to a unique vector representation, so that the classification features can be recognized and processed by the CNN model, realizing the integration of non-numerical information into the overall matching algorithm operation.

[0093] The ink parameters and screen parameters are cleverly combined into a two-dimensional feature matrix, which is regarded as an "image"-like structure when input into the CNN model. This processing method fully utilizes the advantage of the CNN model in processing image structure data. By constructing such a multi-dimensional feature matrix, the model can better explore the intrinsic connection and matching relationship between the ink parameters and the screen parameters, further improve the ability to make accurate matching predictions based on data, and lay a solid foundation for accurately recommending the appropriate screen mesh count.

[0094] In the above example, this SQL relational database solution, through reasonable data field design, powerful dynamic update function, and scientific data standardization and preprocessing, comprehensively ensures that in the photovoltaic glass silk-screen printing process, it can provide efficient, accurate and actual production-oriented data support for the precise matching of ink particles and silk screen mesh count, thus helping the entire silk-screen printing process to achieve better quality and production efficiency.

[0095] Further optionally, it is assumed that the mesh matching model includes at least the following structures: a preprocessing layer, a derivative layer, a prediction layer, and a fusion layer. Based on this structure, the ink particle data and the screen properties are combined, and the mesh matching model is used to predict the target screen mesh whose matching degree with the target ink sample reaches a preset condition, such as Figure 2 As shown, including: 201. Extracting the ink features to be analyzed and the screen features to be analyzed from the ink particle data and the screen properties through a preprocessing layer, and projecting the ink features to be analyzed and the screen features to be analyzed into a corresponding feature conversion space to obtain a corresponding first matching feature map; 202. Selecting a derivative prediction branch that matches the ink feature to be analyzed through the derivative layer, and predicting the first matching feature map through the derivative prediction branch to obtain a first predicted screen mesh count; the derivative prediction branch is constructed based on a derivative knowledge base that matches the ink attribute characteristics; 203. Predicting the ink particles and the screen mesh count in the first matching feature map through a prediction layer to obtain a second predicted screen mesh count; the processing method of the prediction layer is obtained based on preset conditions, candidate screen samples, and historical measurement data and historical analysis data of ink samples; 204. Adaptively fuse the first predicted screen mesh count and the second predicted screen mesh count through a fusion layer to obtain the target screen mesh count.

[0096] The following is a detailed description of each step of the above mesh matching model implementation example: In 201, the pre-processing layer plays a fundamental and key role in the entire mesh matching model. It first receives information from ink particle data and screen properties, which covers multi-dimensional data such as ink particle distribution data (such as average particle diameter, standard deviation, etc.), toner ratio, glass powder ratio, toner type, and screen mesh number.

[0097] The pre-processing layer extracts the ink features and screen features to be analyzed based on its design function. This extraction process is like screening and refining the original data, picking out the key features that are valuable for subsequent analysis and prediction. For example, for the particle distribution data of ink particles, it may focus on extracting core features that can reflect the concentration degree and distribution range of particle size; for component-related features such as the toner ratio, it clarifies key information such as its specific proportion value.

[0098] After extracting the features, the preprocessing layer will further project the ink features to be analyzed and the screen features to be analyzed into the corresponding feature conversion space, thereby obtaining the corresponding first matching feature map. This projection process utilizes the advantages of convolutional neural networks (CNN) in feature processing, and transforms features of different dimensions and properties into a feature space that is more convenient for subsequent analysis and processing through a specific mapping method. The first matching feature map formed can present the relationship between these features in a structured and graphical form, providing a suitable data foundation for subsequent derivative layer, prediction layer and other operations, just like building a unified "stage" for subsequent analysis, allowing different data to interact and analyze on this "stage" according to the set rules.

[0099] In 202, the derivative layer has a unique mechanism of action in the entire model structure. It is built based on the derivative knowledge base matching the ink property characteristics. This derivative knowledge base is a collection of knowledge accumulated through a large number of ink property studies and actual application case analysis, and contains various experiences and rules for matching screen mesh counts under different ink properties.

[0100] After receiving the first matching feature map from the preprocessing layer, the derivative layer will first select the matching derivative prediction branch in the derivative knowledge base based on the characteristics of the ink to be analyzed. For example, if the toner type of the ink belongs to a certain type and the glass powder ratio is within a certain range, the derivative layer can find the prediction branch path summarized in the past practice for this type of ink with similar properties from the knowledge base.

