Intelligent generation method and device for end cap disassembly machine based on AI and point cloud recognition

Abstract

The application discloses an AI and point cloud recognition-based intelligent generation method and device for a special machine for disassembling an end cover. The method comprises the following steps: obtaining three-dimensional product data of a product to be operated, and performing point cloud processing on the three-dimensional product data to obtain point cloud data; segmenting the point cloud data to obtain a plurality of segmented point cloud components, and performing assembly relationship reasoning on the plurality of segmented point cloud components to construct a component assembly relationship graph; identifying an end cover component and an adjacent component of the end cover component from the plurality of segmented point cloud components based on the component assembly relationship graph, and extracting key parameters for the end cover component and the adjacent component; and generating a core component model of a special machine for disassembling an end cover based on the key parameters, and assembling the core component model to generate the special machine for disassembling an end cover. The application solves the technical problem of low design efficiency of a special machine for disassembling an end cover in the prior art.

Description

Technical Field

[0001] This invention relates to the field of intelligent automotive design, and more specifically, to a method and apparatus for intelligent generation of end caps based on AI and point cloud recognition. Background Technology

[0002] In manufacturing fields such as aerospace, automotive, motors, and new energy batteries, the installation and removal of end caps are critical processes. To improve production efficiency, dedicated end cap removal machines are typically designed. Existing end cap removal machines generally consist of a positioning fixture, a clamping mechanism, and an end cap removal mechanism. Their basic working principle is as follows: the positioning fixture supports the product, the clamping mechanism fixes the workpiece, and then the appropriate tools are used to remove the end cap.

[0003] However, due to differences in size, structure, and materials among various products, existing technologies often require manual redesign and modeling of specialized aircraft. This approach has the following drawbacks: First, it is inefficient, as engineers typically need to model from scratch, making it difficult to meet the demands of rapid response. Second, the design process is highly repetitive, lacking the ability to intelligently reuse and modify common components and similar structures. Third, design quality is highly dependent on the designer's experience, posing a risk of human error and resulting in insufficient stability and reliability of the output solution. Finally, existing technologies lack intelligent adaptation methods, making it difficult to quickly and accurately generate suitable specialized aircraft parts based on different product structural characteristics, thus limiting the degree of automation.

[0004] Therefore, existing technologies have significant limitations when facing the need for rapid adaptation to multiple product categories. There is an urgent need for a dedicated machine for generating end caps that can improve design efficiency, reduce reliance on manual labor, and possess intelligent adaptability. Summary of the Invention

[0005] This invention provides a method and apparatus for intelligently generating end cap removal machines based on AI and point cloud recognition, which at least solves the technical problem of low design efficiency of end cap removal machines in the prior art.

[0006] According to one aspect of the present invention, a method for intelligently generating a cap removal machine based on AI and point cloud recognition is provided, comprising: acquiring three-dimensional product data of a product to be operated, and performing point cloud processing on the three-dimensional product data to obtain point cloud data; segmenting the point cloud data to obtain multiple segmented point cloud components, and performing assembly relationship reasoning on the multiple segmented point cloud components to construct a component assembly relationship diagram; based on the component assembly relationship diagram, identifying the cap component and its adjacent components from the multiple segmented point cloud components, and extracting key parameters for the cap component and its adjacent components; generating a core component model of the cap removal machine based on the key parameters, and assembling the core component model to generate the cap removal machine.

[0007] According to another aspect of the present invention, an intelligent generation device for a cap removal machine based on AI and point cloud recognition is also provided, comprising: a point cloud processing module configured to acquire three-dimensional product data of the product to be operated, and to perform point cloud processing on the three-dimensional product data to obtain point cloud data; a construction module configured to segment the point cloud data to obtain multiple segmented point cloud components, and to perform assembly relationship reasoning on the multiple segmented point cloud components to construct a component assembly relationship diagram; an extraction module configured to identify the cap component and its adjacent components from the multiple segmented point cloud components based on the component assembly relationship diagram, and to extract key parameters for the cap component and its adjacent components; and a generation module configured to generate a core component model of the cap removal machine based on the key parameters, and to assemble the core component model to generate the cap removal machine.

