Workpiece edge sealing control method and system based on artificial intelligence

Abstract

The invention provides a workpiece edge sealing control method and system based on artificial intelligence, and relates to the technical field of workpiece packaging, and the method comprises the steps: collecting visual pressure data and edge sealing defect data of a workpiece and an edge sealing belt in an edge sealing process; analyzing a corresponding relation between pressure abnormal data and edge sealing defect data in the visual pressure data, and dynamically correcting a workpiece feeding posture according to the corresponding relation; judging the coverage area of the adhesive layer according to the adhesion trace of the detection liquid, and marking a defect area; preprocessing the trace image through a U-Net model, and fitting the boundary of the defect area; constructing a pressure conduction model, analyzing a stress distribution rule corresponding to the defect area, and calculating the fine adjustment amount of the angle of the material guide plate; pID control is used for adjusting the position of the blocking piece in real time to change the edge sealing tension, and the edge sealing tension is matched with the material guiding angle and the pressure of the pressing wheel. According to the method, a full-process closed-loop optimization system is constructed, so that the method has remarkable advantages in detection precision, control efficiency and quality stability, and the edge sealing defect rate is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of workpiece edge sealing control, and in particular to a workpiece edge sealing control method and system based on artificial intelligence. BACKGROUND

[0002] Traditional workpiece edge sealing control relies on manual experience, and the core defects are the double lack of data and intelligence. Pressure sensing only relies on the feeling of the operator or simple instruments, and cannot capture the subtle fluctuations of the contact between the pressure wheel and the workpiece. Commonly, defects such as glue overflow and virtual sticking are caused by abnormal pressure, and problems are not found until the finished product is detected, and the abnormal recognition is seriously lagging.

[0003] The quality is related to the experience accumulation, and there is no systematic database to support the corresponding relationship between pressure abnormality and defects. The adjustment of the guide plate is mostly trial-and-error operation, and the angle deviation is often repeatedly corrected, which is extremely low in efficiency. The glue layer detection relies on naked eye observation, the defect area positioning is fuzzy, the coordinates and area of the insufficient coverage cannot be quantified, and the parameters cannot be accurately optimized in reverse. Each edge sealing parameter is independently adjusted, the pressure wheel pressure, the guide angle and the edge sealing belt tension lack a linkage mechanism, and there is no intelligent algorithm such as PID control for real-time adaptation, and the parameter adaptability is poor. The manual operation difference is large, the quality stability is insufficient in batch production, the defect rate fluctuates obviously, a closed-loop optimization system cannot be formed, and the high-precision edge sealing demand cannot be met. SUMMARY

[0004] The present application provides a workpiece edge sealing control method based on artificial intelligence, which solves the defects of insufficient quality stability in batch production and obvious defect rate fluctuation in the prior art.

[0005] The present application provides a workpiece edge sealing control method based on artificial intelligence, which includes: S1: collecting visible pressure data and edge sealing defect data of the workpiece and the edge sealing belt in the edge sealing process.

[0006] S2: analyzing the corresponding relationship between the pressure abnormality data in the visible pressure data and the edge sealing defect data, dynamically correcting the workpiece feeding posture according to the corresponding relationship, and obtaining a corrected edge sealing workpiece.

[0007] S3: fully contacting the corrected edge sealing workpiece with the sponge of the infiltration detection liquid, judging the glue layer coverage range according to the detection liquid trace, marking the defect area and collecting the trace image.

[0008] S4: preprocessing the trace image by using a U-Net model, fitting the boundary of the defect area by using a least square method, and outputting the coordinates and area of the defect area.

[0009] S5: constructing a pressure conduction model, inputting the coordinates and area of the defect area into the pressure conduction model, simulating and analyzing the stress distribution law corresponding to the defect area, and calculating the guide plate angle fine adjustment amount.

[0010] S6: Install adjustable baffles according to the fine adjustment amount of the guide plate angle, and use PID control to adjust the position of the adjustable baffles in real time to change the sealing tension, so as to match the sealing tension with the guide angle and the pressure of the pressure roller, and to achieve complete control of the sealing of the workpiece.

[0011] According to the workpiece edge sealing control method based on artificial intelligence provided by the present invention, the specific steps for obtaining the pressure change trend in step S1 are as follows: S11: Collects contact pressure fluctuations and converts the pressure change signal into a mechanical deformation signal.

[0012] S12: Adjust the lever fulcrum position and lever arm length according to the mechanical deformation signal, amplify the small deformation, and output the mechanical displacement.

[0013] S13: Visualize and quantify the mechanical displacement to obtain visual pressure data.

[0014] According to the workpiece edge sealing control method based on artificial intelligence provided by the present invention, the specific steps for adjusting the workpiece feeding posture in step S2 are as follows: S21: Establish a comparative database based on pressure curves under different working conditions from visible pressure data and historical edge sealing quality data.

[0015] S22: Based on the comparison database, use the correlation analysis model to analyze the characteristic parameters of abnormal fluctuations in the pressure change trend, match the characteristic parameters with the edge sealing defect types in the edge sealing defect data, calculate the defect incidence rate of different pressure anomaly modes, and output the correspondence between pressure anomalies and edge sealing defects.

[0016] S23: Analyze the workpiece posture under the tilted state of the feed end guide plate according to the corresponding relationship, and output the correlation model between the tilt angle of the guide plate and the workpiece posture deviation.

