Dispensing and curing method and apparatus

Abstract

This invention discloses a dispensing curing method and dispensing equipment, relating to the field of dispensing technology. The dispensing curing method includes: acquiring the position information of a workpiece at the current dispensing station; calculating a first offset between the actual position of the workpiece and a standard position; performing coordinate fusion calculation between the first offset and a preset standard dispensing trajectory to generate a first compensated dispensing trajectory; controlling a needle to perform a dispensing operation along the first compensated dispensing trajectory to obtain a first dispensing path; after the dispensing operation is completed, controlling a curing device to perform a curing operation on the first dispensing path; after the curing operation is completed, acquiring an image of the first dispensing path; calculating the deviation value between the first dispensing path and the standard dispensing path; and performing trend analysis based on the accumulated deviation values ​​of historical dispensing stations to generate a second compensated dispensing trajectory. This method solves the problem in existing dispensing systems that cannot track dynamic offset trends during continuous processing.

Description

Technical Field

[0001] This invention relates to the field of dispensing technology, and in particular to a dispensing curing method and dispensing equipment. Background Technology

[0002] In precision manufacturing processes such as dispensing adhesive for smart speaker speakers, traditional processing methods suffer from fundamental defects such as a lack of visual guidance and separation of dispensing and curing processes. These defects can easily lead to product displacement during transit, affecting sealing and sound quality. To address these fundamental defects, existing technologies are increasingly incorporating visual guidance mechanisms. These mechanisms acquire the workpiece position and perform preliminary trajectory compensation before dispensing, while simultaneously combining this with the curing operation to improve the stability of each processing step.

[0003] However, existing vision-guided dispensing trajectory compensation methods still have significant limitations: they typically only perform static, single-shot position corrections based on a single visual measurement. In actual continuous automated manufacturing processes (such as multi-station flow processing or long-cycle continuous operations), processing equipment is often affected by complex physical factors such as mechanical thermal deformation and wear of moving mechanisms, resulting in systematic positional offsets with a continuously evolving trend. Because existing static single-shot compensation cannot sense, record, and track this dynamic offset trend that crosses stations or accumulates over time, the dispensing system lacks adaptive correction capabilities throughout the entire processing cycle. Ultimately, this leads to increasingly larger deviations between the dispensing trajectory of subsequent processing stations and the theoretical trajectory, making it impossible to guarantee the consistency of adhesive path quality in continuous production. Summary of the Invention

[0004] The main objective of this invention is to propose a dispensing curing method and dispensing equipment, which aims to solve the problem that existing dispensing systems cannot track dynamic offset trends during continuous processing.

[0005] To achieve the above objectives, the present invention provides a dispensing and curing method, comprising: Obtain the position information of the workpiece at the current dispensing station; Calculate the first offset between the actual position of the workpiece and the standard position, and perform coordinate fusion calculation between the first offset and the preset standard dispensing trajectory to generate the first compensated dispensing trajectory. Control the needle to perform dispensing operation along the first compensated dispensing trajectory to obtain the first dispensing path; After the dispensing operation is completed, the curing device is controlled to perform a curing operation on the first dispensing path; After the curing operation is completed, an image of the first dispensing path is acquired; Calculate the deviation between the first dispensing path and the standard dispensing path; A trend analysis is performed based on the accumulated deviation values ​​of historical dispensing stations to generate a second compensation dispensing trajectory. The workpiece is switched to the next dispensing station. The second compensated dispensing trajectory is used as the execution reference trajectory of the next dispensing station. The needle is controlled to perform dispensing operation along the second compensated dispensing trajectory to obtain a second dispensing path. The second dispensing path is used as the new first dispensing path. The process then returns to the step of controlling the curing device to cure the first dispensing path.

[0006] In one embodiment, the step of obtaining the position information of the workpiece at the current dispensing station includes: The industrial camera is controlled to take an initial picture of the current dispensing station, and the acquired initial image is matched with a template based on preset geometric features. Determine whether the preset geometric feature is successfully matched in the initial image; If the matching fails, a mode switching command is generated to control the smart camera to take a second picture of the current dispensing station. The pre-trained deep learning recognition model is invoked to extract features from the image obtained by the second capture, and the high-dimensional feature coordinates of the workpiece are output as the position information.

[0007] In one embodiment, after the step of outputting the high-dimensional feature coordinates of the workpiece as the position information, the method further includes: Extract the confidence score of the high-dimensional feature coordinates; The confidence score is compared with a preset feature confidence threshold; If the confidence score is lower than the feature confidence threshold, it is determined to be a visual recognition anomaly, the control device is shut down and a manual intervention signal is triggered.

[0008] In one embodiment, the step of performing coordinate fusion calculation between the first offset and a preset standard dispensing trajectory to generate a first compensated dispensing trajectory includes: The control line laser sensor projects line structured light onto the surface of the workpiece at the current dispensing station, and synchronously moves the line laser sensor to acquire multiple frames of line laser stripe images; The multiple frames of the line laser stripe images are stitched together and subjected to three-dimensional calculation to generate three-dimensional point cloud data containing the three-dimensional morphological features of the workpiece surface; Extract multiple discrete points within the current dispensing trajectory planning area from the three-dimensional point cloud data, and calculate the surface normal vector perpendicular to the workpiece surface at each discrete point. Obtain the standard needle movement posture vector of the needle in the initial state, perform a rotation matrix transformation on the standard needle movement posture vector according to the direction of the surface normal vector, and combine the transformed posture vector with the first offset to generate the first compensated dispensing trajectory adapted to the current surface.

[0009] In one embodiment, the step of controlling the needle to perform the dispensing operation along the first compensated dispensing trajectory includes: Extract the Z-axis coordinate components of the surface normal vector in the vertical direction; Based on the preset fixed glue breakage distance, the Z-axis coordinate components are added to the fixed glue breakage distance to generate a real-time following height curve for guiding the Z-axis height change; During the process of controlling the needle to move along the horizontal plane, the Z-axis drive motor is subjected to high-frequency interpolation control according to the real-time following height curve; The actual encoder position of the Z-axis drive motor is read in real time, and the actual encoder position is compared with the real-time following height curve in a closed-loop deviation comparison.

[0010] In one embodiment, the step of performing trend analysis on the accumulated deviation values ​​from historical dispensing stations to generate a second compensated dispensing trajectory includes: Extract the X-axis and Y-axis deviation components of the accumulated deviation values ​​from historical dispensing stations in a preset rectangular coordinate system; Based on the rotation angle of the corresponding dispensing station, the X-axis deviation component and the Y-axis deviation component are synthesized by trigonometric functions and converted into a deviation vector with spatial directionality. The deviation vector is projected onto a polar coordinate system with the rotation center as the pole to generate polar coordinate deviation points containing amplitude and polar angle, and a polar coordinate distribution map is constructed in ascending order of rotation angle. The least squares ellipse fitting operation is performed on all polar coordinate deviation points in the polar coordinate distribution diagram. The polar coordinate compensation point is calculated by interpolation on the fitted ellipse curve based on the estimated rotation angle of the next glue station. The polar coordinate compensation point is reverse-calculated into a compensation offset in a rectangular coordinate system. The compensation offset is then fused with a preset standard dispensing trajectory to generate the second compensation dispensing trajectory.

[0011] In one embodiment, after the steps of reversing the polar coordinate compensation point to a compensation offset in a Cartesian coordinate system, and performing coordinate fusion calculations between the compensation offset and a preset standard dispensing trajectory to generate the second compensation dispensing trajectory, the method further includes: Count the total number of historical workstations that have performed dispensing operations and successfully obtained polar coordinate deviation points; Calculate the ratio of the total number of historical workstations to the preset total number of glue dispensing workstations for the workpiece, and use the ratio as the data confidence level of the polar coordinate distribution map. The data confidence level is compared with a preset allowable overlap threshold in real time; If the data confidence level is less than the allowable superposition threshold, the second compensation dispensing trajectory is discarded, and the execution trajectory of the next dispensing station is generated only based on the first offset; if the data confidence level is greater than or equal to the allowable superposition threshold, the second compensation dispensing trajectory is output as the execution trajectory.