[0008] In this embodiment of the invention, three-dimensional product data of the product to be operated is acquired, and the three-dimensional product data is processed into point cloud data. The point cloud data is segmented to obtain multiple segmented point cloud components, and assembly relationship reasoning is performed on the multiple segmented point cloud components to construct a component assembly relationship diagram. Based on the component assembly relationship diagram, the end cap component and its adjacent components are identified from the multiple segmented point cloud components, and key parameters are extracted for the end cap component and its adjacent components. Based on the key parameters, a core component model of the end cap removal machine is generated, and the core component model is assembled to generate the end cap removal machine. This solution solves the technical problem of low design efficiency of end cap removal machines in the prior art. Attached Figure Description

[0009] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0010] Figure 1 This is a schematic diagram of an optional end cap removal machine according to an embodiment of the present invention;

[0011] Figure 2 This is a flowchart of an optional method for intelligent generation of end cap removal machines based on AI and point cloud recognition according to an embodiment of the present invention.

[0012] Figure 3 This is a flowchart of another optional method for intelligent generation of end cap removal machines based on AI and point cloud recognition according to an embodiment of the present invention;

[0013] Figure 4 This is a flowchart of an optional method for generating a core component model of a special-purpose aircraft with a disassembled end cap, according to an embodiment of the present invention.

[0014] Figure 5 This is a schematic diagram of an optional intelligent generation device for end cap removal based on AI and point cloud recognition according to an embodiment of the present invention.

[0015] Figure 6 A schematic diagram of the structure of a computer device suitable for implementing embodiments of the present disclosure is shown. Detailed Implementation

[0016] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0017] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0018] According to an embodiment of the present invention, a method for intelligent generation of end caps based on AI and point cloud recognition is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0019] Specialized machine for removing end caps, such as Figure 1 As shown, it includes a positioning fixture 26, a positioning clamping mechanism 22, and a cap removal actuator assembly 24. The cap removal actuator assembly 24 may include a head cap removal mechanism 241 and a tail cap removal mechanism 242.

[0020] The head end cap removal mechanism 241 includes an end cap clamping mechanism, a head end cap removal fixture, a motor, a servo module, a feed slide, and a guide rail. The mechanism drives the feed slide along the guide rail through the servo module, thereby causing the end cap clamping mechanism and the removal fixture to achieve precise displacement, so as to complete the clamping and removal operation of one end cap of the product.

[0021] The tail end cap removal mechanism 242 includes an end cap clamping mechanism, a tail end cap removal fixture, a motor, a servo module, a feed slide, and a guide rail. Its structure corresponds to that of the head mechanism, and it also uses servo drive to achieve precise removal and installation of the other end cap.

[0022] The positioning and clamping mechanism 22 includes a clamping cylinder, a clamping positioning block, and a feed slide. This mechanism uses the clamping cylinder to push the clamping positioning block, thereby achieving reliable positioning and clamping of the product and ensuring its stability during the cap removal process.

[0023] Figure 2 This is a method for intelligent generation of end cap removal machines based on AI and point cloud recognition according to embodiments of the present invention, such as... Figure 2 As shown, the method includes the following steps:

[0024] Step S202: Obtain the three-dimensional product data of the product to be operated, and perform point cloud processing on the three-dimensional product data to obtain point cloud data.

[0025] When the 3D product data is a CAD model, a geometric sampling algorithm is used to convert the surface mesh of the CAD model into point cloud data. When the 3D product data is 3D scan data, a statistical filtering method is used to remove outlier noise points in the 3D scan data. Voxel mesh downsampling is used to reduce the data density of the 3D scan data after removing outlier noise points. Principal component analysis is used to calculate the point cloud normal vector of the 3D scan data after reducing the data density to obtain the point cloud data.

[0026] Step S204: Segment the point cloud data to obtain multiple segmented point cloud components, and perform assembly relationship reasoning on the multiple segmented point cloud components to construct a component assembly relationship diagram.

[0027] The point cloud data is divided into multiple independent point cloud components using a point cloud segmentation algorithm. These segmented components include: a region growing method based on geometric features, a random sampling consistency method, or a point cloud segmentation network method based on deep learning. Based on the geometric contact relationships, mating features, and normal vector consistency of the segmented components, assembly constraints between the segmented components are inferred. An assembly relationship diagram is generated based on these constraints to characterize the structural hierarchy and assembly relationships of the product to be operated on.

[0028] Step S206: Based on the component assembly relationship diagram, identify the end cap component and its adjacent components from the multiple segmented point cloud components, and extract key parameters for the end cap component and its adjacent components.

