Robot adaptive glue coating control method and system for variable cross-section sealing surface

Abstract

The application provides a robot adaptive glue coating control method and system for variable cross-section sealing surfaces, and relates to the technical field of adaptive control. The method comprises the following steps: extracting sealing surface features of an oil pump to be sealed to obtain sealing surface boundary features; after performing coating isomer identification, executing dynamic reconstruction of a standard glue coating track to generate a continuous glue coating path; taking a non-coating area as a glue coating constraint domain to obtain a path curvature sequence, a path concave-convex feature sequence and a path width sequence; performing spatial alignment and recombination to obtain a plurality of feature correlation control sections; searching a plurality of control parameter vectors in a glue coating process parameter knowledge base; dynamically fusing to generate a glue coating control sequence; and during execution, performing PID real-time feedback compensation according to glue line cross-section thickness feedback. The technical problem of the prior art that the glue coating quality has large fluctuations in the manual glue coating operation mode, which leads to an increase in the failure rate and an increase in the rework rate and the rework cost caused by the glue coating problem is solved.

Description

Technical Field

[0001] This invention relates to the field of adaptive control technology, and in particular to a robot adaptive adhesive application control method and system for variable cross-section sealing surfaces. Background Technology

[0002] Variable cross-section sealing surfaces often have irregular geometric shapes, including curved surfaces, different bending angles, and local geometric changes. On irregular sealing surfaces, the amount of adhesive applied may be over-applied or under-applied due to changes in the path, resulting in poor sealing effect. It may even lead to adhesive overflow or failure to effectively cover the sealing surface. With the advancement of technology, the precision requirements for sealing adhesive are becoming higher and higher, especially in the field of high-end manufacturing, where the requirements for the thickness and uniformity of the adhesive layer are more stringent.

[0003] Currently, the oil pump coating process in the workshop still uses manual coating. The specific process is as follows: each coating operation by a worker takes approximately 32 seconds, followed immediately by the assembly stage, which takes approximately 73 seconds. The total time for this process is 105 seconds, exceeding the standard production line cycle time of 80 seconds. Frequent reliance on mobile personnel to maintain production rhythm has become a key bottleneck restricting overall line efficiency. Furthermore, due to the instability of manual coating, 105 units experienced oil leaks from January to August, resulting in 105 reworks due to oil leaks caused by coating issues. A total of 37,232 units were installed and tested, representing a failure rate of 0.28%.

[0004] Overall, the efficiency of this manual gluing operation mode is far below the requirements of the production line, resulting in a serious lag in the production cycle. Because manual gluing cannot be done continuously and efficiently, there are time differences between processes, which seriously affects the rhythm of the production line. In order to maintain the rhythm of the production line, the workshop must rely on the support of mobile personnel, which further reduces production efficiency and increases the complexity of personnel management and production scheduling. Due to the instability of manual gluing, the gluing quality fluctuates greatly, which leads to rework and repair, increasing the cost of products.

[0005] It should be noted that the information disclosed in this background section is intended only to enhance the understanding of the overall background of the present invention, and should not be construed as an admission or in any way implying that the information constitutes prior art known to those skilled in the art. Summary of the Invention

[0006] In response to the above-mentioned defects or improvement needs of existing technologies, this invention provides a robot adaptive glue application control method and system for variable cross-section sealing surfaces. This solves the technical problem that existing technologies use manual glue application operations, which result in large fluctuations in glue application quality, leading to increased failure rates, rework rates, and rework costs caused by glue application problems.

[0007] The specific technical solution is as follows:

[0008] According to a first aspect of the present invention, a robot adaptive adhesive application control method for a variable cross-section sealing surface is provided, the method comprising:

[0009] The sealing surface features of the oil pump to be sealed upon entering the sealing station are extracted to obtain the sealing surface boundary features. Based on these boundary features, coating heterogeneity identification is performed, and a standard adhesive application trajectory is dynamically reconstructed to generate a continuous adhesive application path. Using the uncoated area as the adhesive application constraint domain, the sealing surface boundary features are analyzed along the continuous adhesive application path to obtain a path curvature sequence, a path concavity / convexity feature sequence, and a path width sequence. The path curvature sequence, path concavity / convexity feature sequence, and path width sequence are spatially aligned and recombined to obtain multiple feature-associated control segments. These multiple feature-associated control segments are used as search keys to retrieve multiple control parameter vectors from the adhesive application process parameter knowledge base. The multiple control parameter vectors are dynamically fused to generate an adhesive application control sequence. During the process of controlling the adhesive application robot to execute the adhesive application control sequence along the continuous adhesive application path on the sealing surface of the oil pump to be sealed, real-time PID feedback compensation is performed based on the adhesive line cross-sectional thickness feedback.

[0010] In one embodiment, the sealing surface features of the oil pump to be sealed upon entering the sealing station are extracted to obtain the sealing surface boundary features, including:

[0011] After the oil pump to be sealed is transported to the sealing station via a positioning tray, a laser scanner integrated at the end of the adhesive applicator robot is triggered to perform a high-speed 3D scan of the sealing surface, obtaining high-density 3D point cloud data. Based on the model code of the oil pump to be sealed, the oil pump reference geometric model is retrieved from the oil pump model library. Based on the high-density 3D point cloud data, the sealing surface is reconstructed on the oil pump reference geometric model to obtain the sealing surface geometric model. The sealing surface geometric model is then subjected to mesh topological segmentation to extract the sealing surface boundary features.

[0012] In one embodiment, the sealing surface boundary features include distributed curvature gradient, distributed concave-convex height field, and distributed cross-sectional width spectrum.

[0013] In one implementation, after identifying coating heterogeneity based on the boundary features of the sealing surface, a standard adhesive application trajectory is dynamically reconstructed to generate a continuous adhesive application path, including:

[0014] The standard adhesive application trajectory is retrieved from the preset trajectory knowledge base based on the model code of the oil pump to be sealed. After spatially superimposing the distributed curvature gradient and distributed cross-sectional width spectrum, dynamic partitioning is performed based on regional risk judgment rules to obtain distributed steady-state region, distributed transition region, and distributed abrupt change region. The distributed concave-convex height field is traversed to locate distributed groove regions with a concavity depth greater than a preset depth threshold. A preset buffer distance is expanded along the surface normal of the distributed groove region to obtain distributed coating taboo region. By projecting the distributed coating taboo region, distributed transition region, and distributed abrupt change region onto the standard adhesive application trajectory, the taboo truncated segment sequence, transition optimized segment sequence, and abrupt reconstructed segment sequence are located. Feature constraint reconstruction is performed on the taboo truncated segment sequence, transition optimized segment sequence, and abrupt reconstructed segment sequence to output safe avoidance trajectory sequence, optimized transition trajectory sequence, and gradient correction trajectory sequence. Based on the boundary nodes of the taboo truncated segment sequence, transition optimized segment sequence, and abrupt reconstructed segment sequence, C1 continuous spatiotemporal synchronous fusion of the safe avoidance trajectory sequence, optimized transition trajectory sequence, and gradient correction trajectory sequence is performed on the standard adhesive application trajectory to generate the continuous adhesive application path.

[0015] In one embodiment, using the uncoated area as the adhesive constraint domain, the boundary features of the sealing surface are analyzed along the continuous adhesive application path to obtain a path curvature sequence, a path concavity / convexity feature sequence, and a path width sequence, including:

[0016] The sealing surface geometric model is subjected to restricted area feature identification to locate the vector restricted area polygon; a safe adhesive boundary distance is set based on adhesive overflow risk analysis; along the surface normal of the sealing surface geometric model, the normal expansion of the vector restricted area polygon is applied to the adhesive safety boundary distance to generate the uncoated area; a fixed arc length interval is set to locate the sampling point sequence on the continuous adhesive application path; using the uncoated area as the adhesive constraint domain, the distributed curvature gradient, distributed concave-convex height field, and distributed cross-sectional width spectrum are aligned and segmented along the sampling point sequence to obtain the path curvature sequence, path concave-convex feature sequence, and path width sequence.