[0012] In one embodiment, the step of performing trend analysis on the accumulated deviation values ​​from historical dispensing stations to generate a second compensated dispensing trajectory includes: Extract the X-axis and Y-axis deviation components of the accumulated deviation values ​​from historical dispensing stations in a preset rectangular coordinate system; Based on the rotation angle of the corresponding dispensing station, the X-axis deviation component and the Y-axis deviation component are synthesized by trigonometric functions and converted into a deviation vector with spatial directionality. The deviation vector is projected onto a polar coordinate system with the rotation center as the pole to generate polar coordinate deviation points containing amplitude and polar angle, and a polar coordinate distribution map is constructed in ascending order of rotation angle. Extract the amplitude difference and time interval between the polar coordinate deviation points of two adjacent dispensing stations from the polar coordinate distribution map; Divide the amplitude difference by the time interval to calculate the real-time rate of change of the deviation amplitude over time; Retrieve the real-time change rate of multiple recent dispensing stations and determine whether the real-time change rate exhibits a monotonic change pattern of continuous increase or continuous decrease. If a monotonic variation pattern is observed, a univariate linear regression equation is established based on the monotonic variation pattern regarding time and deviation amplitude. The estimated execution time of the next dispensing station is substituted into the univariate linear regression equation to calculate the predicted deviation amplitude. The predicted deviation amplitude is then fused with the preset standard dispensing trajectory to generate the second compensated dispensing trajectory.

[0013] In one embodiment, the step of controlling the curing device to perform a curing operation on the first dispensing path includes: After detecting the signal that the first dispensing path has been completed, the UV curing lamp is controlled to move directly above the first dispensing path and emit UV curing light according to the first preset power and first preset duration calibrated by the previous peel force test. The irradiation stops when the first preset duration is reached, so that the adhesive reaches the preset surface dry state. After irradiation stops, the UV curing lamp is controlled to move out of the visual detection field of view, and the step of acquiring the image of the first dispensing path is triggered; After image acquisition is completed, during the physical gap when the workpiece is switching workstations, the UV curing lamp is controlled to perform secondary penetration irradiation on the first dispensing path at a second preset power greater than the first preset power. The curing time is used to control whether the adhesive curing is complete. When the duration of the secondary penetration irradiation reaches the second preset time, the adhesive curing is determined to be complete, and the step of controlling the workpiece to switch to the next adhesive station is allowed.

[0014] The present invention also provides a dispensing device, wherein the dispensing device performs dispensing processing using the dispensing curing method described above.

[0015] The dispensing and curing method provided by this invention solves the problem that existing dispensing systems cannot track dynamic offset trends during continuous processing by introducing a cyclical mechanism of "dispensing-curing-re-inspection-trend prediction-re-dispensing". Specifically, a first offset is used to achieve basic positioning, obtain a first dispensing path, and cure it; after curing, the solid glue line is image-re-inspected to obtain the deviation value; the historically accumulated deviation value is used for trend analysis to proactively generate a second compensating dispensing trajectory that includes compensation prediction; in the next station, the second compensating dispensing trajectory is directly used for dispensing and the cycle repeats. Through the above settings, the dispensing system is endowed with the ability to self-evolve and correct errors, eliminating the cumulative effect of systematic offsets such as mechanical thermal deformation and fixture wear in multi-station flow, and achieving a high degree of consistency in glue path quality in long-cycle continuous production. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present 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 only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from the structures shown in these drawings without creative effort.

[0017] Figure 1 This is a schematic flowchart of a first embodiment of the dispensing and curing method of the present invention; Figure 2 This is a schematic flowchart of a second embodiment of the dispensing and curing method of the present invention; Figure 3 This is a schematic flowchart of a third embodiment of the dispensing and curing method of the present invention; Figure 4 This is a schematic flowchart of a fourth embodiment of the dispensing and curing method of the present invention; Figure 5 This is a schematic flowchart of the fifth embodiment of the dispensing and curing method of the present invention; Figure 6 This is a schematic flowchart of the sixth embodiment of the dispensing and curing method of the present invention; Figure 7 This is a schematic flowchart of the seventh embodiment of the dispensing and curing method of the present invention; Figure 8 This is a schematic flowchart of the eighth embodiment of the dispensing and curing method of the present invention.

[0018] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0019] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0020] Please refer to Figure 1 , Figure 1 This is a schematic flowchart of a first embodiment of a dispensing and curing method according to the present invention. In one embodiment, the dispensing and curing method proposed by the present invention includes the following steps: S100: Obtain the position information of the workpiece at the current dispensing station.

[0021] It should be noted that the current dispensing station refers to the working position on the multi-station turntable that is currently aligned with the dispensing needle; the workpiece position information refers to the actual spatial coordinate data of the workpiece in the machine coordinate system. The core purpose of obtaining this information is to provide basic data for subsequent elimination of material loading deviations.

[0022] In this step, there are two implementation methods: In the first implementation method, after receiving the station arrival signal, the main control unit directly retrieves the position information cached during the previous station switch as the position information of the current station to save cycle time; In the second implementation method, after receiving the station arrival signal, the main control unit re-triggers the vision detection module to take real-time pictures of the workpiece in the current stationary state to obtain new position information, so as to eliminate the fixture micro-motion error caused by the rotation of the turntable.

[0023] S200. Calculate the first offset between the actual position of the workpiece and the standard position, and perform coordinate fusion calculation between the first offset and the preset standard dispensing trajectory to generate the first compensated dispensing trajectory.

[0024] It should be noted that the first offset refers to the translation difference in two-dimensional or three-dimensional space between the actual position of the workpiece and the theoretically programmed standard position; the standard dispensing trajectory is the theoretical absolute spatial needle path; and the first compensated dispensing trajectory is the actual execution path after superimposing the position offset. Integrating the offset into the standard trajectory aims to correct the workpiece's incoming or loading position error.

[0025] In this step, there are two implementation methods: In the first implementation method, the main control unit uses a simple vector addition method to directly superimpose the first offset onto each discrete coordinate point of the standard dispensing trajectory to generate the first compensated dispensing trajectory; In the second implementation method, the main control unit uses a spline interpolation algorithm to smoothly introduce the first offset at the start and end points of the standard dispensing trajectory to generate a smooth first compensated dispensing trajectory without sudden speed changes.

[0026] S300: Control the needle to perform a dispensing operation along the first compensated dispensing trajectory to obtain the first dispensing path.

[0027] It should be noted that the first dispensing path is the actual physical adhesive line left on the surface of the workpiece after the pointer head follows the first compensation dispensing trajectory. Performing this step marks the completion of the static compensation stage, forming a physical entity for subsequent re-inspection.

[0028] In this step, there are two implementation methods: In the first implementation method, the main control unit controls the needle to travel along the first compensation dispensing trajectory at a constant speed in one go to obtain a continuous first dispensing path; In the second implementation method, the main control unit automatically reduces the needle speed in areas with drastic curvature changes and increases the needle speed in flat areas to obtain a first dispensing path with uniform thickness.

[0029] S400. After the dispensing operation is completed, the curing device is controlled to perform a curing operation on the first dispensing path.

[0030] It should be noted that the curing process refers to the process of causing a cross-linking reaction in liquid adhesive by irradiating it with light of a specific wavelength, turning it into a solid state. Curing is performed immediately after dispensing to prevent the adhesive from flowing or shifting during subsequent workstation changes or inspections.

[0031] In this step, there are two implementation methods: In the first implementation method, the main control unit controls the curing device to follow the movement of the needle and perform a follow-up curing operation while curing; In the second implementation method, after the main control unit controls the needle to complete the entire journey and leave, the curing device performs an overall static curing operation on the first dispensing path separately.

[0032] S500: After the curing operation is completed, obtain an image of the first dispensing path.

