Hot melt adhesive smearing thickness monitoring method based on deep learning

Abstract

The invention provides a hot melt adhesive smearing thickness monitoring method based on deep learning. The method comprises the steps that a use scene stress model is constructed according to stress distribution data, the adhesive layer thickness is input to the use scene stress model, the thickness reservation after curing is obtained through output, and a thickness reservation adjusting instruction is generated; if the actual thickness of the crease stress concentration area is lower than a preset thickness threshold value, spraying parameters are dynamically adjusted according to the thickness difference value, the crease stress concentration area is sprayed through the adjusted spraying parameters, and target glue layer thickness distribution is obtained; an adhesive layer cracking risk value corresponding to the target adhesive layer thickness distribution is obtained through a preset cracking risk assessment model, and if the adhesive layer cracking risk value is lower than a preset risk threshold value, a target production line parameter adjustment scheme is generated according to a risk assessment result; and executing the parameter adjustment scheme, and updating parameters of glue gun spraying and curing equipment in real time to obtain stable brochure sealing quality.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of information technology, and in particular to a hot melt adhesive coating thickness monitoring method based on deep learning. BACKGROUND

[0002] The brochure packaging production line is an important part of the modern printing and packaging industry. Its core task is to ensure the structural stability and durability of brochures during folding, bonding, and long-term use. The hot melt adhesive coating process, as a key step, directly affects the quality of brochures. Especially when users repeatedly flip through them, the adhesive layer needs to withstand stress changes at different opening angles to ensure that it does not crack or peel off. However, existing methods often struggle to accurately adapt to changes in brochure folding angles when dynamically adjusting production line parameters, leading to uneven adhesive layer thickness distribution and affecting the reliability of finished products and user experience. In particular, when the folding angle of a brochure design is adjusted, the adhesive layer thickness is difficult to dynamically compensate according to the expected stress area, and weak areas may form at stress concentration points. For example, when a user opens a brochure significantly, the adhesive layer at the crease may crack due to stress concentration, affecting its appearance and functionality. In addition, the thickness allowance of the adhesive layer after solidification is usually based on static experience and lacks accurate prediction of dynamic stress in use scenarios, resulting in adhesive layers that are either too thick or insufficient in certain areas, increasing material costs or reducing bonding effectiveness. If the coating amount cannot be accurately predicted and adjusted, the adhesive layer may be too thin in high-stress areas, leading to cracking. For example, when a brochure is adjusted from a 60-degree fold to a 120-degree fold, the adhesive layer thickness at the stress concentration point of the crease needs to increase, but existing equipment cannot calculate and implement this change in real time. If the allowance cannot be dynamically optimized according to the actual use scenario, the adhesive layer may not be able to adapt to stress changes when a user repeatedly opens the brochure, leading to quality issues. Therefore, how to dynamically optimize the compensation coating amount of the adhesive layer in the expected stress area when the folding angle of the production line is adjusted, and accurately control the thickness allowance of the adhesive layer at the stress concentration point of the crease after solidification, has become a key issue in improving the quality of brochure packaging. SUMMARY

[0003] The present application provides a hot melt adhesive coating thickness monitoring method based on deep learning, mainly including:

[0004] Obtain different folding angles of brochures and stress sizes corresponding to different folding angles to obtain a stress distribution map. Determine the stress area through stress distribution map analysis, identify the stress size of the stress area, and obtain the position and stress intensity of the crease stress concentration area.

[0005] The stress distribution data of the glue layer is obtained by determining the stress distribution data of the glue layer according to the stress distribution data, determining the hot melt adhesive compensation coating amount according to the stress distribution data of the glue layer, and determining the glue layer thickness according to the hot melt adhesive compensation coating amount;

[0006] The stress distribution data is used to construct a use scene stress model, the glue layer thickness is input into the use scene stress model, and the thickness reservation amount after solidification is output to generate a thickness reservation adjustment instruction;

[0007] The solidification time and temperature are adjusted through the thickness reservation adjustment instruction, the glue layer surface image of the crease area is collected, the glue layer surface image of the crease area is processed, the thickness distribution measurement data is obtained, the crease stress concentration area is determined according to the thickness distribution measurement data, and the actual thickness of the crease stress concentration area is obtained.

[0008] If the actual thickness of the crease stress concentration area is lower than the preset thickness threshold, the spraying parameters are dynamically adjusted according to the thickness difference, the crease stress concentration area is sprayed through the adjusted spraying parameters, and the target glue layer thickness distribution is obtained.

[0009] The target glue layer thickness distribution corresponds to a glue layer cracking risk value through a preset cracking risk evaluation model, and if the glue layer cracking risk value is lower than a preset risk threshold, a target production line parameter adjustment scheme is generated according to the risk evaluation result;

[0010] The parameter adjustment scheme is executed, the glue gun spraying and solidification equipment parameters are updated in real time, and stable brochure packaging quality is obtained.

[0011] Further, the different folding angles of the brochure and the stress sizes corresponding to the different folding angles are obtained to obtain a stress distribution atlas, the stress area is determined through stress distribution atlas analysis, the stress size of the stress area is identified, the position and stress intensity of the crease stress concentration area are obtained, and the position and stress intensity of the crease stress concentration area are obtained.

[0012] The stress sensors are arranged at the key folding positions of the brochure, the stress data in the folding angle range is collected, the stress data is converted into stress values, and the stress values corresponding to each angle are recorded; a two-dimensional stress distribution atlas is generated based on the stress values, the horizontal axis is the folding angle, the vertical axis is the stress size, a continuous stress change curve is generated through interpolation processing; the stress gradient is calculated based on the continuous stress change curve, the area where the stress gradient exceeds the threshold is identified as the stress concentration area; for the stress concentration area, the corresponding folding angle range and maximum stress value in the stress distribution atlas are extracted, and the position range and stress intensity of the crease stress concentration area are determined.

[0013] Further, the stress concentration area of the fold is determined according to the position and stress intensity, and the stress of the target stress area is determined. The stress of the target stress area is analyzed and processed to obtain stress distribution data of the adhesive layer. The compensation amount of the hot melt adhesive is determined according to the obtained stress distribution data of the adhesive layer. The thickness of the adhesive layer is determined according to the compensation amount of the hot melt adhesive, including:

[0014] Based on the position and stress intensity of the stress concentration area of the fold, the stress transfer coefficient is calculated, and the stress of the adhesive layer is generated based on the stress transfer coefficient and the stress intensity. The load bearing capacity is determined based on the stress of the adhesive layer, and the high load area is divided. Based on the high load area, a grid unit is generated, the stress balance of each unit is calculated, and the stress distribution data of the adhesive layer is generated. Based on the maximum stress value in the stress distribution data of the adhesive layer, the compensation amount of the hot melt adhesive is calculated. Based on the compensation amount of the hot melt adhesive and the area, the unit area coating amount is generated. Based on the unit area coating amount and the solidification shrinkage, the thickness of the adhesive layer is generated.

[0015] Further, the stress distribution data is used to construct a use scenario stress model, and the thickness of the adhesive layer is input into the use scenario stress model to output the reserved amount of thickness after solidification, and generate a thickness reservation adjustment instruction, including:

[0016] Based on the stress distribution data, the stress change data is fitted to generate the use scenario stress model. The stress response value under different adhesive layer thicknesses is calculated based on the use scenario stress model to generate the reserved amount of thickness after solidification. The thickness reservation adjustment instruction is generated based on the reserved amount of thickness after solidification.