[0029] Based on the component assembly relationship diagram, geometric features are extracted from the multiple segmented point cloud components. These geometric features include at least one of the following: outer contour shape, circular hole distribution, thickness, and symmetry. Based on these geometric features, a deep learning classification model is used to identify the end cap component and adjacent components from the multiple segmented point cloud components, and the type of the end cap component is determined. The type of the end cap component includes bolt-fixed end caps, snap-fit ​​end caps, or press-fit end caps. Key parameters of the end cap component and adjacent components are extracted. These key parameters include at least one of the following: geometric parameters, assembly constraint parameters, and mechanical requirement parameters. The geometric parameters include at least one of the following: end diameter, thickness, length, chamfer, position coordinates, and installation depth. The assembly constraint parameters include fit tolerances and connection relationships. The mechanical requirement parameters include at least one of the following: disassembly force and clamping force.

[0030] Step S208: Based on the key parameters, generate a core component model of the end cap removal machine, and assemble the core component model to generate the end cap removal machine.

[0031] Based on the key parameters, the reference surface and support points of the product to be operated are determined, and a positioning fixture is generated based on the reference surface and the support points. Based on the key parameters, combined with the force requirements of the end cap of the product to be operated and the overall weight of the product to be operated, the clamping force is calculated. Based on the clamping force, the required specifications, quantity, and layout of the drive components are deduced, and a positioning clamping mechanism is generated based on the specifications, quantity, and layout of the drive components. Based on the type of the end cap assembly and the key parameters, the core component model of the end cap removal machine is generated.

[0032] This embodiment also provides another intelligent generation method for end cap removal machines based on AI and point cloud recognition. This method realizes a complete process from inputting the 3D product data of the product to be operated to the automatic generation of the end cap removal machine, covering a closed-loop chain from input (3D product data) → data processing → AI recognition and reasoning → parameter extraction → generating design schemes → verification feedback → output results. This overcomes the limitations of traditional manual design, which is characterized by high repetition, low efficiency, and reliance on experience. Compared with existing technologies, this method achieves automation, intelligence, and closed-loop design of end cap removal machines by integrating point cloud processing, artificial intelligence recognition, knowledge reasoning, and virtual simulation. The specific process is as follows: Figure 3 As shown, it includes the following steps:

[0033] Step S302: Perform point cloudification processing on the 3D product data of the product to be operated on.

[0034] The input 3D product data includes both CAD models and 3D scan data. When the input is a CAD model, an adaptive geometric sampling algorithm based on curvature (e.g., dynamic Poisson disk sampling) is used. The sampling density is adjusted according to the curvature gradient of the end cap and connecting parts (e.g., increasing the sampling rate of thread edges in high curvature areas), converting its surface mesh into point cloud data P={pi∣pi∈R3,i=1,...,N}, where P represents the point cloud geometry and N is the number of sampling points. This adaptive sampling can retain connection boundary information and avoid the density unevenness problem caused by traditional uniform sampling. When the input is 3D scan data, preprocessing is performed first. Statistical OR is used to remove outlier noise points, and topology-based voxel mesh downsampling is used (combined with the initial estimate of the connection region by GNN to dynamically adjust the voxel size). Principal component analysis (PCA) is then used to calculate the point cloud normal vectors, thus providing basic point cloud data for subsequent segmentation and analysis. Through the above methods, a point cloud with enhanced boundary confidence can be output, ensuring the integrity of the end cap connection topology.

[0035] Step S304: Perform component segmentation and assembly relationship reasoning on the point cloud data.

[0036] A graph convolutional neural network (GNN) segmentation algorithm based on topological constraints is employed to segment the overall point cloud into multiple independent point cloud components, i.e., the segmented point cloud components. This GNN model uses points / supervoxels as nodes and Euclidean distance plus curvature weights as edges. It utilizes mechanical connection priors (such as the helical curvature constraint of bolted connections) as a self-attention mechanism. During training, a topological consistency loss function (combined with L1 loss for assembly constraints) is used to achieve precise boundary separation between the end cap and connecting components. Subsequently, based on the geometric contact relationships, mating features (such as coaxiality and coplanarity), and normal vector consistency between point cloud components, assembly constraints (such as thread matching or weld contact) between each point cloud component are inferred. Based on these assembly constraints, a component assembly relationship graph is generated to characterize the structural hierarchy and assembly relationships of the product to be operated on.

[0037] In other embodiments, in addition to GNN, region growing algorithms or model fitting methods (such as RANSAC) based on traditional geometric features (such as curvature and normals) can also be used as auxiliary methods, but GNN is the main method to improve the segmentation accuracy of complex connections.

[0038] Step S306: Based on the component segmentation results, identify the end cap component and its adjacent components, and determine the end cap type.