[0017] In one implementation, the plurality of feature-associated control segments are used as search keys to retrieve multiple control parameter vectors in the adhesive coating process parameter knowledge base, including:

[0018] Based on the spatial alignment and aggregation of multiple path sampling points in the sampling point sequence, the path curvature sequence, the path concavity / convexity feature sequence, and the path width sequence, multiple feature-associated control segments are obtained. Each feature-associated control segment includes the mean curvature value within the segment, the range of concavity / convexity height within the segment, and the mean width within the segment. S1: The mean width within the first segment of the first feature-associated control segment is used as a primary search key to match the first dispensing pressure in the adhesive coating process parameter knowledge base. S2: The first dispensing pressure is used to narrow the parameter space of the adhesive coating process parameter knowledge base to obtain a primary limited knowledge base. S3: The first concavity / convexity value within the first segment of the first feature-associated control segment is... The height range is used as a secondary search key to match the first coating head tilt angle in the primary knowledge base; S4: The first coating head tilt angle is used to narrow the parameter space of the primary knowledge base to obtain a secondary knowledge base; S5: The mean curvature value within the first segment of the first feature-associated control segment is used as a tertiary search key to match the first moving speed in the secondary knowledge base; wherein, the first dispensing pressure, the first coating head tilt angle, and the first moving speed constitute the first control parameter vector; Steps S1 to S5 are executed by analogy, using the multiple feature-associated control segments as search keys to retrieve the multiple control parameter vectors in the coating process parameter knowledge base.

[0019] In one implementation, after spatially superimposing the distributed curvature gradient and the distributed cross-sectional width spectrum, dynamic partitioning is performed based on regional risk criterion rules to obtain distributed steady-state regions, distributed transition regions, and distributed abrupt change regions, including:

[0020] The distributed curvature gradient and distributed cross-sectional width spectrum are mapped to the UV parameter coordinate system of the sealing surface geometric model, and the eigenvalue grid is aligned using a bilinear interpolation algorithm to obtain a unified feature grid. The unified feature grid is then traversed using the regional risk criterion rule to identify grid attributes and output attribute-identified grids. The attribute-identified grids are separated to obtain discrete steady-state grids, discrete transition grids, and discrete abrupt change grids. The watershed algorithm is used to perform regional aggregation processing on the discrete transition grids and discrete abrupt change grids to output the distributed transition region and the distributed abrupt change region. The distributed transition region and the distributed abrupt change region are removed from the unified feature grid to obtain the distributed steady-state region.

[0021] In one embodiment, during the process of controlling the glue-applying robot to execute the glue-applying control sequence along the continuous glue-applying path on the sealing surface of the oil pump to be sealed, PID real-time feedback compensation is performed based on the glue line cross-sectional thickness feedback, including:

[0022] A laser triangulation thickness sensor integrated 5mm behind the glue-applying head of the glue-applying robot is used to dynamically acquire the thickness of the glue line cross-section; the target cross-section thickness and the glue line cross-section thickness are dynamically compared to output the real-time thickness deviation; real-time compensation actions are matched according to the deviation characteristics and amplitude of the real-time thickness deviation; after the real-time compensation action is executed by a PID controller and the PID real-time feedback compensation is performed to the preset glue-applying path length, the glue line cross-section thickness is re-checked, and the glue-applying control closed-loop feedback compensation is performed according to the re-check result.

[0023] In one embodiment, the control parameter vector includes moving speed, dispensing pressure, and applicator head tilt angle.

[0024] According to a second aspect of the present invention, a robot adaptive adhesive application control system for a variable cross-section sealing surface is provided, the system comprising:

[0025] The system comprises the following modules: a feature extraction module for extracting sealing surface features from the oil pump entering the sealing station, obtaining sealing surface boundary features; a trajectory reconstruction module for dynamically reconstructing a standard adhesive application trajectory based on the sealing surface boundary features after identifying coating heterogeneity, generating a continuous adhesive application path; a feature parsing module for parsing the sealing surface boundary features along the continuous adhesive application path using the non-coated area as the adhesive application constraint domain, obtaining a path curvature sequence, a path concavity / convexity feature sequence, and a path width sequence; a control segment acquisition module for spatially aligning and recombining the path curvature sequence, path concavity / convexity feature sequence, and path width sequence to obtain multiple feature-associated control segments; a control parameter retrieval module for using the multiple feature-associated control segments as retrieval keys to retrieve multiple control parameter vectors from the adhesive application process parameter knowledge base; a control sequence generation module for dynamically fusing the multiple control parameter vectors to generate an adhesive application control sequence; and a feedback compensation module for controlling the adhesive application robot to perform real-time PID feedback compensation based on the adhesive line cross-sectional thickness feedback during the execution of the adhesive application control sequence along the continuous adhesive application path on the sealing surface of the oil pump.

[0026] Beneficial effects of the embodiments of the present invention:

[0027] By extracting the sealing surface features of the oil pump entering the sealing station, the boundary features of the sealing surface can be accurately captured, ensuring a full understanding of the shape changes of the sealing surface during adhesive application. Based on the sealing surface boundary features, the standard adhesive application trajectory can be dynamically reconstructed, effectively identifying and adapting to the geometric changes of the sealing surface, avoiding irregularities in the adhesive application process, and preventing over-coating or uneven adhesive application. By analyzing the non-coated areas on the adhesive application path and using these areas as the constraint domain for adhesive application, the robot can avoid adhesive-prohibited areas on complex sealing surfaces, thereby reducing the risk of adhesive overflow or insufficient application. By extracting the path curvature sequence, path concavity and convexity feature sequence, and path width sequence, the robot can consider factors such as path curvature, changes, and surface irregularities during the execution of the adhesive application path, thus adjusting the adhesive application strategy in real time during actual adhesive application to cope with different sealing surface conditions. By spatially aligning and reconstructing the path curvature sequence and path width sequence, the robot can effectively address these issues. The concave-convex feature sequence and path width sequence form multiple feature-related control segments, providing a precise basis for subsequent glue application strategy adjustments and parameter matching, ensuring that the glue application process can more meticulously adapt to each changing sealing surface. By using multiple feature-related control segments as search keys, relevant control parameter vectors can be quickly and accurately retrieved from the glue application process parameter knowledge base, effectively providing customized parameters for different glue application needs. Through the dynamic fusion of multiple control parameter vectors, a glue application control sequence suitable for the current glue application path and sealing surface is generated. Dynamic fusion helps to automatically adjust the glue application parameters according to the geometric changes of the sealing surface, ensuring real-time adaptation and optimization of the glue application process. When there is a deviation in the glue line thickness, the PID controller automatically compensates according to the thickness deviation. Through real-time feedback compensation, the glue application process becomes a closed-loop control system, which can effectively reduce glue application errors and ensure the stability and consistency of glue application quality.

[0028] Of course, implementing any product or method of the present invention does not necessarily require achieving all of the advantages described above at the same time. Attached Figure Description

[0029] 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 based on these drawings without creative effort.

[0030] Figure 1 A schematic diagram of the robot adaptive adhesive application control method for variable cross-section sealing surfaces provided by the present invention is shown.

[0031] Figure 2 A schematic diagram of the robot adaptive adhesive application control system for variable cross-section sealing surfaces provided by the present invention is shown.

[0032] Figure labeling: Feature extraction module 10, trajectory reconstruction module 20, feature parsing module 30, control segment acquisition module 40, control parameter retrieval module 50, control sequence generation module 60, feedback compensation module 70. Detailed Implementation

[0033] To facilitate understanding of the present invention, a more complete description of the invention will be given below with reference to the accompanying drawings, which illustrate preferred embodiments of the invention. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein; rather, these embodiments are provided so that the disclosure of the invention will be more thorough and complete.

[0034] Furthermore, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0035] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," and "counterclockwise," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0036] Unless otherwise expressly stated, throughout the specification and claims, the term "comprising" or its variations such as "including" or "comprises" shall be understood to include the stated elements or components without excluding other elements or other components.

[0037] The present invention provides a robot adaptive glue application control method and system for variable cross-section sealing surfaces, which solves the technical problem that the existing technology uses a manual glue application operation mode, resulting in large fluctuations in glue application quality, leading to increased failure rate, increased rework rate, and increased rework costs caused by glue application problems.

[0038] Example 1: See Figure 1 The present invention provides an adaptive adhesive application control method for robots facing variable cross-section sealing surfaces, the method comprising:

[0039] Y100: Extract the sealing surface features of the oil pump to be sealed upon entering the sealing station to obtain the sealing surface boundary features.

[0040] The oil pump to be sealed is transported to the sealing station via a positioning tray, triggering a laser scanner integrated at the end of the adhesive applicator robot to perform a high-precision 3D scan of the sealing surface, obtaining high-density 3D point cloud data of the sealing surface. Based on the oil pump model code, the reference geometric model of the oil pump is retrieved from the oil pump model library, and the geometric shape of the sealing surface is reconstructed using the point cloud data. The geometric model of the sealing surface is then meshed and topologically segmented to extract the boundary features of the sealing surface, including distributed curvature gradient, distributed concave-convex height field, and distributed cross-sectional width spectrum, which are used for subsequent path planning and adhesive applicator control.