[0017] S24: Real-time monitoring of pressure change trends, calibration based on the correlation model to ensure the guide plate reaches the target angle, mechanical guidance to correct the workpiece feeding posture, and outputting the corrected edge-sealing workpiece. According to the workpiece edge-sealing control method based on artificial intelligence provided by this invention, the specific steps in step S22 for obtaining the correspondence between pressure anomalies and edge-sealing defects are as follows: Based on the comparative database, feature engineering is performed on the abnormal fluctuation data in the pressure change trend to construct a standardized feature vector.

[0018] The feature vectors are matched with the edge sealing defect types in the comparison database, and the support and confidence of the association rules are calculated.

[0019] The association rules with confidence levels higher than a preset threshold are selected, and the frequency and probability of pressure anomalies in historical data for each association rule are statistically analyzed to obtain the correspondence between pressure anomalies and edge sealing defects.

[0020] According to the workpiece edge sealing control method based on artificial intelligence provided by the present invention, the specific steps for acquiring the image of the detection liquid adhesion trace in step S3 are as follows: S31: Adjust the placement angle of the sponge to bring the workpiece being corrected and sealed into contact with the sponge soaked in the testing liquid.

[0021] S32: Determine the coverage area of ​​the sealing adhesive layer by detecting the interruption and narrowness of the liquid adhesion trace, and mark the adhesive layer defect area at the corresponding position on the workpiece.

[0022] S33: Collect images of defect marks in the adhesive layer defect area of ​​the workpiece and traces of the detection liquid adhesion on the sponge.

[0023] According to the workpiece edge sealing control method based on artificial intelligence provided by the present invention, the specific steps in step S4 of outputting the precise coordinates and area of ​​the insufficiently covered area are as follows: S41: The U-Net image segmentation model is used to extract deep features of the liquid adhesion trace image, the decoder restores the spatial information, segments the adhesive layer coverage area, and outputs the adhesive layer coverage feature mask.

[0024] S42: Compare the adhesive layer coverage feature mask with the standard adhesive layer coverage image feature pixel by pixel, calculate the difference between the pixel gray value and the texture feature, determine whether the difference exceeds the preset standard, and mark it as a suspected defect area to form a preliminary defect feature map.

[0025] S43: Use the least squares method to perform curve fitting on the edge pixels of the suspected defect area in the preliminary defect feature map to construct a continuous and smooth defect boundary contour, thus obtaining the defect boundary contour.

[0026] S44: Based on the defect boundary contour and image pixels, calculate the precise coordinates of the insufficiently covered area in the workpiece coordinate system through a coordinate transformation algorithm, count the number of pixels within the boundary to calculate the actual area, and output the precise coordinates and area of ​​the insufficiently covered area.

[0027] According to the workpiece edge sealing control method based on artificial intelligence provided by the present invention, the specific steps for forming a preliminary defect feature map in step S42 are as follows: Obtain the adhesive layer coverage feature mask and the standard adhesive layer coverage image features, and then perform grayscale processing on the adhesive layer coverage feature mask and the standard adhesive layer coverage image.

[0028] The Euclidean distance of pixel grayscale values ​​is calculated, and the texture feature vector of each pixel neighborhood is extracted using the directional gradient histogram algorithm. The texture difference is calculated using the cosine similarity formula to obtain the joint pixel difference.

[0029] The dynamic difference threshold is determined based on the difference distribution data of historical qualified samples. The pixel joint difference degree is compared with the dynamic difference threshold, and pixels with difference degree exceeding the threshold are screened. All pixels exceeding the threshold are integrated into suspected defect areas to generate a preliminary defect feature map.

[0030] According to the workpiece edge sealing control method based on artificial intelligence provided by the present invention, the specific steps for obtaining the angle fine-tuning amount of the feed end guide plate in step S5 are as follows: S51: Standardize the coordinates and area of ​​the defect region into geometric input parameters that the model can recognize, and output the quantitative features of the defect region.

[0031] S52: Input the quantitative features of the defect area into the pressure transmission model, calculate the force distribution data of the pressure roller corresponding to the defect area, and output the numerical matrix of the force distribution.

[0032] S53: Establish formulas for the guide plate angle and pressure roller force distribution based on the force distribution numerical matrix, analyze the influence of angle change on the force compensation of defect area, and output the correlation curve between angle adjustment and force improvement.

[0033] S54: Based on the correlation curve and with the optimization objective of uniform stress in the defect area, the optimal solution for fine-tuning the guide plate angle is obtained through a nonlinear programming algorithm, and the fine-tuning amount of the guide plate angle is output.

[0034] According to the workpiece edge sealing control method based on artificial intelligence provided by the present invention, the specific steps in step S6 for forming the edge sealing parameter optimization mechanism are as follows: S61: Install an adjustable baffle along the edge banding conveyor path according to the fine adjustment amount of the guide plate angle, and calibrate the initial position of the adjustable baffle.

[0035] S62: Set the target tension value of the sealing strip according to the initial position of the baffle, collect the actual value and the target value and input them into the PID controller to calculate the tension deviation.

[0036] S63: Adjust the position according to the tension deviation, dynamically change the contact pressure between the sealing strip and the baffle, and correct the tension deviation in real time.

[0037] S63: Monitor the tension adaptation effect and edge sealing quality feedback, transmit the adjustment data back to the system in real time, iteratively optimize the PID control parameters and baffle adjustment logic, and construct the workpiece edge sealing control mechanism.