[0039] Multimodal geometric features are extracted from the segmented point cloud components, including the shape of the point cloud outline, the distribution of circular holes, thickness, and symmetry, as well as 2D projected texture features (such as weld lines or snap-fit ​​groove edges). These multimodal features are then input into a deep learning classification model trained on historical case data (e.g., PointNet++ fused with CLIP-inspired cross-modal aligned GNN) to identify the end cap assembly and its adjacent components, and to determine the type of the end cap assembly. The types of end cap assemblies include bolt-fixed end caps, snap-fit ​​end caps, and press-fit end caps. This application employs physical prior aids (such as connection stress labels generated by finite element analysis) for zero-sample generalization, which improves adaptability to unlabeled industrial data.

[0040] In other embodiments, rule-based matching auxiliary strategies can also be used to identify end cap components. For example, if the algorithm detects evenly distributed annular holes around the component, it can directly determine it as a "bolt-fixed end cap" according to preset rules, serving as an effective supplement and alternative to pure AI classification.

[0041] Step S308: Extract key parameters for the end cap assembly and its adjacent components.

[0042] For the identified end cap assembly and its adjacent components, a multimodal parameter extraction model based on graph neural networks (GNN) is used to extract their key parameters. These key parameters include geometric parameters, assembly constraint parameters, and mechanical requirement parameters. Geometric parameters include end diameter, thickness, length, chamfer, position coordinates, and installation depth. Assembly constraint parameters include fit tolerances and connection relationships. Mechanical requirement parameters include disassembly force and clamping force, which are calculated based on material properties and the assembly relationships indicated by the component assembly relationship diagram. In other embodiments, the mechanical requirement parameters can also be adaptively calculated by fusing point cloud geometry, material properties, and assembly relationships using GNN. A prediction loss function (MSE + physical constraint loss) trained on historical disassembly datasets is used to achieve robust estimation of uncertainties (such as weld fatigue). In this embodiment, confidence intervals for the model output parameters are extracted to ensure downstream design accuracy.

[0043] Step S310: Generate the core component model of the end cap removal machine.

[0044] First, based on the product's reference surface and support points, a suitable fixture model (such as a bottom support, V-block, or locating pin) is intelligently matched or generated from the fixture knowledge base. Second, combining the end cap's force requirements and the product's overall weight, the required clamping force is calculated. Based on the clamping force, the specifications, quantity, and layout of the drive components are deduced, thereby generating a positioning and clamping mechanism. Drive components may include clamping cylinders, feed slides, and clamping positioning blocks. Finally, a cap removal actuator is generated based on the type and key parameters of the end cap assembly. This cap removal actuator includes a power and feed module, an execution end module, and auxiliary and support modules. The power and feed module typically uses a precision feed module driven by a servo motor to provide the high-precision, controllable linear motion required for the disassembly process. The execution end module is a special tool head (such as an electric wrench head or a snap-lock separator) that is intelligently matched or generated according to the end cap type, used to directly interact with the end cap and complete the disassembly operation. The auxiliary and support module includes a clamping mechanism (such as a three-jaw chuck) for stabilizing the gripping of the end cap, a guide mechanism (such as a slide rail) to ensure motion accuracy, and a support structure (such as a column) to provide overall rigidity.

[0045] In other embodiments, the AI ​​model can prioritize intelligent matching and selection from a standard parts library, and only initiate generative design to create new parts when no matching standard parts are available, thereby improving the standardization and economy of the solution.

[0046] Step S312: Automatically assemble the core component model and perform functional verification.

[0047] After generating the core components of the special-purpose machine, each component is automatically assembled with the equipment's geometric model. The assembly relationships are derived by a geometric constraint solver to ensure that the positioning fixture, clamping mechanism, and cover removal actuator are precisely matched with the equipment.

[0048] After assembly, a virtual simulation module is run to verify the functionality of the machine. This includes kinematic simulation-based verification of the cap removal path's rationality and motion interference detection, force analysis-based verification of clamping and disassembly forces, and stability analysis-based verification of operational stability. If verification fails, the reasons for the failure are analyzed (e.g., motion interference, insufficient clamping force), and these reasons are fed back to the AI ​​generation module as optimization suggestions. The AI ​​generation module uses reinforcement learning or parameter updates to achieve adaptive optimization and design iteration of the model until a fully verified cap removal machine solution is generated.

[0049] Step S314, output the result.