[0033] It should be noted that obtaining the image of the first adhesive path refers to taking a picture of the actual adhesive line using a vision sensor after the adhesive has initially cured and set. Because the adhesive is fluid, it is necessary to take a picture after curing to obtain the true final shape of the adhesive line as a basis for detection.

[0034] In this step, there are two implementation methods: In the first implementation method, the main control unit controls a fixed industrial camera to take a single-view frontal photograph of the first dispensing path to obtain an image; In the second implementation method, the main control unit controls a multi-angle camera array or a three-dimensional scanning sensor to take multi-view photographs or three-dimensional scans of the first dispensing path to obtain an image containing height information.

[0035] S600. Calculate the deviation between the first dispensing path and the standard dispensing path.

[0036] It should be noted that the deviation value refers to the difference between the actual first dispensing path and the theoretical standard dispensing path in terms of spatial position, width, height, and other dimensions. This deviation value is calculated to quantify the actual error in this dispensing process, serving as a data source for subsequent trend analysis.

[0037] In this step, there are two implementation methods: In the first implementation method, the main control unit uses an image edge extraction algorithm to calculate only the two-dimensional positional deviation value of the center line of the first dispensing path relative to the center line of the standard dispensing path; In the second implementation method, the main control unit uses a three-dimensional point cloud registration algorithm to calculate the three-dimensional shape deviation value of the volume shape of the first dispensing path relative to the standard dispensing path.

[0038] S700: Based on the accumulated deviation values ​​of historical dispensing stations, perform trend analysis to generate a second compensation dispensing trajectory.

[0039] It should be noted that the cumulative deviation value of historical dispensing stations refers to the set of deviation data generated by the current workpiece at each previously completed station; trend analysis refers to using mathematical statistical methods to find the pattern of deviation changes with station or time; the second compensation dispensing trajectory is a new execution route for the next station, which is corrected in advance based on the predicted pattern. Trend analysis achieves the leap from "post-event measurement" to "pre-event prediction".

[0040] In this step, there are two implementation methods: In the first implementation method, the main control unit uses a polar coordinate ellipse fitting algorithm to predict the periodic deviation trend caused by fixture eccentricity based on the deviation value of historical workstations, and generates a second compensation dispensing trajectory; In the second implementation method, the main control unit uses a linear regression algorithm to predict the linear drift trend caused by equipment thermal deformation based on the deviation value of historical workstations, and generates a second compensation dispensing trajectory.

[0041] S800: Control the workpiece to switch to the next dispensing station, use the second compensated dispensing trajectory as the execution reference trajectory of the next dispensing station, control the needle to perform dispensing operation along the second compensated dispensing trajectory to obtain the second dispensing path, use the second dispensing path as the new first dispensing path, and return to the execution step S400.

[0042] It should be noted that the second dispensing path is a new solid glue line obtained after executing the second compensated dispensing trajectory with trend prediction; returning to the execution step S400 means that the process of "dispensing-curing-re-inspection-prediction" is solidified into a continuous closed-loop control cycle until all workstations have completed processing.

[0043] In this step, there are two implementation methods: In the first implementation method, the main control unit directly replaces the first compensation trajectory of the next station with the second compensation dispensing trajectory as the sole reference for dispensing, thus obtaining the second dispensing path; In the second implementation method, the main control unit performs weighted fusion of the second compensation dispensing trajectory and the first offset newly measured at the next station as the execution reference trajectory for dispensing, thus obtaining the second dispensing path.

[0044] The dispensing and curing method provided by this invention introduces a cyclical mechanism of "dispensing-curing-re-inspection-trend prediction-re-dispensing," which overcomes the limitation of existing technologies in tracking dynamic offset trends during continuous processing. Specifically, a first offset is used to achieve basic positioning, obtain the first dispensing path, and cure it. After curing, the solid glue line is image-re-inspected to obtain the deviation value. The historically accumulated deviation value is used for trend analysis to proactively generate a second compensating dispensing trajectory that includes compensation prediction. In the next workstation, this second compensating dispensing trajectory is directly used for dispensing, and the cycle repeats. Through the above settings, the dispensing system is endowed with the ability to self-evolve and correct errors, eliminating the cumulative effect of systematic offsets such as mechanical thermal deformation and fixture wear in multi-station flow, and achieving a high degree of consistency in glue path quality in long-cycle continuous production.

[0045] Please refer to Figure 2 , Figure 2 This is a schematic flowchart of a second embodiment of the dispensing and curing method of the present invention. In this embodiment, step S100 includes: S110: Control the industrial camera to take an initial picture of the current dispensing station, and perform template matching of preset geometric features on the acquired initial image.

[0046] It should be noted that the preset geometric features refer to the theoretical binary image contour of a certain positioning hole or edge of the workpiece that is stored in the system beforehand; template matching refers to the process of sliding this theoretical contour in the initial image and calculating the similarity to find the actual position of the workpiece. This step utilizes traditional vision algorithms to achieve high-speed positioning.

[0047] In this step, there are two implementation methods: In the first implementation method, the main control unit controls the industrial camera to take a global exposure picture and performs integer-pixel template matching based on the normalized cross-correlation algorithm in the image memory; In the second implementation method, after the main control unit controls the industrial camera to take a picture, it first extracts the Harris corner features of the image and then performs sub-pixel feature matching based on the corner descriptor.

[0048] S120. Determine whether the preset geometric feature is successfully matched in the initial image.

[0049] It should be noted that determining whether a match is successful is to confirm whether the current image quality is sufficient to support reliable results from traditional algorithms. Traditional algorithms will typically fail if the workpiece surface has severe reflections or oil stains.

[0050] In this step, there are two implementation methods: In the first implementation method, the main control unit compares the calculated similarity score with a fixed global threshold. If the score is higher than the threshold, the matching is considered successful. In the second implementation method, the main control unit divides the image into multiple sub-regions and determines whether each sub-region is successfully matched. Only when all the core regions are successfully matched is the overall matching considered successful.

[0051] S130. If the matching fails, a mode switching command is generated to control the smart camera to take a second picture of the current dispensing station.

[0052] It should be noted that the mode switching command is a hardware or software control signal that triggers the backup vision solution; the intelligent camera is a high-end camera with a built-in artificial intelligence computing unit. Degrading to an artificial intelligence algorithm when the traditional algorithm fails is a redundant design to ensure uninterrupted production line operation.

[0053] In this step, there are two implementation methods: In the first implementation method, the main control unit generates a hardware-level level switching command to hard-trigger the exposure pin of the smart camera to take a second picture; In the second implementation method, the main control unit sends a software communication protocol packet to the smart camera via Ethernet to soft-trigger the smart camera to take a second picture.

[0054] S140. Call the pre-trained deep learning recognition model to extract features from the image obtained by the second shooting, and output the high-dimensional feature coordinates of the workpiece as the position information.

[0055] It should be noted that the deep learning recognition model is a convolutional neural network trained on a large number of defect samples; high-dimensional feature coordinates refer to the coordinate representation of multi-dimensional feature vectors with strong anti-interference capabilities extracted through multiple layers of networks after being mapped back to physical space. This utilizes artificial intelligence models to overcome the bottleneck of traditional vision's sensitivity to complex backgrounds.

[0056] In this step, there are two implementation methods: In the first implementation method, the edge computing chip built into the smart camera directly runs the deep learning recognition model and packages the calculated high-dimensional feature coordinates and sends them to the main control unit; In the second implementation method, the smart camera only sends the original images captured twice to the main control unit, and the main control unit calls local computing resources to run the deep learning recognition model to extract the high-dimensional feature coordinates.

[0057] In this embodiment, a dual-modal switching strategy of "industrial camera initial screening + intelligent camera recalculation" is adopted to achieve both high efficiency of the main cycle and strong robustness under operating conditions. The template matching of the industrial camera can output position information at the millisecond level under most normal operating conditions; once the matching is determined to be unsuccessful, the system seamlessly switches to the intelligent camera based on the deep learning recognition model to extract high-dimensional feature coordinates. The generalization ability of deep learning to complex backgrounds breaks through the physical bottleneck of traditional vision in terms of anti-reflection and anti-fouling, ensuring the absolute reliability of the position information acquisition process.

