A photovoltaic cell printing screen burst early warning system and method

Abstract

The application discloses a photovoltaic cell printing screen burst early warning system and method. The system comprises: an image detection module for periodically collecting images of the printing screen to generate real-time screen images; a tension detection module for periodically detecting the tension of multiple points on the printing screen to generate screen tension values at the points; a first displacement detection module for periodically detecting the height of the four corners of the printing screen to generate screen corner displacement amounts; a second displacement detection module for periodically detecting the depth of the down pressure of the scraper to generate scraper down pressure depth values; and a screen burst prediction module for generating a screen burst risk probability and a remaining service life according to the real-time screen images, the screen tension values at the points, the screen corner displacement amounts and the scraper down pressure depth values. The application realizes early identification of screen degradation precursor characteristics before screen burst occurs, enables the screen to be safely taken offline when the printing life reaches the maximum value, and avoids slurry loss, equipment pollution and production line downtime losses caused by screen burst.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic cell technology, and in particular to a system and method for early warning of printing plate explosion in photovoltaic cell printing screens. Background Technology

[0002] In the photovoltaic cell manufacturing process, screen printing is a critical step that determines cell conversion efficiency and production costs. The printing screen, as a precision mold for transferring conductive pastes (such as silver or aluminum paste), directly affects print quality. With the accumulation of printing cycles, the screen gradually exhibits fatigue deterioration phenomena such as decreased screen tension, mesh geometric deformation, edge warping, and the propagation of micro-cracks, eventually leading to sudden screen bursting (i.e., "screen bursting"). When screen bursting occurs, a large amount of paste spills out, resulting in a loss of 1-2 kg of paste (costing several thousand yuan) in a single incident. Simultaneously, it contaminates the printing table, squeegee, and subsequent cells, leading to large-scale waste and requiring machine shutdown for cleaning and screen replacement, severely impacting production efficiency and yield.

[0003] Current screen printing plate burst monitoring technology employs an inductive sensor at the initial position of the print head and a light curtain sensor between the screen and the printing table, connected in series to form a trigger circuit. When screen bursts, screen fragments enter the gap between the screen and the table, triggering the light curtain sensor. Simultaneously, the print head, losing proper support, undergoes abnormal displacement, triggering the inductive sensor. The signals from both sensors, connected in series, send an alarm to the control system and stop the printing press. This technology requires bursting and fragments to have already splattered, and the sensor response depends on damage to the screen's physical structure. Therefore, even with a short time between bursting and shutdown, ink leakage is unavoidable, leading to irreversible material loss and equipment contamination. Furthermore, this technology does not monitor early signs of progressive degradation such as screen tension decay, line deformation, edge warping, and microcrack propagation, lacking predictive capabilities and failing to guide operators to proactively replace the screen before bursting occurs, thus failing to maximize the actual lifespan of the screen. Furthermore, this technology outputs binary alarm signals, which do not contain detailed information such as risk level, remaining lifespan, and fault location, and cannot support advance scheduling of maintenance plans or full lifecycle trend analysis of network plate status.

[0004] In summary, existing webpage overload monitoring technologies are essentially reactive, lacking predictive capabilities, trend analysis, and remaining lifespan estimation functions, and thus cannot achieve early warning and accurate replacement. Summary of the Invention

[0005] This invention provides a screen printing stencil bursting early warning system and method for photovoltaic cell printing stencils. It enables early identification of screen deterioration precursors before bursting occurs, allowing the screen to be safely removed from the production line when it reaches its maximum printing lifespan. This fundamentally avoids ink loss, equipment contamination, and production line downtime losses caused by bursting.

[0006] In a first aspect, the present invention provides a screen printing stencil explosion early warning system for photovoltaic cells, comprising: The image detection module is used to periodically acquire images of the printing screen and generate real-time screen images; The tension detection module is used to periodically detect the tension at multiple points on the printing screen and generate the screen tension value at each point. The first displacement detection module is used to periodically detect the height of the four corners of the printing screen and generate the displacement of the four corners of the screen. The second displacement detection module is used to periodically detect the downward pressure depth of the scraper and generate the scraper downward pressure depth value; The screen printing failure prediction module is connected to the image detection module, the tension detection module, the first displacement detection module, and the second displacement detection module, respectively. It is used to generate the probability of screen printing failure and the remaining service life based on the real-time screen printing image, the screen printing tension value at each point, the displacement of the four corners of the screen printing, and the depth of the squeegee.

[0007] Optionally, the printing screen includes a screen and a frame for fixing the screen; the screen includes an effective printing area; The image detection module includes an image sensor located above the printing screen, and the imaging area of ​​the image sensor covers the effective printing area. The tension detection module includes multiple tension sensors, all of which are located on the connection structure between the wire mesh and the frame. The first displacement detection module includes a plurality of first displacement sensors, which are located above the printing screen and their vertical projections are respectively located in the four corner areas of the printing screen. The second displacement detection module includes a second displacement sensor, which is located on the scraper holder.

[0008] Optionally, it also includes an early warning output module, which is connected to the peak page prediction module; the early warning output module is also communicatively connected to an external device. The "hot item prediction" module is also used to generate a risk level based on the probability of a hot item being predicted. The early warning output module: Used to execute the corresponding alarm action based on the risk level; It is also used to send webpage information to the external device; the webpage information includes the risk level, the probability of a webpage becoming overloaded, and the remaining service life.

[0009] Optionally, it may also include a data storage module; The data storage module is connected to the image detection module, the tension detection module, the first displacement detection module, the second displacement detection module, and the bursting prediction module, respectively, and is used to store the real-time screen printing image, the screen printing tension value at each point, the displacement of the four corners of the screen printing, the depth of the scraper, the probability of bursting risk, and the remaining service life.

[0010] Secondly, the present invention also provides a method for early warning of printing plate bursts in photovoltaic cell printing screens, which is executed using the printing plate burst warning system for photovoltaic cell printing screens as described in any one of the first aspects, wherein the printing plate burst warning method includes: Acquire real-time screen images periodically collected by the image detection module; Obtain the screen tension values ​​at various points periodically generated by the tension detection module; Obtain the displacement of the four corners of the screen periodically generated by the first displacement detection module; Obtain the scraper pressing depth value periodically generated by the second displacement detection module; Based on the real-time screen printing image, the screen printing tension value at each point, the displacement of the four corners of the screen printing, and the depth of the squeegee, the probability of screen bursting and the remaining service life are generated.

[0011] Optionally, based on the real-time screen printing image, the screen printing tension values ​​at each point, the displacement of the four corners of the screen printing, and the depth of the squeegee press, a probability of screen bursting and the remaining service life are generated, including: A multimodal fusion network based on an attention mechanism or a weighted stitching fusion method is used to fuse the real-time screen image, the screen tension value at each point, the displacement of the four corners of the screen, and the depth of the scraper to generate a risk feature vector. The risk feature vectors are arranged in chronological order to form a time series, which is then input into a time series prediction model trained based on historical high-risk data to generate the high-risk probability and the remaining lifespan.

[0012] Optionally, a multimodal fusion network based on an attention mechanism or a weighted stitching fusion method is used to fuse the real-time screen image, the screen tension values ​​at each point, the displacement of the four corners of the screen, and the depth of the scraper press to generate a risk feature vector, including: Feature extraction is performed on the real-time screen image to obtain line offset, warping height value, abnormal gray area and total crack length; Feature extraction is performed on the screen tension values ​​at each point to obtain the average screen tension value, screen attenuation rate, screen standard deviation, and local anomaly score at each point of the screen. Feature extraction is performed on the displacement of the four corners of the screen to obtain the flatness deviation of the screen; Based on an attention-based multimodal fusion network or a weighted splicing fusion method, the line offset, the warping height, the abnormal gray area, the total crack length, the average tension of the screen printing plate, the screen printing plate attenuation rate, the screen printing plate standard deviation, the local anomaly scores of each point on the screen printing plate, the screen printing plate flatness deviation, and the squeegee pressing depth are fused to generate the risk feature vector.