[0101] Then, by selecting a good derivative prediction branch, the first matching feature map is predicted to obtain the first predicted screen mesh count. This process makes full use of existing knowledge and experience, and can quickly give a preliminary screen mesh count prediction result based on the properties of the ink itself, which is equivalent to giving a possible matching answer from the perspective of specific knowledge accumulated in the past, providing a reference and comparison basis for subsequent more comprehensive prediction analysis, so that the entire prediction process has both rapid judgment based on experience and continuous optimization combined with subsequent analysis of other layers.

[0102] In 203, the processing method of the prediction layer has a solid basis, which is trained based on preset conditions, historical measurement data of candidate screen samples and ink samples, and historical analysis data. During the model training stage, a large number of historical matching records containing particle parameters (such as average particle diameter, toner ratio, glass powder ratio, toner type) and corresponding screen mesh counts are collected as input data, and the output target is the optimal screen mesh count or mesh count range.

[0103] For different prediction target situations, appropriate loss functions are used to optimize training. For example, when predicting a single mesh size, it is treated as a classification problem and the cross entropy loss function is used; if the mesh size range is predicted, it is treated as a regression problem and the mean square error (MSE) loss function is used. Through such targeted training, the prediction layer learns the complex relationship between particle parameters and screen mesh size.

[0104] After receiving the first matching feature map from the preprocessing layer, the prediction layer will use the trained method to predict the ink particles and screen mesh count, thereby obtaining the second predicted screen mesh count. This prediction process is based on the rules mined from a large amount of historical data, and gives a prediction result that conforms to the actual matching situation in the past from a data-driven perspective, providing another important reference dimension for the final comprehensive conclusion of the accurate target screen mesh count, which complements and verifies the prediction results of the derivative layer.

[0105] In 204, the core task of the fusion layer is to adaptively fuse the first predicted screen mesh count and the second predicted screen mesh count obtained previously. Since the first predicted screen mesh count is derived from the empirical perspective based on the derived knowledge base, and the second predicted screen mesh count is obtained through data mining based on historical data training, each has its own advantages and focuses.

[0106] The fusion layer will adopt an adaptive fusion strategy based on their respective characteristics and the actual situation of the current input data. For example, when the input ink particle data is highly similar to a typical case in the derived knowledge base, the weight of the first predicted screen count in the fusion may be appropriately increased; and if the overall data characteristics of the current input are more consistent with a common matching pattern in the historical data, the weight of the second predicted screen count may be increased accordingly.

[0107] Through such an adaptive fusion process, the advantages of both experience-based and data-driven are fully integrated, and finally a more accurate and reliable target screen mesh number is obtained, which achieves a precise match between particle data and screen mesh number, and provides a strong decision-making basis for selecting the appropriate screen mesh number in the photovoltaic glass screen printing process, thereby improving printing quality, increasing production efficiency and reducing costs.

[0108] In this way, the entire mesh matching model fully utilizes the advantages of convolutional neural networks in feature extraction, complex relationship capture, etc. through such a layered collaborative, focused structure and processing method, and combines knowledge and experience with historical data to achieve efficient and accurate screen mesh prediction. At the same time, with the help of mechanisms such as dynamic updates, it can continuously adapt to changes in new inks and screen specifications, and always maintain good practicality and adaptability.

[0109] It is further important to understand that the fusion layer first evaluates the reliability of the derivative layer and the prediction layer. For example, if the data in the derivative knowledge base is strictly screened expert experience, and the current input ink features are very consistent with the typical cases in the knowledge base, then the reliability of the derivative layer is high and it will be given a relatively high weight during fusion. For the prediction layer, if the amount of historical data it is based on is large and the feature distribution is highly similar to the current input ink and screen data, it will also receive a higher weight.

[0110] The weights are not fixed, but will be adjusted dynamically according to the specific situation of the input data. For example, when encountering a new type of ink, its particle distribution data is very different from the previous ones, and there is no very matching case in the derived knowledge base, but there are records of similar particle distribution changes in the historical data of the prediction layer, then the weight of the prediction layer may be automatically increased during the fusion process to better adapt to this new situation.