[0050] After the above process is completed, the system outputs a 3D assembly file (such as in STEP format) for the end cap removal machine, a bill of materials (BOM) for the machine (including positioning fixtures, clamping mechanisms, and end cap removal parts), and process parameter files (including clamping pressure, bolt torque, disassembly path, etc.). It can also output interface files for interfacing with the manufacturing execution system, enabling rapid conversion from design to manufacturing.

[0051] Through the above methods, this invention achieves end-to-end full-process automation, freeing manual labor from tedious and repetitive design work, enabling "one-click generation" and significantly improving design efficiency. By integrating historical expert experience and data patterns into decision-making through artificial intelligence, and combining virtual verification and closed-loop feedback mechanisms, human error is avoided, ensuring the feasibility and rationality of the output solution. At the same time, this method has self-learning and expansion capabilities, and can be continuously optimized to adapt to more types of products, showing broad application prospects.

[0052] This application also provides an intelligent generation method for a special end cap removal machine based on artificial intelligence and point cloud recognition. The main difference between this method and the previous embodiments lies in step S310, namely the generation process of the core component model of the end cap removal machine. Therefore, this embodiment will focus on describing the generation method of the core component model. This method relies on artificial intelligence and point cloud recognition technology. After completing the identification of the end cap assembly and its adjacent components and the extraction of key parameters, it gradually generates a positioning fixture, a positioning clamping mechanism, and an end cap removal execution mechanism according to a predetermined process, thereby forming a complete core component model of the end cap removal machine.

[0053] Specifically, such as Figure 4 As shown, the method includes the following steps:

[0054] S402, generate a positioning fixture.

[0055] A positioning fixture is generated based on the reference surface and support points of the product to be operated. The determination of the reference surface depends on the overall geometric features of the product. The principal component analysis method is used to automatically identify the geometric surface suitable as the positioning reference, such as the product's flat shell, end cap mounting base surface, or symmetrical plane. The support points are extracted from the protruding parts or symmetrical features of the product's bottom to ensure the stability of the positioning.

[0056] After determining the reference plane and support points, a jig knowledge base is invoked for matching. This knowledge base stores various standardized jig models, including bottom supports, V-blocks, and locating pins. Using a geometric feature alignment algorithm, the reference plane of the product to be operated is compared one by one with the jig templates in the knowledge base. When the matching error is within a preset threshold, the corresponding standard jig model is directly invoked for combination, thus achieving rapid configuration. If no suitable standard jig model is found, the process proceeds to the personalized jig generation stage.

[0057] The process of generating personalized jigs mainly relies on point cloud curvature fitting and parametric modeling. When the reference surface is a complex curved surface, the curvature radius distribution of the reference surface is first calculated using point cloud curvature analysis. Then, based on this curvature parameter, a jig structure with a corresponding contact surface is constructed to ensure a high degree of fit between the jig and the product surface. For example, for the housing of a new energy vehicle motor, its outer shell is often a complex curved surface. If a traditional planar jig is used, it is difficult to provide stable support. However, the automatic fitting method in this embodiment can generate a jig support surface that matches the curvature of the housing, effectively improving positioning accuracy and stability.

[0058] Furthermore, during the jig generation process, the stress distribution and displacement of the jig under preset clamping and supporting forces are simulated and calculated. Materials with sufficient strength and stiffness are selected from the material library to ensure that the generated jig will not deform or break under stress. This ensures that the jig not only matches the product geometrically but also fully verifies its mechanical properties.

[0059] This embodiment adopts a personalized fixture generation method based on point cloud curvature fitting, and verifies the mechanical properties through finite element analysis, ensuring the fixture's dual adaptability in geometry and mechanics.

[0060] Step S404: Generate the positioning and clamping mechanism.

[0061] The clamping force required for the product to be operated is calculated, and the specifications, quantity and layout of the drive components are deduced based on the clamping force, thereby generating a reasonable positioning and clamping mechanism.

[0062] Specifically, firstly, based on the geometric parameters and assembly constraint parameters of the end cap assembly and its adjacent components, a force model of the end cap during disassembly is established. The minimum disassembly force required for the end cap is calculated through finite element analysis, and the minimum clamping force is derived by considering the overall weight of the product, the distribution of support points, and the coefficient of friction.