[0038] The present invention also provides an artificial intelligence-based workpiece edge sealing control system, including: a pressure acquisition module, which acquires visual pressure data of the contact between the workpiece and the edge sealing strip during the edge sealing process to obtain the pressure change trend.

[0039] The attitude adjustment module correlates pressure change trends with historical edge banding quality data, filters the correspondence between pressure anomalies and edge banding defects, and dynamically adjusts the workpiece feeding attitude of the mechanical structure based on the correspondence.

[0040] The adhesive layer defect initial inspection module works by fully contacting the dynamically corrected edge-sealing workpiece with a sponge soaked in the detection liquid, determining the adhesive layer coverage area based on the adhesion marks of the detection liquid, marking the defect area, and acquiring trace images.

[0041] The defect analysis module preprocesses the trace image using the U-Net model, extracts adhesive layer features, compares them pixel-level with the standard image, fits the defect boundary using the least squares method, and outputs the coordinates and area of ​​the insufficiently covered area.

[0042] The fine-tuning calculation module uses the coordinates and area of ​​the adhesive layer defect area as the core input, and combines the pressure transmission model of the pressure roller to analyze the influence of the force distribution and calculate the fine-tuning amount of the guide plate angle.

[0043] The closed-loop control module, by adding an adjustable baffle based on the fine adjustment of the guide plate angle, adjusts the position of the baffle in real time through PID control to change the tension of the sealing strip, so that the tension is matched with the guide angle and the pressure of the pressure roller, thus forming a sealing parameter optimization mechanism.

[0044] This invention provides an artificial intelligence-based workpiece edge sealing control method. By constructing a complete chain of data acquisition, intelligent analysis, dynamic adjustment, precise detection, and closed-loop optimization, the beneficial effects achieved are as follows: Precise pressure sensing provides reliable data support for subsequent analysis. The method collects pressure data through customized spring pressure plates, selects suitable pressure plates by combining the characteristics of the pressure roller and material properties, ensures accurate mapping between deformation and pressure through pre-compression calibration, and then achieves clear capture of minute pressure fluctuations through lever amplification and visualization. This avoids data distortion problems from the source and lays a solid foundation for pressure change trend analysis.

[0045] It possesses strong intelligent correlation analysis capabilities, enabling precise localization of defect causes. Based on the gradient boosting tree algorithm, a matching model is constructed, integrating Apriori correlation analysis and anomaly detection technology. It mines the strong correlation between pressure anomalies and edge sealing defects from historical data, and combines finite element simulation to deduce the mechanical causes, breaking through the limitations of traditional experience-based judgment. This makes the adjustment of guide plate parameters more scientifically based and significantly improves the pertinence of feeding posture correction.

[0046] The system combines automation and high precision in defect detection, ensuring accurate quality assessment. The U-Net model is used to achieve precise segmentation of the adhesive layer area. Through pixel-level comparison of grayscale and texture features and least-squares boundary fitting, visual inspection errors are controlled to a minimum. The system automatically outputs defect coordinates and area, replacing subjective biases in manual inspection and providing quantitative data for subsequent parameter fine-tuning, thus improving the precision of quality control.

[0047] Parameter adjustment is closed-loop, constructing a dynamic optimization mechanism. Using defect data as input, the fine-tuning amount of the guide plate is calculated through an energy conservation model and pressure transmission analysis. This, combined with PID control of adjustable baffles, achieves dynamic tension adaptation, forming a closed-loop link of pressure monitoring, attitude adjustment, quality inspection, and parameter optimization. This linkage mechanism ensures that the guide angle, pressure roller pressure, and edge sealing tape tension are mutually adapted, continuously optimizing the edge sealing effect.

[0048] The method exhibits outstanding adaptability and robustness, making it suitable for various working conditions. It incorporates parameters such as workpiece material and edge banding characteristics into data processing, and eliminates interference from environmental and process differences through standardization and normalization. Model training relies on historical data from multiple working conditions, and cross-validation eliminates accidental correlations, ensuring stable operation of the system under different workpiece and material combinations, thus reducing changeover and adjustment costs.

[0049] In summary, this method deeply integrates artificial intelligence algorithms with mechanical control, realizing the digital and intelligent upgrade of the edge banding process, effectively improving the stability of edge banding quality, reducing manual intervention and defect rate, and providing an efficient and reliable technical solution for industrial edge banding production. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0051] Figure 1 This is a flowchart illustrating the steps of an artificial intelligence-based workpiece edge sealing control method provided in an embodiment of the present invention. Figure 2 This is a flowchart of a workpiece edge sealing control method based on artificial intelligence provided in an embodiment of the present invention; Figure 3 This is a block diagram of an artificial intelligence-based workpiece edge sealing control system provided in an embodiment of the present invention. Detailed Implementation

[0052] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0053] The following is combined Figures 1-3 This invention describes a workpiece edge sealing control method and system based on artificial intelligence.

[0054] Figure 1 This invention provides an artificial intelligence-based workpiece edge sealing control method.

[0055] like Figures 1-3 As shown in the figure, an embodiment of the present invention provides a workpiece edge sealing control method based on artificial intelligence, which mainly includes the following steps: S1: Collect visible pressure data of the contact between the workpiece and the sealing strip during the sealing process to obtain the pressure change trend.