[0058] Please refer to Figure 3 , Figure 3 This is a schematic flowchart of a third embodiment of the dispensing and curing method of the present invention. In this embodiment, after step S140, the method further includes: S150. Extract the confidence score of the high-dimensional feature coordinates.

[0059] It should be noted that the confidence score is a quantitative indicator calculated by the deep learning recognition model through the output layer, representing the probability of the current output coordinates being accurate. Even when using artificial intelligence models, they may output incorrect coordinates that are forcibly fitted when faced with completely non-standard and abnormal materials, so it is necessary to extract their confidence scores.

[0060] In this step, there are two implementation methods: In the first implementation method, the main control unit directly reads the probability value of the target category in the output layer of the deep learning model as the confidence score; In the second implementation method, the main control unit calculates the Mahalanobis distance between the output high-dimensional feature vector and multiple template vectors in the standard feature library, and uses the negative exponent normalized result of the distance as the confidence score.

[0061] S160. The confidence score is compared with a preset feature confidence threshold.

[0062] It should be noted that the feature confidence threshold is a manually set safety baseline. Comparing continuous probability scores to this threshold is to determine whether the AI ​​model's confidence in the current outcome meets the safety standards for allowing the action to be performed.

[0063] In this step, there are two implementation methods: In the first implementation method, the main control unit uses a single fixed value as the feature confidence threshold and directly compares the confidence score with the fixed value; In the second implementation method, the main control unit dynamically retrieves the differentiated feature confidence threshold corresponding to the current workpiece model from the database for comparison.

[0064] S170. If the confidence score is lower than the feature confidence threshold, it is determined to be a visual recognition anomaly, the control device is stopped and a manual intervention signal is triggered.

[0065] It should be noted that visual recognition anomalies refer to situations where the system can no longer be certain of the authenticity of the current location data; controlling the equipment to stop and triggering a manual intervention signal is the ultimate safety precaution to prevent needle collisions or batch scrapping caused by performing dispensing based on "blind guessing" coordinates.

[0066] In this step, there are two implementation methods: In the first implementation method, after the main control unit determines the abnormality, it immediately cuts off the enable signal of the motion control module, causing all servo axes to brake and stop instantly, and emits an audible intervention signal through a buzzer; In the second implementation method, after the main control unit determines the abnormality, it controls the current workpiece to skip the dispensing process and directly transfer it to the manual review station, while a red flashing intervention prompt box pops up on the screen, and the main body of the production line does not stop.

[0067] In this embodiment, a safety circuit breaker mechanism for visual recognition is constructed by introducing a confidence score comparison. The confidence score of high-dimensional feature coordinates is extracted to quantitatively evaluate the credibility of the deep learning model output; by comparing it with the feature credibility threshold, an insurmountable bottom line is established; if it falls below the threshold, a shutdown or manual intervention is triggered, fundamentally preventing collision disasters or batch scrapping caused by dispensing based on low confidence coordinates, and providing the ultimate backup protection for extreme boundary conditions.

[0068] Please refer to Figure 4 , Figure 4 This is a schematic flowchart of a fourth embodiment of the dispensing and curing method of the present invention. In this embodiment, step S200 includes: S210: Control the line laser sensor to project line structured light onto the surface of the workpiece at the current dispensing station, and synchronously move the line laser sensor to acquire multiple frames of line laser stripe images.

[0069] It should be noted that line structured light refers to a single laser line projected by a line laser; synchronous motion acquisition refers to continuously capturing images of the laser line modulated by the undulations of the workpiece surface as the line laser sensor or the workpiece undergoes relative displacement. This is the hardware foundation for acquiring the three-dimensional physical morphology of the workpiece.

[0070] In this step, there are two implementation methods: In the first implementation method, the workpiece remains stationary, and the main control unit controls the line laser sensor to translate along the planned direction with the robotic arm, scanning the workpiece surface line by line to acquire multiple frames of images; In the second implementation method, the line laser sensor is fixed, and the main control unit controls the turntable to drive the workpiece to rotate, and completes the spiral scanning to acquire multiple frames of images in conjunction with the sensor's linear field of view.

[0071] S220. The multiple frames of the line laser stripe images are stitched together and three-dimensionally calculated to generate three-dimensional point cloud data containing the three-dimensional morphological features of the workpiece surface.

[0072] It should be noted that 3D solution refers to the mathematical process of calculating the actual physical height based on the pixel offset of the laser line in the camera image using the principle of triangulation; 3D point cloud data is a collection of countless points with spatial coordinates. This data digitizes the undulations of the workpiece surface.

[0073] In this step, there are two implementation methods: In the first implementation method, the main control unit uses an algorithm based on laser centerline extraction to directly extract the sub-pixel center on the two-dimensional image and calculate the three-dimensional coordinates, and then stitches them together in time sequence to generate a point cloud; In the second implementation method, the main control unit uses a structured light grating decoding algorithm to perform phase unfolding calculation on multiple frames of stripe images with phase differences, and reconstruct high-density three-dimensional point cloud data.

[0074] S230. Extract multiple discrete points within the current dispensing trajectory planning area from the three-dimensional point cloud data, and calculate the surface normal vector perpendicular to the workpiece surface at each discrete point.

[0075] It should be noted that discrete points refer to spatial coordinate points sampled at fixed intervals along the planned trajectory; the surface normal vector refers to the three-dimensional direction vector perpendicular to the local tangent plane at a discrete point. Calculating the normal vector is to determine the true tilt angle of the workpiece surface at each point.

[0076] In this step, there are two implementation methods: In the first implementation method, for each discrete point, the main control unit extracts the neighboring points within a fixed radius around it, fits the local micro-tangent plane using the least squares method, and calculates the normal vector of the tangent plane as the surface normal vector; In the second implementation method, the main control unit adopts a surface reconstruction algorithm based on the moving least squares method to directly calculate the continuously changing surface normal vector at each discrete point on the point cloud model.

[0077] S240. Obtain the standard needle movement posture vector of the needle in the initial state, perform a rotation matrix transformation on the standard needle movement posture vector according to the direction of the surface normal vector, and combine the transformed posture vector with the first offset to generate the first compensated dispensing trajectory adapted to the current surface.

[0078] It should be noted that the standard needle posture vector is usually the direction vector when the needle tip is vertically downward; the rotation matrix transformation refers to the process of rotating the needle tip's posture to be parallel to the normal vector using spatial geometric matrix formulas. Combining the posture transformation result with the first offset generates a surface-fitting trajectory that corrects not only the position but also the angle.

[0079] In this step, there are two implementation methods: In the first implementation method, the main control unit uses the Rodriguez rotation formula to rotate the standard needle movement posture vector around a rotation axis perpendicular to the normal vector by a specified angle, and after completing the posture transformation, it is combined with the first offset to generate the trajectory; In the second implementation method, the main control unit uses the quaternion rotation algorithm to convert the standard needle movement posture vector into a quaternion, multiply it with the normal vector quaternion to complete the posture transformation, and then combine it with the first offset to generate the trajectory.

[0080] In this embodiment, by introducing line laser scanning and normal vector attitude calculation, six-axis linkage trajectory compensation based on real three-dimensional topography is realized. Three-dimensional point cloud data is acquired and surface normal vectors are extracted, accurately quantifying the local spatial geometry of the workpiece surface. A rotation matrix transformation forces the needle attitude to conform to the surface normal in real time, eliminating interference from dispensing angles on complex curved surfaces. The attitude transformation result is combined with a first offset to generate a first compensated dispensing trajectory, thus increasing the compensation dimension from the traditional two-dimensional to three-dimensional, including attitude angles, solving the angle interference problem in dispensing on irregular curved surfaces.

[0081] Please refer to Figure 5 , Figure 5 This is a schematic flowchart of a fifth embodiment of the dispensing and curing method of the present invention. In this embodiment, step S300 includes: S310. Extract the Z-axis coordinate components of the surface normal vector in the vertical direction.