[0013] Optionally, feature extraction is performed on the real-time screen image to obtain line offset, warping height value, abnormal grayscale area, and total crack length, including: Calculate the pixel offset of each line in the real-time screen printing image and the screen printing reference image; Based on the first preset conversion coefficient, the pixel offset is converted into the actual size offset to obtain the line offset.

[0014] Optionally, feature extraction is performed on the real-time screen image to obtain line offset, warping height value, abnormal grayscale area, and total crack length, including: Based on the real-time screen image, a three-dimensional shape of the printing screen is established, and a current screen height map is generated; The current screen height map is compared with the reference screen height map to obtain the height difference of the edge grid lines of the printing screen, and the height difference is used as the warpage height value.

[0015] Optionally, feature extraction is performed on the real-time screen image to obtain line offset, warping height value, abnormal grayscale area, and total crack length, including: The grayscale difference is obtained by performing a difference operation between the real-time screen printing image and the screen printing reference image; Count the number of pixels in the region where the grayscale difference is greater than a preset grayscale threshold; The number of pixels is converted into the actual area according to the second preset conversion coefficient to obtain the abnormal grayscale area.

[0016] Optionally, feature extraction is performed on the screen tension values ​​at each point to obtain the average screen tension value, screen attenuation rate, screen standard deviation, and local anomaly scores at each point on the screen, including: The deviation between the screen tension value at each point and the preset reference value is calculated to obtain the local anomaly score for each point of the screen.

[0017] Optionally, feature extraction is performed on the displacement of the four corners of the screen to obtain the screen flatness deviation, including: Based on the displacement of the four corners of the screen, calculate the height difference between the two corners on the two diagonals of the printing screen. The maximum value among the height differences is taken as the screen flatness deviation.

[0018] The photovoltaic cell printing screen explosion early warning system and method provided by this invention utilizes an image detection module to acquire images of the printing screen, a tension detection module to detect the tension at multiple points on the printing screen, a first displacement detection module to detect the height of the four corners of the printing screen, and a second displacement detection module to detect the downward pressure depth of the squeegee. The generated real-time screen image, the screen tension value at each point, the displacement of the four corners of the screen, and the squeegee downward pressure depth value are input into the explosion prediction module for data fusion processing to obtain the explosion risk probability and remaining service life. This enables early identification of screen deterioration precursors before explosion occurs, allowing the screen to be safely removed from the production line when it reaches its maximum printing life, fundamentally avoiding pulp loss, equipment contamination, and production line downtime losses caused by explosion.

[0019] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a schematic diagram of the structure of a photovoltaic cell printing screen explosion early warning system provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of another photovoltaic cell printing screen explosion early warning system provided in an embodiment of the present invention; Figure 3 A top view schematic diagram of a printing screen provided in an embodiment of the present invention; Figure 4 A flowchart illustrating a method for early warning of printing plate bursts in photovoltaic cells, provided in an embodiment of the present invention; Figure 5 A flowchart illustrating another method for early warning of printing plate explosion in photovoltaic cells provided in an embodiment of the present invention; Figure 6 This is a flowchart illustrating another method for early warning of printing plate explosion in photovoltaic cells, provided as an embodiment of the present invention. Detailed Implementation

[0022] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, the accompanying drawings show only the parts relevant to the present invention, and not all of the structures.

[0023] The terminology used in the embodiments of this invention is for the purpose of describing specific embodiments only and is not intended to limit the invention. It should be noted that directional terms such as "upper," "lower," "left," and "right" described in the embodiments of this invention are used to describe the angles shown in the accompanying drawings and should not be construed as limiting the embodiments of this invention. Furthermore, in the context, it should be understood that when referring to an element being formed "on" or "below" another element, it can be formed not only directly on or below the other element, but also indirectly on or below it through intermediate elements. The terms "first," "second," etc., are used for descriptive purposes only and do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0024] The term "comprising" and its variations as used in this invention are open-ended, meaning "including but not limited to". The term "based on" means "at least partially based on". The term "one embodiment" means "at least one embodiment".

[0025] It should be noted that the concepts of "first" and "second" mentioned in this invention are only used to distinguish the corresponding contents and are not used to limit the order or interdependence.

[0026] It should be noted that the terms "a" and "a plurality of" used in this invention are illustrative rather than restrictive. Those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0027] Figure 1 This is a schematic diagram of a screen printing stencil explosion early warning system for photovoltaic cells, provided in an embodiment of the present invention. Figure 1As shown, the screen printing stencil bursting early warning system includes an image detection module 1, a tension detection module 2, a first displacement detection module 3, a second displacement detection module 4, and a bursting prediction module 5. The image detection module 1 periodically acquires images of the printing stencil to generate real-time stencil images; the tension detection module 2 periodically detects the tension at multiple points on the printing stencil to generate tension values ​​at each point; the first displacement detection module 3 periodically detects the height of the four corners of the printing stencil to generate corner displacement values; the second displacement detection module 4 periodically detects the squeegee's downward pressure depth to generate squeegee downward pressure depth values; the bursting prediction module 5 is connected to the image detection module 1, tension detection module 2, first displacement detection module 3, and second displacement detection module 4, and is used to generate the bursting risk probability and remaining service life based on the real-time stencil images, tension values ​​at each point, corner displacement values, and squeegee downward pressure depth values.

[0028] Specifically, when printing photovoltaic cells using a printing screen, the squeegee repeatedly presses down and scrapes across the surface of the printing screen, causing the screen to be subjected to periodic stretching and friction, resulting in deformation of the printing screen. Therefore, after multiple printings, the printing screen may experience fatigue deterioration phenomena such as screen tension attenuation, grid line geometric deformation, edge warping, or microcrack propagation, leading to screen bursting.

[0029] Based on this, refer to Figure 1 The image detection module 1 can periodically acquire images of the printing screen to generate real-time screen images. This acquisition cycle can be performed after each printing cycle or after multiple printing cycles, for example, 10-50 times. Furthermore, by analyzing the real-time screen images, the line offset, edge warping height, abnormal grayscale area, and total crack length on the printing screen can be determined. Line offset refers to the degree of geometric positional shift of the grid lines during printing due to uneven screen tension distribution. Excessive line offset may lead to grid breakage or uneven line width. Edge warping height refers to the localized lifting or sinking of the outermost grid lines due to screen fatigue or emulsion layer aging. Excessive edge warping height may cause uneven edge printing pressure, leading to screen bursting. Abnormal grayscale area refers to areas of uneven grayscale distribution on the printing screen surface caused by ink penetration, doctor blade wear, or foreign matter adhesion. The presence of abnormal grayscale areas indicates possible ink leakage, contamination, or wear. The total crack length refers to the extension length of cracks generated by repeated stress on the screen or emulsion layer. When the total crack length increases, it indicates that the printing screen is about to burst.