[0041] Y200: After identifying coating heterogeneity based on the boundary features of the sealing surface, perform dynamic reconstruction of the standard adhesive application trajectory to generate a continuous adhesive application path.

[0042] Based on the boundary features of the sealing surface, heterogeneity identification is performed in the region, identifying irregular areas with different shapes and width variations. This is because variable cross-section sealing surfaces have different geometric structures, such as local depressions, protrusions, and width variations. By comparing the identified heterogeneous features with the standard adhesive application trajectory, the standard adhesive application trajectory is dynamically reconstructed to adapt to different sealing surface features. The reconstructed trajectory can better adapt to the geometric changes of the sealing surface, ensuring the continuity of the adhesive application path. Based on the reconstructed adhesive application trajectory, a continuous adhesive application path is generated. This ensures that the robot can smoothly follow the path and apply adhesive evenly throughout the entire adhesive application process.

[0043] Y300: Using the uncoated area as the adhesive constraint domain, analyze the boundary features of the sealing surface along the continuous adhesive application path to obtain the path curvature sequence, the path concavity and convexity feature sequence, and the path width sequence.

[0044] The uncoated area refers to the absolute avoidance area for adhesive application generated through restricted area feature identification and adhesive overflow risk analysis. It is formed by the normal expansion safety boundary of restricted areas such as bolt holes and deep grooves on the sealing surface. It serves as a hard constraint boundary for the robot's adhesive application path planning, ensuring that the adhesive application head maintains a safe distance from the risk area.

[0045] Using this region as a rigid constraint boundary, when analyzing the three-dimensional geometric features of the sealing surface along the continuous adhesive application path, the boundary curvature gradient is extracted point by point along the path to form a path curvature sequence, the concave-convex height difference is extracted point by point to form a path concave-convex feature sequence, and the cross-sectional width change is extracted point by point to form a path width sequence.

[0046] Y400: Spatial alignment and recombination of the path curvature sequence, path concave-convex feature sequence, and path width sequence to obtain multiple feature-associated control segments.

[0047] Spatial alignment technology is used to align the path curvature sequence, path convexity / concave feature sequence, and path width sequence data to the same reference coordinate system. This alignment process ensures that the various feature data are consistent in spatial location, facilitating subsequent feature association and control. The aligned data is then reorganized to form multiple feature association control segments. Each control segment represents a small area on the adhesive application path, containing normalized information such as the curvature, convexity / concave features, and width of that area. By integrating these feature data into control segments, each segment of the path can be controlled more precisely.

[0048] Y500: Use the multiple feature-associated control segments as search keys to retrieve multiple control parameter vectors in the glue coating process parameter knowledge base.

[0049] By using multiple feature-associated control segments as search keys, the robot locates corresponding control parameter vectors in a pre-established knowledge base of adhesive application process parameters. Each control parameter vector contains adhesive application control parameters for the corresponding feature segment, such as dispensing pressure, dispensing head tilt angle, and robot movement speed. Through this feature-parameter mapping method, the adhesive application robot can automatically adjust the adhesive application parameters based on the actual characteristics of the path. Automatic matching from the knowledge base reduces manual adjustments to adhesive application parameters, improves production efficiency, and lowers the risk of human error.

[0050] Y600: Dynamically fuses the multiple control parameter vectors to generate an adhesive application control sequence.

[0051] Dynamically fusing multiple control parameter vectors means combining parameters such as dispensing pressure, moving speed, and dispensing head angle of adjacent connecting path segments. For example, based on the characteristics of the dispensing path and the real-time feedback from the dispensing robot, an appropriate weighting strategy is selected to adjust these parameters. For different areas of the dispensing path, the fusion process can adapt to path changes, ensuring the continuity and consistency of control parameters throughout the dispensing process. After parameter fusion, a continuous dispensing control sequence is generated, including the specific actions to be performed on each segment of the path during the dispensing process. By dynamically fusing different control parameter vectors, the continuity and consistency of parameter adjustments during the dispensing process are ensured, sudden parameter jumps are avoided, and the instability of dispensing quality is reduced.

[0052] Y700: During the process of controlling the glue-applying robot to execute the glue-applying control sequence along the continuous glue-applying path on the sealing surface of the oil pump to be sealed, real-time PID feedback compensation is performed based on the feedback of the glue line cross-sectional thickness.

[0053] The glue-applying robot integrates a laser triangulation thickness sensor at its end effector to dynamically monitor the cross-sectional thickness of the glue line during the application process. This sensor acquires real-time glue line thickness data and compares it with a preset target thickness. By comparing the target and actual thicknesses, the real-time thickness deviation can be calculated. This deviation is used to adjust control parameters during the glue-applying process for real-time compensation.

[0054] Based on thickness deviation, a PID controller is used for real-time compensation. The PID controller dynamically adjusts parameters such as the dispensing pressure, moving speed, and dispensing head tilt angle of the glue-applying robot through feedback and regulation to ensure that the glue line thickness reaches the preset value. After the glue-applying robot adjusts according to PID control, the glue line thickness is measured again to ensure that the compensated effect meets expectations. If a deviation still exists, PID adjustment continues until the glue line thickness stabilizes within the preset range.

[0055] In one implementation, the sealing surface features of the oil pump to be sealed upon entering the sealing station are extracted to obtain the sealing surface boundary features, including:

[0056] Y110: After the oil pump to be sealed is transported to the sealing station via the positioning tray, the laser scanner integrated at the end of the adhesive applicator robot is triggered to perform a high-speed three-dimensional scan of the sealing surface to obtain high-density three-dimensional point cloud data; Y120: According to the model code of the oil pump to be sealed, the oil pump reference geometric model is retrieved from the oil pump model library; Y130: Based on the high-density three-dimensional point cloud data, the sealing surface is reconstructed on the oil pump reference geometric model to obtain the sealing surface geometric model; Y140: The sealing surface geometric model is subjected to mesh topological segmentation to extract the sealing surface boundary features.

[0057] The oil pump to be sealed is transported to the sealing station via a positioning tray. This process ensures the pump is precisely fixed in the designated position for subsequent scanning. When the pump arrives at the sealing station, the laser scanner integrated into the end effector of the adhesive applicator robot is activated. Using high-speed 3D scanning technology, the laser scanner performs a comprehensive scan of the pump's sealing surface. The scanning process involves emitting a laser beam and measuring the distance from each point to the laser, thereby obtaining a large number of 3D data points, i.e., point cloud data. This high-density point cloud data can accurately capture the details and subtle changes of the sealing surface. The resulting high-density 3D point cloud data contains the geometric information of the pump's sealing surface in space, forming the basis for subsequent processing.

[0058] Because of the mass production of the oil pumps to be sealed, the corresponding model reference geometric model of the oil pump can be found in the preset oil pump model library according to the model code. This geometric model is the standard design document of the oil pump, which contains important geometric features such as the shape of the oil pump, the dimensions of each part, and the curvature.

[0059] The high-density 3D point cloud data obtained from the scan was compared with the retrieved reference geometric model of the oil pump. The reference geometric model provides the general shape of the sealing surface, while the point cloud data provides more refined 3D features of the sealing surface. Through data fusion and comparison, details not captured in the reference model were supplemented. Surface fitting was performed using the point cloud data to reconstruct the actual geometric model of the sealing surface. This process involves converting the point cloud data into a smooth surface, eliminating noise and errors, and ensuring that the reconstructed geometric model reflects the shape of the sealing surface as accurately as possible. A sealing surface geometric model was obtained, which accurately represents the shape of the sealing surface of the oil pump to be sealed and possesses high-precision geometric information.

[0060] The sealing surface geometry model is transformed into a mesh model. Meshing decomposes the continuous geometric surface into small triangular or quadrilateral elements, which are connected by shared edges or vertices. Meshing facilitates subsequent calculations and analysis. Topological analysis is used to segment the mesh model, identifying the geometric features of different regions, such as areas with large curvature changes, smooth areas, local convexities, and depressions. These regions have different impacts on the adhesive application process and require separate handling. The sealing surface boundary features are extracted from the meshed geometric model, including distributed curvature gradients, distributed convexity / concave height fields, and distributed cross-sectional width spectra. These boundary features directly affect the adhesive application trajectory planning and process control.

[0061] In one implementation, the sealing surface boundary features include distributed curvature gradient, distributed concave-convex height field, and distributed cross-sectional width spectrum.