[0056] S11: A spring pressure plate is installed at the contact position of the pressure roller to sense contact pressure fluctuations and convert pressure changes into mechanical deformation signals. The installation of the spring pressure plate at the pressure roller contact position involves first selecting a pressure plate with an elastic coefficient linear range suitable for the contact curvature of the pressure roller and the compressive strength characteristics of the sealing material. This ensures that the pressure plate can produce reproducible deformation within the pressure range of the sealing operation. Next, the pressure plate is fixed to the preset installation position on the pressure roller bracket using positioning pins and adjusting shims, ensuring that the force-bearing surface of the pressure plate is coplanar with the contact point between the pressure roller and the workpiece, preventing pressure transmission deviation. Finally, pre-pressure calibration is performed. By applying gradient standard pressure data and feeding back the deformation of the pressure plate, it is ensured that it can accurately sense dynamic pressure fluctuations during the sealing process and synchronously convert real-time pressure changes into continuous mechanical deformation signals.

[0057] S12: The mechanical deformation signal is used to adjust the lever fulcrum position and lever arm length, so that the small deformation of the pressure plate is amplified after being transmitted by the lever, and output as mechanical displacement. After acquiring the mechanical deformation signal output by the spring pressure plate, the amplitude and frequency of the deformation signal are collected in real time by the displacement sensor, converted into an electrical signal and transmitted to the control module. The control module then analyzes the current pressure fluctuation amplitude based on the strength of the deformation signal to determine whether the lever amplification ratio needs to be adjusted. Subsequently, the control module sends a drive command to the fulcrum adjustment mechanism, which drives the fulcrum seat to slide along the guide rail through the stepper motor, realizing the dynamic adjustment of the lever fulcrum position. At the same time, the linkage lever extension mechanism changes the length of the resistance arm to match the lever amplification factor with the deformation signal intensity. After the lever fulcrum and lever arm length are adapted and adjusted, the lever input end and the pressure plate deformation output end are flexibly connected by a ball joint to ensure efficient transmission of deformation force without additional resistance. Then, a transmission accuracy test is performed. A small deformation of known amplitude is applied to verify the linear relationship between the lever output displacement and the input deformation, eliminating errors caused by transmission gaps. Finally, under stable transmission conditions, the small deformation of the pressure plate is amplified by a preset factor after being transmitted by the lever, continuously outputting a precisely captured mechanical displacement.

[0058] S13: Set a visual indicator at the output end of the amplified mechanical displacement. Changes in data displacement will establish a correspondence between the amplified mechanical displacement and the original pressure fluctuations, and be transformed into a pressure change trend.

[0059] A lightweight pointer-type indicator is installed at the output end of the amplified mechanical displacement to ensure that the pointer can move synchronously with the displacement without generating inertial interference. Then, a transparent scale with micron-level graduations is placed in the corresponding area of ​​the pointer's movement trajectory. The scale range of the scale is adapted to the maximum output displacement of the lever, and pressure conversion reference values ​​are marked. Subsequently, the system is calibrated using a standard pressure source to establish a one-to-one mapping relationship between the displacement scale corresponding to different pressure values ​​and the original pressure fluctuations. Finally, by observing the position change of the pointer on the scale in real time, the displacement data is converted into an intuitive pressure change trend, providing a basis for judging the uniformity of fit.

[0060] S2: An intelligent matching model is constructed based on the gradient boosting tree algorithm. The pressure change trend is correlated with historical edge sealing quality data for training. The correspondence between pressure anomalies and edge sealing defects is screened through an anomaly detection algorithm. The parameters of the guide plate are determined through a multi-objective optimization model, and the workpiece feeding posture is dynamically adjusted in conjunction with the mechanical structure.

[0061] S21: Collect pressure change trend data generated by the current edge banding operation, and combine it with the pressure curves and corresponding edge banding effects under different working conditions in the historical edge banding quality data to establish a database containing pressure fluctuation ranges, defect types, and process parameters. Based on the pressure change trend data, preprocess the currently collected pressure curves to remove high-frequency noise caused by equipment vibration and anomalies caused by data acquisition delays, ensuring that the pressure fluctuation characteristics truly reflect the bonding state. Extract the pressure curves corresponding to different workpiece materials and edge banding types from the historical edge banding quality data, as well as the corresponding defect data such as glue overflow and glue separation. Finally, according to the working condition classification standard, structure and store the preprocessed current data and historical data, supplement process parameter remarks information, and establish a comparison database containing pressure characteristics, defect types, and working condition parameters.

[0062] S22: Based on the established comparison database, the characteristic parameters of abnormal fluctuations in the pressure change trend are mined using the correlation analysis model. These parameters are then matched with the edge sealing defect types in historical data. The defect incidence rates corresponding to different pressure anomaly patterns are statistically analyzed, and the correspondence between pressure anomalies and edge sealing defects with strong correlations is selected.

[0063] Based on the comparative database, feature engineering was performed on the abnormal fluctuation data in the pressure change trend to extract key parameters that can accurately characterize pressure anomalies, such as peak value, duration, and fluctuation frequency. Normalization was used to eliminate the dimensional differences between different parameters, and a standardized feature vector with a unified dimension that can be directly used for analysis was constructed.

[0064] Using the constructed standardized feature vectors, the Apriori algorithm in the association analysis model is called to match the feature vectors with the various edge sealing defect types recorded in the comparison database one by one. The system traverses all feature combinations and solves the support and confidence of the corresponding association items, quantifying the association strength between different feature vectors and edge sealing defects.