[0030] Further, refer to Figure 1Simultaneously, the tension detection module 2 can periodically detect the tension at multiple points on the printing screen, generating the screen tension value for each point. Similarly, this detection cycle can be performed after each printing cycle, or it can be performed after multiple printing cycles. It is understood that when detecting the tension at multiple points on the printing screen, these points should be evenly distributed across the entire area of ​​the printing screen, rather than concentrated in a single area. This allows for the determination of the average screen tension, screen attenuation rate, screen standard deviation, and local anomaly scores for each point based on the screen tension values ​​at each point. The average screen tension value refers to the arithmetic mean of the screen tension values ​​at each point, reflecting the overall tension of the printing screen. A low average screen tension value may cause the printing screen to loosen, leading to increased grid line deformation during printing and a higher risk of screen breakage. Screen attenuation rate refers to the decrease in average tension value per unit printing pass. It reflects the rate at which the tension of the printing screen decreases. When the attenuation rate is too high, it indicates accelerated fatigue of the printing screen, which may lead to screen bursting. Screen standard deviation refers to the standard deviation of the screen tension value at each point, reflecting the spatial uniformity of the printing screen tension. An excessively large screen standard deviation may lead to local tension concentration, easily causing cracks at that location. Local anomaly score at each point of the screen refers to the deviation of the screen tension value at each point from the preset reference value. It is used to quantify the degree of local anomaly in the tension at each point. An excessively high local anomaly score indicates that the screen at that point has experienced a sudden tension change, which is prone to becoming the initiation point of screen bursting.

[0031] Further, refer to Figure 1 Simultaneously, the first displacement detection module 3 can periodically detect the height of the four corners of the printing screen, generating the displacement of the four corners. Similarly, this detection cycle can be performed after each printing cycle, or it can be performed after multiple printing cycles. After obtaining the displacement of the four corners, the flatness deviation of the printing screen can be further determined based on the displacement. The flatness deviation refers to the degree of height difference between the four corners of the printing screen. Excessive flatness deviation will lead to uneven printing pressure distribution, local stress concentration, and is very likely to cause screen bursting and printing defects.

[0032] Further, refer to Figure 1 Simultaneously, the second displacement detection module 4 can periodically detect the squeegee's downward pressure depth and generate a squeegee downward pressure depth value. This detection cycle can be performed on the squeegee's downward pressure depth during each printing process. The squeegee downward pressure depth value refers to the distance the squeegee moves downward from its initial position during the printing process, representing the depth to which the squeegee presses into the surface of the printing screen. Excessive downward pressure depth will cause the screen to bear excessive tensile stress, thereby accelerating screen fatigue and emulsion layer wear, and even directly tearing the screen and causing it to burst.

[0033] Further, refer to Figure 1 The screen printing failure prediction module 5 is connected to the image detection module 1, tension detection module 2, first displacement detection module 3, and second displacement detection module 4, respectively. It receives real-time screen printing images, screen tension values ​​at various points, screen corner displacements, and squeegee depth values. It then fuses these data to obtain line offset, edge warping height, abnormal grayscale area, total crack length, average screen tension, screen attenuation rate, screen standard deviation, local anomaly scores at various points on the screen, screen flatness deviation, and squeegee depth values. Based on this, it generates the screen printing failure risk probability and remaining service life. The screen printing failure risk probability P can be any value between 0 and 1. A higher value indicates a greater probability of screen burn-out. Remaining lifespan represents the number of printing cycles the screen has remaining; more remaining cycles mean a lower probability of burn-out. Based on this, operators can determine after which to replace the screen based on the probability of burn-out and remaining lifespan, thus preventing burn-out during printing and providing an early warning function.

[0034] It should be noted that the acquisition cycle of the image detection module, the detection cycle of the tension detection module, the detection cycle of the first displacement detection module, and the detection cycle of the second displacement detection module can be the same or different. That is, the image detection module, tension detection module, first displacement detection module, and second displacement detection module can detect the printing screen according to their respective preset sampling frequencies, which can be the same or different. In this case, the processing frequency for data fusion and calculation by the burst prediction module can be the greatest common divisor of the sampling frequencies, and correspondingly, the processing cycle is the least common multiple of the detection cycles. Within a processing cycle, if a module has no newly acquired data in that cycle, the most recently acquired data from that module can be used as the input for the current cycle.

[0035] This invention utilizes an image detection module to acquire images of the printing screen, a tension detection module to detect the tension at multiple points on the printing screen, a first displacement detection module to detect the height of the four corners of the printing screen, and a second displacement detection module to detect the depth of the squeegee's downward pressure. The generated real-time screen image, the screen tension values ​​at each point, the displacement of the four corners of the screen, and the squeegee's downward pressure value are input into a bursting prediction module for data fusion processing to obtain the probability of bursting risk and the remaining service life. This enables the early identification of screen deterioration precursors before bursting occurs, allowing the screen to be safely removed from the production line when it reaches its maximum printing life, fundamentally avoiding pulp loss, equipment contamination, and production line downtime losses caused by bursting.

[0036] Furthermore, the screen breakage prediction module 5 can also compare the scraper pressing depth value detected by the second displacement detection module 4 with a preset depth threshold. When the scraper pressing depth value exceeds the threshold, it indicates that the stretching of the screen by the scraper has entered a dangerous range. At this time, the screen tension is often close to the point of sudden change, which can easily cause local abnormalities or even screen breakage. In this case, the screen breakage prediction module 5 can output an over-threshold alarm signal separately.

[0037] Optionally, Figure 2 This is a schematic diagram of another photovoltaic cell printing screen explosion early warning system provided in an embodiment of the present invention. Figure 3 This is a top view schematic diagram of a printing screen provided in an embodiment of the present invention, with reference to the above. Figures 1-3 The printing screen 6 includes a screen 61 and a frame 62 for fixing the screen 61. The screen 61 includes an effective printing area. The image detection module 1 includes an image sensor 11, which is located above the printing screen 6 and its imaging area covers the effective printing area. The tension detection module 2 includes multiple tension sensors 21, all located on the connection structure between the screen 61 and the frame 62. The first displacement detection module 3 includes multiple first displacement sensors 31, which are located above the printing screen 6 and their vertical projections are located in the four corner areas of the printing screen 6. The second displacement detection module 4 includes a second displacement sensor 41, which is located on the squeegee holder.

[0038] Specifically, when printing on the photovoltaic cell 10 using the printing screen 6, the photovoltaic cell 10 is located on the machine base 9, the printing screen 6 is located above the photovoltaic cell 10, and the squeegee 8 is located on the squeegee holder 7 and is movable. During the printing process, the squeegee 8 presses down on the printing screen 6, causing the screen 61 to elastically stretch, squeezing and transferring the ink through the grid lines to the surface of the photovoltaic cell 10 below, thus completing the printing. The area of ​​the screen 61 where the ink can be printed onto the photovoltaic cell 10 is the effective printing area, and other areas are ineffective printing areas.

[0039] refer to Figures 1-3 The image detection module 1 includes an image sensor 11, which is located above the printing screen 6. The imaging area of ​​the image sensor 11 covers the effective printing area. For example, the image sensor 11 can be a global shutter CMOS industrial camera with a resolution of ≥5 million pixels and a frame rate of ≥30fps, equipped with a high-uniformity coaxial light source. It can also be a CCD camera or a line scan camera. The image sensor 11 periodically acquires images of the printing screen 6, generates real-time screen images, and can transmit the real-time screen images in a compressed format to the printout prediction module 5.

[0040] Further, refer to Figures 1-3 The tension detection module 2 includes multiple tension sensors 21, all located on the connection structure between the screen 61 and the frame 62. Each tension sensor 21 can detect the tension of the screen 61 at its location and generate a tension value for each point. To ensure that the tension value at each point accurately reflects the tension experienced by the entire printing screen 6, the multiple tension sensors 21 can be evenly distributed across various areas of the printing screen 6. For example, the tension sensors 21 can be strain gauge-type miniature pressure sensors or piezoresistive miniature pressure sensors with an accuracy ≥0.05N. Figure 3 Eight tension sensors 21 can be arranged, located at the four corners and the midpoints of the four sides of the printing screen 6. Based on the tension values ​​detected by the tension sensors 21 at the corners and midpoints, the tension value of the central area of ​​the printing screen 6 can be calculated using a simplified mechanical model. Thus, the tension values ​​generated by the multiple tension sensors 21 are input into the burst prediction module 5, which can then determine the tension on the entire printing screen 6. It should be noted that this application does not limit the number or location of the tension sensors 21; the number can be increased to 16 or reduced to 5. Furthermore, when contact tension sensors 21 cannot be arranged on the printing screen 6, non-contact laser tension detection heads can be used. The tension sensors 21 can be fiber optic strain sensors, non-contact laser Doppler vibrometers, or ultrasonic tension detection devices.