[0062] Distributed curvature gradients describe the rate of curvature change in different regions of the sealing surface. Regions with large curvature gradients represent abrupt changes or significant geometric features on the surface. These regions have a significant impact on the adhesive application process and usually require special handling, such as adjusting the application speed or pressure. Distributed unevenness height field represents the distribution of unevenness height in local areas of the sealing surface. Uneven areas affect the uniformity of adhesive application; therefore, the application trajectory or process parameters need to be adjusted based on the unevenness. Distributed cross-sectional width spectrum describes the width variation at different locations on the sealing surface. Regions with large width variations may require special application strategies to ensure that the application path covers the entire sealing surface, avoiding missed or over-application. These three features describe the geometric information of the sealing surface, serving as the basis for adhesive trajectory planning.

[0063] In one implementation, after identifying coating heterogeneity based on the boundary features of the sealing surface, a standard adhesive application trajectory is dynamically reconstructed to generate a continuous adhesive application path, including:

[0064] Y210: Retrieve the standard adhesive application trajectory from the preset trajectory knowledge base according to the model code of the oil pump to be sealed; Y220: After spatially superimposing the distributed curvature gradient and distributed cross-sectional width spectrum, dynamically partition the area based on regional risk judgment rules to obtain distributed steady-state region, distributed transition region, and distributed abrupt change region; Y230: Traverse the distributed concave-convex height field to locate distributed groove regions with a concavity depth greater than a preset depth threshold; Y240: Expand along the surface normal of the distributed groove region by a preset buffer distance to obtain the distributed coating forbidden region; Y250: By dividing the distributed coating forbidden region, distributed transition region, and distributed abrupt change region... The region is projected onto the standard adhesive application trajectory to locate the forbidden truncated segment sequence, the transition optimization segment sequence, and the abrupt reconstruction segment sequence; Y260: The forbidden truncated segment sequence, the transition optimization segment sequence, and the abrupt reconstruction segment sequence are reconstructed using feature constraints to output a safe avoidance trajectory sequence, an optimized transition trajectory sequence, and a gradient correction trajectory sequence; Y270: Based on the boundary nodes of the forbidden truncated segment sequence, the transition optimization segment sequence, and the abrupt reconstruction segment sequence, the C1 continuous spatiotemporal synchronous fusion of the safe avoidance trajectory sequence, the optimized transition trajectory sequence, and the gradient correction trajectory sequence is performed on the standard adhesive application trajectory to generate the continuous adhesive application path.

[0065] Based on the model code of the oil pump to be sealed, a standard adhesive application trajectory matching the model is retrieved from the pre-set adhesive application trajectory knowledge base. The standard adhesive application trajectory is pre-set according to the geometric characteristics and adhesive application requirements of similar models of oil pumps or equipment. It includes the basic planning of the adhesive application path, such as the starting point, the ending point, and the curvature of the path. The standard trajectory provides a preliminary reference path for the adhesive application process.

[0066] The distributed curvature gradient and distributed cross-sectional width spectrum of each region on the sealing surface are spatially superimposed, that is, these two features are mapped onto the geometric model of the sealing surface. The purpose is to comprehensively consider the changes in curvature and width on the sealing surface and form a unified feature description.

[0067] Based on the spatial distribution of these two characteristics, a regional risk criterion rule is used to divide the sealing surface into zones. This risk criterion rule is based on some preset thresholds, such as excessive curvature or excessive width variation, to define these zones. In the distributed steady-state zone, the curvature and width of the sealing surface change relatively little, making the adhesive application process relatively stable, and control parameters do not require significant adjustments. In the distributed transition zone, the curvature and width change significantly, making adhesive application more difficult and requiring adjustments to the adhesive application parameters based on the specific changes. In the distributed abrupt change zone, the curvature or width changes drastically, typically where the geometry of the sealing surface changes abruptly; the adhesive application robot needs special handling to avoid over-coating or under-coating.

[0068] By traversing the distributed unevenness height field of the sealing surface, i.e., the height of the depressions or protrusions in each region, analysis of the height field can identify areas with significant height variations, especially the depressions. If the depression depth of a certain region exceeds a preset threshold, that region is designated as a distributed groove area. These areas have greater depression depths, which can affect the uniformity of adhesive application and therefore require special attention.

[0069] For identified distributed groove areas, expansion is performed along the normal direction of the surface (the direction perpendicular to the surface). This expansion aims to create a buffer zone for the groove area. A preset buffer distance, the size of which depends on the flow characteristics of the adhesive and the working accuracy of the applicator head, is used to expand this buffer zone in the normal direction of the groove area, ensuring that adhesive does not enter these sensitive areas during the application process. Distributed coating taboo areas are then identified; these areas are where the applicator robot is prohibited from applying adhesive. These taboo areas are typically extensions of the groove areas, ensuring that adhesive is not mistakenly applied to these locations.

[0070] The distributed coating forbidden zone, distributed transition zone, and distributed abrupt change zone are projected onto the standard coating trajectory. The projection operation maps information from these regions to the spatial location of the coating trajectory, determining which parts of the standard trajectory are affected. Based on the projection results, several distinct trajectory segments are identified: the forbidden truncation segment is located in the forbidden coating zone; the coating robot should avoid these areas and therefore needs to be truncated from the standard coating trajectory; the transition optimization segment is located in the transition zone, requiring path optimization to ensure coating uniformity and avoid inconsistent coating thicknesses during coating; the abrupt change reconstruction segment is located in the abrupt change zone, requiring special adjustment of the coating head's movement to adapt to the abrupt changes in the sealing surface geometry.

[0071] Constraint reconstruction is performed on different trajectory segments to optimize the adhesive application path, making it smoother and more continuous, and able to handle complex sealing surface geometries. Specifically, for trajectory segments in the restricted area, it is ensured that these segments are correctly removed from the adhesive application path to prevent the adhesive head from entering the restricted area; for trajectory segments in the transition area, the path is optimized to ensure a smooth transition in the adhesive application process; for trajectory segments in the abrupt change area, a more complex reconstruction is performed, including smoothing the curves of abrupt changes to avoid uneven adhesive application caused by abrupt adjustments to the adhesive application path.

[0072] During the reconstruction process, all trajectory segments are ensured to meet specific constraints, such as position continuity, velocity synchronization, and normal acceleration constraints, to generate optimized adhesive application paths, namely, a safe avoidance trajectory sequence, an optimized transition trajectory sequence, and a gradient correction trajectory sequence.

[0073] Multiple optimized trajectory segments are fused spatiotemporally and synchronously using C1 continuity (first derivative continuity). C1 continuity requires that the trajectory path not only be continuous, but its derivative (i.e., tangent) must also be continuous. This ensures smooth movement of the dispensing head, without abrupt turns or speed changes, thus avoiding unevenness in the dispensing process. During spatiotemporal fusion, the trajectories are synchronized in time and space. This means that when the robot executes along the fused path, parameters such as speed and dispensing volume of each path segment transition smoothly and meet the actual requirements of the dispensing process. Based on the boundary nodes of taboo truncation segments, transition optimization segments, and abrupt reconstruction segments, these trajectory segments are fused using C1 continuity to generate the final continuous dispensing path. This continuous dispensing path will be the robot's operational trajectory when actually performing the dispensing task.

[0074] In one implementation, the uncoated area is used as the adhesive application constraint domain. The boundary features of the sealing surface are analyzed along the continuous adhesive application path to obtain a path curvature sequence, a path concavity / convexity feature sequence, and a path width sequence, including:

[0075] Y310: Identify restricted area features on the sealing surface geometric model and locate the vector restricted area polygon; Y320: Set the adhesive safety boundary distance based on the adhesive overflow risk analysis; Y330: Apply normal expansion of the adhesive safety boundary distance to the vector restricted area polygon along the surface normal of the sealing surface geometric model to generate the uncoated area; Y340: Set a fixed arc length interval to locate the sampling point sequence on the continuous adhesive application path; Y350: Using the uncoated area as the adhesive constraint domain, align and segment the distributed curvature gradient, distributed concave-convex height field, and distributed cross-sectional width spectrum along the sampling point sequence to obtain the path curvature sequence, path concave-convex feature sequence, and path width sequence.

[0076] The geometric model of the sealing surface is analyzed to identify areas where adhesive cannot be applied. These areas are unsuitable for adhesive application due to factors such as geometry and surface conditions; common no-go zones include bolt holes and deep grooves. These areas are then transformed into vector polygons. Specifically, boundary information is extracted from the sealing surface geometry model to mark areas where adhesive application is prohibited. These no-go zone polygons can be represented as a set of coordinates, clearly identifying which areas are hazardous. The resulting vector no-go zone polygons, representing the hazardous areas where adhesive cannot be applied, will serve as constraints during adhesive application path planning.