[0063] The search is performed in a drive component library based on the minimum clamping force. This library contains various specifications of clamping cylinders, feed slides, and clamping positioning blocks. Existing methods often use a single mechanical constraint as the basis for component selection, which may lead to component redundancy. In this embodiment, mechanical parameters are processed based on clamping force to obtain candidate drive components that meet the rated output requirements; spatial layout processing is performed based on the product's three-dimensional geometric data to obtain a combination of components that can fit within a limited installation area; dynamic characteristic processing is performed based on the operation cycle of end cap disassembly to obtain a set of components that meet the response time requirements; energy efficiency processing is performed based on energy consumption parameters to obtain candidate components with the minimum power consumption while meeting functional requirements; and stability simulation is performed based on the structural stress model to obtain a component layout that will not resonate during operation. Through the above multi-stage processing, the specifications, quantity, and layout of drive components that meet the clamping requirements are finally generated.

[0064] After the specifications, quantity, and layout of the driving components are determined, a corresponding three-dimensional geometric model is generated based on the above. The assembly with the positioning fixture is automatically completed through the geometric constraint solver to ensure that the clamping positioning block can achieve uniform force and accurate positioning when it comes into contact with the product.

[0065] Step S406: Generate the cap-removing actuator.

[0066] The end cap removal mechanism is the core function of the entire end cap removal machine, and its role is to perform disassembly operations on different types of end cap assemblies. This embodiment takes the head end cap removal mechanism as an example.

[0067] Specifically, the power and feed modules are generated first. Based on the requirement for high-precision linear motion during disassembly, a precision feed module driven by a servo motor is selected. Displacement signal processing is performed based on encoder feedback to obtain high-resolution displacement and velocity information. Closed-loop control processing is then performed to obtain linear motion output that meets the high-precision feed requirements. Next, based on the servo controller parameter configuration, control signal processing is performed for the position loop, velocity loop, and force control loop to obtain an adjustable set of control parameters. Impedance calculation is then performed to obtain the target stiffness parameters for the servo output. During the end cap contact stage, real-time force feedback processing is performed based on contact sensor signals to obtain the contact force threshold. Low-stiffness output processing is then performed to obtain a compliant motion state, thereby reducing the force applied and avoiding damage to the end cap. During the disassembly stage, a mechanical model is performed based on the disassembly force requirements to obtain the target output force parameters. High-stiffness output processing is then performed to obtain a rigid motion state, ensuring that the output force is sufficient to complete the end cap disassembly operation. Finally, threshold identification is performed based on real-time sensor data to obtain the operating condition signal. Then, control parameters are switched based on this signal to achieve a smooth transition output from low to high stiffness. Through this phased processing, the resulting power and feed module ensures both safety during the contact phase and reliability and efficiency during the disassembly phase when the end cap is in contact or being disassembled.

[0068] Next, the execution end module is generated. Intelligent matching is performed based on the end cap type. When the end cap is bolt-fixed, an electric wrench head is generated, and its specifications are customized according to the bolt diameter and distribution pattern. When the end cap is snap-fit, a snap-fit ​​separator is generated, and the number and angle of the separator's claw arms are automatically adjusted to adapt to the snap-fit ​​distribution position. When the end cap is press-fit, a press-fit separation tool head is generated, and its contact area is optimized through mechanical simulation to prevent localized stress concentration from damaging the parts.

[0069] Then, auxiliary and support modules are generated. Appropriate clamping mechanisms, guiding structures, and support structures are automatically selected based on the end cap's geometry and weight. For example, for large-diameter end caps, a large-aperture three-jaw chuck is preferentially selected, and its installation position is determined through geometric constraint reasoning; simultaneously, a double-rail support structure is added to ensure that tipping does not occur during disassembly.

[0070] To improve the reliability of the actuator, in some other embodiments, dynamic path planning and force control simulation can be coupled together on the generated execution tool.

[0071] For example, after the execution tool is generated, spatial modeling is performed based on the geometric parameters of the end cap, the installation depth, and the assembly relationships of surrounding components to obtain a three-dimensional constraint environment for the product structure. Path search is then performed based on this constraint environment to obtain a preliminary disassembly path. In this embodiment, the preliminary disassembly path can be represented by multiple key motion nodes in a discretized manner, each containing position coordinates and attitude angle information.

[0072] After obtaining the initial disassembly path, simulation processing can be performed based on the force control model to obtain the force curve required for path execution. In this embodiment, the force control simulation combines the disassembly force requirements of the end cap, the clamping force constraints, and the dynamic characteristics of the feed module to calculate the force output parameters of each motion node. Then, based on the force output parameters of each motion node, an energy consumption model is used to perform energy consumption assessment processing to obtain the energy consumption distribution of path execution.