[0041] Further, refer to Figures 1-3 The first displacement detection module 3 includes multiple first displacement sensors 31, which are located above the printing screen 6, and their vertical projections are respectively located in the four corner areas of the printing screen 6. Each first displacement sensor 31 is used to detect the distance of the corner position area of ​​the printing screen 6 directly below it relative to itself, thereby obtaining the height value of the corner position area, generating the screen corner displacement, and inputting the screen corner displacement to the screen breakage prediction module 5. Meanwhile, all the first displacement sensors 31 are installed at the same horizontal height, so when the screen corner displacements are inconsistent, it indicates that the corresponding corner position area of ​​the printing screen 6 has shifted. For example, the first displacement sensor 31 can be a laser displacement sensor, a high-precision inductive displacement sensor, a capacitive displacement sensor, or a magnetostrictive displacement sensor, with an accuracy ≤1μm. It should be noted that, to improve the accuracy of the screen breakage prediction module 5 in determining the flatness deviation of the printing screen 6, the first displacement sensor 31 can also be set in the center position area of ​​the four sides of the printing screen 6; this application does not limit this.

[0042] Further, refer to Figures 1-3The second displacement detection module 4 includes a second displacement sensor 41, which is located on the squeegee holder. When the squeegee 8 presses down on the printing screen 6, the second displacement sensor 41 can detect the moving distance of the squeegee relative to itself, thereby generating a squeegee pressing depth value, which is then input into the screen breakage prediction module 5. For example, the second displacement sensor 41 can be a laser displacement sensor or a high-precision inductive displacement sensor. Alternatively, the second displacement sensor 41 can be integrated into the servo drive system of the squeegee 8, indirectly calculating the squeegee pressing depth value through an electrode encoder.

[0043] This invention achieves comprehensive detection of printing screen image, tension, flatness, and squeegee status by setting image sensors, multiple tension sensors, multiple first displacement sensors, and second displacement sensors in appropriate locations.

[0044] Optionally, the overprint warning system also includes a warning output module, which is connected to the overprint prediction module 5; the warning output module is also communicatively connected to external devices. The overprint prediction module is also used to generate a risk level based on the probability of an overprint risk. The warning output module is used to execute corresponding alarm actions based on the risk level, and is also used to send overprint information to external devices, including the risk level, the probability of an overprint risk, and the remaining service life.

[0045] Specifically, when the bursting prediction module 5 determines the bursting risk probability and remaining service life based on the real-time screen image generated by the sensor, the screen tension value at each point, the displacement of the four corners of the screen, and the squeegee pressing depth, the bursting risk probability and remaining service life will be inconsistent if any of these values ​​change. For the same printing screen 6, the bursting risk probability and remaining service life will also be different after different numbers of uses. Therefore, the bursting prediction module 5 can generate a risk level based on the bursting risk probability so that operators can have a more intuitive understanding of the bursting risk.

[0046] The risk level can be divided into multiple levels according to actual needs. The probability threshold for the bursting risk corresponding to different levels can also be adaptively adjusted according to the printing screen type, paste characteristics, and production line requirements. This application does not limit these aspects. In a specific embodiment, the risk level can be divided into low risk, medium risk, high risk, and extremely high risk. When the bursting risk probability P < 0.4, the risk level is low risk; when the bursting risk probability P satisfies 0.4 ≤ P < 0.7, the risk level is medium risk; when the bursting risk probability P satisfies 0.7 ≤ P < 0.9, the risk level is high risk; and when the bursting risk probability P ≥ 0.9, the risk level is extremely high risk.

[0047] Furthermore, the early warning output module is connected to the overprint prediction module 5. The early warning output module can execute corresponding alarm actions according to the risk level. For example, the early warning output module includes an LED indicator and a buzzer. When the overprint prediction module 5 determines that the risk level is low, the LED indicator lights up green and the buzzer does not sound, indicating that it can be used normally. When the overprint prediction module 5 determines that the risk level is medium, the LED indicator lights up yellow and the buzzer sounds at the first frequency, indicating that attention is needed. When the overprint prediction module 5 determines that the risk level is high, the LED indicator lights up orange and the buzzer sounds at the second frequency, indicating that replacement is imminent. When the overprint prediction module 5 determines that the risk level is extremely high, the LED indicator lights up red and the buzzer sounds at the third frequency, indicating that immediate shutdown and replacement are required.

[0048] Furthermore, the early warning output module also communicates with external devices to send screen printing information, including risk level, probability of screen bursting, and remaining lifespan. It can also send the location and characteristics of abnormal areas, such as abnormal tension attenuation in the lower left corner and abnormal line deformation, based on sensor settings. Additionally, it can send the screen printing number so that operators can associate the abnormality with the printing screen. For example, external devices can be workshop monitoring centers, on-site dashboards, management personnel's mobile phones, manufacturing execution systems, etc., and communication methods can be industrial Ethernet or wireless communication networks, such as Profinet, 5G, or WiFi 6. Based on this, the early warning output module can also communicate with the operator's wearable wristband to issue alarms via vibration.

[0049] Optionally, the screen printing malfunction warning system also includes a data storage module, which is connected to the image detection module, tension detection module, first displacement detection module, second displacement detection module, and malfunction prediction module, respectively. The data storage module is used to store real-time screen printing images, screen printing tension values ​​at each point, screen printing corner displacement, scraper pressing depth, malfunction risk probability, and remaining service life.

[0050] Specifically, the data storage module can create an independent file for each printing screen, containing real-time screen images throughout its entire lifecycle, screen tension values ​​at various points, screen corner displacement, squeegee depth, burst risk probability, and remaining lifespan. This supports the burst prediction module in iterating through the data, making its predictions more accurate and easier for operators to find when determining the burst risk probability and remaining lifespan based on real-time screen images, screen tension values ​​at various points, screen corner displacement, and squeegee depth.

[0051] This invention generates risk levels based on the probability of screen printing failure and executes alarm actions corresponding to those risk levels through an early warning output module. This allows operators to more intuitively understand the status of the printing screen and guides them to proactively replace the screen at the most economical time, effectively reducing the risk of screen printing failure and preventing premature or delayed shutdown. Simultaneously, a data storage module independently stores various data points throughout the entire lifecycle of each printing screen, providing quantitative evidence for process optimization and supplier quality evaluation. It also provides a basis for data iteration in the screen printing failure prediction module, thereby improving the accuracy of the prediction results.

[0052] Based on the same inventive concept, this invention also provides a method for early warning of printing plate explosion in photovoltaic cell printing screens. Figure 4 This is a flowchart illustrating a method for early warning of printing plate bursts in photovoltaic cells, provided by an embodiment of the present invention. Figure 4 As shown, the methods for issuing warnings for pages that have been flooded include: S101. Obtain real-time screen images periodically collected by the image detection module.

[0053] Specifically, when printing photovoltaic cells using a printing screen, the squeegee repeatedly presses down and scrapes across the surface of the printing screen, causing the screen to be subjected to periodic stretching and friction, resulting in deformation of the printing screen. Therefore, after multiple printings, the printing screen may experience fatigue deterioration phenomena such as screen tension attenuation, grid line geometric deformation, edge warping, or microcrack propagation, leading to screen bursting.