[0065] Based on the calculated support and confidence scores, association rules with confidence scores higher than a preset threshold are selected. The frequency and probability of defects occurring in historical data for each rule's corresponding stress anomaly pattern are statistically analyzed. False associations caused by data randomness are eliminated through multiple sets of cross-validation. Finally, a statistically significant correspondence between stress anomalies and edge sealing defects is obtained.

[0066] S23: Based on the correlation between the screened pressure anomalies and edge sealing defects, analyze the mechanical causes of the pressure anomalies. Combined with the force transmission path during workpiece feeding, infer the influence of the current tilt state of the feed guide plate on the workpiece posture, and then determine the tilt direction and angle range that the guide plate needs to be adjusted to ensure that the pressure distribution can be improved after adjustment. First, deduce the mechanical transmission path of the pressure anomalies in reverse, analyze the causes of uneven pressure distribution when the workpiece contacts the edge sealing strip, and determine whether it is caused by the deviation of the feeding posture. Then, simulate the force state of the workpiece under different guide plate tilt angles through finite element method, compare the simulation results with the actual pressure anomaly characteristics, clarify the interference mechanism of the current tilt state of the guide plate on the workpiece posture, and calculate the tilt direction and angle range that the guide plate needs to be adjusted to ensure that the contact pressure distribution can be balanced after adjustment.

[0067] S24: Based on the determined tilt direction and angle parameters of the guide plate, the mechanical adjustment mechanism drives the guide plate to rotate around the hinge point. During the adjustment process, the feedback of pressure change trend is monitored in real time. Through fine-tuning calibration, the guide plate is made to accurately reach the target angle. At the same time, the workpiece positioning device is linked to correct the workpiece feeding posture through mechanical guidance. First, the tilt angle data of the guide plate is collected in real time by the displacement sensor, and the dynamic feedback of pressure change trend is monitored synchronously. Then, the adjustment effect is judged based on the stability of the pressure curve. If there is still local pressure abnormality, the guide plate angle is corrected at the millimeter level through the fine-tuning mechanism. At the same time, the side push roller and positioning pin in the workpiece positioning device are linked to correct the workpiece feeding guide trajectory in real time according to the adjustment angle of the guide plate. Finally, multiple sets of sealing tests are conducted to verify the uniformity of pressure distribution, ensuring that the workpiece feeding posture is accurately corrected through mechanical guidance and eliminating pressure abnormality problems.

[0068] S3: After dynamic correction and edge sealing, the workpiece is brought into full contact with the sponge soaked in the test liquid. The coverage of the edge sealing adhesive layer is determined based on the adhesion traces of the test liquid on the sponge. Defective areas of the adhesive layer are marked, and images of the adhesion traces of the test liquid are collected.

[0069] S31: Adjust the placement angle and height of the sponge to ensure that the workpiece with the sealed edge after dynamic correction is in full contact with the sponge immersed in the testing liquid.

[0070] S32: Observe the coverage width distribution of the test liquid adhesion traces left after the workpiece comes into contact with the sponge. Determine the coverage range of the sealing adhesive layer by the interruption and narrowness of the traces, and mark the adhesive layer defect area at the corresponding position on the workpiece.

[0071] S33: Collect images of the defect markings on the workpiece in the adhesive layer defect area and the traces of the inspection liquid adhering to the sponge.

[0072] S4: The U-Net image segmentation model is used to preprocess the trace image, extract the features of the adhesive layer coverage area, compare the features with the standard image at the pixel level, fit the defect boundary using the least squares method, and automatically output the accurate coordinates and area of ​​the insufficiently covered area.

[0073] S41: The U-Net image segmentation model is used to preprocess the liquid adhesion trace images. The encoder extracts deep features from the image, and the decoder restores spatial information, accurately segmenting the adhesive layer coverage area from the background and outputting an adhesive layer coverage feature mask. The U-Net image segmentation model is used to preprocess the detected liquid adhesion trace images. First, adaptive histogram equalization is performed on the original image to enhance the contrast between the adhesive layer traces and the background. Simultaneously, Gaussian filtering is used to remove shooting noise and improve image quality. Then, the preprocessed image is input into the U-Net model encoder, which gradually extracts deep semantic features through convolution and pooling operations. Finally, the decoder uses upsampling and skip connections to restore spatial information, accurately segmenting the adhesive layer coverage area from the background and outputting a binarized adhesive layer coverage feature mask.

[0074] S42: Compare the adhesive layer coverage feature mask with the preset standard adhesive layer coverage image features pixel by pixel, calculate the difference between pixel grayscale values ​​and texture features, and mark suspected defect areas where the difference exceeds a threshold to form a preliminary defect feature map. First, perform grayscale normalization and resolution alignment on both to eliminate interference from differences in ambient lighting and shooting parameters. Then, calculate the absolute difference of grayscale values ​​pixel by pixel. Simultaneously, use a local binary mode algorithm to extract texture features and calculate similarity to obtain a joint difference matrix. Finally, set a dynamic threshold based on historical data training, mark pixels with differences exceeding the threshold, and form a preliminary defect feature map containing suspected defect areas.

[0075] S43: The least squares method is used to perform curve fitting on the edge pixels of the suspected defect region in the preliminary defect feature map to construct a continuous and smooth defect boundary contour. The least squares method is used to perform curve fitting on the edge pixels of the suspected defect region in the preliminary defect feature map. First, morphological closing operations are used to fill the small holes in the defect region and remove isolated noise points. Then, the Canny edge detection algorithm is used to extract the edge pixel coordinates of the suspected defect region. Finally, the edge pixels are substituted into the least squares fitting formula to construct a continuous and smooth polynomial curve that minimizes the error, resulting in an accurate defect boundary contour.