[0111] The advantage of the derived layer is that it can use existing knowledge and experience to quickly give a preliminary mesh count prediction. For example, for an ink with a specific color powder type and glass powder ratio, the derived layer can directly output a possible mesh count based on previous experience matching the properties of this ink. The prediction layer is trained based on a large amount of historical measurement and analysis data, and it can mine more complex particle-mesh count relationships from the data, such as the optimal mesh count under different particle distribution ranges and color powder ratio combinations.

[0112] The fusion layer combines the two complementary pieces of information. For example, when the first predicted mesh count predicted by the derivative layer is based on the composition characteristics of the ink (the ratio of toner and glass powder), and the second predicted mesh count obtained by the prediction layer focuses more on the particle distribution data, the fusion layer will take these two factors into consideration. If the two prediction results are close within a certain range, it means that both factors support this mesh count range; if the difference is large, the fusion layer will make a comprehensive judgment based on the weight and other relevant factors (such as whether the current production focus is on ink composition matching or better particle permeability), and finally output a more reasonable target mesh count.

[0113] The fusion layer receives feedback on the target mesh count of the final output in actual application. If it is found that the output mesh count leads to poor printing quality in the actual printing process, such as uneven ink transfer or insufficient pattern clarity, the error information will be fed back. Then, the fusion layer will adjust the fusion strategy with the derivative layer and the prediction layer based on the error analysis, such as changing the weight distribution or fusion method, to improve the accuracy of the next prediction.

[0114] As new ink and screen sample data are continuously collected and used for model training and knowledge base updates, the fusion layer will also update the fusion strategy accordingly. For example, when new data shows that the particle shape of a certain ink has an important influence on the selection of screen mesh, and this factor has not been fully considered in the fusion process before, the fusion layer will adjust the strategy and update this factor through the derivative layer or prediction layer or add new considerations in the fusion process, so that the final output target screen mesh can better adapt to the new situation.

[0115] As an optional embodiment, the prediction layer is used to predict the ink particles and the screen mesh count in the first matching feature map to obtain a second predicted screen mesh count, including: A preset convolution operation is performed on the first matching feature map to obtain a second matching feature map; the acquisition process of the second matching feature map is expressed as follows:

[0116] in, is the first matching feature map, is the dilated convolution kernel, Indicates the processing method of the convolution kernel in the feature channel, is the position index value of the convolution kernel in the horizontal and vertical directions, n represents the nth convolution kernel, is the void ratio, is the step length of the dilated convolution kernel operation, is the convolution kernel size, is the second matching feature map after the dilated convolution. represents the corresponding channel identifiers in the first matching feature map and the second matching feature map, and Respectively represent the corresponding row position index and column position index in the first matching feature map and the second matching feature map; The second predicted screen mesh count is obtained by performing a prediction based on the second matching feature map.

[0117] After the feature extraction of the first matching feature map by the previous dilated convolution and the feature integration and nonlinear transformation of the two fully connected layers, the model finally outputs a predicted value for the screen mesh. This predicted value is based on the rules and patterns learned by the model from the input ink particles and other related features. It can be fused with other prediction results (such as the prediction results from the derivative layer) through the fusion layer to obtain a more accurate and comprehensive target screen mesh, which is used to guide the selection of screen mesh in actual screen printing.

[0118] Therefore, the embodiment of the present application can use the void convolution to capture the contextual features of ink particles and screen properties, and then use the powerful feature integration and nonlinear mapping capabilities of the fully connected layer to gradually transform the input features into a reasonable prediction of the screen mesh number, aiming to improve the prediction accuracy and effect of the mesh number matching model in practical applications.

[0119] As an optional embodiment, the generating screen recommendation information based on the target screen mesh count includes: identifying the recommended level of information refinement required for the printing process to be performed; Determine a target recommendation mode based on the level of refinement of the recommendation information; the target recommendation mode includes a precise mode or a compatible mode; The screen recommendation information is generated based on the target recommendation mode; wherein, in the precise mode, the screen recommendation information includes the best screen mesh matching the target ink sample as the target screen mesh; and in the compatible mode, the screen recommendation information includes target screen meshes in multiple dynamic adjustment ranges obtained by interval search and / or dynamic programming.