[0073] In the coupling process of path and force control simulation, interference identification is performed based on kinematic detection algorithm to obtain potential interference points; the node positions of the interference points are corrected based on constraint adjustment method to obtain a corrected interference-free path. For path segments with energy consumption distribution exceeding the threshold, parameter adjustment is performed based on velocity and acceleration curve optimization method to obtain an optimized path that minimizes energy consumption.

[0074] The disassembly path obtained using the above method can avoid interference from surrounding components and meet the force requirements of the disassembly process. The final generated path file can be directly used as motion command input for the actuator, thereby ensuring the stable and reliable operation of the end-effector module and auxiliary modules.

[0075] This application also provides an intelligent end cap removal and generation device based on AI and point cloud recognition, such as... Figure 5 As shown, it includes: a point cloud processing module 52, configured to acquire 3D product data of the product to be operated, and perform point cloud processing on the 3D product data to obtain point cloud data; a construction module 54, configured to segment the point cloud data to obtain multiple segmented point cloud components, and perform assembly relationship reasoning on the multiple segmented point cloud components to construct a component assembly relationship diagram; an extraction module 56, configured to identify the end cap component and its adjacent components from the multiple segmented point cloud components, and extract key parameters for the end cap component and its adjacent components; and a generation module 58, configured to generate a core component model of the end cap removal machine based on the key parameters, and assemble the core component model to generate the end cap removal machine.

[0076] It should be noted that the AI- and point cloud recognition-based intelligent end cap removal machine generation device provided in the above embodiments is only an example of the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the AI- and point cloud recognition-based intelligent end cap removal machine generation device and the AI- and point cloud recognition-based intelligent end cap removal machine generation method embodiment provided in the above embodiments belong to the same concept, and the specific implementation process is detailed in the method embodiment, which will not be repeated here.

[0077] Figure 6 A schematic diagram of a computer device suitable for implementing embodiments of the present disclosure is shown. It should be noted that... Figure 6 The computer device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments disclosed herein.

[0078] like Figure 6 As shown, the computer device includes a central processing unit (CPU) 1001, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage section 1008 into a random access memory (RAM) 1003. The RAM 1003 also stores various programs and data required for system operation. The CPU 1001, ROM 1002, and RAM 1003 are interconnected via a bus 1004. An input / output (I / O) interface 1005 is also connected to the bus 1004.

[0079] The following components are connected to I / O interface 1005: an input section 1006 including a keyboard, mouse, etc.; an output section 1007 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 1008 including a hard disk, etc.; and a communication section 1009 including a network interface card such as a LAN card, modem, etc. The communication section 1009 performs communication processing via a network such as the Internet. A drive 1010 is also connected to I / O interface 1005 as needed. A removable medium 1011, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on drive 1010 as needed so that computer programs read from it can be installed into storage section 1008 as needed.

[0080] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for intelligent generation of end caps using a dedicated machine based on AI and point cloud recognition, characterized in that, include: Acquire the 3D product data of the product to be operated, and perform point cloud processing on the 3D product data to obtain point cloud data; The point cloud data is segmented to obtain multiple segmented point cloud components, and the assembly relationship of the multiple segmented point cloud components is inferred to construct a component assembly relationship diagram. Based on the component assembly relationship diagram, the end cap component and its adjacent components are identified from the multiple segmented point cloud components, and key parameters are extracted for the end cap component and its adjacent components. Based on the aforementioned key parameters, a core component model of the end cap removal machine is generated, and the core component model is assembled to generate the end cap removal machine. The process involves identifying end cap components and their adjacent components from the multiple segmented point cloud components, and extracting key parameters for the end cap components and their adjacent components. This includes: extracting geometric features from the multiple segmented point cloud components based on the component assembly relationship diagram; the geometric features include at least one of the following: outer contour shape, circular hole distribution, thickness, and symmetry; identifying the end cap components and their adjacent components from the multiple segmented point cloud components using a deep learning classification model based on the geometric features, and determining the type of the end cap components, wherein the type of the end cap components includes bolt-fixed end caps, snap-fit ​​end caps, or press-fit end caps; and extracting key parameters for the end cap components and their adjacent components, wherein the key parameters include at least one of the following: geometric parameters, assembly constraint parameters, and mechanical requirement parameters; wherein the geometric parameters include at least one of the following: end diameter, thickness, length, chamfer, position coordinates, and installation depth; the assembly constraint parameters include fit tolerance connection relationships; and the mechanical requirement parameters include at least one of the following: disassembly force and clamping force. The core component model of the end cap removal machine is generated based on the key parameters, including at least one of the following: determining the reference surface and support point of the product to be operated based on the key parameters; generating a positioning fixture based on the reference surface and support point; calculating the clamping force based on the key parameters, combined with the force requirements of the end cap of the product to be operated and the overall weight of the product to be operated; inferring the specifications, quantity, and layout of the drive components required for the product to be operated based on the clamping force; and generating a positioning clamping mechanism based on the specifications, quantity, and layout of the drive components; and generating a cap removal execution mechanism assembly based on the type of the end cap assembly and the key parameters. The core component model includes at least one of the following: the positioning fixture, the positioning clamping mechanism, and the cap removal execution mechanism assembly.