[0082] It should be noted that the Z-axis coordinate component is the absolute height value of the current discrete point in the vertical direction of the machine coordinate system. This value is extracted to determine the true physical height of the workpiece surface at that point, serving as the basis for subsequent calculations of the needle safety distance.

[0083] In this step, there are two implementation methods: In the first implementation method, the main control unit directly reads the Z-axis index value in the discrete point three-dimensional coordinate array as the Z-axis coordinate component; In the second implementation method, the main control unit performs a dot product operation between the surface normal vector and the Z-axis unit vector of the machine coordinate system, and obtains the accurate Z-axis coordinate component through projection calculation.

[0084] S320. Based on the preset fixed glue breakage distance, add the fixed glue breakage distance to the Z-axis coordinate component to generate a real-time following height curve to guide the change of Z-axis height.

[0085] It should be noted that the fixed glue-breaking distance is a constant safety height maintained between the glue outlet of the pointer and the workpiece surface, used to break the glue filament using fluid dynamics to prevent stringing; the real-time following height curve is a sequence of absolute Z-axis position commands that varies with the trajectory after adding the safety distance to the surface height.

[0086] In this step, there are two implementation methods: In the first implementation method, the main control unit directly adds the Z-axis coordinate component to the fixed glue break distance as a scalar to generate a stepped real-time following height curve; In the second implementation method, the main control unit smooths the added height value through a low-pass filtering algorithm to generate a smooth real-time following height curve without sharp peaks.

[0087] S330. During the process of controlling the needle to move along the horizontal plane, the Z-axis drive motor is subjected to high-frequency interpolation control according to the real-time following height curve.

[0088] It should be noted that high-frequency interpolation control refers to the process of sending position commands to the motor driver at an extremely high frequency. This breaks the limitations of traditional low-speed scanning cycles, allowing it to match the needs of rapidly undulating curved surfaces.

[0089] In this step, there are two implementation methods: In the first implementation method, the main control unit adopts a time-synchronized hard real-time interpolation cycle and sends interpolation position messages to the Z-axis servo driver strictly according to a fixed time interval; In the second implementation method, the main control unit adopts a flexible interpolation mechanism based on position thresholds, and triggers an interpolation calculation and transmission of the Z-axis position when the horizontal axis movement reaches a preset distance increment.

[0090] S340. Read the actual encoder position of the Z-axis drive motor in real time, and compare the actual encoder position with the real-time following height curve in a closed-loop deviation comparison.

[0091] It should be noted that closed-loop deviation comparison is a process of comparing and correcting the actual feedback position of the motor with the command curve in real time. A millisecond-level position loop is constructed to eliminate the "fluctuating highs and lows" phenomenon caused by mechanical response lag.

[0092] In this step, there are two implementation methods: In the first implementation method, the main control unit adopts a proportional-integral-derivative control algorithm to adjust the output torque of the Z-axis drive motor in real time according to the closed-loop deviation comparison result; In the second implementation method, the main control unit adopts a feedforward plus feedback composite control algorithm to superimpose the derivative of the real-time following height curve with the feedforward quantity, and then output it to the Z-axis drive motor in combination with the closed-loop deviation comparison result.

[0093] In this embodiment, by combining Z-axis height extraction, glue break distance superposition, and high-frequency closed-loop interpolation, absolute constant segment distance control is achieved during the surface dispensing process. Extracting Z-axis coordinate components and combining them with a fixed glue break distance generates a real-time following height curve, establishing a follow-up reference for the Z-axis. High-frequency interpolation control breaks the limitations of traditional low-speed scanning cycles. By reading the actual encoder position and comparing the closed-loop deviation, the absolute constancy of the glue break distance is ensured, suppressing stringing and cold solder joints from a physical source.

[0094] Please refer to Figure 6 , Figure 6 This is a schematic flowchart of a sixth embodiment of the dispensing and curing method of the present invention. In this embodiment, step S700 includes: S710. Extract the X-axis deviation component and Y-axis deviation component of the accumulated deviation value of the historical dispensing station in a preset rectangular coordinate system.

[0095] It should be noted that the X-axis deviation component and the Y-axis deviation component are two orthogonal displacements projected from the deviation in three-dimensional space onto the horizontal plane. These two components are extracted to analyze the spatial distribution of the deviation using two-dimensional planar geometry methods.

[0096] In this step, there are two implementation methods: In the first implementation method, the main control unit directly reads the pre-calculated X-axis and Y-axis component values ​​from the deviation value structure variable; In the second implementation method, the main control unit retrieves the historical three-dimensional deviation vector and performs an inner product operation with the basis vector of the horizontal plane to parse the X-axis deviation component and Y-axis deviation component in real time.

[0097] S720. Based on the rotation angle of the corresponding dispensing station, the X-axis deviation component and the Y-axis deviation component are synthesized by trigonometric functions and converted into a deviation vector with spatial directionality.

[0098] It should be noted that the rotation angle refers to the absolute angular position of the workpiece when it is rotated to that position by the multi-station turntable; the deviation vector refers to the error arrow with magnitude and direction. By synthesizing trigonometric functions, the two scalars are transformed into vectors in polar coordinates, facilitating subsequent analysis in the rotating coordinate system.

[0099] In this step, there are two implementation methods: In the first implementation method, the main control unit directly uses the conventional formula for converting rectangular coordinates to polar coordinates to perform the synthesis and conversion calculation; In the second implementation method, the main control unit uses a lookup table method combined with coordinate rotation digital computer algorithm hardware acceleration to realize the synthesis and conversion of trigonometric functions.

[0100] S730. Project the deviation vector onto a polar coordinate system with the rotation center as the pole to generate polar coordinate deviation points containing amplitude and polar angle, and construct a polar coordinate distribution map in ascending order of rotation angle.

[0101] It should be noted that polar coordinate deviation points convert the error arrows into distance (amplitude) and angle (polar angle) relative to the rotation center; the polar coordinate distribution plot is a graph depicting how the error changes with the rotation angle. If there is eccentricity, these points will necessarily be arranged approximately in an ellipse.

[0102] In this step, there are two implementation methods: In the first implementation method, the main control unit establishes a dynamic array in memory, stores the data structure of polar coordinate deviation points in chronological order, and draws a scatter plot on the screen as a polar coordinate distribution map; In the second implementation method, the main control unit uses a circular linked list data structure to store the polar coordinate deviation points and uses a graphics rendering engine to refresh and construct the polar coordinate distribution map in real time.

[0103] S740. Perform least squares ellipse fitting operation on all polar coordinate deviation points in the polar coordinate distribution diagram, and interpolate to calculate the polar coordinate compensation point on the fitted ellipse curve based on the estimated rotation angle of the next glue station.

[0104] It should be noted that the least squares ellipse fitting operation is the process of finding the ideal ellipse equation that best matches these points using mathematical optimization techniques; the interpolation calculation is the process of substituting the estimated angle of the next glue station into the ellipse equation to find the corresponding error point.

[0105] In this step, there are two implementation methods: In the first implementation method, the main control unit uses ordinary least squares based on algebraic distance to perform ellipse fitting operation and uses a linear interpolation algorithm to calculate polar coordinate compensation points; In the second implementation method, the main control unit uses direct least squares based on geometric distance to perform ellipse fitting operation and uses a cubic spline interpolation algorithm to calculate polar coordinate compensation points.

[0106] S750. The polar coordinate compensation point is reverse-calculated into a compensation offset in a rectangular coordinate system. The compensation offset is then fused with a preset standard dispensing trajectory to generate the second compensation dispensing trajectory.

[0107] It should be noted that the inverse solution is the process of converting the prediction error in polar coordinates back to the conventional X and Y rectangular coordinate correction values. Integrating this correction value into the standard trajectory completes the generation of the forward-looking second compensated trajectory.

[0108] In this step, there are two implementation methods: In the first implementation method, the main control unit directly calculates the compensation offset using the polar coordinate inverse transformation formula, and then merges it with the standard dispensing trajectory through vector addition to generate the second compensation dispensing trajectory; In the second implementation method, after the main control unit calculates the compensation offset using the inverse transformation, it introduces a low-pass filter to smooth the compensation offset, and finally merges it with the standard dispensing trajectory to generate the second compensation dispensing trajectory.