[0120] It is understandable that the precision mode is mainly suitable for scenarios with high requirements for specific processes, such as photovoltaic glass pattern detail printing. In this case, the printed pattern requires extremely high fineness, and it is necessary to ensure that the ink can pass through the screen accurately to form a high-quality, high-definition pattern. Its purpose is to provide an optimal screen mesh with the highest match to the target ink sample, thereby maximizing the matching accuracy.

[0121] Through in-depth analysis of historical data, error tolerance and optimization algorithms are introduced into the model. For printing processes with high detail requirements, the recommended mesh count will be fine-tuned according to specific needs. For example, when there is a slight change in the ink particle distribution data or the printing environment (such as temperature and humidity) fluctuates, the system can adjust the recommended mesh count based on the pre-set error tolerance and use the optimization algorithm to ensure that the accuracy is not affected.

[0122] In the actual printing process, if there is a slight deviation, the system will make adaptive compensation based on the actual feedback data. For example, if it is found that the clarity of a part of the printed pattern has decreased due to a slight mismatch between the screen mesh and the ink particles, the system will dynamically adjust the screen mesh or other related process parameters such as printing pressure and speed based on the feedback data, thereby reducing the impact of the error on the final effect.

[0123] In the recommendation output module, the precision mode is output in the form of a single recommended mesh number. It will also be presented in a variety of visual output methods, such as clearly listing the recommended mesh number, the corresponding ink particle range and other key parameters in a table form, which is convenient for users to quickly check and compare; it will also use charts, such as generating intuitive line charts, bar charts or heat maps, to show the adaptability of different screen mesh numbers under various ink parameter conditions, helping users better understand why this mesh number is the best choice.

[0124] The compatible mode is suitable for scenarios that require flexible adjustments according to actual conditions during the production process. In actual production, you may encounter fluctuations in ink composition, large changes in the production environment, or situations where the printing accuracy requirements are not top-notch. The purpose of this mode is to recommend an adaptive mesh range, list a variety of available meshes and their advantages and disadvantages, and facilitate users to flexibly adjust according to actual needs to ensure the stability of the printing process and basic printing quality.

[0125] By comprehensively analyzing the characteristics of ink particles and production process requirements, an optimization algorithm (such as interval search or dynamic programming) is used to determine the upper and lower limits of the adaptive mesh count. A suitable mesh count range is calculated based on factors such as ink particle size, distribution, and changes in the production environment. For example, when the average diameter of ink particles fluctuates within a certain range, the interval search algorithm is used to find the mesh count interval that can satisfy the smooth passage of ink through the screen; or a dynamic programming algorithm is used to consider the optimal mesh count strategy under different production stages (such as different drying conditions, printing speeds, etc.) to determine a flexible mesh count range.

[0126] In the recommended mesh range, there will be a certain precision difference between each mesh. This is to ensure that in actual production, when users flexibly select the mesh within this range, they can still meet the basic requirements of the process. For example, in a recommended mesh range of 100-120 mesh, the difference between each mesh can ensure that no matter 100 mesh, 110 mesh or 120 mesh is selected, it can adapt to the changes in ink and production environment to a certain extent, so that the printing effect can reach the basic quality standard.

[0127] In the recommendation output module, the compatible mode is output in the form of a recommended adaptive mesh range. A variety of visual output methods will also be used, such as listing a variety of available meshes and their corresponding characteristics, advantages and disadvantages in a table, so that users can make intuitive comparisons; the adaptability of the mesh range under different conditions will also be displayed through charts, so that users can make choices according to actual conditions in production. At the same time, users can also select the output format (such as PDF, Excel, etc.) through the system interface, and set the level of detail of the displayed content. It supports the export of historical recommendation records to facilitate subsequent tracking and optimization of production processes. In addition, users can also provide real-time feedback on the recommendation results. The system dynamically updates the recommendation results based on the input to meet personalized needs. In special cases, possible alternatives and their adaptability analysis will also be provided.