[0077] Based on the geometric characteristics of the sealing surface, especially areas with significant curvature changes, the risk of adhesive overflow is assessed. Overflow typically occurs when the application path is irregular or the application equipment is not precisely controlled, such as in bends or small curved surfaces. Excessive pressure or improper path control can cause adhesive to overflow. Based on this risk analysis, a safe application boundary distance is set. This boundary is a preset distance range to ensure that adhesive does not overflow or enter restricted areas during the application robot's movement. For example, a larger safety boundary is set around recessed areas, and a relatively smaller safety boundary is set around smooth areas.

[0078] Normal expansion is applied to the identified vector restricted area polygons. Normal expansion means extending the restricted area a certain distance along the normal direction of the sealing surface. This expanded area is actually to provide additional safety space for the adhesive application path, ensuring that the adhesive application robot does not enter the danger zone. Specifically, normal expansion extends along the perpendicular direction of the curved surface, preventing the adhesive application head from entering these restricted areas due to accuracy issues, or preventing adhesive from overflowing into the restricted areas. Through this normal expansion process, a new uncoated area is formed around the original restricted area, providing stricter constraints for subsequent path planning.

[0079] Setting fixed arc length intervals along the continuous adhesive application path means uniform sampling along the path, with each sampling point being the same distance apart, ensuring consistent path control during the adhesive application process. These fixed-interval sampling points form a sampling point sequence, which will serve as the basis for subsequent path optimization.

[0080] Using the uncoated area as the absolute boundary for adhesive application, the three-dimensional feature data of the sealing surface are aligned and segmented along the sampling point sequence. Specifically, the distributed curvature gradient, which characterizes the rate of change of surface curvature, the distributed concave-convex height field, which quantifies the depth of the depression, and the distributed cross-sectional width spectrum, which reflects the width of the sealing strip, are spatially matched and data sliced ​​according to the coordinates of the path sampling points.

[0081] This operation yields a path curvature sequence, representing the time-series set of curvature gradient values ​​at each point on the path; a path concavity / convexity feature sequence, reflecting the range sequence of concavity / convexity heights at each point on the path; and a path width sequence, representing the change sequence of cross-sectional widths at each point on the path. Through this analysis, the path curvature sequence, path concavity / convexity feature sequence, and path width sequence are generated, providing detailed geometric information for subsequent path optimization.

[0082] In one implementation, the multiple feature-associated control segments are used as search keys to retrieve multiple control parameter vectors from the adhesive coating process parameter knowledge base, including:

[0083] Y510: Based on multiple path sampling points in the sampling point sequence, spatial alignment and aggregation of the path curvature sequence, path concavity / convexity feature sequence, and path width sequence are performed to obtain the multiple feature-associated control segments. Each feature-associated control segment includes the mean curvature within the segment, the range of concavity / convexity height within the segment, and the mean width within the segment. S1: Using the mean width within the first segment of the first feature-associated control segment as a primary search key, a first dispensing pressure is matched in the adhesive coating process parameter knowledge base. S2: The parameter space of the adhesive coating process parameter knowledge base is narrowed using the first dispensing pressure to obtain a primary limited knowledge base. S3: The mean width within the first segment of the first feature-associated control segment is... The height range is used as a secondary search key to match the first coating head tilt angle in the primary knowledge base; S4: The first coating head tilt angle is used to narrow the parameter space of the primary knowledge base to obtain a secondary knowledge base; S5: The mean curvature value within the first segment of the first feature-associated control segment is used as a tertiary search key to match the first moving speed in the secondary knowledge base; wherein, the first dispensing pressure, the first coating head tilt angle, and the first moving speed constitute the first control parameter vector; Y520: Steps S1 to S5 are executed by analogy, and the multiple feature-associated control segments are used as search keys to search for the multiple control parameter vectors in the coating process parameter knowledge base.

[0084] The path curvature sequence, path concavity / convexity feature sequence, and path width sequence are spatially aligned and aggregated. Spatial alignment means mapping these feature sequences one-to-one with the path's sampling point sequence to ensure that each feature value matches its corresponding path position. The aggregated feature sequence is then divided into multiple feature-associated control segments. Each control segment consists of several sampling points on the path and is described by the following features: the mean curvature within the segment (the average curvature of all sampling points within the control segment), representing the average curvature of the path segment, used to eliminate manual adhesive application jitter; the maximum concavity / convexity height difference within the segment, representing the degree of path unevenness, used to address oil leakage in deep grooves; a larger range indicates greater adhesive application difficulty requiring special treatment; and the mean width within the segment, representing the average width of the path, which affects the distribution of adhesive during application and is used for precise adhesive control. The resulting multiple feature-associated control segments each describe the geometric features of a specific region on the path, providing a basis for optimizing process parameters and path control.

[0085] Using the average width of the first segment within the first feature-related control segment as the primary search key, the glue coating process parameter knowledge base is queried. The average width affects the amount of glue applied, because a wider path requires more glue, while a narrower path requires less glue. Based on the average width, the knowledge base returns a suitable first dispensing pressure. The magnitude of the dispensing pressure determines the amount of glue output during glue coating. Excessive pressure may cause glue to overflow, while insufficient pressure may result in insufficient glue coating.

[0086] By using the initial dispensing pressure as a constraint, the range of adhesive application process parameters is reduced. In other words, once a suitable dispensing pressure is determined based on the average width of the adhesive application path, the space for adhesive application parameters is narrowed down to the range that meets that pressure. This step effectively filters the adhesive application process parameters from the entire knowledge base into a smaller set that satisfies the requirement of the initial adhesive application pressure on the path. This results in a first-level constrained knowledge base, which is a set of parameters optimized for the average width within the first segment.

[0087] The first segment of the first feature-associated control segment is used as the second-level search key for the height difference between the concavity and convexity of the first segment. This value reflects the surface unevenness of the path segment, that is, the concavity and convexity difference of the path. A large concavity and convexity difference means that the tilt angle of the glue applicator needs to be adjusted to adapt to the surface undulation and ensure the uniformity of glue application. The tilt angle of the glue applicator determines the contact method of the glue applicator at different surface angles, avoiding uneven coating or glue waste.

[0088] Based on the retrieved first applicator head tilt angle, the applicator parameter space in the primary knowledge base is restricted to obtain a more precise parameter range. This step eliminates unsuitable applicator parameters, retaining only those that match the current path features and applicator head tilt angle. By restricting the parameter space, a secondary knowledge base is obtained, which contains applicator control parameters matching the first applicator head tilt angle.

[0089] The mean curvature of the first segment of the first feature-associated control segment is used as the third-level retrieval key. The mean curvature reflects the average curvature of the path. In areas with greater curvature, the dispensing robot needs to reduce its movement speed to improve dispensing accuracy, while in areas with less curvature, it can increase its speed. Based on the mean curvature, the corresponding first movement speed is queried from the second-level limited knowledge base. The choice of movement speed will affect the uniformity of adhesive application and the movement stability of the dispensing head. An appropriate speed can avoid instability or uneven dispensing during the dispensing process.

[0090] The first dispensing pressure, the first dispensing head tilt angle, and the first moving speed constitute the first control parameter vector.

[0091] The retrieval logic in the aforementioned process is applied to other feature-related control segments, and similar retrieval and process parameter matching are performed step by step. For each feature-related control segment, features such as path curvature, concavity and convexity, and width are used to search the knowledge base step by step to obtain a complete control parameter vector. Each control parameter vector contains glue dispensing pressure, glue head tilt angle, and moving speed, which are used to guide the actual operation of the glue dispensing robot.

[0092] In one implementation, after spatially superimposing the distributed curvature gradient and the distributed cross-sectional width spectrum, dynamic partitioning is performed based on regional risk criterion rules to obtain distributed steady-state regions, distributed transition regions, and distributed abrupt change regions, including:

[0093] Y221: Map the distributed curvature gradient and distributed cross-sectional width spectrum to the UV parameter coordinate system of the sealing surface geometric model, and use a bilinear interpolation algorithm to align the eigenvalue grids to obtain a unified feature grid; Y222: Use the regional risk criterion rule to traverse the unified feature grid to determine and identify grid attributes, and output attribute-identified grids; Y223: Separate the attribute-identified grids to obtain discrete steady-state grids, discrete transition grids, and discrete abrupt change grids; Y224: Use the watershed algorithm to perform regional aggregation processing on the discrete transition grids and discrete abrupt change grids, and output the distributed transition region and distributed abrupt change region; Y225: Remove the distributed transition region and distributed abrupt change region from the unified feature grid to obtain the distributed steady-state region.