[0017] Further, the thickness reservation adjustment instruction is used to adjust the solidification time and temperature, and the fold area adhesive layer surface image is collected. The fold area adhesive layer surface image is processed to obtain thickness distribution measurement data, and the fold stress concentration area is determined according to the thickness distribution measurement data, and the actual thickness of the fold stress concentration area is obtained, including:

[0018] Based on the thickness reservation adjustment instruction, the solidification parameter table is queried to generate the adjusted solidification process parameters. Based on the adjusted solidification process parameters, the adhesive layer is solidified, the fold area adhesive layer surface image is collected, and the fold area adhesive layer surface image is processed to generate the thickness distribution measurement data. Based on the thickness distribution measurement data, the thickness change rate is calculated to determine the fold stress concentration area. The actual thickness of the fold stress concentration area is measured by laser ranging technology.

[0019] Further, the fold area adhesive layer surface image is processed to generate the thickness distribution measurement data, including:

[0020] Segmenting the crease area adhesive layer surface image, calculating the thickness variation gradient between adjacent measurement points, identifying the region boundary where the thickness variation gradient exceeds the threshold value; based on the thickness data within the region boundary, identifying the concave position, connecting the concave position to generate the concave region; based on the characteristics of the concave region, identifying the weak belt, generating the thickness distribution measurement data based on the thickness data within the weak belt, determining the dangerous area, and generating the target area list.

[0021] Further, if the actual thickness of the crease stress concentration area is lower than the preset thickness threshold, the spraying parameters are dynamically adjusted according to the thickness difference, the crease stress concentration area is sprayed by the adjusted spraying parameters, and the target adhesive layer thickness distribution is obtained, including:

[0022] Comparing the actual thickness of the crease stress concentration area with the preset thickness threshold, generating a thickness compensation amount; based on the thickness compensation amount, querying a parameter table to generate adjusted spraying parameters; based on the adjusted spraying parameters, controlling the spraying equipment, real-time monitoring the spraying thickness, and generating the target adhesive layer thickness distribution.

[0023] Further, the crease stress concentration area is sprayed by the adjusted spraying parameters to obtain the target adhesive layer thickness distribution, including:

[0024] Based on the spatial distribution position of the crease stress concentration area, a glue gun movement trajectory is generated; differential spraying parameters are set for the glue gun movement trajectory; based on the differential spraying parameters, the glue gun is controlled to perform spraying, the thickness of each area is monitored in real time, the spraying parameters are adjusted based on the thickness deviation, and the target adhesive layer thickness distribution is generated.

[0025] Further, the target adhesive layer thickness distribution corresponding to the adhesive layer cracking risk value is obtained by the preset cracking risk evaluation model, and if the adhesive layer cracking risk value is lower than the preset risk threshold, a target production line parameter adjustment scheme is generated according to the risk evaluation result, including:

[0026] The target adhesive layer thickness distribution is input into the cracking risk evaluation model to generate the adhesive layer cracking risk value; based on the comparison between the adhesive layer cracking risk value and the preset risk threshold, a safety margin coefficient is generated; based on the safety margin coefficient, a parameter adjustment mapping table is queried to generate the target production line parameter adjustment scheme.

[0027] Further, the parameter adjustment scheme is executed to update the glue gun spraying and curing equipment parameters in real time, and a stable brochure packaging quality is obtained, including:

[0028] The parameter adjustment scheme is received by a production line dynamic control system, a spraying speed is converted into a pressure control value, a motor speed control instruction is generated, a heating controller target temperature value is set, and a conveying belt operation cycle is adjusted;The spraying and curing equipment is adjusted based on the converted parameter;The glue layer thickness deviation and the bonding force are monitored to obtain stable brochure packaging quality.

[0029] The technical scheme provided by the embodiment of the application can include the following beneficial effects:

[0030] The application discloses a hot melt adhesive coating thickness monitoring method based on deep learning, which solves the core problem of weak glue layer in stress concentration area and easy cracking. The folding angle and stress distribution data are obtained by a stress sensor, a stress model is constructed, the stress concentration area and its stress intensity are accurately located, and the target glue layer stress is determined. Based on this, the application dynamically optimizes the glue layer thickness distribution through hot melt adhesive compensation coating and thickness reservation adjustment. By using image processing and laser ranging technology, the weak glue layer and micro crack area are identified, the spraying parameters and glue gun trajectory are adjusted, and segmented differential spraying is implemented to ensure uniform and stable glue layer in the crease area. The crack risk evaluation model further outputs the glue layer cracking risk value to guide the dynamic adjustment of the production line parameters, including spraying speed, glue temperature and curing time, and finally realizes stable brochure packaging quality. The application significantly improves the durability and packaging reliability of the glue layer, reduces the folding and cracking risk, and optimizes the production efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0031] Fig. 1 The flowchart of the hot melt adhesive coating thickness monitoring method based on deep learning of the application.

[0032] Fig. 2 The schematic diagram of the hot melt adhesive coating thickness monitoring method based on deep learning of the application. DETAILED DESCRIPTION

[0033] The technical scheme in the embodiments of the application will be described in detail below with reference to the drawings in the embodiments of the application. The described embodiments are only a part of the embodiments of the application.

[0034] As Figs. 1-2 , the hot melt adhesive coating thickness monitoring method based on deep learning of the embodiment can specifically include:

[0035] In step S101, different folding angles of the brochure and corresponding stress sizes of different folding angles are obtained, a stress distribution map is obtained, a stress area is determined through stress distribution map analysis, the stress size of the stress area is identified, and the position and stress intensity of the crease stress concentration area are obtained.

[0036] A plurality of stress sensors are arranged at the key folding positions of the brochure to obtain continuous stress data in the range from 0 degrees to 180 degrees of the folding angle. The stress values are converted from the electrical signals collected by the sensors, and the stress values corresponding to each angle are recorded. A two-dimensional stress distribution map is constructed according to the collected stress values, the horizontal axis represents the folding angle, and the vertical axis represents the stress size. The discrete data points are smoothed by cubic spline interpolation to obtain a continuous stress change curve. The stress gradient is calculated by the first derivative of the stress change curve, and the region with a stress gradient exceeding a preset gradient threshold is identified as a stress concentration region. If the stress value of a region is greater than the average value of all collected stress values multiplied by a preset multiple threshold, the region is determined as a high-risk region of crease formation. For the determined high-risk region, the folding angle range and the maximum stress value corresponding to the region are extracted from the stress distribution map to determine the position range and the corresponding stress intensity value of the crease stress concentration region in the folding angle coordinate system.

[0037] In a possible implementation, the arrangement of the stress sensors needs to consider the material properties and folding mode of the brochure. For a common copper plate paper brochure, the sensors are usually arranged at a position 5 mm away from both sides of the folding line, forming a symmetrical distribution. The sensor adopts a thin film piezoresistive sensor with a thickness of only 0.2 mm, which does not affect the normal folding of the brochure. When the brochure is gradually folded from a flat state, the pressure on the sensor will cause a change in resistance, which is converted into a voltage signal through a Wheatstone bridge circuit, and then digitized stress values are obtained through analog-to-digital conversion.