[0128] As an optional embodiment, online learning adopts an incremental training method, and the model will gradually adapt to the new data instead of performing full training every time it is updated. This incremental update strategy ensures high learning efficiency and avoids the computational overhead caused by training from scratch every time. In addition, in order to ensure the training effect and avoid excessive adjustment of the model, batch updates are performed regularly to ensure the long-term stability of the algorithm. During the online learning process, newly collected user feedback data usually affects the learning process of the model. In order to prevent the new data from having too much impact or deviation on the model, a dynamic weighting mechanism can be set to balance the weights of new and old data. For example, a time decay factor is used so that the newer data has a higher weight in training, but it does not completely replace the historical data, thereby maintaining the stability and anti-interference ability of the model. After each online learning update, the model will be evaluated through the validation set to ensure accuracy during the learning process. If the performance of the model drops significantly, the system will automatically trigger the fallback mechanism and restore it to the last valid model version to avoid adverse effects caused by feedback data quality problems. The online learning system will be equipped with a noise detection mechanism to detect outliers in the data through statistical analysis (such as box plot analysis or z-score algorithm) and remove them from the data set. These outliers may be caused by equipment failure, misoperation or occasional errors, so filtering out these abnormal data can effectively avoid negative impacts on the model. When collecting user feedback, the system will ensure the quality of feedback data by setting reasonable feedback standards. For example, users are required to provide detailed usage environment information, ink formula and screen specifications in the feedback to help filter out invalid or incomplete feedback. Before entering the learning model, all input data (such as particle distribution data, toner ratio, etc.) will be standardized and normalized so that parameters of different dimensions have a unified dimension to avoid some parameters affecting the model training process due to large or small scales. For missing or incomplete user feedback data, the system will use interpolation methods (such as mean interpolation, linear interpolation, etc.) to automatically fill in missing values ​​to ensure data integrity. In addition, the system can also automatically generate reasonable supplementary data based on the changing trend of feedback data to avoid data deviation caused by missing values. Furthermore, the algorithm automatically identifies and removes duplicate data to ensure that each piece of feedback data is independent. In addition, the data will be checked for consistency to avoid affecting the learning effect due to inconsistent data formats or input errors.

[0129] In the embodiment of the present application, the screen mesh number is intelligently matched based on the ink sample and recommended to the user, achieving accurate matching of ink particles and screen mesh number, solving the printing quality problem of improper matching of ink particle size and screen mesh number in the photovoltaic glass screen printing process in the related technology, and significantly improving the printing quality. At the same time, it helps to reduce the production debugging costs and resource waste caused by repeated debugging, and greatly improves the production efficiency of the entire photovoltaic glass screen printing.

[0130] Based on the same implementation principle, the embodiment of the present application also provides a screen printing machine, which includes at least the following units: An acquisition module, used for acquiring an ink sample to be analyzed as a target ink sample; A first acquisition module is used to acquire ink particle data in the target ink sample; the ink particle data at least includes: particle distribution data, toner ratio, glass powder ratio, and toner type; The second collection module is used to collect the screen properties of the candidate screen sample; the screen properties at least include: the screen material and screen mesh number of the candidate screen sample; A prediction module, for combining the ink particle data with the screen properties, and predicting a target screen mesh whose matching degree with the target ink sample meets a preset condition through a mesh matching model; The recommendation module is used to generate screen recommendation information based on the target screen mesh count and send the screen recommendation information to the user to assist the user in completing the screen selection process.

[0131] For similarities, please refer to the relevant description of the above embodiments, and the specific implementation methods of each unit in the above device will not be expanded for the time being.

[0132] Based on the same implementation principle, the embodiment of the present application also provides an electronic device, Figure 3 is a structural block diagram of the electronic device 300, such as Figure 3 As shown, the electronic device 300 includes: a processor 301, a memory 302, a communication interface 303, a communication bus 304 and a controller 305; wherein the processor 301, the memory 302 and the communication interface 303 communicate with each other via the communication bus 304; the memory 302 is used to store computer programs; the processor 301 is used to execute the programs stored in the memory 302 to realize the corresponding processing functions; the communication bus 304 is used for communication between the electronic device 300 and other devices; the controller 305 is used to realize the screen mesh recommendation method based on screen printing described in the above method embodiment.

[0133] In the embodiment of the present application, the communication bus 304 may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The communication bus 304 may be divided into an address bus, a data bus, a control bus, etc., and the specific form is not limited. For ease of representation, Figure 3 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0134] The memory 302 may include a random access memory (RAM) or a non-volatile memory (non-volatile memory), such as at least one disk memory. Optionally, the memory may also be at least one storage device located far away from the processor 301 .

[0135] The processor 301 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other editable logic devices, discrete gate or transistor logic devices, discrete hardware components, and the specific form can be determined according to actual needs.