[0109] In this embodiment, accurate prediction and compensation for rotational eccentricity errors are achieved through mapping from rectangular coordinates to polar coordinates and least-squares ellipse fitting. Components are extracted and converted into deviation vectors, quantifying the spatial properties of the error; a polar coordinate distribution map is constructed, and ellipse fitting is performed on the deviation points, revealing the essential physical laws of fixture eccentricity from the chaotic error data; by interpolating to calculate polar coordinate compensation points and inversely solving to generate the second compensation point adhesive trajectory, the system is endowed with a geometric model-based advanced correction capability, solving the problem of periodic deviation accumulation caused by multi-station rotation.

[0110] Please refer to Figure 7 , Figure 7 This is a schematic flowchart of a seventh embodiment of a dispensing and curing method according to the present invention. In this embodiment, after step S750, the method further includes: S760: Count the total number of historical workstations that have performed dispensing operations and successfully obtained polar coordinate deviation points.

[0111] It should be noted that the total number of historical workstations refers to the cumulative number of workstations where the current workpiece has successfully completed dispensing and the polar coordinate deviation point has been calculated, starting from the first workstation. This value is collected to assess whether the amount of data currently used for fitting the ellipse is sufficient.

[0112] In this step, there are two implementation methods: In the first implementation method, the main control unit obtains the total number of historical workstations by reading the value of a global integer counter; In the second implementation method, the main control unit counts the total number of historical workstations by traversing the number of valid polar coordinate deviation point records associated with the current workpiece in the historical database.

[0113] S770. Calculate the ratio of the total number of historical workstations to the preset total number of dispensing workstations for the workpiece, and use the ratio as the data confidence level of the polar coordinate distribution map.

[0114] It should be noted that the preset total number of dispensing stations refers to the total number of locations on the entire turntable where the workpiece needs to be dispensed with adhesive; the data confidence level is a percentage value between 0 and 1, which directly reflects the statistical reliability of the fitted ellipse equation.

[0115] In this step, there are two implementation methods: In the first implementation method, the main control unit uses a simple division operation to calculate the ratio and directly retains two decimal places as the data confidence level; In the second implementation method, after calculating the ratio, the main control unit introduces a weighting coefficient based on the location area of ​​the workstation and uses the weighted ratio as the data confidence level.

[0116] S780. The data confidence level is compared with the preset allowable superposition threshold in real time.

[0117] It should be noted that the allowed overlay threshold is a manually set safety threshold percentage. Comparing this threshold is to determine whether the currently predicted second compensation dispensing trajectory is eligible for actual execution.

[0118] In this step, there are two implementation methods: In the first implementation method, the main control unit directly compares the data confidence level with a fixed allowable superposition threshold (such as 50%); In the second implementation method, the main control unit dynamically retrieves the corresponding allowable superposition threshold from the configuration table for comparison according to the accuracy level requirements of the current product.

[0119] S790. If the data confidence level is less than the allowable superposition threshold, the second compensation dispensing trajectory is discarded, and the execution trajectory of the next dispensing station is generated only based on the first offset; if the data confidence level is greater than or equal to the allowable superposition threshold, the second compensation dispensing trajectory is output as the execution trajectory.

[0120] It should be noted that discarding the second compensation dispensing trajectory is a mistake-proofing and degradation strategy, meaning that when data is insufficient, it reverts to the basic static compensation mode; the output as the execution trajectory means that the prediction model officially takes over and realizes trend compensation.

[0121] In this step, there are two implementation methods: In the first implementation method, the main control unit implements the two-choice logic through conditional branch statements, directly overriding the corresponding trajectory variables; In the second implementation method, the main control unit switches flag bits through state machine mode, and different state flag bits correspond to calling different trajectory generation functions to determine the execution trajectory.

[0122] In this embodiment, a safe startup mechanism for the trend prediction algorithm is constructed by introducing data confidence comparison. The total number of historical workstations is counted and the proportion is calculated as the data confidence level, which quantitatively evaluates the statistical basis of ellipse fitting. By comparing with the allowable superposition threshold, a hard threshold for model entry is established. If the value is below the threshold, the predicted trajectory is discarded and the system returns to basic compensation. This effectively prevents the disastrous consequence of "the more you compensate, the more biased it becomes" caused by insufficient data sampling in the first few workstations, which would result in a distorted ellipse. This ensures absolute safety during the system startup phase.

[0123] Please refer to Figure 8 , Figure 8 This is a schematic flowchart of an eighth embodiment of a dispensing and curing method according to the present invention. In this embodiment, step S700 includes: S711. Extract the X-axis deviation component and Y-axis deviation component of the accumulated deviation value of the historical dispensing station in a preset rectangular coordinate system. Based on the rotation angle of the corresponding dispensing station, the X-axis deviation component and the Y-axis deviation component are synthesized by trigonometric functions and converted into a deviation vector with spatial directionality. S721. Project the deviation vector onto a polar coordinate system with the rotation center as the pole to generate polar coordinate deviation points containing amplitude and polar angle, and construct a polar coordinate distribution map in ascending order of rotation angle. S731. Extract the amplitude difference and time interval between the polar coordinate deviation points of two adjacent dispensing stations in the polar coordinate distribution map.

[0124] It should be noted that the amplitude difference refers to the change in the magnitude of the error between two adjacent workstations; the time interval refers to the actual time span between the gluing actions performed at these two workstations. These two values ​​are extracted to calculate the rate at which the error evolves over time.

[0125] In this step, there are two implementation methods: In the first implementation method, the main control unit directly calculates the amplitude difference using the array index difference and reads the system timestamp difference as the time interval; In the second implementation method, the main control unit extracts adjacent points from the polar coordinate distribution map through a sliding window and obtains a high-precision time interval through a high-speed timer.

[0126] S741. Divide the amplitude difference by the time interval to calculate the real-time rate of change of the deviation amplitude over time.

[0127] It should be noted that the real-time rate of change is equivalent to a "speedometer" for error evolution, reflecting how quickly the equipment deviation expands or contracts. This is a core indicator for identifying linear drift trends.

[0128] In this step, there are two implementation methods: In the first implementation method, the main control unit directly calculates the real-time rate of change using a simple division operation; In the second implementation method, after the division operation, the main control unit performs a median filtering process on the calculation result to eliminate the influence of occasional pulse interference on the rate of change.

[0129] S751. Retrieve the real-time change rate of the most recent consecutive dispensing stations, and determine whether the real-time change rate shows a monotonic change pattern of continuous increase or continuous decrease.

[0130] It should be noted that a monotonic change pattern refers to a data sequence in which the rate of change exhibits a strictly increasing or strictly decreasing trend. If such a pattern is observed, it indicates that the error is not random fluctuation, but rather there is a definite direction of thermal deformation or wear drift.

[0131] In this step, there are two implementation methods: In the first implementation method, the main control unit identifies the monotonically increasing or decreasing pattern by simply judging the size relationship of adjacent data; In the second implementation method, the main control unit performs first-order difference calculation on multiple consecutive real-time change rates and identifies the monotonically changing pattern by judging the sign consistency of the difference sequence.

[0132] S761. If a monotonic variation pattern is observed, a univariate linear regression equation is established based on the monotonic variation pattern regarding time and deviation amplitude. The estimated execution time of the next dispensing station is substituted into the univariate linear regression equation to calculate the predicted deviation amplitude. The predicted deviation amplitude is then fused with the preset standard dispensing trajectory to generate the second compensated dispensing trajectory.

[0133] It should be noted that the univariate linear regression equation, i.e., the linear equation (y=kx+b), uses historical velocity patterns to extrapolate and predict future error magnitudes. Integrating the predicted deviation magnitude into the standard trajectory generates a second compensation trajectory for thermal drift.