[0038] Specifically, the measurement of the folding angle is realized by installing an angle encoder on the folding axis, and stress data collection is triggered once every 1 degree of change. This synchronous collection method ensures the accuracy of the correspondence between the angle and the stress. The collected discrete data points need to be processed to form a continuous analysis basis.

[0039] It should be noted that cubic spline interpolation has unique advantages in processing stress data. This method constructs a cubic polynomial between adjacent data points to ensure the second-order continuity of the curve. During the folding process of the brochure, the stress changes have the characteristics of continuity, and the cubic spline interpolation can accurately reflect this smooth transition. The stress change curve after interpolation presents obvious peak characteristics, and the stress concentration phenomenon usually occurs in the range from 60 degrees to 120 degrees of the folding angle.

[0040] In an embodiment, the calculation of the stress gradient adopts the central difference method. For any point on the curve, the gradient value of the point is obtained by calculating the stress difference of the adjacent points before and after the point divided by the angle difference. When the gradient value exceeds the preset threshold value, it indicates that the stress of the region changes sharply. At the same time, by calculating the average stress value of all acquisition points, the average stress value is multiplied by the preset multiple threshold value as the judgment standard. This double judgment mechanism improves the recognition accuracy.

[0041] Exemplarily, when the folding angle of a certain brochure is 90 degrees, the stress value reaches the peak value, which is 2.3 times the average stress, and the stress gradient value of the point is significantly higher than that of other regions. By marking on the stress distribution map, it can be seen that the angle range forms an obvious stress concentration zone. This visual expression helps the designers to optimize the crease position, and by adjusting the paper thickness or adding a pre-crease treatment, the stress concentration degree can be effectively reduced, and the service life of the brochure can be prolonged.

[0042] In step S102, the stress of the target stress area is determined according to the position and stress intensity of the stress concentration area of the crease, the load bearing capacity of the stress of the target stress area is analyzed and processed, the stress distribution data of the glue layer is obtained, the hot melt adhesive compensation coating amount is determined according to the obtained stress distribution data of the glue layer, and the thickness of the glue layer is determined according to the hot melt adhesive compensation coating amount.

[0043] According to the position and stress intensity of the stress concentration area of the crease, the stress transfer coefficient is calculated by multiplying the ratio of the elastic modulus of the material to the elastic modulus of the glue layer by the ratio of the area of the crease area to the contact area of the glue layer. The stress intensity is multiplied by the stress transfer coefficient to obtain the stress value of the target stress area. The load bearing capacity of the stress value of the glue layer is determined. If the stress value of the glue layer exceeds the preset bearing threshold value, it is marked as a high load area. By establishing a grid element in the high load area and solving the stress balance equation of each element, the stress distribution data of the glue layer is obtained. According to the difference between the maximum stress value in the stress distribution data of the glue layer and the yield stress of the glue layer material, the hot melt adhesive compensation coating amount is determined by multiplying the ratio of the bulk modulus of the glue layer material to the density. The hot melt adhesive compensation coating amount is divided by the area of the high load area to obtain the coating amount per unit area. According to the product of the coating amount per unit area and the solidification shrinkage rate of the hot melt adhesive, the thickness of the glue layer is determined.

[0044] In one possible implementation, the calculation of the stress transfer coefficient is based on the stress transfer principle in material mechanics. When stress concentration occurs at the brochure fold, this stress is transmitted to the adhesive layer through the paper material. The elastic modulus of the material reflects the ability of the material to resist deformation, and the elastic modulus of the paper is usually 3 to 5 GPa, while the elastic modulus of the hot melt adhesive is about 0.1 to 0.3 GPa. The ratio of the two indicates the degree of attenuation of the stress when it is transmitted between different materials. The ratio of the area of the fold area to the contact area of the adhesive layer reflects the stress dispersion effect. The larger the contact area, the smaller the stress per unit area.

[0045] Specifically, when the fold stress intensity is 100 MPa, the stress transfer coefficient is 12 by calculating the elastic modulus ratio of 15 and the area ratio of 0.8, and the stress value of the adhesive layer is 1200 MPa. This value directly affects the subsequent load capacity determination.

[0046] It should be noted that the load bearing capacity determination adopts the yield criterion of the material. The preset load threshold is usually set to 0.7 times the yield strength of the adhesive material, which can leave a certain safety margin. When the stress value of the adhesive layer exceeds this threshold, it indicates that there is a risk of failure in this area. The establishment process of the grid element divides the high load area into multiple small elements, each element has a size of about 0.5 mm square. By solving the force balance equation of each element, the stress distribution inside the adhesive layer can be obtained.

[0047] In one embodiment, the adhesive stress distribution data shows a characteristic of gradually decaying from the contact surface to the inside. The maximum stress value usually appears on the surface in direct contact with the paper, and the stress inside the adhesive layer decreases with increasing depth. The yield stress of the adhesive material is the critical value at which permanent deformation begins to occur, and for commonly used EVA hot melt adhesive, this value is about 8 to 12 MPa.

[0048] For example, when the maximum stress value is 15 MPa and the yield stress is 10 MPa, the difference is 5 MPa. The ratio of bulk modulus to density reflects the compression characteristics of the material, and for hot melt adhesive, this ratio is about 1000 cubic meters per kilogram. The compensation coating amount obtained by multiplication is 5000 cubic meters per square meter, which guides the actual coating operation.

[0049] It can be understood that the hot melt adhesive will shrink in volume during the curing process, and the shrinkage rate is usually between 5% and 15%. The unit area coating amount divided by the high load area, and then considering the effect of curing shrinkage, the finally determined adhesive thickness can compensate for the effect of stress concentration. This thickness calculation method ensures that the adhesive layer has sufficient buffering capacity when subjected to fold stress, effectively prolongs the service life of the brochure, and avoids the phenomenon of adhesive cracking or falling off due to repeated folding.

[0050] In step S103, a use scenario stress model is constructed according to the stress distribution data, the thickness of the adhesive layer is input into the use scenario stress model, the thickness reserve after curing is output, and a thickness reserve adjustment instruction is generated.

[0051] According to the maximum stress value and the stress concentration area position in the stress distribution data, the stress change data under different folding times is fitted by the least square method to construct a function relationship between the stress value and the folding times as the use scenario stress model, and the model parameters are obtained. The stress response values under different thicknesses of the adhesive layer are calculated by using the model parameters, and the safety factor is defined as the ratio of the yield strength of the adhesive layer material to the current stress value. When the safety factor is lower than a preset threshold, it is determined that there is a cracking risk, and a corresponding relationship between the thickness of the adhesive layer and the cracking risk is established. The initial thickness of the adhesive layer is input into the corresponding relationship between the thickness of the adhesive layer and the cracking risk to obtain the corresponding cracking risk value. If the risk value exceeds a preset risk threshold, the thickness of the adhesive layer is gradually increased until the risk value decreases to below the threshold, and the thickness reserve adjustment instruction containing the reserve value and the increase or decrease identifier is output.

[0052] In a possible implementation, the least square fitting process is based on the actually measured stress data points. When the brochure experiences different folding times, the stress value presents regular changes. The stress value is low at the initial folding, and as the folding times increase, the material fatigue accumulation causes the stress value to gradually increase. By collecting stress data of multiple nodes such as 100 times, 500 times and 1000 times of folding, a discrete data set is formed. The least square method determines the optimal function parameters by minimizing the sum of squares of errors between the measured values and the fitted curve. The common fitting function form is a power function relationship, in which the folding times are taken as the independent variable and the stress value is taken as the dependent variable.