[0076] S44: Based on the defect boundary contour and image pixels, the precise coordinates of the under-covered area in the workpiece coordinate system are calculated using a coordinate transformation algorithm. The actual area is calculated by counting the number of pixels within the boundary, and the precise coordinates and area of ​​the under-covered area are automatically output. Based on the defect boundary contour and image pixels, the transformation relationship between the image coordinate system and the workpiece coordinate system is first established using camera calibration parameters to determine the conversion ratio between pixels and actual physical dimensions. Then, based on the pixel coordinates of the defect boundary contour, the precise coordinates of the four vertices of the under-covered area in the workpiece coordinate system are calculated using a coordinate transformation algorithm. Finally, the effective number of pixels within the boundary contour is counted, and the actual area is calculated using the conversion ratio. The precise coordinates and area data of the under-covered area are automatically output through the data interface.

[0077] S5: Using the coordinates and area of ​​the insufficient adhesive layer coverage area as the core input parameters, analyze the influence of the pressure transmission model of the edge banding machine's pressure roller on the force distribution of the workpiece, and calculate the angle fine adjustment of the feed end guide plate.

[0078] S51: Collect the precise coordinates and area data of areas with insufficient adhesive layer coverage, and input them as core parameters into the edge banding machine pressure analysis system. The system calls the preset pressure transmission model of the pressure roller to initially associate the defect data with the pressure distribution of the pressure roller.

[0079] S52: Simulate the force distribution of the workpiece under different pressure values ​​based on the associated information, locate the key stress points caused by uneven pressure transmission leading to insufficient adhesive layer coverage through comparative analysis, and determine the influence law between the pressure roller pressure and the workpiece posture.

[0080] S53: Based on the law of pressure transmission, combined with auxiliary parameters such as workpiece material and edge banding characteristics, the angle calculation module is activated to deduce the angle fine adjustment amount of the feed end guide plate in reverse, so as to ensure that the workpiece is subjected to uniform force after fine adjustment in order to improve the adhesive layer coverage problem.

[0081] The total elastic potential energy of the workpiece in contact with the edge banding is equal to the integral of the elastic potential energy of each contact element. The energy loss before fine-tuning is:

[0082] In the formula, E total The total elastic potential energy of the contact system is determined by the output power of the edge banding machine's pressure roller and the contact time, where A is the contact area. For contact stress distribution, For strain distribution, ΔE represents the elastic potential energy density per unit volume, and ΔE represents the energy loss measure, reflecting the degree of uneven force distribution.

[0083] The guide plate angle adjustment corrects the energy distribution by changing the workpiece contact posture, and its angle fine-tuning and energy loss satisfy the following:

[0084] In the formula, Δθ is the fine adjustment amount of the guide plate angle, k is the angle-to-energy conversion coefficient, which is calibrated by the equipment structural parameters, θ0 is the initial angle of the guide plate, cosθ0 is the effective component of the angle correction, and E w The workpiece's elastic modulus. Let b be the moment of inertia of the workpiece cross section. w h is the workpiece width. w E represents the workpiece thickness. t The elastic modulus of the edge banding. Let b be the moment of inertia of the edge banding section. t h represents the width of the edge banding. t This refers to the thickness of the edge banding tape. .

[0085] S6: Based on the fine-tuning amount of the guide plate angle, an adjustable baffle is installed next to the edge sealing belt conveyor path to change the conveying tension of the edge sealing belt. The position of the baffle is adjusted in real time through the PID control algorithm to change the conveying tension of the edge sealing belt, so that the conveying state of the edge sealing belt is adapted to the fine-tuned guide angle and pressure roller pressure, forming an edge sealing parameter optimization mechanism.

[0086] Based on the fine-tuning of the guide plate angle, intelligent adjustable baffles are precisely installed next to the edge banding conveyor path. The initial position and installation accuracy of the baffles are calibrated to ensure that the movement trajectory of the baffles matches the conveying direction of the edge banding, thus laying a mechanical foundation for tension adjustment.

[0087] Based on the initial position of the baffle, and combined with the material guide angle and pressure roller pressure parameters, the target tension value of the sealing strip is set. The actual value and target value collected in real time by the tension sensor are input into the PID controller to calculate the tension deviation.

[0088] Based on the deviation adjustment signal output by the PID controller, the baffle actuator is driven to make fine adjustments to its position, dynamically changing the contact pressure between the sealing strip and the baffle, correcting the tension deviation in real time, and making the tension precisely matched with the guide angle and the pressure roller pressure.

[0089] Continuously monitor the tension adaptation effect and edge sealing quality feedback, and transmit the adjustment data back to the parameter optimization system in real time to iteratively optimize the PID control parameters and baffle adjustment logic, forming a closed-loop linkage edge sealing parameter optimization mechanism.

[0090] like Figure 3 As shown, the present invention also provides an artificial intelligence-based workpiece edge sealing control system, comprising: The pressure acquisition module collects visual pressure data of the contact between the workpiece and the sealing strip during the sealing process, and obtains the pressure change trend.

[0091] The attitude adjustment module correlates pressure change trends with historical edge banding quality data, filters the correspondence between pressure anomalies and edge banding defects, and dynamically adjusts the workpiece feeding attitude of the mechanical structure based on the correspondence.