[0134] In this step, there are two implementation methods: In the first implementation method, the main control unit uses the least squares method to fit the linear equation, and directly substitutes the estimated execution time into the equation to calculate the predicted deviation amplitude, and then performs coordinate fusion to generate the second compensation dispensing trajectory; In the second implementation method, the main control unit adds a forgetting factor when fitting the linear equation, assigns a larger weight to the closer workstation, and generates the second compensation dispensing trajectory after calculating the predicted deviation amplitude.

[0135] In this embodiment, dynamic prediction and compensation for time-drift errors such as equipment thermal deformation are achieved by extracting the real-time rate of change and establishing a univariate linear regression equation. The real-time rate of change is obtained by calculating the amplitude difference and time interval, which keenly captures the dynamic speed of error evolution; the monotonic change pattern eliminates the interference of random disturbances; the univariate linear regression equation is established and the estimated time is substituted into the calculation of the predicted deviation amplitude. A simple linear extrapolation model is used to achieve low-cost, high-reliability proactive interception of slow-drift errors, further enriching the trend analysis dimensions of the system.

[0136] The present invention provides a ninth embodiment of a dispensing and curing method. In this embodiment, step S400 includes: S410. After detecting the signal that the first dispensing path has been completed, control the UV curing lamp to move directly above the first dispensing path and emit curing light according to the first preset power and first preset duration calibrated by the peel force test in the early stage. Stop irradiation when the cumulative irradiation reaches the first preset duration, so that the glue reaches the surface dry state.

[0137] It should be noted that this invention uses a UV lamp to cure the adhesive. The UV lamp power, time, and other parameters for the initial curing are determined by testing the degree of curing through a preliminary peel force test. The first preset power refers to a curing power with lower energy. The surface dry state refers to a specific physical state in which a layer of "skin" forms on the surface of the adhesive, but the interior remains liquid. Low-power pre-curing aims to shape the surface without interfering with subsequent visual inspection.

[0138] In this step, there are two implementation methods: In the first implementation method, the main control unit uses a fixed first preset power and strictly accumulates the irradiation time through an internal timer. After the first preset time is reached, the power is directly cut off to stop irradiation. In the second implementation method, the main control unit monitors the aging degree of the UV lamp in real time and dynamically fine-tunes the actual output power according to the aging attenuation coefficient to ensure that the total UV energy actually output within the first preset time is consistent with the energy calibrated in the previous peel force test. Irradiation is stopped after the surface is dry.

[0139] S420. After irradiation stops, control the UV curing lamp to move out of the visual detection field of view and trigger the step of acquiring the image of the first dispensing path.

[0140] It should be noted that removing the image from the visual inspection field of view is to eliminate overexposure or occlusion interference caused by the strong light or shadow of the UV curing lamp on subsequent image capture. Triggering the image acquisition step signifies that the process has entered the re-inspection stage.

[0141] In this step, there are two implementation methods: In the first implementation method, the main control unit controls the UV curing lamp to move along the horizontal slide rail away from the area directly above, triggering the industrial camera to take a picture; In the second implementation method, the main control unit controls the UV curing lamp to flip upward around the hinge and move out of the field of view, triggering the smart camera to take a picture.

[0142] S430. After image acquisition is completed, during the physical gap when the workpiece is switching positions, the UV curing lamp is controlled to perform secondary penetration irradiation on the first dispensing path at a second preset power greater than the first preset power.

[0143] It should be noted that the physical gap refers to the movement time consumed by the turntable mechanism in rotating and moving the workpiece away from its current position; the second preset power refers to the high-energy curing power; and the secondary penetration irradiation is the process of using high energy to penetrate the surface dry layer and cure the internal liquid adhesive. Utilizing the idle time for secondary curing does not increase the cycle time.

[0144] In this step, there are two implementation methods: In the first implementation method, the main control unit immediately controls the UV curing lamp to irradiate at full power with the second preset power the instant it receives the signal that the turntable starts to rotate; In the second implementation method, the main control unit controls the UV curing lamp to output the second preset power in a stepped manner for secondary penetration irradiation in the middle section of the physical gap according to the speed curve of the turntable rotation.

[0145] S440. The curing time is accumulated to control whether the adhesive curing is complete. When the duration of the secondary penetration irradiation reaches the second preset time, the adhesive curing is determined to be complete, and the step of controlling the workpiece to switch to the next adhesive station is allowed.

[0146] It should be noted that subsequent production controls the completion of adhesive curing through UV curing time; the second preset time refers to the calibrated time required for complete internal curing of the adhesive under high-power irradiation. Time-based control avoids the complex installation and data processing delays associated with temperature sensors, thus improving cycle time.

[0147] In this step, there are two implementation methods: In the first implementation method, the main control unit starts a high-precision timer at the same time as starting the secondary penetration irradiation. When the timer reaches the second preset duration, it directly determines that the curing is complete and allows the switching step to be executed; In the second implementation method, the main control unit combines the real-time data of the ambient temperature sensor to fine-tune and compensate the second preset duration (such as appropriately extending it when the ambient temperature is low). When the compensated cumulative irradiation time reaches the target value, it determines that the curing is complete and allows the switching.

[0148] In this embodiment, a four-step curing strategy of "low-power surface drying - removal from field of view - high-power penetration through gaps - time-controlled release" effectively integrates curing and inspection. Initially, peel force testing calibrates the power and time for the first curing step. Low-power irradiation achieves surface drying, preventing adhesive flow deformation and creating conditions for shadowless photography. Secondary penetration irradiation using physical gaps achieves zero cycle time occupancy. Subsequent production controls adhesive curing completion by accumulating curing time, avoiding the limitations of temperature profile monitoring. This improves system response speed and stability while ensuring curing quality, effectively reducing flow defects caused by incomplete curing.

[0149] Furthermore, this invention also proposes a dispensing device that uses the dispensing and curing method described above for dispensing processing. The specific structure of this dispensing and curing method is as described in the foregoing embodiments. Since this dispensing device employs all the technical solutions of all the foregoing embodiments, it possesses at least all the beneficial effects brought about by the technical solutions of the foregoing embodiments, which will not be elaborated upon here.

[0150] It should be noted that the dispensing equipment can integrate a three-axis robot system, a rotary fixture unit, a vision recognition system, a dispensing system, a UV curing system, a CCD inspection system, and an MES interface module in its hardware architecture, serving as the physical execution carrier for the aforementioned method steps. The three-axis robot system acts as the main control center, responsible for overall motion control and process management. The dispensing system, mounted on the three-axis robot arm, performs the dispensing operation by controlling the needle along a compensation trajectory. The UV curing system, mounted on the robot arm's Z-axis, performs the step of controlling the curing device to cure the dispensing path. The rotary fixture unit, driven by a servo motor, supports multi-angle positioning at up to 16 rotational positions, corresponding to the step of controlling the workpiece to switch to the next dispensing station. At the perception and detection level, the vision recognition system (including industrial and intelligent cameras) performs the visual recognition steps of acquiring workpiece position information and taking initial and secondary photographs. The CCD inspection system performs the detection steps of acquiring dispensing path images and calculating deviation values. The MES interface module supports product information traceability and the optimized scheduling of process parameters (such as standard dispensing trajectories).

[0151] The above are merely preferred embodiments of the present invention and do not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct / indirect applications in other related technical fields, are included within the patent protection scope of the present invention.