[0094] Distributed curvature gradients and distributed cross-sectional width spectra are transformed into data in the UV parametric coordinate system through mapping. This allows for unified processing and analysis of features from different regions. The UV coordinate system is used to parametrically represent surfaces, converting three-dimensional geometric data into a two-dimensional representation. To ensure a uniform distribution of eigenvalues ​​in the coordinate system, a bilinear interpolation algorithm is used to align the curvature gradients and width spectra to a grid. Bilinear interpolation is a weighted averaging method based on neighboring points, which smoothly fills each point on the grid with discrete eigenvalues, thus obtaining a uniform feature grid. The resulting uniform feature grid contains the curvature and width eigenvalues ​​mapped to the UV coordinate system.

[0095] Based on the risks during the adhesive application process, such as adhesive overflow and uneven application, regional risk judgment rules are established. These rules are based on the geometric characteristics of different regions, such as curvature and width variations. By analyzing the feature mesh, it is determined which regions have higher risks. Each mesh cell in the unified feature mesh is traversed, and the region type of that mesh cell is determined according to the judgment rules, including steady state, transition, or abrupt change. The mesh cells are then attribute-labeled, resulting in an attribute-labeled mesh, where each mesh cell is marked as a different region type.

[0096] Based on the attribute identifiers of the mesh elements, mesh elements marked as steady-state regions are selected, resulting in discrete steady-state meshes. These regions exhibit gradual geometrical changes, making the adhesive application process relatively stable and posing a low risk. Based on the attribute identifiers of the mesh elements, mesh elements marked as transition regions are selected, resulting in discrete transition meshes. The geometrical characteristics of these regions show a gradual change trend, requiring adjustments to the adhesive application strategy to address the gradually changing surface properties. Based on the attribute identifiers of the mesh elements, mesh elements marked as abrupt change regions are selected, resulting in discrete abrupt change meshes. These regions exhibit drastic changes along the path, such as a sharp increase in curvature or a dramatic change in width. Applying adhesive to these regions carries a higher risk, such as adhesive overflow or uneven application, requiring special attention.

[0097] The watershed algorithm is an image segmentation technique used to divide similar regions in an image into different parts. Its core idea is to treat different regions in the image as terrain, simulating water flow to gradually merge adjacent similar regions. For transitional and abrupt change regions along the path, the watershed algorithm can aggregate these regions based on boundary information. Through the watershed algorithm, these discrete grid regions are merged into larger, coherent distributed transition regions (continuous transition regions) and distributed abrupt change regions (continuous abrupt change regions).

[0098] The distributed transition region and the distributed mutation region are removed from the unified feature mesh, leaving the part with small geometric changes, namely the distributed steady-state region. The distributed steady-state region refers to the area with small curvature changes and gentle width changes. The coating conditions in these areas are relatively stable, and the coating robot can operate according to standard control parameters without much adjustment.

[0099] In one implementation, during the process of controlling the glue-applying robot to execute the glue-applying control sequence along the continuous glue-applying path on the sealing surface of the oil pump to be sealed, real-time PID feedback compensation is performed based on the glue line cross-sectional thickness feedback, including:

[0100] Y710: Dynamically acquire the thickness of the adhesive line cross-section using a laser triangulation thickness sensor integrated 5mm behind the adhesive application head of the adhesive application robot; Y720: Dynamically compare the target cross-section thickness with the adhesive line cross-section thickness and output the real-time thickness deviation; Y730: Perform real-time compensation action matching based on the deviation characteristics and amplitude of the real-time thickness deviation; Y740: After executing the real-time compensation action using a PID controller and performing PID real-time feedback compensation to the preset adhesive application path length, re-check the adhesive line cross-section thickness and perform closed-loop feedback compensation for adhesive application control based on the re-check result.

[0101] A laser triangulation thickness sensor is integrated 5mm behind the glue-applying head of the glue-applying robot. This sensor, based on the principle of triangulation, can accurately measure the thickness of the glue line. This precise location helps to monitor the glue layer thickness in a timely and accurate manner. The sensor monitors the glue line thickness changes in real time and feeds the data back to the control system. This allows for continuous updates to the glue line thickness information during the glue-applying process, providing real-time data on the glue line cross-sectional thickness as a basis for subsequent compensation and adjustments.

[0102] The target cross-sectional thickness is a preset standard value, set according to the characteristics of the sealing surface and the adhesive application path. The adhesive line cross-sectional thickness is obtained by dynamically collecting data through a laser sensor and comparing the two in real time. Based on the difference between the target cross-sectional thickness and the adhesive line cross-sectional thickness, the real-time thickness deviation is output. This deviation is used to determine whether the adhesive application is too thick or too thin, thereby triggering the corresponding compensation action.

[0103] The deviation characteristics of real-time thickness deviation are analyzed. When the real-time thickness deviation is positive, it indicates that the glue line is too thick, i.e., glue overflow has occurred, and the amount of glue needs to be reduced. When the deviation is negative, it indicates that the glue line is too thin, i.e., insufficient glue application, and the amount of glue applied needs to be increased. Based on the deviation characteristics and deviation amplitude, control parameters in the glue application process, such as glue application pressure, moving speed, and glue head height, are adjusted to correct the glue line thickness. For example, when the real-time thickness deviation e(t) > 0.1 mm, the glue application process triggers a deceleration and pressure reduction mode, i.e., the moving speed is increased and the glue application pressure is reduced, thereby reducing the amount of glue and controlling the glue line thickness. When e(t) < -0.1 mm, the glue application process triggers a pressure increase and lifting mode, i.e., the moving speed is reduced, the glue application pressure is increased, and the glue head height is raised, to increase the amount of glue and ensure uniform glue application.

[0104] A PID (Proportional-Integral-Derivative) controller is a feedback controller used in automation control. It adjusts control parameters by calculating the current error. Once a suitable compensation value is calculated, the glue-applying robot adjusts the glue-applying parameters accordingly. During the glue-applying process, the glue line thickness is continuously monitored, and the glue-applying path length is periodically rechecked. After each segment of the path is glued, the glue line thickness is rechecked and compared with the target value. If the glue thickness still deviates significantly after the path length recheck, the PID controller readjusts the glue-applying parameters through a feedback mechanism until the thickness error reaches an acceptable range. This process is a closed-loop control process, meaning the glue-applying robot continuously adjusts the glue-applying control based on real-time feedback information to ensure the glue-applying quality reaches the ideal state.

[0105] In one implementation, the control parameter vector includes moving speed, dispensing pressure, and applicator head tilt angle.

[0106] The moving speed of the glue-applying robot directly affects the amount of glue deposited during the application process. If the robot moves too fast, the glue may not have enough time to fully deposit, resulting in a thin coating. Conversely, if the speed is too slow, too much glue may deposit, causing an overly thick coating or glue overflow. Dispensing pressure controls the glue flow rate. Excessive pressure may cause glue overflow, while insufficient pressure will result in insufficient glue flow and uneven application. The tilt angle of the dispensing head directly affects the glue distribution on the surface. An appropriate tilt angle allows the glue to evenly cover the target surface, while an inappropriate angle can lead to glue accumulation or uneven distribution.

[0107] Example 2: Based on the same inventive concept as the robot adaptive adhesive application control method for variable cross-section sealing surfaces in the foregoing examples, this invention provides a robot adaptive adhesive application control system for variable cross-section sealing surfaces. See [link to example]. Figure 2 As shown, the system includes:

[0108] The feature extraction module 10 is used to extract the sealing surface features of the oil pump to be sealed upon entering the sealing station, and obtain the sealing surface boundary features; the trajectory reconstruction module 20 is used to perform coating heterogeneity identification based on the sealing surface boundary features, and then perform standard adhesive application trajectory dynamic reconstruction to generate a continuous adhesive application path; the feature parsing module 30 is used to parse the sealing surface boundary features along the continuous adhesive application path using the non-coated area as the adhesive application constraint domain, and obtain the path curvature sequence, path concavity and convexity feature sequence, and path width sequence; the control segment acquisition module 40 is used to spatially align and reorganize the path curvature sequence, path concavity and convexity feature sequence, and path width sequence to obtain multiple feature-associated control segments; the control parameter retrieval module 50 is used to use the multiple feature-associated control segments as retrieval keys to retrieve multiple control parameter vectors in the adhesive application process parameter knowledge base; the control sequence generation module 60 is used to dynamically fuse the multiple control parameter vectors to generate an adhesive application control sequence; the feedback compensation module 70 is used to control the adhesive application robot to perform real-time PID feedback compensation based on the adhesive line cross-sectional thickness feedback during the execution of the adhesive application control sequence along the continuous adhesive application path on the sealing surface of the oil pump to be sealed.