[0136] Accordingly, an embodiment of the present application further provides a computer-readable storage medium storing a computer program, which, when executed, can implement each step that can be executed by an electronic device in the above method embodiment.

[0137] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for recommending screen mesh count based on screen printing, characterized in that: The method comprises: obtaining an ink sample to be analyzed as a target ink sample; Collecting ink particle data in the target ink sample; the ink particle data at least includes: particle distribution data, toner ratio, glass powder ratio, toner type; Collecting screen attributes from the candidate screen sample; the screen attributes at least include: screen material and screen mesh number of the candidate screen sample; Combining the ink particle data with the screen properties, predicting a target screen mesh whose matching degree with the target ink sample reaches a preset condition through a mesh matching model; Silk screen recommendation information is generated based on the target silk screen mesh count, and the silk screen recommendation information is sent to the user to assist the user in completing the silk screen selection process.

2. The method according to claim 1, characterized in that The collecting of ink particle data in the target ink sample comprises: Randomly selecting a portion of ink samples from the ink samples to be analyzed as the target ink samples; Performing component analysis on the target ink sample to obtain the toner ratio, glass powder ratio, and toner type of the target ink sample; the analysis method at least includes spectrophotometry; Setting particle measurement parameters of the target ink sample based on the toner ratio, glass powder ratio, and toner type of the target ink sample; Using particle measurement parameters, obtaining scattered light intensity data of ink particles in the target ink sample at different angles; the acquisition method at least includes multi-angle dynamic light scattering method; The target ink sample is subjected to a particle size analysis based on the scattered light intensity data to obtain the particle distribution data.

3. The method according to claim 2, characterized in that The step of setting the particle measurement parameters of the target ink sample based on the toner ratio, glass powder ratio, and toner type of the target ink sample includes: Selecting a measurement mode that matches the characteristics of the toner ratio, glass powder ratio, and toner type of the target ink sample, and configuring the particle measurement parameters used by the measuring device in the selected measurement mode; wherein the measurement mode is a static light scattering mode or a dynamic light scattering mode; The performing particle size analysis on the target ink sample based on the scattered light intensity data to obtain the particle distribution data comprises: Performing a particle size analysis on the target ink sample based on the scattered light intensity data to obtain a particle size distribution curve of ink particles in the target ink sample; The particle distribution data is obtained based on the particle size distribution curve.

4. The method according to claim 3, characterized in that Before using the particle measurement parameters to obtain the scattered light intensity data of the ink particles in the target ink sample at different angles, the method further includes: Selecting a calibration ink sample having a similar composition to the target ink sample as a reference sample; The accuracy of the measuring device is calibrated based on the reference sample, and the reference sample deviation data obtained by the accuracy calibration is set as the compensation parameter in the particle measurement parameter.

5. The method according to claim 3, characterized in that: The step of obtaining the particle distribution data based on the particle size distribution curve comprises: Extracting the original particle size measurement data of multiple measurements from the particle size distribution curve; marking each curve distribution interval with the number of ink particles or the ink particle hit frequency corresponding to the current distribution interval; Based on the original particle size measurement data, the particle distribution data corresponding to each distribution interval is calculated; wherein the particle distribution data corresponding to each distribution interval includes at least: the average particle diameter, the particle standard deviation, and the particle diameter distribution range corresponding to each distribution interval.

6. The method according to claim 5, characterized in that The original particle size measurement data includes: the interval particle distribution center value and the particle hit frequency in each distribution interval; the interval particle distribution center value is the position of the geometric center of the particle distribution shape; Among them, the average diameter of the particles corresponding to the i-th distribution interval is calculated using the following formula: ;in, is the center value of the interval particle distribution in the i-th distribution interval, is the particle hit frequency corresponding to the i-th distribution interval; The particle standard deviation corresponding to the i-th distribution interval is calculated using the following formula: ;in, is the average diameter of particles in the i-th distribution interval, is the center value of the interval particle distribution in the i-th distribution interval, is the corresponding particle hit frequency in the i-th distribution interval.

7. The method according to claim 5, characterized in that The original particle size measurement data also includes: the minimum particle diameter and the maximum particle diameter in each distribution interval; The particle diameter distribution range corresponding to the i-th distribution interval is calculated using the following formula: ;in, is the maximum particle diameter in the i-th distribution interval, is the minimum particle diameter in the i-th distribution interval.