[0054] Based on this, real-time screen printing images can be acquired periodically by the image detection module. The acquisition cycle of the image detection module can be either acquiring images of the printing screen after each printing cycle, or acquiring images of the printing screen after multiple printing cycles. For example, multiple acquisition cycles can be 10-50 times. Furthermore, by analyzing the real-time screen printing images, the line offset, edge warping height, abnormal grayscale area, and total crack length on the printing screen can be determined at that time.

[0055] S102. Obtain the tension values ​​of the screen printing plate at each point periodically generated by the tension detection module.

[0056] Simultaneously, the tension values ​​of each point on the screen, periodically generated by the tension detection module, can be obtained. Similarly, this detection cycle can be performed after each printing cycle, or it can be performed after multiple printing cycles. It is understood that when detecting the tension at multiple points on the printing screen, these points should be evenly distributed across the entire area of ​​the screen, rather than concentrated in a single area. This allows for the determination of the average screen tension, screen attenuation rate, screen standard deviation, and local anomaly scores at each point based on the tension values ​​at each point.

[0057] S103. Obtain the displacement of the four corners of the screen periodically generated by the first displacement detection module.

[0058] Simultaneously, the screen corner displacements periodically generated by the first displacement detection module can be obtained. Similarly, this detection cycle can be either detecting the height of the four corners of the printing screen after each printing cycle, or detecting the height of the four corners of the printing screen again after multiple printing cycles. After obtaining the screen corner displacements, the flatness deviation of the printing screen can be further determined based on these displacements.

[0059] S104. Obtain the scraper pressing depth value periodically generated by the second displacement detection module.

[0060] Simultaneously, the squeegee pressing depth value generated periodically by the second displacement detection module can be obtained. This detection cycle allows for the detection of the squeegee pressing depth during each printing operation. The squeegee pressing depth value refers to the distance the squeegee moves downward from its initial position during the printing process, representing the depth to which the squeegee presses into the surface of the printing screen. Excessive pressing depth can cause the screen to bear excessive tensile stress, thereby accelerating screen fatigue and emulsion layer wear, and even directly tearing the screen, leading to screen bursting.

[0061] S105. Based on the real-time screen printing image, the screen tension value at each point, the displacement of the four corners of the screen printing, and the depth of the scraper, generate the probability of screen bursting and the remaining service life.

[0062] Specifically, the screen breakage prediction module receives real-time screen images, screen tension values ​​at various points, screen corner displacements, and squeegee depth values. It then fuses these data to obtain line offset, edge warping height, abnormal grayscale area, total crack length, average screen tension, screen attenuation rate, screen standard deviation, local anomaly scores at various points on the screen, screen flatness deviation, and squeegee depth values. Based on this, it generates the screen breakage risk probability and remaining service life. The screen breakage risk probability P can be any value between 0 and 1. A higher value indicates a greater probability of screen burn-out. Remaining lifespan represents the number of printing cycles the screen has remaining; more remaining cycles mean a lower probability of burn-out. Based on this, operators can determine after which to replace the screen based on the probability of burn-out and remaining lifespan, thus preventing burn-out during printing and providing an early warning function.

[0063] This invention acquires real-time screen printing images generated by an image detection module, screen printing tension values ​​at various points generated by a tension detection module, screen printing corner displacements generated by a first displacement detection module, and squeegee pressing depth values ​​generated by a second displacement detection module. The acquired data is then fused to generate the probability of screen bursting and the remaining lifespan. This allows for the early identification of screen deterioration precursors before bursting occurs, ensuring the screen is safely removed from the printing line when it reaches its maximum printing lifespan. This fundamentally avoids ink loss, equipment contamination, and production line downtime losses caused by screen bursting.

[0064] Optionally, Figure 5 This is a flowchart illustrating another method for early warning of screen breakage in photovoltaic cell printing screens provided by an embodiment of the present invention. This embodiment is a refinement of the above embodiment. Specifically, for step S105, generating the probability of screen breakage risk and the remaining service life based on the real-time screen image, the screen tension value at each point, the displacement of the four corners of the screen, and the squeegee pressing depth value, it can be further refined as follows: A multimodal fusion network based on an attention mechanism or a weighted stitching fusion method is used to fuse real-time screen printing images, screen printing tension values ​​at each point, screen printing corner displacements, and squeegee pressing depth values ​​to generate a risk feature vector. The risk feature vectors are arranged in chronological order to form a time series, which is then input into a time series prediction model trained based on historical high-risk data to generate the probability of high-risk data and the remaining lifespan.

[0065] For parts not described in detail in this embodiment, please refer to the foregoing embodiments, such as... Figure 5 As shown, the method for issuing early warnings of overloaded content provided in this embodiment includes: S201. Obtain real-time screen images periodically collected by the image detection module.

[0066] S202. Obtain the tension values ​​of the screen printing plate at each point periodically generated by the tension detection module.

[0067] S203. Obtain the displacement of the four corners of the screen generated periodically by the first displacement detection module.

[0068] S204. Obtain the scraper pressing depth value periodically generated by the second displacement detection module.

[0069] S205. A multimodal fusion network based on an attention mechanism or a weighted stitching fusion method is used to fuse real-time screen printing images, screen printing tension values ​​at each point, screen printing corner displacements, and squeegee pressing depth values ​​to generate a risk feature vector.

[0070] Specifically, after acquiring real-time screen printing images, screen tension values ​​at various points, screen corner displacements, and scraper depth, a multimodal fusion network based on an attention mechanism or a weighted splicing fusion method can be used to fuse multiple data points and generate a risk feature vector. This risk feature vector can be a fixed-dimensional vector, meaning its length remains constant regardless of changes in the real-time screen printing image, screen tension values ​​at various points, screen corner displacements, and scraper depth. The attention-based multimodal fusion network can highlight the features most relevant to the screen printing failure in the current state through adaptive weights, while the weighted splicing fusion method can quickly integrate multimodal information using preset weights; both can effectively characterize the health status of the screen printing. In an optional embodiment, multi-source data fusion based on Kalman filtering, probabilistic fusion based on Bayesian networks, or decision-level fusion based on DS evidence theory can also be used to fuse the sensor-detected data.

[0071] S206. The risk feature vectors are arranged in chronological order to form a time series, which is then input into a time series prediction model trained based on historical high-risk data to generate the probability of high-risk data and the remaining lifespan.

[0072] Specifically, since risk is not an instantaneous event but a continuously evolving process, the risk feature vectors generated in each calculation cycle are arranged in chronological order to form a time series of risk feature vectors. This time series reflects the evolution trajectory of the screen printing plate's health status from its initial use to the present moment. Furthermore, this time series is input into a pre-trained time series prediction model. The time series prediction module can output the probability of screen printing plate failure and its remaining lifespan. The probability of failure ranges from 0 to 1, representing the likelihood of the screen printing plate failing within a certain period. The remaining lifespan represents the number of printing cycles the screen printing plate can still function normally. Thus, early warning of failure and prediction of remaining lifespan are achieved. For example, the time series prediction module can employ Long Short-Term Memory (LSTM), Transformer time series models, Temporal Convolutional Networks (TCN), or traditional machine learning models, including Random Forest, Gradient Boosting Tree, Support Vector Regression, and XGBoost.