[0053] Specifically, the construction of the use scenario stress model needs to consider the actual use environment. The brochure is folded about 5000 times in a year on average, 10 to 20 times a day. The model parameters include the initial stress coefficient, the stress growth rate and the material degradation coefficient. These parameters are automatically determined through the fitting process, reflecting the stress evolution law under specific materials and structures.

[0054] It should be noted that the definition of the safety factor is crucial to the cracking risk evaluation. The yield strength of the adhesive layer material is the critical stress value at which the material begins to produce irreversible deformation. The current stress value is calculated by the model and represents the actual stress under a specific thickness and folding times. The safety factor is equal to the yield strength divided by the current stress value, and this ratio directly reflects the safety margin of the adhesive layer. When the safety factor decreases to below 1.5, it indicates that the adhesive layer is close to a dangerous state.

[0055] In an embodiment, the corresponding relationship between the adhesive layer thickness and the cracking risk presents a nonlinear feature. When the thickness increases from 0.1 mm to 0.2 mm, the cracking risk can be reduced by 50%, but when the thickness increases from 0.2 mm to 0.3 mm, the risk reduction is only 20%. This marginal effect diminishing phenomenon indicates that there is an optimal thickness range. By establishing this corresponding relationship, the service life under different thicknesses can be accurately predicted.

[0056] For example, the initial design of the adhesive layer thickness is 0.15 mm, and the calculated cracking risk value is 0.7 after inputting this value. The preset risk threshold is usually set to 0.3, which means that a failure probability of 30% is acceptable as the upper limit. Since 0.7 exceeds 0.3, the system automatically increases the thickness by 0.02 mm each time. When the thickness reaches 0.21 mm, the risk value decreases to 0.28, meeting the requirements. The thickness allowance after curing is 0.21 mm.

[0057] It can be understood that the generation of the thickness reservation adjustment instruction takes into account the actual production process. The difference calculation shows that the thickness needs to be increased by 0.06 mm. The increase in the instruction identifies the adjustment of the coating parameters on the production line to ensure that the product reaches the target thickness.

[0058] In step S104, the curing time and temperature are adjusted by the thickness reservation adjustment instruction, the crease area adhesive layer surface image is collected, the crease area adhesive layer surface image is image-processed to obtain thickness distribution measurement data, the crease stress concentration area is determined according to the thickness distribution measurement data, and the actual thickness of the crease stress concentration area is obtained.

[0059] By querying the corresponding relationship table of the curing time and the thickness through the numerical value in the thickness reservation adjustment instruction, the corresponding curing time extension amount is obtained, and the temperature reduction value is calculated according to the thickness increase amount, to obtain the adjusted curing process parameters. After the adhesive layer is cured using the adjusted curing process parameters, the industrial camera is used to collect the crease area adhesive layer surface image, the contrast enhancement and Gaussian filtering are used to remove noise, and the pretreated surface image is obtained. The gradient calculation is performed on the pretreated surface image, the pixel points with a gray value change exceeding a preset threshold are extracted, the thickness change area is identified according to the pixel point distribution density, the thickness distribution measurement data is obtained through the conversion relationship between the pixel coordinates and the actual coordinates. According to the thickness difference between adjacent measurement points in the thickness distribution measurement data, the thickness change rate is calculated, the continuous area with a change rate exceeding a preset threshold is determined as the crease stress concentration area, and the actual thickness of the crease stress concentration area is measured by point-by-point scanning of the area using a laser ranging sensor.

[0060] In one possible implementation, the correspondence table of solidification time and thickness is established based on a large amount of experimental data. The solidification process of the hot melt adhesive involves cross-linking reaction of molecular chain segments, and the increase in thickness means that the internal heat dissipation speed decreases. When the adhesive layer thickness increases from 0.2 mm to 0.3 mm, the solidification time needs to be extended from the original 30 seconds to 45 seconds. This correspondence presents a nonlinear feature, and the accurate parameter adjustment value can be quickly obtained through table lookup. The calculation of temperature reduction value is based on the principle of heat conduction. When the thickness increases by 0.1 mm, the temperature decreases by 2 to 3 degrees Celsius, avoiding the situation that the surface layer is quickly solidified while the inside is still in a flowing state.

[0061] Specifically, the adjusted solidification process parameters directly affect the final performance of the adhesive layer. The adhesive layer after solidification has stable physical properties, and at this time, image acquisition can obtain true thickness distribution information. The industrial camera selects a device with a resolution of 2048x1536 pixels, and cooperates with a ring LED light source to provide uniform illumination. Contrast enhancement processing adjusts the brightness distribution range of the image, so that the gray difference of the thickness change area is more obvious. The Gaussian filter uses a 5x5 filter kernel to effectively remove random noise in the image acquisition process.

[0062] It should be noted that gradient calculation is a key step to identify thickness changes. In a digital image, the gray value reflects the height information of the adhesive layer surface. The gray difference value of adjacent pixels divided by the pixel spacing is the gradient value. When the thickness of the adhesive layer changes, the surface will form a slope or step structure, which is represented as a gradual or sudden change in gray value in the image. The gradient threshold is set to 10 gray levels per pixel, and the area exceeding this value is marked as a thickness change area.

[0063] In one embodiment, the conversion of pixel coordinates to actual coordinates needs to be calibrated in advance. By placing a standard scale plate on the imaging plane, a mapping relationship between pixel distance and actual distance is established. Assuming that the camera field of view covers an area of 50x40 mm, each pixel corresponds to an actual size of about 0.024 mm. The thickness distribution measurement data is stored in matrix form, and each element represents the relative thickness value of the corresponding position.

[0064] For example, in the crease area, the thickness change rate of 20 consecutive measurement points exceeds 15%, and these points form a strip area with a length of about 5 mm. The laser ranging sensor uses the principle of triangulation, and the laser beam is irradiated to the adhesive layer surface at a fixed angle, and the reflected light is received by a position sensitive detector. The accurate distance value is calculated according to the shift amount of the light spot position. The sensor scans the entire stress concentration area at a step interval of 0.1 mm, and the data of each measurement point includes the horizontal and vertical coordinates and the thickness value. The actual thickness data obtained has a precision of 0.01 mm.

[0065] The thickness distribution measurement data obtained after image processing of the surface of the adhesive layer in the crease area is regionally segmented, the region boundary with the largest thickness change gradient is identified, the recessed position where the thickness of the adhesive layer is obviously lower than the surrounding area is marked, the weak belt formed by repeated bending during the folding of the brochure is located, the dangerous area where micro-cracks or voids appear in the adhesive layer is confirmed, a spatial distribution map of the stress concentration area of the crease is established, and a list of target areas that need to be reinforced is formed.