[0092] The adhesive layer defect initial inspection module works by fully contacting the dynamically corrected edge-sealing workpiece with a sponge soaked in the detection liquid, determining the adhesive layer coverage area based on the adhesion marks of the detection liquid, marking the defect area, and acquiring trace images.

[0093] The defect analysis module preprocesses the trace image using the U-Net model, extracts adhesive layer features, compares them pixel-level with the standard image, fits the defect boundary using the least squares method, and outputs the coordinates and area of ​​the insufficiently covered area.

[0094] The fine-tuning calculation module uses the coordinates and area of ​​the adhesive layer defect area as the core input, and combines the pressure transmission model of the pressure roller to analyze the influence of the force distribution and calculate the fine-tuning amount of the guide plate angle.

[0095] The closed-loop control module, by adding an adjustable baffle based on the fine adjustment of the guide plate angle, adjusts the position of the baffle in real time through PID control to change the tension of the sealing strip, so that the tension is matched with the guide angle and the pressure of the pressure roller, thus forming a sealing parameter optimization mechanism.

[0096] This invention provides an artificial intelligence-based workpiece edge sealing control method. By constructing a complete chain of data acquisition, intelligent analysis, dynamic adjustment, precise detection, and closed-loop optimization, the beneficial effects achieved are as follows: Precise pressure sensing provides reliable data support for subsequent analysis. The method collects pressure data through customized spring pressure plates, selects suitable pressure plates by combining the characteristics of the pressure roller and material properties, ensures accurate mapping between deformation and pressure through pre-compression calibration, and then achieves clear capture of minute pressure fluctuations through lever amplification and visualization. This avoids data distortion problems from the source and lays a solid foundation for pressure change trend analysis.

[0097] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of embodiments.

[0098] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. These modifications or substitutions do not cause the essence of the corresponding technical solutions to depart from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An artificial intelligence-based workpiece edge sealing control method, characterized by, Comprise: S1: collect visible pressure data and edge sealing defect data of workpiece and edge sealing tape in the edge sealing process; S2: analyze the corresponding relationship between the abnormal pressure data in the visible pressure data and the edge sealing defect data, dynamically correct the workpiece feeding posture according to the corresponding relationship, and obtain the corrected edge sealing workpiece; S3: the corrected edge sealing workpiece is fully contacted with the sponge of the infiltration detection liquid, the coverage range of the adhesive layer is judged according to the trace of the detection liquid, the defect area is marked and the trace image is collected; S4: the trace image is preprocessed by U-Net model, the boundary of the defect area is fitted by using least square method, and the coordinates and area of the defect area are output; S5: construct a pressure conduction model, input the coordinates and area of the defect area into the pressure conduction model, simulate and analyze the stress distribution law corresponding to the defect area, and calculate the angle adjustment amount of the guide plate; S6: according to the angle adjustment amount of the guide plate, install adjustable baffle, use PID control to adjust the position of the adjustable baffle in real time, and then change the edge sealing tension, adapt the edge sealing tension to the guide angle and pressure roller pressure, and control the workpiece edge sealing.

2. The workpiece edge sealing control method based on artificial intelligence according to claim 1, characterized in that, In step S1, the specific steps of obtaining visible pressure data are: S11: collect contact pressure fluctuation, convert pressure change signal into mechanical deformation signal; S12: adjust the lever fulcrum position and force arm length according to the mechanical deformation signal, amplify the micro deformation, and output mechanical displacement; S13: the mechanical displacement is visualized and quantified to obtain the visible pressure data. 3.The workpiece edge sealing control method based on artificial intelligence according to claim 1, wherein, In step S2, the specific steps of adjusting the workpiece feeding posture are: S21: establish a comparison database according to the pressure curves in different working conditions in the visible pressure data and historical edge sealing quality data; S22: according to the comparison database, use correlation analysis model to analyze the characteristic parameters of abnormal fluctuation in pressure change trend, match the characteristic parameters with the edge sealing defect types in the edge sealing defect data, calculate the defect occurrence rate of different pressure abnormal modes, and output the corresponding relationship between pressure abnormality and edge sealing defect; S23: according to the corresponding relationship, analyze the workpiece posture of the inclined state of the feeding end guide plate, output the correlation model of the inclination angle of the guide plate and the deviation of the workpiece posture; S24: real-time monitor the pressure change trend, calibrate the guide plate to reach the target angle according to the correlation model, correct the workpiece feeding posture through mechanical guidance, and output the corrected edge sealing workpiece.

4. The workpiece edge sealing control method based on artificial intelligence according to claim 3, characterized in that, In step S22, the specific steps of obtaining the corresponding relationship between pressure abnormality and edge sealing defect are: According to the comparison database, the feature engineering processing is carried out on the abnormal fluctuation data in the pressure change trend, and the standardized feature vector is constructed; Match the feature vector with the edge sealing defect type in the comparison database, calculate the support and confidence of the association rule; Screen out the association rules with confidence higher than the preset threshold, count the defect occurrence frequency and probability of each association rule pressure abnormality in historical data, and obtain the corresponding relationship between pressure abnormality and edge sealing defect.

5. The workpiece edge sealing control method based on artificial intelligence according to claim 1, characterized in that, In step S3, the specific steps of collecting the trace image of the detection liquid are: S31: adjust the placing angle of the sponge, contact the corrected edge sealing workpiece with the sponge of the infiltration detection liquid; S32: Determine the coverage range of the sealing adhesive layer by detecting the interruption and narrowness of the detection liquid attachment trace, and mark the adhesive layer defect area at the corresponding position of the workpiece; S33: Collect the defect mark of the adhesive layer defect area of the workpiece and the trace image of the detection liquid attachment trace on the sponge.