[0073] This invention employs a multimodal fusion network with an attention mechanism or a weighted splicing fusion method to perform multimodal fusion of real-time screen images, screen tension values ​​at various points, screen corner displacements, and squeegee pressing depth values ​​to generate a risk feature vector. This risk feature vector is then input into a time series prediction model trained based on historical burst screen data, ultimately outputting the burst screen risk probability and the remaining lifespan in units of printing cycles. This achieves a leap from post-event burst screen alarms to pre-event risk probability and remaining lifespan prediction, enabling risk identification and tiered warnings within hours or thousands of printing cycles before a burst screen occurs. This can reduce burst screen-related slurry losses by more than 80%, reduce unplanned downtime, increase screen utilization by 20% to 30%, and save hundreds of thousands of yuan in costs per production line per year. Meanwhile, by using multimodal fusion to comprehensively assess the risk of grid failure, rather than relying on a single parameter threshold, the risk characteristics are more comprehensive and the predictions are more reliable. The time series model utilizes the trends, speeds and dependencies in historical grid failure data to quantify the grid degradation process, output the continuous probability of grid failure risk and remaining service life, support full life cycle management, maximize grid utilization, and reduce consumable costs and maintenance losses.

[0074] Optionally, Figure 6 This is a flowchart illustrating another method for early warning of screen printing defects in photovoltaic cells, provided by an embodiment of the present invention. This embodiment is a refinement of the above embodiment. Specifically, for step S205, the multimodal fusion network based on an attention mechanism or a weighted splicing fusion method, the real-time screen image, the screen tension value at each point, the displacement of the four corners of the screen, and the squeegee pressing depth value are fused to generate a risk feature vector. This can be further refined as follows: Feature extraction is performed on real-time screen printing images to obtain line offset, warping height value, abnormal gray area and total crack length; Feature extraction is performed on the screen tension values ​​at each point to obtain the average screen tension value, screen attenuation rate, screen standard deviation, and local anomaly scores at each point of the screen. Feature extraction is performed on the displacement at the four corners of the screen to obtain the flatness deviation of the screen. Based on an attention-based multimodal fusion network or weighted splicing fusion method, the risk feature vector is generated by fusing line offset, warping height, abnormal gray area, total crack length, average tension value of the screen, screen attenuation rate, screen standard deviation, local anomaly scores at various points of the screen, screen flatness deviation, and squeegee pressing depth.

[0075] For parts not described in detail in this embodiment, please refer to the foregoing embodiments, such as... Figure 6 As shown, the method for issuing early warnings of overloaded content provided in this embodiment includes: S301. Acquire real-time screen images periodically collected by the image detection module.

[0076] S302. Obtain the tension values ​​of the screen printing plate at each point periodically generated by the tension detection module.

[0077] S303. Obtain the displacement of the four corners of the screen periodically generated by the first displacement detection module.

[0078] S304. Obtain the scraper pressing depth value periodically generated by the second displacement detection module.

[0079] S305. Extract features from the real-time screen image to obtain line offset, warping height value, abnormal gray area and total crack length.

[0080] Specifically, after acquiring the real-time screen image, feature extraction can be performed to obtain line offset, warpage height, abnormal grayscale area, and total crack length. Line offset refers to the degree of geometric positional shift of the grid lines during printing due to uneven screen tension distribution. Excessive line offset may lead to grid breakage or uneven line width. Warpage height refers to the localized lifting or sinking of the outermost grid lines due to screen fatigue or emulsion layer aging. Excessive warpage height may cause uneven edge printing pressure, leading to screen bursting. Abnormal grayscale area refers to areas of uneven grayscale distribution on the screen surface caused by ink penetration, squeegee wear, or foreign matter adhesion. The presence of abnormal grayscale areas indicates potential ink leakage, contamination, or wear. Total crack length refers to the extension length of cracks generated by repeated stress on the screen or emulsion layer. Accumulated crack length indicates that the screen is about to burst.

[0081] In an optional embodiment, the line offset can be obtained by calculating the pixel offset of each line in the real-time screen image and the screen reference image, and then converting the pixel offset into the actual size offset according to the first preset conversion coefficient.

[0082] Specifically, before the printing screen is used for the first time, a reference image of the printing screen can be pre-stored. Then, after the printing screen is used, the real-time screen image is compared with the reference image to calculate the pixel offset of each line in the real-time screen image and the reference image. The pixel offset reflects the displacement of the lines caused by the deformation of the screen. Then, according to a first preset conversion coefficient, the pixel offset is converted into the actual size offset to obtain the line offset. The first preset conversion coefficient can be obtained by calibration using a calibration plate or known dimensional features of the screen.

[0083] In another alternative embodiment, a three-dimensional shape of the printing screen can be established based on the real-time screen image to generate a current screen height map; then the current screen height map is compared with the reference screen height map to obtain the height difference of the edge grid lines of the printing screen, and the height difference is used as the warp height value.

[0084] Specifically, before the printing screen is used for the first time, a reference screen height map can be pre-stored. Then, after the printing screen is used, the three-dimensional shape of the printing screen is reconstructed based on the real-time screen image, generating the current screen height map. Further, the current screen height map is compared with the reference screen height map, and the height difference between the two at the edge grid lines of the printing screen is calculated. This height difference is the warping height value, reflecting the degree of local rise or fall of the screen edge grid lines relative to their initial flat state during use. If the difference is too large, it indicates that the screen edge has warped and deformed, posing a risk of screen breakage. For example, the three-dimensional shape reconstruction can use the fringe projection method.

[0085] In another optional embodiment, a difference operation can be performed between the real-time screen image and the screen reference image to obtain a grayscale difference; then, the number of pixels in the region where the grayscale difference is greater than a preset grayscale threshold can be counted; finally, according to a second preset conversion coefficient, the number of pixels can be converted into the actual area to obtain the abnormal grayscale area.

[0086] Specifically, before the printing screen is used for the first time, a reference image of the screen can be pre-stored. Then, after the printing screen is used, a difference operation is performed between the real-time screen image and the reference image to obtain a difference image. The grayscale value of each pixel in the difference image is the grayscale difference between the real-time screen image and the reference image at that location, reflecting the degree of grayscale change on the surface at that location. Then, the number of pixels in areas where the grayscale difference is greater than a preset grayscale threshold is counted. Finally, according to a second preset conversion coefficient, the number of pixels is converted into actual area to obtain the abnormal grayscale area. The second preset conversion coefficient is a preset actual physical area corresponding to a unit pixel. The abnormal grayscale area reflects the severity of grayscale anomalies on the screen surface caused by ink leakage, ink contamination, or wear; the larger the abnormal area, the more severe the defects on the screen surface.

[0087] In another optional embodiment, crack identification can be performed on the non-grid area of ​​the real-time screen image to obtain preliminary cracks; then, the preliminary cracks that exist in multiple consecutive frames of the real-time screen image are determined as real cracks; finally, the sum of the pixel lengths of the real cracks with a length greater than or equal to a preset threshold is calculated to obtain the total crack length.

[0088] Specifically, after acquiring the real-time screen printing image, to avoid misidentifying the grid lines of the printing screen as cracks, the grid line regions and non-grid line regions in the real-time screen printing image are first identified, and crack identification is performed on the non-grid line regions of the real-time screen printing image to obtain preliminary cracks. Then, to avoid misidentifying random noise, dust, or scratches as cracks, multiple frames of images continuously acquired at the same location can be acquired, and the preliminary cracks appearing in each frame are confirmed as real cracks. Finally, the sum of the pixel lengths of real cracks with a length greater than or equal to a preset threshold is calculated to obtain the total crack length, where the preset threshold can be twice the grid line width. For example, crack identification can employ edge enhancement and skeletonization processing methods.

[0089] It should be noted that before extracting features from the real-time screen printing image to obtain line offset, warping height, abnormal grayscale area, and total crack length, the real-time screen printing image can be preprocessed. This includes using median filtering to remove noise, using histogram equalization to enhance contrast, and registering the real-time screen printing image with the screen printing reference image. This can effectively preserve edge information, improve the contrast of cracks and grayscale anomalies, and eliminate positional offset, thereby improving the accuracy and stability of feature extraction.