Claims

1. A dispensing and curing method, characterized in that, include: Obtain the position information of the workpiece at the current dispensing station; Calculate the first offset between the actual position of the workpiece and the standard position, and perform coordinate fusion calculation between the first offset and the preset standard dispensing trajectory to generate the first compensated dispensing trajectory. Control the needle to perform dispensing operation along the first compensated dispensing trajectory to obtain the first dispensing path; After the dispensing operation is completed, the curing device is controlled to perform a curing operation on the first dispensing path; After the curing operation is completed, an image of the first dispensing path is acquired; Calculate the deviation between the first dispensing path and the standard dispensing path; A second compensation dispensing trajectory is generated by performing trend analysis on the accumulated deviation values ​​of historical dispensing stations. The trend analysis includes using mathematical statistical methods to find the pattern of the deviation values ​​changing with the dispensing station or time. The step of generating the second compensation dispensing trajectory includes: predicting the compensation offset of the next dispensing station according to the found pattern, and performing coordinate fusion calculation between the predicted compensation offset and the preset standard dispensing trajectory to generate the second compensation dispensing trajectory. The workpiece is switched to the next dispensing station. The second compensated dispensing trajectory is used as the execution reference trajectory of the next dispensing station. The needle is controlled to perform dispensing operation along the second compensated dispensing trajectory to obtain a second dispensing path. The second dispensing path is used as the new first dispensing path. The process then returns to the step of controlling the curing device to cure the first dispensing path.

2. The dispensing and curing method as described in claim 1, characterized in that, The step of obtaining the position information of the workpiece at the current dispensing station includes: The industrial camera is controlled to take an initial picture of the current dispensing station, and the acquired initial image is matched with a template based on preset geometric features. Determine whether the preset geometric feature is successfully matched in the initial image; If the matching fails, a mode switching command is generated to control the smart camera to take a second picture of the current dispensing station. The pre-trained deep learning recognition model is invoked to extract features from the image obtained by the second capture, and the high-dimensional feature coordinates of the workpiece are output as the position information.

3. The dispensing and curing method as described in claim 2, characterized in that, After the step of outputting the high-dimensional feature coordinates of the workpiece as the position information, the method further includes: Extract the confidence score of the high-dimensional feature coordinates; The confidence score is compared with a preset feature confidence threshold; If the confidence score is lower than the feature confidence threshold, it is determined to be a visual recognition anomaly, the control device is shut down and a manual intervention signal is triggered.

4. The dispensing and curing method as described in claim 1, characterized in that, The step of performing coordinate fusion calculation between the first offset and the preset standard dispensing trajectory to generate the first compensated dispensing trajectory includes: The control line laser sensor projects line structured light onto the surface of the workpiece at the current dispensing station, and synchronously moves the line laser sensor to acquire multiple frames of line laser stripe images; The multiple frames of the line laser stripe images are stitched together and subjected to three-dimensional calculation to generate three-dimensional point cloud data containing the three-dimensional morphological features of the workpiece surface; Extract multiple discrete points within the current dispensing trajectory planning area from the three-dimensional point cloud data, and calculate the surface normal vector perpendicular to the workpiece surface at each discrete point. Obtain the standard needle movement posture vector of the needle in the initial state, perform a rotation matrix transformation on the standard needle movement posture vector according to the direction of the surface normal vector, and combine the transformed posture vector with the first offset to generate the first compensated dispensing trajectory adapted to the current surface.

5. The dispensing and curing method as described in claim 4, characterized in that, The steps of controlling the needle to perform the dispensing operation along the first compensated dispensing trajectory include: Extract the Z-axis coordinate components of the surface normal vector in the vertical direction; Based on the preset fixed glue breakage distance, the Z-axis coordinate components are added to the fixed glue breakage distance to generate a real-time following height curve for guiding the Z-axis height change; During the process of controlling the needle to move along the horizontal plane, the Z-axis drive motor is subjected to high-frequency interpolation control according to the real-time following height curve; The actual encoder position of the Z-axis drive motor is read in real time, and the actual encoder position is compared with the real-time following height curve in a closed-loop deviation comparison.

6. The dispensing and curing method according to any one of claims 1 to 5, characterized in that, The step of performing trend analysis on the accumulated deviation values ​​from historical dispensing stations to generate a second compensated dispensing trajectory includes: Extract the X-axis and Y-axis deviation components of the accumulated deviation values ​​from historical dispensing stations in a preset rectangular coordinate system; Based on the rotation angle of the corresponding dispensing station, the X-axis deviation component and the Y-axis deviation component are synthesized by trigonometric functions and converted into a deviation vector with spatial directionality. The deviation vector is projected onto a polar coordinate system with the rotation center as the pole to generate polar coordinate deviation points containing amplitude and polar angle, and a polar coordinate distribution map is constructed in ascending order of rotation angle. The least squares ellipse fitting operation is performed on all polar coordinate deviation points in the polar coordinate distribution diagram. The polar coordinate compensation point is calculated by interpolation on the fitted ellipse curve based on the estimated rotation angle of the next glue station. The polar coordinate compensation point is reverse-calculated into a compensation offset in a rectangular coordinate system. The compensation offset is then fused with a preset standard dispensing trajectory to generate the second compensation dispensing trajectory.

7. The dispensing and curing method as described in claim 6, characterized in that, After the steps of reversing the polar coordinate compensation point to a compensation offset in a Cartesian coordinate system, and performing coordinate fusion calculations between the compensation offset and a preset standard dispensing trajectory to generate the second compensation dispensing trajectory, the method further includes: Count the total number of historical workstations that have performed dispensing operations and successfully obtained polar coordinate deviation points; Calculate the ratio of the total number of historical workstations to the preset total number of glue dispensing workstations for the workpiece, and use the ratio as the data confidence level of the polar coordinate distribution map. The data confidence level is compared with a preset allowable overlap threshold in real time; If the data confidence level is less than the allowable superposition threshold, the second compensation dispensing trajectory is discarded, and the execution trajectory of the next dispensing station is generated only based on the first offset; if the data confidence level is greater than or equal to the allowable superposition threshold, the second compensation dispensing trajectory is output as the execution trajectory.

8. The dispensing and curing method according to any one of claims 1 to 5, characterized in that, The step of performing trend analysis on the accumulated deviation values ​​from historical dispensing stations to generate a second compensated dispensing trajectory includes: Extract the X-axis and Y-axis deviation components of the accumulated deviation values ​​from historical dispensing stations in a preset rectangular coordinate system; Based on the rotation angle of the corresponding dispensing station, the X-axis deviation component and the Y-axis deviation component are synthesized by trigonometric functions and converted into a deviation vector with spatial directionality. The deviation vector is projected onto a polar coordinate system with the rotation center as the pole to generate polar coordinate deviation points containing amplitude and polar angle, and a polar coordinate distribution map is constructed in ascending order of rotation angle. Extract the amplitude difference and time interval between the polar coordinate deviation points of two adjacent dispensing stations from the polar coordinate distribution map; Divide the amplitude difference by the time interval to calculate the real-time rate of change of the deviation amplitude over time; Retrieve the real-time change rate of multiple recent dispensing stations and determine whether the real-time change rate exhibits a monotonic change pattern of continuous increase or continuous decrease. If a monotonic variation pattern is observed, a univariate linear regression equation is established based on the monotonic variation pattern regarding time and deviation amplitude. The estimated execution time of the next dispensing station is substituted into the univariate linear regression equation to calculate the predicted deviation amplitude. The predicted deviation amplitude is then fused with the preset standard dispensing trajectory to generate the second compensated dispensing trajectory.

9. The dispensing and curing method according to any one of claims 1 to 5, characterized in that, The steps of the curing control device performing the curing operation on the first dispensing path include: After detecting the signal that the first dispensing path has been completed, the UV curing lamp is controlled to move directly above the first dispensing path and emit UV curing light according to the first preset power and first preset duration calibrated by the previous peel force test. The irradiation stops when the first preset duration is reached, so that the adhesive reaches the preset surface dry state. After irradiation stops, the UV curing lamp is controlled to move out of the visual detection field of view, and the step of acquiring the image of the first dispensing path is triggered; After image acquisition is completed, during the physical gap when the workpiece is switching workstations, the UV curing lamp is controlled to perform secondary penetration irradiation on the first dispensing path at a second preset power greater than the first preset power. The curing time is used to control whether the adhesive curing is complete. When the duration of the secondary penetration irradiation reaches the second preset time, the adhesive curing is determined to be complete, and the step of controlling the workpiece to switch to the next adhesive station is allowed.

10. A dispensing device, characterized in that, The dispensing equipment performs dispensing processing using the dispensing and curing method as described in any one of claims 1 to 9.

Images