8. The method according to claim 1, characterized in that After collecting the screen attributes in the candidate screen samples, the method further includes: The ink particle data and the screen properties are stored in a dynamic data space, and corresponding update parameters are configured to achieve dynamic update; The dynamic data space is used to dynamically store historical measurement data and historical analysis data of a variety of candidate screen samples and ink samples, so as to provide the mesh matching model with screen mesh prediction analysis.

9. The method according to claim 1, characterized in that: The mesh matching model at least includes the following structures: a preprocessing layer, a derivative layer, a prediction layer, and a fusion layer; The combining of the ink particle data and the screen properties and predicting the target screen mesh whose matching degree with the target ink sample reaches a preset condition through a mesh matching model comprises: Extracting the ink features to be analyzed and the screen features to be analyzed from the ink particle data and the screen properties through the preprocessing layer, and projecting the ink features to be analyzed and the screen features to be analyzed into the corresponding feature conversion space to obtain a corresponding first matching feature map; Through the derivative layer, a derivative prediction branch matching the ink feature to be analyzed is selected, and the first matching feature map is predicted through the derivative prediction branch to obtain a first predicted screen mesh count; the derivative prediction branch is constructed based on a derivative knowledge base matching the ink attribute characteristics; The prediction layer predicts the ink particles and the screen mesh count in the first matching feature map to obtain a second predicted screen mesh count; the processing method of the prediction layer is based on preset conditions, candidate screen samples, and historical measurement data and historical analysis data of ink samples; The first predicted screen mesh count and the second predicted screen mesh count are adaptively fused through a fusion layer to obtain the target screen mesh count.

10. The method according to claim 9, characterized in that The prediction layer predicts the ink particles and the screen mesh count in the first matching feature map to obtain a second predicted screen mesh count, including: A preset convolution operation is performed on the first matching feature map to obtain a second matching feature map; the acquisition process of the second matching feature map is expressed as follows: ； in, is the first matching feature map, is the dilated convolution kernel, Indicates the processing method of the convolution kernel in the feature channel, is the position index value of the convolution kernel in the horizontal and vertical directions, n represents the nth convolution kernel, is the void ratio, is the step length of the dilated convolution kernel operation, is the convolution kernel size, is the second matching feature map after the dilated convolution. represents the corresponding channel identifiers in the first matching feature map and the second matching feature map, and Respectively represent the corresponding row position index and column position index in the first matching feature map and the second matching feature map; The second predicted screen mesh count is obtained by performing a prediction based on the second matching feature map.

11. The method according to claim 1, characterized in that: The generating of screen recommendation information based on the target screen mesh count includes: identifying the recommended level of information refinement required for the printing process to be performed; Determine a target recommendation mode based on the level of refinement of the recommendation information; the target recommendation mode includes a precise mode or a compatible mode; The screen recommendation information is generated based on the target recommendation mode; wherein, in the precise mode, the screen recommendation information includes the best screen mesh matching the target ink sample as the target screen mesh; and in the compatible mode, the screen recommendation information includes target screen meshes in multiple dynamic adjustment ranges obtained by interval search and / or dynamic programming.

12. A screen printing machine, characterized in that: The screen printing machine comprises at least the following units, wherein An acquisition module, used for acquiring an ink sample to be analyzed as a target ink sample; A first acquisition module, used for acquiring ink particle data in the target ink sample; The ink particle data at least includes: particle distribution data, toner ratio, glass powder ratio, toner type; The second collection module is used to collect the screen properties of the candidate screen sample; the screen properties at least include: the screen material and screen mesh number of the candidate screen sample; A prediction module, for combining the ink particle data with the screen properties, and predicting a target screen mesh whose matching degree with the target ink sample meets a preset condition through a mesh matching model; The recommendation module is used to generate screen recommendation information based on the target screen mesh count and send the screen recommendation information to the user to assist the user in completing the screen selection process.

13. An electronic device, characterized in that: include: Memory for storing computer software programs; A processor is used to read and execute the computer software program, thereby implementing the screen mesh recommendation method based on screen printing as described in any one of claims 1 to 11.

14. A non-transitory computer-readable storage medium, characterized in that: The storage medium stores a computer software program, and when the computer software program is executed by the processor, the method for recommending screen mesh count based on screen printing according to any one of claims 1 to 11 is implemented.

Images