[0066] The thickness distribution measurement data is regionally segmented, the thickness change gradient is obtained by calculating the thickness difference between adjacent measurement points, the continuous point set with a gradient value exceeding a preset threshold is identified, and the region boundary with the largest thickness change gradient is determined. According to the thickness data within the region boundary, the average thickness of each measurement point and its eight adjacent measurement points arranged in a grid is calculated. If the thickness of a certain point is lower than a preset proportion threshold of the average value, the point is marked as a recessed position, and adjacent recessed positions are connected to form a recessed area. By analyzing the length-width ratio and continuity characteristics of the recessed area, the recessed area with a length-width ratio exceeding a preset value and a belt-shaped distribution along the folding line direction is identified as a weak belt. The ratio of the minimum thickness in the weak belt to the thickness of the normal area is calculated. If the ratio is lower than a preset safety threshold, it is confirmed as a dangerous area where micro-cracks or voids appear in the adhesive layer. The center coordinates, coverage area and thickness ratio data of the dangerous area are recorded in the spatial distribution map. According to the thickness ratio from low to high, the risk level is determined, and a list of target areas that need to be reinforced is formed according to the risk level.

[0067] In a possible implementation, the region segmentation processing adopts a gradient-based segmentation method. The thickness distribution measurement data is stored in a matrix form, and each element represents the thickness value of a measurement point. The distance between adjacent measurement points is usually 0.5 mm, forming a regular grid structure. The thickness change gradient is obtained by calculating the thickness difference in the horizontal and vertical directions. When the gradient value in a certain direction exceeds the threshold of 0.02 mm per mm, it indicates that there is a significant thickness change at this position. These points exceeding the threshold are connected to form the region boundary, outlining the profile of the thickness abnormal area.

[0068] Specifically, the eight adjacent measurement points arranged in a grid include four forward adjacent points and four diagonal adjacent points. This neighborhood definition is derived from the eight-connected concept in image processing, which can fully reflect the thickness distribution around the measurement point. When calculating the average value, each adjacent point has the same weight. The preset proportion threshold is usually set to 0.85, which means that when the thickness of a certain point is lower than 85% of the average value of the surrounding area, it is identified as a recess. This relative determination method can adapt to the different thickness differences of different areas.

[0069] It should be noted that the aspect ratio is a key indicator for determining the weak zone. During repeated folding of the brochure, stress concentration causes the adhesive layer to form an elongated damage area along the folding line direction. When the length-to-width ratio of the recessed area exceeds 5:1, it can be basically determined that it is a weak zone caused by folding rather than a random defect. The calculation of the thickness ratio requires first determining the reference thickness of the normal area, which is usually selected as a position away from the crease area and with uniform thickness distribution as the reference.

[0070] In one embodiment, the identification of the dangerous area is based on the principles of material mechanics. When the adhesive layer thickness decreases to below 60% of the normal value, its carrying capacity will decrease sharply. The preset safety threshold of 0.6 is based on this critical point. Areas below this threshold are prone to crack propagation during subsequent use. The spatial distribution map uses a two-dimensional coordinate system, with the lower left corner of the brochure as the origin, the horizontal axis representing the width direction, and the vertical axis representing the height direction.

[0071] For example, the center coordinates of a certain dangerous area are 25 mm in the horizontal direction and 80 mm in the vertical direction, the coverage area is 15 mm2, and the thickness ratio is 0.55. These data completely describe the location, size, and severity of the area. The risk level is divided into three levels according to the thickness ratio: below 0.5 is high risk, 0.5 to 0.6 is medium risk, and 0.6 to 0.7 is low risk. The target area list is sorted according to this level, with high-risk areas being prioritized. This quantitative evaluation method provides a clear priority and target for subsequent reinforcement, avoiding material waste caused by blind reinforcement, while ensuring that critical parts are adequately protected.

[0072] In step S105, if the actual thickness of the crease stress concentration area is lower than the preset thickness threshold, the spraying parameters are dynamically adjusted according to the thickness difference, the crease stress concentration area is sprayed by the adjusted spraying parameters, and the target adhesive layer thickness distribution is obtained.

[0073] The actual thickness of the crease stress concentration area is compared with the preset thickness threshold. If the actual thickness is lower than the preset thickness threshold, the difference between the preset thickness threshold and the actual thickness is calculated to obtain the thickness compensation amount. According to the thickness compensation amount, the preset thickness and parameter correspondence table is queried to obtain the spraying flow coefficient and speed coefficient corresponding to each millimeter of thickness. The thickness compensation amount is multiplied by the corresponding coefficients to obtain the spraying flow increase value and the spraying speed reduction ratio. The standard spraying flow is added to the flow increase value, and the standard spraying speed is multiplied by the speed reduction ratio to obtain the adjusted spraying parameters. The adjusted spraying parameters are used to control the spraying equipment to spray the crease stress concentration area layer by layer, and the current spraying thickness is monitored in real time by the laser thickness sensor. The difference between the current thickness and the initial thickness is used as the cumulative amount, and the spraying is stopped when the cumulative amount reaches the thickness compensation amount to obtain the target adhesive layer thickness distribution.

[0074] In one possible implementation, the determination of the preset thickness threshold is based on the material mechanical properties and the usage requirements. For the glue layer of a brochure, the thickness threshold is usually set as the minimum thickness that can withstand the expected number of folds without cracks. When the actual measured thickness is 0.18 mm and the preset threshold is 0.25 mm, the thickness compensation amount is 0.07 mm. This difference directly reflects the increased glue layer thickness that needs to be added to ensure that the crease area has sufficient strength.

[0075] Specifically, the thickness and parameter correspondence table is an empirical relationship table established through a large amount of experimental data. The table records the adjustment value of the spraying parameters required for different thickness increments. The spraying flow coefficient represents the glue flow that needs to be increased by 1 mm of thickness, which is usually 15 to 20 ml per minute. The speed coefficient reflects the adjustment proportion of the spraying moving speed. The greater the thickness requirement, the slower the spraying speed, so as to ensure that the glue is fully deposited. When 0.07 mm of thickness needs to be compensated, the table lookup gives a flow increase of 1.4 ml per minute and a speed reduction to 0.8 times the original speed.

[0076] It should be noted that the standard spraying parameters are the reference values under normal production conditions. The standard spraying flow is usually 50 ml per minute, and the standard spraying speed is 100 mm per second. After adjustment, the new spraying flow becomes 51.4 ml per minute, and the spraying speed becomes 80 mm per second. This parameter combination can deposit more glue in a unit area to achieve accurate compensation of the thickness.

[0077] In one embodiment, the layer-by-layer spraying uses a spiral path or a parallel line path to cover the target area. Each layer has a thickness of about 0.02 mm, and the target thickness is achieved by stacking multiple layers. The laser thickness sensor is installed vertically next to the spray head at a fixed distance from the spray head. The laser beam emitted by the sensor irradiates the surface of the glue layer, and the thickness at the current position is calculated by measuring the time difference of the reflected light. This non-contact measurement method does not interfere with the uncured glue layer.

[0078] For example, for an area with an initial thickness of 0.18 mm, the laser thickness sensor collects data every 0.5 seconds during the spraying process. After the first layer is sprayed, the measured thickness is 0.20 mm, and the cumulative amount is 0.02 mm. The second layer is continued to be sprayed, and the thickness reaches 0.22 mm, with a cumulative amount of 0.04 mm. When the fourth layer is completed, the measured thickness is 0.25 mm, and the cumulative amount reaches the compensation target of 0.07 mm. The control system immediately stops spraying.