6. The workpiece edge sealing control method based on artificial intelligence according to claim 1, characterized in that, In step S4, the specific steps for outputting the precise coordinates and area of the insufficient coverage area are as follows: S41: Extract deep features of the liquid attachment trace image using a U-Net image segmentation model, restore spatial information with a decoder, segment the adhesive layer coverage area, and output an adhesive layer coverage feature mask; S42: Compare the adhesive layer coverage feature mask with the standard adhesive layer coverage image feature pixel by pixel, calculate the difference degree of pixel gray value and texture feature, and determine whether the difference degree exceeds the preset standard. If yes, mark as a suspected defect area and form a preliminary defect feature map; S43: Use the least squares method to perform curve fitting on the edge pixel points of the suspected defect area of the preliminary defect feature map, construct a continuous and smooth defect boundary contour, and obtain the defect boundary contour; S44: According to the defect boundary contour and image pixels, calculate the precise coordinates of the insufficient coverage area in the workpiece coordinate system through a coordinate conversion algorithm, count the number of pixels within the boundary to calculate the actual area, and output the precise coordinates and area of the insufficient coverage area.

7. The workpiece edge sealing control method based on artificial intelligence according to claim 6, characterized in that, In step S42, the specific steps for forming a preliminary defect feature map are as follows: Obtain the adhesive layer coverage feature mask and the standard adhesive layer coverage image feature, and perform gray processing on the adhesive layer coverage feature mask and the standard adhesive layer coverage image; Calculate the Euclidean distance of the pixel gray value, extract the texture feature vector of each pixel neighborhood using the histogram of oriented gradients algorithm, calculate the texture difference through the cosine similarity formula, and obtain the joint difference degree of the pixels; Determine the dynamic difference threshold value according to the difference distribution data of historical qualified samples, compare the joint difference degree of the pixels with the dynamic difference threshold value, filter the pixels with a difference degree exceeding the threshold value, integrate all the pixels exceeding the threshold value into a suspected defect area, and generate a preliminary defect feature map. 8.The workpiece edge sealing control method based on artificial intelligence according to claim 1, wherein, In step S5, the specific steps for obtaining the angle fine tuning amount of the feed end guide plate are as follows: S51: Standardize the defect area coordinates and area into geometric input parameters recognizable by the model, and output the defect area quantitative features; S52: Input the defect area quantitative features into the pressure conduction model, calculate the pressure roller stress distribution data corresponding to the defect area, and output the stress distribution numerical matrix; S53: Establish the guide plate angle and pressure roller stress distribution formula according to the stress distribution numerical matrix, analyze the influence law of angle change on defect area stress compensation, and output the correlation curve of angle adjustment amount and stress improvement degree; S54: According to the correlation curve, take the stress uniformization of the defect area as the optimization target, solve the optimal solution of the guide plate angle fine tuning through the nonlinear programming algorithm, and output the guide plate angle fine tuning amount. 9.The workpiece edge sealing control method based on artificial intelligence according to claim 1, wherein, In step S6, the specific steps for forming the sealing parameter optimization mechanism are as follows: S61: Install an adjustable baffle beside the sealing tape conveying path according to the guide plate angle fine tuning amount, and calibrate the initial position of the adjustable baffle; S62: Set the edge band target tension value according to the initial position of the baffle, collect the actual value and the target value input the PID controller, and calculate the tension deviation; S63: According to the tension deviation, the position is fine-tuned, the contact pressure of the edge band and the baffle is dynamically changed, and the tension deviation is corrected in real time; S63: Monitor the tension adaptation effect and edge sealing quality feedback, and real-time feedback the adjustment data to the system, iteratively optimize the PID control parameters and baffle adjustment logic, and build a workpiece edge sealing control mechanism.

10. An artificial intelligence-based workpiece edge sealing control system employing an artificial intelligence-based workpiece edge sealing control method according to any one of claims 1 to 9, characterized by The control system comprises: A pressure collection module is configured to collect the contact visible pressure data of the workpiece and the edge band during the edge sealing process, and obtain the pressure change trend; The attitude adjustment module is configured to associate the pressure change trend with historical edge sealing quality data, screen the corresponding relationship between pressure abnormalities and edge sealing defects, and dynamically adjust the workpiece feeding attitude according to the corresponding relationship; The glue layer defect preliminary inspection module is configured to fully contact the dynamically corrected edge sealing workpiece with the sponge of the immersion detection liquid, judge the glue layer coverage range according to the detection liquid trace, mark the defect area and collect the trace image; The defect analysis module is configured to preprocess the trace image by U-Net model, extract the glue layer features, compare with the standard image pixel level, fit the defect boundary by least square method, and output the coordinates and area of the insufficient coverage area; The fine-tuning amount calculation module is configured to take the coordinates and area of the glue layer defect area as the core input, analyze the stress distribution influence combined with the pressure roller pressure conduction model, and calculate the guide plate angle fine-tuning amount; The closed-loop control module is configured to install an adjustable baffle according to the guide plate angle fine-tuning amount, and adjust the baffle position in real time through PID control to change the edge band tension, so that the tension, guide angle and pressure roller pressure are adapted, and an edge sealing parameter optimization mechanism is formed.

Images