[0090] S306. Extract features from the screen tension values ​​at each point to obtain the average screen tension value, screen attenuation rate, screen standard deviation, and local anomaly scores at each point of the screen.

[0091] Specifically, after obtaining the screen tension values ​​at each point, feature extraction can be performed on these values ​​to obtain the average screen tension value, screen attenuation rate, screen standard deviation, and local anomaly scores for each point on the screen. The average screen tension value is the arithmetic mean of the screen tension values ​​at each point, reflecting the overall tension of the printing screen. A low average screen tension value may cause the screen to loosen, leading to increased grid line deformation during printing and a higher risk of screen breakage. The screen attenuation rate is the decrease in average tension value per unit of printing, reflecting the rate of tension decay. An excessively high decay rate indicates accelerated screen fatigue, potentially leading to screen breakage. The screen standard deviation is the standard deviation of the screen tension values ​​at each point, reflecting the spatial uniformity of the screen tension distribution. An excessively large standard deviation may cause localized tension concentration, easily leading to cracks at those locations. The local anomaly score of each point of the screen refers to the deviation of the screen tension value at each point from the preset reference value. It is used to quantify the degree of local anomaly of the tension at each point. If the local anomaly score is too high, it means that the screen has undergone a sudden change in tension at that point, which is prone to becoming the initiation point of screen cracking.

[0092] In an optional embodiment, the average screen tension value can be obtained by calculating the arithmetic mean of the screen tension values ​​at each point. Further, the screen attenuation rate is obtained by calculating the ratio of the change in the average screen tension value generated over multiple consecutive periods to the number of periods. Specifically, the average screen tension values ​​corresponding to multiple consecutive detection periods can be selected, and the change in the average tension value between the first and last periods can be calculated. This change is then divided by the number of selected periods to obtain the decrease in the average tension value per unit period, which is the screen attenuation rate. The screen attenuation rate is used to quantify the rate at which tension degrades with increasing printing cycles. A higher attenuation rate indicates faster screen aging and a shorter remaining usable lifespan. Further, the standard deviation of the screen tension values ​​at each point is calculated to obtain the screen standard deviation.

[0093] Further, in another optional embodiment, a local anomaly score for each point on the screen is obtained by calculating the deviation between the screen tension value at each point and a preset reference value. Specifically, the local anomaly score for each point on the screen is obtained by calculating the deviation between the actual tension value at each tension sensor point and the preset reference value. The preset reference value can be the global average of all points, the initial tension value at that point when it is online, or the average of the surrounding neighboring points. The larger the calculated deviation value, the higher the degree to which the tension at that point deviates from the normal distribution, and the more severe the local degradation. The local anomaly score for each point on the screen is used to accurately locate weak areas on the screen. An excessively high score indicates that there may be defects such as screen loosening, microcracks, or delamination at that location, which can easily become the starting point for screen breakage.

[0094] S307. Extract features from the four corner displacements of the screen printing plate to obtain the flatness deviation of the screen printing plate.

[0095] Specifically, after obtaining the displacement of the four corners of the screen, feature extraction can be performed on the displacement to obtain the screen flatness deviation. The screen flatness deviation refers to the degree of height difference between the four corners of the printing screen. Excessive flatness deviation will lead to uneven printing pressure distribution and local stress concentration, which can easily cause screen bursting and printing defects.

[0096] In an optional embodiment, the height difference between the two corners on the two diagonals of the printing screen is calculated based on the displacement of the four corners of the screen; then the maximum value of the height difference is taken as the flatness deviation of the screen.

[0097] Specifically, based on the displacement of the four corners of the screen measured by the first displacement sensor in the four corner areas, the height difference between the two corner positions along the two diagonals of the printing screen is calculated, i.e., the height difference of one set of diagonals and the height difference of another set of diagonals. Then, the larger of these two height differences is taken as the screen flatness deviation. This screen flatness deviation is used to quantify the degree of relative height difference between the four corner positions of the screen. The larger the deviation, the worse the screen flatness, the more serious the local sinking or warping, the more uneven the squeegee pressure distribution during printing, and the higher the risk of screen bursting.

[0098] S308. A multimodal fusion network based on an attention mechanism or a weighted splicing fusion method is used to fuse line offset, warping height, abnormal gray area, total crack length, average tension value of the screen, screen attenuation rate, screen standard deviation, local anomaly scores at various points of the screen, screen flatness deviation, and squeegee pressing depth to generate a risk feature vector.

[0099] Specifically, the image detection module, tension detection module, first displacement detection module, and second displacement detection module can detect the printing screen according to their respective preset sampling frequencies. These sampling frequencies can be the same or different. Furthermore, to effectively integrate the aforementioned multi-source data, the screen breakage prediction module can adopt a unified data processing cycle, which can be the greatest common divisor of the sampling frequencies. Within a processing cycle, if a module has no newly acquired data, the most recently acquired data from that module is used as the input for the current cycle.

[0100] Based on this, the line offset, warping height, abnormal gray area, total crack length, average tension value of the screen, screen attenuation rate, screen standard deviation, local anomaly score of each point of the screen, screen flatness deviation and scraper pressing depth value obtained in the current processing cycle are used as inputs. The risk feature vector is generated by using a multimodal fusion network based on attention mechanism or a weighted splicing fusion method.

[0101] S309. The risk feature vectors are arranged in chronological order to form a time series, which is then input into a time series prediction model trained based on historical high-risk data to generate the probability of high-risk data and the remaining lifespan.

[0102] This invention extracts targeted features from multi-source raw data such as images, tension, and displacement to obtain quantitative indicators reflecting different dimensions of the screen printing plate, including geometric deformation, three-dimensional warping, surface contamination, microcrack damage, overall and local tension status, flatness deviation, and scraper process parameters. These features are then aggregated into a unified risk feature vector through attention mechanisms or weighted splicing fusion methods, integrating multi-dimensional information such as visual, mechanical, and kinematic features. This overcomes the limitations of relying on a single sensor or feature to comprehensively assess the health status of the screen printing plate.

[0103] In one practical embodiment, the image sensor used in the printing plate burst warning system is a Sony IMX250 with a resolution of 2448×2048 and a frame rate of 75fps; the tension sensor is arranged in a 3×3 array with eight points, an accuracy of 0.05N, and a sampling frequency of once per printed battery cell; the first and second displacement sensors are both laser displacement sensors with an accuracy of 0.5μm. The printing plate is numbered #001, with dimensions of 210mm×210mm, an initial average tension of 28N, and tension deviation at each point within ±1N. The initial image template has been pre-stored in the system.

[0104] During the first 150,000 printing cycles of the screen printing plate, the bursting warning system monitored that all characteristic parameters remained stable, and the bursting risk probability remained below 0.3, without triggering any warnings. After the 150,000th printing cycle, analysis of images captured by the image sensor revealed a slight distortion of 0.06mm in a key grid line at the lower left corner of the screen printing plate. Simultaneously, the tension value measured by the tension sensor at the lower left corner decreased from 28N to 24N, with a tension decay rate of 0.08N / thousand cycles, significantly higher than the average decay rate of 0.03N / thousand cycles at other points. The first displacement sensor also showed a warping height of 0.03mm at the lower left corner.

[0105] Based on this, the printout prediction module fuses and analyzes the aforementioned multi-source features, calculating the current printout risk probability to be 0.55, which falls under the medium risk level. It also predicts the remaining lifespan of the printing screen to be approximately 30,000 prints. The printout warning system then controls the on-site LED indicator to turn solid yellow and simultaneously sends a message to the monitoring center: "Screen #001, medium risk (55%), tension attenuation and line deformation in the lower left corner; inspection within 12 hours recommended."