[0079] It can be understood that this real-time monitoring and dynamic control method avoids material waste and uneven thickness caused by excessive spraying. The target glue layer thickness distribution presents a smooth transition feature, with no sudden change from the normal area to the reinforced area, ensuring uniform stress transmission.

[0080] According to the spatial distribution position of the crease stress concentration area, the moving track of the glue gun is adjusted, the stopping time in the weak area of the glue layer is increased, the multi-layer superposition spraying treatment is performed on the micro crack position, the filling depth of the glue solution in the recessed area is controlled, the sufficient glue layer coverage thickness in the dangerous area is ensured, the differential reinforcement in different areas is realized through the segmented spraying mode, and the uniform and stable glue layer thickness distribution is formed in the crease stress concentration area.

[0081] According to the spatial distribution position coordinates of the crease stress concentration area, the center points of the weak area, the start and end points of the micro crack, and the boundary points of the recessed area are taken as the key control points, the control points are connected in turn through the shortest path connection algorithm, and the moving track of the glue gun is generated. For the generated moving track of the glue gun, the stopping time in the weak area is set to be a preset multiple of the standard time, the superposition number at the micro crack position is determined according to the crack length, the maximum depth value of the recessed area is multiplied by the area to obtain the required glue filling volume, and the differential spraying parameter set of each area is formed. The differential spraying parameter set is used to control the glue gun to perform segmented spraying, the spraying is maintained according to the corresponding stopping time in each area, the spraying is repeated according to the superposition number, and the flow is adjusted according to the filling volume, so that the targeted reinforcement treatment of the dangerous area is completed. The glue layer thickness values in each area are monitored through the laser thickness sensor, the measured thickness is compared with the target thickness, the spraying speed and flow of the current area are adjusted according to the difference until the thickness deviation of each point in the crease stress concentration area is within the preset range, and the uniform and stable glue layer thickness distribution is formed.

[0082] In a possible implementation, the shortest path connection algorithm is based on the solution idea of the traveling salesman problem in graph theory. The selection of the key control points reflects the characteristics of different types of defects: the weak area usually presents a sheet-like distribution, and its geometric center is selected as the representative point; the micro crack extends in a linear manner, and the start and end points identify the complete range of the crack; and the boundary points of the recessed area outline the profile to be filled. The algorithm calculates the Euclidean distance between each point to construct a distance matrix, and then uses a greedy strategy to start from the starting point, selects the nearest unvisited point as the next target each time, and iterates until all control points are traversed.

[0083] Specifically, the setting of the stopping time is based on the flow and penetration characteristics of the glue solution. The standard stopping time is usually 0.5 seconds, and in the weak area, the stopping time is set to 1.5 to 2 times the standard time because more glue solution needs to penetrate into the material. The determination of the superposition number takes into account the depth and width of the crack, and one superposition spraying is performed for every millimeter of crack length to ensure that the crack is fully filled. The volume calculation of the recessed area uses the trapezoidal rule to approximate the irregular recess as a combination of multiple trapezoidal columns, and the relationship between the maximum depth value and the average depth determines the actual filling amount required.

[0084] It is necessary to point out that the formation process of the differentiated spraying parameter set embodies the concept of precise control. Each parameter is related to a specific physical process: the residence time affects the amount of glue solution deposited per unit area, the number of superimpositions determines the cumulative thickness in the vertical direction, and the filling volume ensures complete filling of the recessed area. These parameters work together to achieve targeted treatment of different defect types.

[0085] In one embodiment, the execution process of segmented spraying is similar to the layer-by-layer manufacturing of 3D printing. The glue gun moves according to the predetermined trajectory, automatically reduces the moving speed and keeps spraying when it reaches the weak area, and the glue solution spreads uniformly under the action of gravity and surface tension. For micro-crack locations, the glue gun moves back and forth on the same path, each spraying forming a coverage layer of about 0.03 millimeters, and after multiple layers are accumulated, the crack is completely closed. The filling of the recessed area uses a spiral path from the outside to the inside, ensuring that the glue solution flows from the edge to the center, avoiding the formation of bubbles.

[0086] For example, real-time monitoring by laser thickness measurement sensors provides feedback for dynamic adjustment. The sensor collects thickness data at a frequency of 10 Hz, and the thickness value of each measurement point is compared with the target thickness at that location. When the measured thickness is 0.22 millimeters and the target thickness is 0.25 millimeters, a thickness difference of 0.03 millimeters is calculated, and accordingly the current spraying flow is increased by 12% and the moving speed is reduced by 8%. This closed-loop control ensures the uniformity of the final thickness distribution, with the thickness deviation of each measurement point controlled within ±5% of the target value, achieving reliable reinforcement of the crease area.

[0087] Step S106, obtaining a glue layer cracking risk value corresponding to the target glue layer thickness distribution through a preset cracking risk evaluation model, if the glue layer cracking risk value is lower than a preset risk threshold, generating a target production line parameter adjustment scheme according to the risk evaluation result.

[0088] The target glue layer thickness distribution data is input into the cracking risk evaluation model, the ratio of the minimum value in the thickness distribution to the preset reference thickness value is extracted as the thickness attenuation rate, the thickness attenuation rate is multiplied by the fatigue limit coefficient of the glue layer material to obtain the glue layer cracking risk value. If the glue layer cracking risk value is lower than the preset risk threshold, the difference between the preset risk threshold and the cracking risk value is calculated as a safety margin coefficient, and the corresponding spraying speed reduction ratio, glue temperature increase value and solidification time increase ratio are found in the pre-established parameter adjustment mapping table according to the safety margin coefficient. The spraying speed reduction ratio found is multiplied by the current spraying speed to obtain a new spraying speed parameter, the glue temperature increase value is added to the current temperature to obtain a new glue temperature parameter, and the solidification time increase ratio is multiplied by the current solidification time to obtain a new solidification time parameter. The three new parameters are integrated to generate a target production line parameter adjustment scheme.

[0089] In one possible implementation, the core of the cracking risk assessment model lies in quantifying the stress-bearing capacity of the adhesive layer. The preset reference thickness value is usually set to 0.3 millimeters, which is the optimal thickness based on a large number of experiments. When the target adhesive layer thickness distribution data shows that the minimum value in a certain area is 0.21 millimeters, the thickness attenuation rate is 0.21 divided by 0.3, resulting in 0.7. The fatigue limit coefficient of the adhesive material reflects the durability of the material under cyclic stress, and for hot melt adhesive material, this coefficient is usually between 0.6 and 0.8. Multiplying the thickness attenuation rate of 0.7 by the fatigue limit coefficient of 0.7 gives a cracking risk value of 0.49.

[0090] Specifically, the calculation of the safety margin coefficient embodies the principle of conservative design. The preset risk threshold is generally set to 0.3, indicating that a failure probability of 30% is the upper limit of what is acceptable. When the actual risk value is 0.49, it exceeds the threshold, indicating that production parameters need to be adjusted. If the risk value is 0.25, which is below the threshold, the safety margin coefficient is 0.3 minus 0.25, resulting in 0.05. The larger this coefficient, the more adequate the safety margin, allowing for moderate optimization of production efficiency.