[0109] In one implementation, the feature extraction module 10 is used to:

[0110] After the oil pump to be sealed is transported to the sealing station via a positioning tray, a laser scanner integrated at the end of the adhesive applicator robot is triggered to perform a high-speed 3D scan of the sealing surface, obtaining high-density 3D point cloud data. Based on the model code of the oil pump to be sealed, the oil pump reference geometric model is retrieved from the oil pump model library. Based on the high-density 3D point cloud data, the sealing surface is reconstructed on the oil pump reference geometric model to obtain the sealing surface geometric model. The sealing surface geometric model is then subjected to mesh topological segmentation to extract the sealing surface boundary features.

[0111] In one implementation, the sealing surface boundary features include distributed curvature gradient, distributed concave-convex height field, and distributed cross-sectional width spectrum.

[0112] In one implementation, the trajectory reconstruction module 20 is used for:

[0113] The standard adhesive application trajectory is retrieved from the preset trajectory knowledge base based on the model code of the oil pump to be sealed. After spatially superimposing the distributed curvature gradient and distributed cross-sectional width spectrum, dynamic partitioning is performed based on regional risk judgment rules to obtain distributed steady-state region, distributed transition region, and distributed abrupt change region. The distributed concave-convex height field is traversed to locate distributed groove regions with a concavity depth greater than a preset depth threshold. A preset buffer distance is expanded along the surface normal of the distributed groove region to obtain distributed coating taboo region. By projecting the distributed coating taboo region, distributed transition region, and distributed abrupt change region onto the standard adhesive application trajectory, the taboo truncated segment sequence, transition optimized segment sequence, and abrupt reconstructed segment sequence are located. Feature constraint reconstruction is performed on the taboo truncated segment sequence, transition optimized segment sequence, and abrupt reconstructed segment sequence to output safe avoidance trajectory sequence, optimized transition trajectory sequence, and gradient correction trajectory sequence. Based on the boundary nodes of the taboo truncated segment sequence, transition optimized segment sequence, and abrupt reconstructed segment sequence, C1 continuous spatiotemporal synchronous fusion of the safe avoidance trajectory sequence, optimized transition trajectory sequence, and gradient correction trajectory sequence is performed on the standard adhesive application trajectory to generate the continuous adhesive application path.

[0114] In one implementation, the feature parsing module 30 is used for:

[0115] The sealing surface geometric model is subjected to restricted area feature identification to locate the vector restricted area polygon; a safe adhesive boundary distance is set based on adhesive overflow risk analysis; along the surface normal of the sealing surface geometric model, the normal expansion of the vector restricted area polygon is applied to the adhesive safety boundary distance to generate the uncoated area; a fixed arc length interval is set to locate the sampling point sequence on the continuous adhesive application path; using the uncoated area as the adhesive constraint domain, the distributed curvature gradient, distributed concave-convex height field, and distributed cross-sectional width spectrum are aligned and segmented along the sampling point sequence to obtain the path curvature sequence, path concave-convex feature sequence, and path width sequence.

[0116] In one implementation, the control parameter retrieval module 50 is used for:

[0117] Based on the spatial alignment and aggregation of multiple path sampling points in the sampling point sequence, the path curvature sequence, the path concavity / convexity feature sequence, and the path width sequence, multiple feature-associated control segments are obtained. Each feature-associated control segment includes the mean curvature value within the segment, the range of concavity / convexity height within the segment, and the mean width within the segment. S1: The mean width within the first segment of the first feature-associated control segment is used as a primary search key to match the first dispensing pressure in the adhesive coating process parameter knowledge base. S2: The first dispensing pressure is used to narrow the parameter space of the adhesive coating process parameter knowledge base to obtain a primary limited knowledge base. S3: The first concavity / convexity value within the first segment of the first feature-associated control segment is... The height range is used as a secondary search key to match the first coating head tilt angle in the primary knowledge base; S4: The first coating head tilt angle is used to narrow the parameter space of the primary knowledge base to obtain a secondary knowledge base; S5: The mean curvature value within the first segment of the first feature-associated control segment is used as a tertiary search key to match the first moving speed in the secondary knowledge base; wherein, the first dispensing pressure, the first coating head tilt angle, and the first moving speed constitute the first control parameter vector; Steps S1 to S5 are executed by analogy, using the multiple feature-associated control segments as search keys to retrieve the multiple control parameter vectors in the coating process parameter knowledge base.

[0118] In one implementation, the trajectory reconstruction module 20 is used for:

[0119] The distributed curvature gradient and distributed cross-sectional width spectrum are mapped to the UV parameter coordinate system of the sealing surface geometric model, and the eigenvalue grid is aligned using a bilinear interpolation algorithm to obtain a unified feature grid. The unified feature grid is then traversed using the regional risk criterion rule to identify grid attributes and output attribute-identified grids. The attribute-identified grids are separated to obtain discrete steady-state grids, discrete transition grids, and discrete abrupt change grids. The watershed algorithm is used to perform regional aggregation processing on the discrete transition grids and discrete abrupt change grids to output the distributed transition region and the distributed abrupt change region. The distributed transition region and the distributed abrupt change region are removed from the unified feature grid to obtain the distributed steady-state region.

[0120] In one implementation, the feedback compensation module 70 is used for:

[0121] A laser triangulation thickness sensor integrated 5mm behind the glue-applying head of the glue-applying robot is used to dynamically acquire the thickness of the glue line cross-section; the target cross-section thickness and the glue line cross-section thickness are dynamically compared to output the real-time thickness deviation; real-time compensation actions are matched according to the deviation characteristics and amplitude of the real-time thickness deviation; after the real-time compensation action is executed by a PID controller and the PID real-time feedback compensation is performed to the preset glue-applying path length, the glue line cross-section thickness is re-checked, and the glue-applying control closed-loop feedback compensation is performed according to the re-check result.

[0122] In one implementation, the control parameter vector includes moving speed, dispensing pressure, and applicator head tilt angle.

[0123] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

[0124] The foregoing description of specific exemplary embodiments of the invention is for illustrative and explanatory purposes. These descriptions are not intended to limit the invention to the precise forms disclosed, and it will be apparent that many changes and variations can be made in accordance with the foregoing teachings. The exemplary embodiments were chosen and described in order to explain the specific principles of the invention and its practical application, thereby enabling those skilled in the art to implement and utilize various different exemplary embodiments of the invention, as well as various different choices and variations. The scope of the invention is intended to be defined by the claims and their equivalents.

Claims

1. A robot adaptive adhesive application control method for variable cross-section sealing surfaces, characterized in that, The method includes: The sealing surface features of the oil pump to be sealed are extracted upon entering the sealing station to obtain the boundary features of the sealing surface. After identifying coating heterogeneity based on the boundary features of the sealing surface, dynamic reconstruction of the standard adhesive application trajectory is performed to generate a continuous adhesive application path. Using the uncoated area as the adhesive constraint domain, the boundary features of the sealing surface are analyzed along the continuous adhesive application path to obtain the path curvature sequence, the path concavity and convexity feature sequence, and the path width sequence. Spatial alignment and recombination of the path curvature sequence, path concavity / convexity feature sequence, and path width sequence yields multiple feature-associated control segments; Using the multiple feature-associated control segments as search keys, multiple control parameter vectors are retrieved from the glue coating process parameter knowledge base; The multiple control parameter vectors are dynamically fused to generate an adhesive application control sequence; During the process of controlling the glue-applying robot to execute the glue-applying control sequence along the continuous glue-applying path on the sealing surface of the oil pump to be sealed, PID real-time feedback compensation is performed based on the feedback of the glue line cross-sectional thickness.