[0106] Upon receiving the warning, management personnel can organize technicians to conduct an on-site inspection of the printing screen during a production break (approximately the 160,000th print). This confirms that the lower left corner of the screen shows obvious fatigue but has not yet broken. Based on the warning information, technicians replace the screen in advance, thus preventing a screen breakage accident. Actual testing showed that the printing screen was used 160,000 times, with a theoretical maximum lifespan of approximately 180,000 prints, representing a utilization rate of 88.9%, and no ink leakage occurred throughout the entire process. In contrast, traditional screens without predictive mechanisms typically break suddenly around 180,000 prints, resulting in a loss of approximately 1.5 kg of ink per breakage, as well as equipment contamination and production line downtime. Therefore, the screen breakage warning system provided by this invention has significant advantages in early warning, guiding precise screen replacement, and avoiding economic losses.

[0107] Note that the above description is merely a preferred embodiment of the present invention and the technical principles employed. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and various obvious changes, readjustments, combinations, and substitutions can be made without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the above embodiments, and may include many other equivalent embodiments without departing from the concept of the present invention, the scope of which is determined by the scope of the appended claims.

Claims

1. A screen printing stencil explosion early warning system for photovoltaic cells, characterized in that, include: The image detection module is used to periodically acquire images of the printing screen and generate real-time screen images; The tension detection module is used to periodically detect the tension at multiple points on the printing screen and generate the screen tension value at each point. The first displacement detection module is used to periodically detect the height of the four corners of the printing screen and generate the displacement of the four corners of the screen. The second displacement detection module is used to periodically detect the downward pressure depth of the scraper and generate the scraper downward pressure depth value; The screen printing failure prediction module is connected to the image detection module, the tension detection module, the first displacement detection module, and the second displacement detection module, respectively. It is used to generate the probability of screen printing failure and the remaining service life based on the real-time screen printing image, the screen printing tension value at each point, the displacement of the four corners of the screen printing, and the depth of the squeegee.

2. The overprint warning system according to claim 1, characterized in that, The printing screen includes a screen and a frame for fixing the screen; the screen includes an effective printing area; The image detection module includes an image sensor located above the printing screen, and the imaging area of ​​the image sensor covers the effective printing area. The tension detection module includes multiple tension sensors, all of which are located on the connection structure between the wire mesh and the frame. The first displacement detection module includes a plurality of first displacement sensors, which are located above the printing screen and their vertical projections are respectively located in the four corner areas of the printing screen. The second displacement detection module includes a second displacement sensor, which is located on the scraper holder.

3. The overprint warning system according to claim 1, characterized in that, It also includes an early warning output module, which is connected to the peak page prediction module; the early warning output module is also communicatively connected to external devices. The "hot item prediction" module is also used to generate a risk level based on the probability of a hot item being predicted. The early warning output module: Used to execute the corresponding alarm action based on the risk level; It is also used to send webpage information to the external device; the webpage information includes the risk level, the probability of a webpage becoming overloaded, and the remaining service life.

4. The overprint warning system according to claim 1, characterized in that, It also includes a data storage module; The data storage module is connected to the image detection module, the tension detection module, the first displacement detection module, the second displacement detection module, and the bursting prediction module, respectively, and is used to store the real-time screen printing image, the screen printing tension value at each point, the displacement of the four corners of the screen printing, the depth of the scraper, the probability of bursting risk, and the remaining service life.

5. A method for early warning of printing plate bursts in photovoltaic cell printing screens, characterized in that, The system is implemented using the photovoltaic cell printing screen printing plate explosion warning system according to any one of claims 1-4, wherein the explosion warning method includes: Acquire real-time screen images periodically collected by the image detection module; Obtain the screen tension values ​​at various points periodically generated by the tension detection module; Obtain the displacement of the four corners of the screen periodically generated by the first displacement detection module; Obtain the scraper pressing depth value periodically generated by the second displacement detection module; Based on the real-time screen printing image, the screen printing tension value at each point, the displacement of the four corners of the screen printing, and the depth of the squeegee, the probability of screen bursting and the remaining service life are generated.

6. The method for early warning of content piracy according to claim 5, characterized in that, Based on the real-time screen printing image, the screen tension values ​​at each point, the displacement of the four corners of the screen printing, and the depth of the squeegee press, a probability of screen bursting and the remaining service life are generated, including: A multimodal fusion network based on an attention mechanism or a weighted stitching fusion method is used to fuse the real-time screen image, the screen tension value at each point, the displacement of the four corners of the screen, and the depth of the scraper to generate a risk feature vector. The risk feature vectors are arranged in chronological order to form a time series, which is then input into a time series prediction model trained based on historical high-risk data to generate the high-risk probability and the remaining lifespan.

7. The method for early warning of content piracy according to claim 6, characterized in that, A multimodal fusion network based on an attention mechanism or a weighted stitching fusion method fuses the real-time screen image, the screen tension values ​​at each point, the displacement of the four corners of the screen, and the depth of the scraper press to generate a risk feature vector, including: Feature extraction is performed on the real-time screen image to obtain line offset, warping height value, abnormal gray area and total crack length; Feature extraction is performed on the screen tension values ​​at each point to obtain the average screen tension value, screen attenuation rate, screen standard deviation, and local anomaly score at each point of the screen. Feature extraction is performed on the displacement of the four corners of the screen to obtain the flatness deviation of the screen; Based on an attention-based multimodal fusion network or a weighted splicing fusion method, the line offset, the warping height, the abnormal gray area, the total crack length, the average tension of the screen printing plate, the screen printing plate attenuation rate, the screen printing plate standard deviation, the local anomaly scores of each point on the screen printing plate, the screen printing plate flatness deviation, and the squeegee pressing depth are fused to generate the risk feature vector.

8. The method for early warning of content piracy according to claim 7, characterized in that, Feature extraction is performed on the real-time screen printing image to obtain line offset, warping height value, abnormal grayscale area, and total crack length, including: Calculate the pixel offset of each line in the real-time screen printing image and the screen printing reference image; Based on the first preset conversion coefficient, the pixel offset is converted into the actual size offset to obtain the line offset.

9. The method for early warning of content piracy according to claim 7, characterized in that, Feature extraction is performed on the real-time screen printing image to obtain line offset, warping height value, abnormal grayscale area, and total crack length, including: Based on the real-time screen image, a three-dimensional shape of the printing screen is established, and a current screen height map is generated; The current screen height map is compared with the reference screen height map to obtain the height difference of the edge grid lines of the printing screen, and the height difference is used as the warpage height value.

10. The method for early warning of content explosion according to claim 7, characterized in that, Feature extraction is performed on the real-time screen printing image to obtain line offset, warping height value, abnormal grayscale area, and total crack length, including: The grayscale difference is obtained by performing a difference operation between the real-time screen printing image and the screen printing reference image; Count the number of pixels in the region where the grayscale difference is greater than a preset grayscale threshold; The number of pixels is converted into the actual area according to the second preset conversion coefficient to obtain the abnormal grayscale area.

11. The method for early warning of content explosion according to claim 7, characterized in that, Feature extraction is performed on the screen tension values ​​at each point to obtain the average screen tension value, screen attenuation rate, screen standard deviation, and local anomaly scores at each point on the screen, including: The deviation between the screen tension value at each point and the preset reference value is calculated to obtain the local anomaly score for each point of the screen.

12. The method for early warning of content explosion according to claim 7, characterized in that, Feature extraction is performed on the displacement of the four corners of the screen to obtain the screen flatness deviation, including: Based on the displacement of the four corners of the screen, calculate the height difference between the two corners on the two diagonals of the printing screen. The maximum value among the height differences is taken as the screen flatness deviation.

Images