[0091] It should be noted that the parameter adjustment mapping table is an empirical database established through orthogonal experiments. The table divides the safety margin coefficient into multiple intervals, each corresponding to a set of optimized parameters. When the safety margin coefficient is 0.05, the table shows that the spraying speed can be increased to 1.1 times the original speed to improve production efficiency, the glue temperature can be reduced by 3 degrees Celsius to save energy, and the curing time can be shortened to 0.95 times the original time to speed up the production rhythm. This mapping relationship ensures that efficiency is improved while quality is guaranteed.

[0092] In one embodiment, the implementation process of parameter adjustment needs to consider the response characteristics of the equipment. The current spraying speed is 80 millimeters per second, and after multiplying by the reduction ratio of 1.1, the new speed is 88 millimeters per second. The glue temperature is reduced from the original 165 degrees Celsius to 162 degrees Celsius, which is still within the optimal flow range of the glue. The curing time is shortened from 60 seconds to 57 seconds, which is achieved by adjusting the speed of the conveyor belt.

[0093] For example, after implementing parameter adjustment, a certain production line increased its daily output from 8000 brochures to 8800, while reducing energy consumption by 5%. More importantly, through precise risk assessment and parameter optimization, the product pass rate remained above 99.5%. This dynamic parameter adjustment method based on risk assessment achieves a balance between quality control and production efficiency. The generated target production line parameter adjustment scheme contains a complete set of process parameters, which can be directly input into the production control system for execution, avoiding errors and delays caused by manual adjustment.

[0094] Step S107, the parameter adjustment scheme is executed to update the glue gun spraying and curing equipment parameters in real time, and stable brochure packaging quality is obtained.

[0095] The parameter adjustment scheme is received by the production line dynamic control system, the spraying speed parameter is converted into a corresponding pressure control value according to the linear relationship between the spraying flow and the pressure, the motor speed control instruction is directly generated according to the speed parameter, the glue liquid temperature parameter is taken as the target temperature value of the heating controller, and the curing time parameter is converted into the running period of the conveying belt. The spraying pressure is adjusted to the corresponding value according to the converted pressure control value, the glue gun moving speed is set through the motor speed control instruction, the target temperature value is input into the heating controller to adjust the heating power, and the conveying mechanism speed is adjusted according to the running period of the conveying belt, so that each equipment cooperates according to the new parameters. Continuous production is carried out by using the updated equipment parameters, and the thickness deviation of the glue layer and the bonding force of the glue layer and the paper of each batch of brochures are monitored through the thickness measuring device and the tensile testing device respectively. When the thickness deviation of a continuous preset batch is within a preset range and the bonding force exceeds a preset strength threshold, stable brochure packaging quality is obtained.

[0096] In a possible implementation, the production line dynamic control system serves as the center of the entire production process and is responsible for accurately converting the optimized parameters into control signals recognizable by each equipment. The linear relationship between the spraying flow and the pressure is based on the principle of fluid mechanics. When the spraying flow needs to be increased by 20%, the pressure needs to be increased by about 44% according to the Bernoulli equation. The conversion table built in the system records the pressure values corresponding to different flows. When receiving the target flow parameter of 60 milliliters per minute, the table is automatically looked up to obtain the pressure that needs to be set, which is 0.8 megapascals.

[0097] Specifically, the generation of the motor speed control instruction involves kinematics calculation. The linear moving speed of the glue gun is associated with the rotating speed of the driving motor through the transmission ratio. If the pitch of the transmission mechanism is 5 millimeters, when a moving speed of 100 millimeters per second is needed, the rotating speed of the motor should be set to 1200 revolutions per minute. The heating controller adopts a PID control algorithm to compare the target temperature value with the real-time temperature, and adjusts the heating power to stabilize the temperature within a range of ±1 degree Celsius of the set value.

[0098] It should be noted that the key to the cooperative operation of the equipment lies in the timing synchronization. The running period of the conveying belt determines the residence time of the brochure at each station. When the curing time parameter is 45 seconds, considering that the length of the curing cavity is 3 meters, the conveying belt speed should be adjusted to 4 meters per minute. In this way, each brochure is ensured to stay in the curing area for exactly 45 seconds. Each equipment is kept synchronized through a unified clock signal to avoid quality problems caused by timing misalignment.

[0099] In one embodiment, the thickness measuring device adopts the principle of laser triangulation, with a measuring point set every 50 mm, covering the entire crease area. The measurement accuracy reaches 0.01 mm, which can accurately detect the slight changes in the thickness of the adhesive layer. The thickness deviation is obtained by calculating the difference between the measured value and the target value, and the preset range is usually ±0.02 mm. The tensile test device simulates the peeling force in actual use by measuring the bonding force of the adhesive layer to the paper by vertical stretching.

[0100] For example, in actual production, the detection data of the first batch of 100 brochures shows that the average thickness of the adhesive layer is 0.251 mm, the standard deviation is 0.015 mm, and the thickness deviation of all measuring points is within the range of ±0.02 mm. The tensile test shows that the average bonding force is 3.2 N / cm, which exceeds the preset strength threshold of 2.5 N / cm. The data of the next 5 batches all meet the requirements, indicating that the production process has stabilized.

[0101]

[0102] It can be understood that this real-time monitoring and feedback mechanism ensures the consistency of production quality. By converting abstract parameters into specific device control instructions and verifying the effect through online detection, a complete quality control closed loop is formed. Stable packaging quality not only reflects in numerical indicators, but more importantly, ensures that each brochure can withstand repeated folding tests in actual use, prolonging the service life of the product.

[0103] It should be noted that the above-mentioned is only a few specific embodiments of the present application. Obviously, the present application is not limited to the above embodiments, but can also have many variations. All variations that can be directly derived or inferred from the disclosure of the present application by those of ordinary skill in the art should be considered as falling within the scope of the present application.​

Claims

1. A method for monitoring the thickness of hot melt adhesive application based on deep learning, characterized in that, The method includes: Obtain the stress magnitude corresponding to different folding angles of the brochure, and obtain a stress distribution map. Analyze the stress distribution map to determine the stress area, identify the stress magnitude in the stress area, and obtain the location and stress intensity of the stress concentration area of ​​the crease. The stress magnitude of the adhesive layer in the target stress area is determined based on the location and stress intensity of the stress concentration area of ​​the crease. The load-bearing capacity of the adhesive layer in the target stress area is analyzed to obtain the stress distribution data of the adhesive layer. The amount of hot melt adhesive compensation is determined based on the obtained stress distribution data of the adhesive layer. The thickness of the adhesive layer is determined based on the amount of hot melt adhesive compensation. Based on the stress distribution data, construct a stress model for the application scenario, input the adhesive layer thickness into the stress model for the application scenario, output the thickness allowance after curing, and generate a thickness allowance adjustment instruction. The curing time and temperature are adjusted by the thickness reservation adjustment command. At the same time, the surface image of the adhesive layer in the crease area is acquired. The surface image of the adhesive layer in the crease area is processed to obtain the thickness distribution measurement data. The crease stress concentration area is determined based on the thickness distribution measurement data, and the actual thickness of the crease stress concentration area is obtained. If the actual thickness of the stress concentration area of ​​the crease is lower than the preset thickness threshold, the spraying parameters are dynamically adjusted according to the thickness difference. The adjusted spraying parameters are then used to spray the stress concentration area of ​​the crease to obtain the target adhesive layer thickness distribution. The cracking risk value of the adhesive layer corresponding to the target adhesive layer thickness distribution is obtained through a preset cracking risk assessment model. If the cracking risk value of the adhesive layer is lower than the preset risk threshold, an adjustment plan for the target production line parameters is generated based on the risk assessment results. By implementing the parameter adjustment scheme, the parameters of the glue gun spraying and curing equipment are updated in real time to obtain stable brochure packaging quality.