2. The robot adaptive adhesive application control method for variable cross-section sealing surfaces as described in claim 1, characterized in that, The sealing surface features of the oil pump entering the sealing station are extracted to obtain the boundary features of the sealing surface, including: After the oil pump to be sealed is transported to the sealing station via the positioning tray, it triggers the laser scanner integrated at the end of the glue-applying robot to perform a high-speed three-dimensional scan of the sealing surface and obtain high-density three-dimensional point cloud data. Based on the model code of the oil pump to be sealed, retrieve the reference geometric model of the oil pump from the oil pump model library; Based on the high-density three-dimensional point cloud data, the sealing surface is reconstructed in the oil pump reference geometric model to obtain the sealing surface geometric model. The geometric model of the sealing surface is subjected to mesh topological segmentation to extract the boundary features of the sealing surface.

3. The robot adaptive adhesive application control method for variable cross-section sealing surfaces as described in claim 2, characterized in that, The sealing surface boundary features include distributed curvature gradient, distributed concave-convex height field, and distributed cross-sectional width spectrum.

4. The robot adaptive adhesive application control method for variable cross-section sealing surfaces as described in claim 3, characterized in that, After identifying coating heterogeneity based on the boundary features of the sealing surface, a standard adhesive application trajectory is dynamically reconstructed to generate a continuous adhesive application path, including: The standard adhesive application trajectory is retrieved from the preset trajectory knowledge base based on the model code of the oil pump to be sealed. After spatially superimposing the distributed curvature gradient and the distributed cross-sectional width spectrum, dynamic partitioning is performed based on the regional risk criterion rule to obtain the distributed steady-state region, the distributed transition region and the distributed abrupt change region. Traverse the distributed concave-convex height field to locate distributed groove regions where the concavity depth is greater than a preset depth threshold; Expand a preset buffer distance along the curved normal of the distributed groove region to obtain the distributed coating forbidden zone; By projecting the distributed coating forbidden region, distributed transition region and distributed mutation region onto the standard coating trajectory, the forbidden truncated fragment sequence, the transition optimized fragment sequence and the mutation reconstructed fragment sequence are located; The taboo truncated fragment sequence, the transition optimization fragment sequence, and the mutation reconstruction fragment sequence are reconstructed by feature constraints to output a safe avoidance trajectory sequence, an optimized transition trajectory sequence, and a gradient correction trajectory sequence. Based on the boundary nodes of the forbidden truncated segment sequence, the transition optimized segment sequence, and the mutation reconstructed segment sequence, the C1 continuous spatiotemporal synchronous fusion of the safe avoidance trajectory sequence, the optimized transition trajectory sequence, and the gradient correction trajectory sequence is performed on the standard adhesive application trajectory to generate the continuous adhesive application path.

5. The robot adaptive adhesive application control method for variable cross-section sealing surfaces as described in claim 3, characterized in that, Using the uncoated area as the adhesive constraint domain, the boundary features of the sealing surface are analyzed along the continuous adhesive application path to obtain a path curvature sequence, a path concavity / convexity feature sequence, and a path width sequence, including: The restricted area feature is identified by performing restricted area feature identification on the geometric model of the sealing surface, and the restricted area polygon is located by vector. Set the safe boundary distance for glue application based on glue overflow risk analysis; Along the surface normal of the sealing surface geometry model, apply a normal expansion of the adhesive safety boundary distance to the vector restricted area polygon to generate the uncoated area; A fixed arc length interval is set to locate the sampling point sequence along the continuous adhesive application path; Using the uncoated area as the adhesive constraint domain, the distributed curvature gradient, distributed concavity and convexity height field, and distributed cross-sectional width spectrum are aligned and segmented along the sampling point sequence to obtain the path curvature sequence, path concavity and convexity feature sequence, and path width sequence.

6. The robot adaptive adhesive application control method for variable cross-section sealing surfaces as described in claim 5, characterized in that, Using the aforementioned feature-associated control segments as search keys, multiple control parameter vectors are retrieved from the adhesive coating process parameter knowledge base, including: Based on multiple path sampling points in the sampling point sequence, the spatial alignment and aggregation of the path curvature sequence, the path concavity and convexity feature sequence, and the path width sequence are performed to obtain the multiple feature association control segments, wherein each feature association control segment includes the mean curvature value within the segment, the range of concavity and convexity height within the segment, and the mean width within the segment. S1: Use the average width of the first segment of the first feature-associated control segment as the first-level search key, and match the first dispensing pressure in the glue coating process parameter knowledge base; S2: The parameter space of the coating process parameter knowledge base is narrowed by the first dispensing pressure to obtain a first-level limited knowledge base; S3: Use the height difference of the first segment of the first feature-associated control segment as the secondary search key, and match the first glue applicator tilt angle in the primary knowledge base; S4: The parameter space of the first-level limited knowledge base is narrowed by using the tilt angle of the first dispensing head to obtain the second-level limited knowledge base; S5: Use the average curvature value within the first segment of the first feature-associated control segment as the third-level retrieval key, and match the first moving speed in the second-level limited knowledge base; The first dispensing pressure, the first dispensing head tilt angle, and the first moving speed constitute the first control parameter vector. By analogy, steps S1 to S5 are executed, and the multiple feature-associated control segments are used as search keys to retrieve the multiple control parameter vectors in the glue coating process parameter knowledge base.

7. The robot adaptive adhesive application control method for variable cross-section sealing surfaces as described in claim 4, characterized in that, After spatially superimposing the distributed curvature gradient and distributed cross-sectional width spectrum, dynamic partitioning is performed based on regional risk criterion rules to obtain distributed steady-state regions, distributed transition regions, and distributed abrupt change regions, including: The distributed curvature gradient and distributed cross-sectional width spectrum are mapped to the UV parameter coordinate system of the sealing surface geometric model, and the eigenvalue grid is aligned using a bilinear interpolation algorithm to obtain a unified feature grid. The unified feature grid is traversed using the aforementioned regional risk criterion rule to determine and identify grid attributes, and the attribute-identified grid is output. Separate the attribute-identifying mesh to obtain discrete steady-state mesh, discrete transition mesh, and discrete abrupt change mesh; The watershed algorithm is used to perform region aggregation processing on the discrete transition grid and the discrete mutation grid, and the distributed transition region and the distributed mutation region are output. The distributed steady-state region is obtained by removing the distributed transition region and the distributed mutation region from the unified feature grid.

8. The robot adaptive adhesive application control method for variable cross-section sealing surfaces as described in claim 1, characterized in that, During the process of controlling the glue-applying robot to execute the glue-applying control sequence along the continuous glue-applying path on the sealing surface of the oil pump to be sealed, real-time PID feedback compensation is performed based on the feedback of the glue line cross-sectional thickness, including: Dynamic acquisition of adhesive line cross-sectional thickness is performed by using a laser triangulation thickness sensor integrated 5mm at the rear end of the adhesive application head of the adhesive application robot. The target cross-sectional thickness and the adhesive line cross-sectional thickness are dynamically compared, and the real-time thickness deviation is output. Real-time compensation actions are matched based on the deviation characteristics and magnitude of the real-time thickness deviation; After the real-time compensation action is performed by using a PID controller and the PID real-time feedback compensation is performed to the preset glue application path length, the glue line cross-sectional thickness is re-checked, and the glue application control closed-loop feedback compensation is performed based on the re-check result.

9. The robot adaptive adhesive application control method for variable cross-section sealing surfaces as described in claim 1, characterized in that, The control parameter vector includes moving speed, dispensing pressure, and dispensing head tilt angle.

10. A robot adaptive adhesive application control system for variable cross-section sealing surfaces, characterized in that, For implementing the method steps of any one of claims 1 to 9, including: The feature extraction module is used to extract the sealing surface features of the oil pump to be sealed as it enters the sealing station, and obtain the sealing surface boundary features. The trajectory reconstruction module is used to perform dynamic reconstruction of the standard adhesive application trajectory based on the boundary features of the sealing surface after identifying the coating heterogeneity, and to generate a continuous adhesive application path. The feature parsing module is used to analyze the boundary features of the sealing surface along the continuous adhesive application path, with the uncoated area as the adhesive constraint domain, to obtain the path curvature sequence, the path concavity and convexity feature sequence, and the path width sequence. The control segment acquisition module is used to spatially align and reconstruct the path curvature sequence, path concave-convex feature sequence, and path width sequence to obtain multiple feature-associated control segments. The control parameter retrieval module is used to retrieve multiple control parameter vectors from the adhesive coating process parameter knowledge base by using the multiple feature-associated control segments as retrieval keys. A control sequence generation module is used to dynamically fuse the multiple control parameter vectors to generate an adhesive application control sequence; The feedback compensation module is used to control the glue-applying robot to perform real-time PID feedback compensation based on the thickness of the glue line cross-section during the glue-applying control sequence along the continuous glue-applying path on the sealing surface of the oil pump to be sealed.

Images