2. The method for monitoring the thickness of hot melt adhesive application based on deep learning according to claim 1, characterized in that, The process involves obtaining stress magnitudes corresponding to different folding angles of the brochure, generating a stress distribution map, analyzing the stress distribution map to determine the stress-bearing areas, identifying the stress magnitudes in these areas, and determining the location and intensity of stress concentration areas at the creases. Stress sensors are placed at key folding locations in the brochure to collect stress data within the folding angle range. The stress data is converted into stress values, and the stress values ​​corresponding to each angle are recorded. A two-dimensional stress distribution map is generated based on the stress values, with the horizontal axis representing the folding angle and the vertical axis representing the stress magnitude. A continuous stress variation curve is generated through interpolation. The stress gradient is calculated based on the continuous stress variation curve, and the region where the stress gradient exceeds a threshold is identified as a stress concentration region. For the stress concentration region, the corresponding folding angle range and maximum stress value in the stress distribution map are extracted to determine the location range and stress intensity of the crease stress concentration region.

3. The method for monitoring the thickness of hot melt adhesive application based on deep learning according to claim 1, characterized in that, The process of determining the stress magnitude of the adhesive layer in the target stress area based on the location and stress intensity of the stress concentration area of ​​the crease, performing load-bearing capacity analysis on the stress magnitude of the adhesive layer in the target stress area to obtain adhesive layer stress distribution data, determining the hot melt adhesive compensation application amount based on the obtained adhesive layer stress distribution data, and determining the adhesive layer thickness based on the hot melt adhesive compensation application amount includes: Based on the location and stress intensity of the stress concentration area of ​​the crease, a stress transfer coefficient is calculated, and a stress value of the adhesive layer is generated based on the stress transfer coefficient and the stress intensity. The load-bearing capacity is determined based on the stress value of the adhesive layer, and high-load areas are divided. Mesh cells are generated based on the high-load areas, and the stress balance of each cell is calculated to generate the stress distribution data of the adhesive layer. Based on the maximum stress value in the stress distribution data of the adhesive layer, the hot melt adhesive compensation application amount is calculated. Based on the hot melt adhesive compensation application amount and the area, the application amount per unit area is generated, and the adhesive layer thickness is generated based on the application amount per unit area and the curing shrinkage rate.

4. The method for monitoring the thickness of hot melt adhesive application based on deep learning according to claim 1, characterized in that, The process involves constructing a usage scenario stress model based on stress distribution data, inputting the adhesive layer thickness into the usage scenario stress model, outputting the thickness allowance after curing, and generating a thickness allowance adjustment command, including: Based on the stress distribution data, fit the stress change data to generate the stress model for the application scenario; based on the stress model for the application scenario, calculate the stress response value under different adhesive layer thicknesses to generate the thickness allowance after curing; based on the thickness allowance after curing, generate the thickness allowance adjustment instruction.

5. The method for monitoring the thickness of hot melt adhesive application based on deep learning according to claim 1, characterized in that, The process involves adjusting the curing time and temperature using a thickness pre-set adjustment command, simultaneously acquiring images of the adhesive layer surface in the crease area, processing these images to obtain thickness distribution measurement data, determining the crease stress concentration area based on the thickness distribution measurement data, and obtaining the actual thickness of the crease stress concentration area. This includes: Based on the thickness reservation adjustment instruction, the curing parameter table is queried to generate the adjusted curing process parameters; the adhesive layer is cured based on the adjusted curing process parameters, and the surface image of the adhesive layer in the crease area is acquired. The surface image of the adhesive layer in the crease area is processed to generate the thickness distribution measurement data; the thickness change rate is calculated based on the thickness distribution measurement data to determine the crease stress concentration area; the actual thickness of the crease stress concentration area is measured using laser ranging technology.

6. The method for monitoring the thickness of hot melt adhesive application based on deep learning according to claim 5, characterized in that, The process of processing the surface image of the adhesive layer in the crease area to generate the thickness distribution measurement data includes: The surface image of the adhesive layer in the crease area is segmented, the thickness change gradient between adjacent measurement points is calculated, and the boundary of the region where the thickness change gradient exceeds a threshold is identified. Based on the thickness data within the boundary of the region, the location of the depression is identified, and the depression locations are connected to generate a depression region. Based on the characteristics of the depression region, weak zones are identified, and the thickness distribution measurement data is generated based on the thickness data within the weak zones. Dangerous areas are determined, and a list of target areas is generated.

7. The method for monitoring the thickness of hot melt adhesive application based on deep learning according to claim 1, characterized in that, If the actual thickness of the stress concentration area of ​​the crease is lower than a preset thickness threshold, the spraying parameters are dynamically adjusted based on the thickness difference. The adjusted spraying parameters are then used to spray the stress concentration area of ​​the crease to obtain the target adhesive layer thickness distribution, including: The actual thickness of the stress concentration area of ​​the crease is compared with a preset thickness threshold to generate a thickness compensation amount; the parameter table is queried based on the thickness compensation amount to generate adjusted spraying parameters; the spraying equipment is controlled based on the adjusted spraying parameters to monitor the spraying thickness in real time and generate the target adhesive layer thickness distribution.

8. The method for monitoring the thickness of hot melt adhesive application based on deep learning according to claim 1, characterized in that, The step of spraying the crease stress concentration area with adjusted spraying parameters to obtain the target adhesive layer thickness distribution includes: The glue gun movement trajectory is generated based on the spatial distribution of the stress concentration area of ​​the crease; differentiated spraying parameters are set for the glue gun movement trajectory; the glue gun is controlled to perform spraying based on the differentiated spraying parameters, the thickness of each area is monitored in real time, the spraying parameters are adjusted based on the thickness deviation, and the target adhesive layer thickness distribution is generated.

9. The method for monitoring the thickness of hot melt adhesive application based on deep learning according to claim 1, characterized in that, The cracking risk value of the adhesive layer corresponding to the target adhesive layer thickness distribution is obtained through a preset cracking risk assessment model. If the cracking risk value of the adhesive layer is lower than a preset risk threshold, a target production line parameter adjustment plan is generated based on the risk assessment results, including: Input the target adhesive layer thickness distribution into the cracking risk assessment model to generate the adhesive layer cracking risk value; compare the adhesive layer cracking risk value with a preset risk threshold to generate a safety margin coefficient; query the parameter adjustment mapping table based on the safety margin coefficient to generate the target production line parameter adjustment plan.

10. The method for monitoring the thickness of hot melt adhesive application based on deep learning according to claim 1, characterized in that, The execution of the parameter adjustment scheme, which updates the parameters of the glue gun spraying and curing equipment in real time, results in stable brochure packaging quality, including: The production line dynamic control system receives the parameter adjustment scheme, converts the spraying speed into a pressure control value, generates a motor speed control command, sets the target temperature value of the heating controller, and adjusts the conveyor belt running cycle; it adjusts the spraying and curing equipment based on the converted parameters; and it monitors the adhesive layer thickness deviation and bonding force to obtain stable brochure packaging quality.

Images