2. The method according to claim 1, characterized in that, The three-dimensional product data is processed into point cloud data, including: When the three-dimensional product data is a CAD model, a geometric sampling algorithm is used to convert the surface mesh of the CAD model into the point cloud data; When the three-dimensional product data is three-dimensional scan data, statistical filtering is used to remove outlier noise points in the three-dimensional scan data. Voxel grid downsampling is used to reduce the data density of the three-dimensional scan data after removing outlier noise points. Principal component analysis is used to calculate the point cloud normal vector of the three-dimensional scan data after reducing the data density, so as to obtain the point cloud data.

3. The method according to claim 1, characterized in that, The point cloud data is segmented to obtain multiple segmented point cloud components, and assembly relationship reasoning is performed on the multiple segmented point cloud components to construct a component assembly relationship diagram, including: The point cloud data is divided into multiple independent point cloud components using a point cloud segmentation algorithm, which are the multiple segmented point cloud components. The point cloud segmentation algorithm includes: a region growing method based on geometric features, a random sampling consistency method, or a point cloud segmentation network method based on deep learning. Based on the geometric contact relationship, mating features, and normal vector consistency of the multiple segmented point cloud components, the assembly constraints between the multiple segmented point cloud components are inferred, and the component assembly relationship diagram is generated according to the assembly constraints to characterize the structural hierarchy and assembly relationship of the product to be operated.

4. The method according to claim 1, characterized in that, After generating the end cap removal machine, the method further includes at least one of the following: Kinematic simulation was used to check whether the cap removal path was reasonable and whether there was any motion interference. Based on force analysis, verify whether the clamping force and disassembly force meet the requirements; Stability analysis was used to test whether the stability of the end cap removal machine during operation met the requirements.

5. A smart end cap removal and production device based on AI and point cloud recognition, characterized in that, include: The point cloudification processing module is configured to acquire the three-dimensional product data of the product to be operated on, and to perform point cloudification processing on the three-dimensional product data to obtain point cloud data. The construction module is configured to segment the point cloud data to obtain multiple segmented point cloud components, and to perform assembly relationship reasoning on the multiple segmented point cloud components to construct a component assembly relationship diagram. The extraction module is configured to identify the end cap component and its adjacent components from the plurality of segmented point cloud components based on the component assembly relationship diagram, and extract key parameters for the end cap component and its adjacent components. The generation module is configured to generate core component models of the end cap removal machine and assemble the core component models to generate the end cap removal machine. The device is further configured to: extract geometric features from the plurality of segmented point cloud components based on the component assembly relationship diagram, wherein the geometric features include at least one of the following: outer contour shape, circular hole distribution, thickness, and symmetry; based on the geometric features, identify the end cap component and the adjacent components from the plurality of segmented point cloud components using a deep learning classification model, and determine the type of the end cap component, wherein the type of the end cap component includes a bolt-fixed end cap, a snap-fit ​​end cap, or a press-fit end cap; extract key parameters of the end cap component and the adjacent components, wherein the key parameters include at least one of the following: geometric parameters, assembly constraint parameters, and mechanical requirement parameters, wherein the geometric parameters include at least one of the following: end diameter, thickness, length, chamfer, position coordinates, and installation depth; the assembly constraint parameters include fit tolerance connection relationships; and the mechanical requirement parameters include at least one of the following: disassembly force and clamping force. The device is further configured to: determine the reference surface and support point of the product to be operated based on the key parameters; generate a positioning fixture based on the reference surface and the support point; calculate the clamping force based on the key parameters, combined with the force requirements of the end cap of the product to be operated and the overall weight of the product to be operated; deduce the specifications, quantity, and layout of the drive components required for the product to be operated based on the clamping force; and generate a positioning clamping mechanism based on the specifications, quantity, and layout of the drive components; and generate a cap removal execution mechanism assembly based on the type of the end cap assembly and the key parameters; wherein the core component model includes at least one of the following: the positioning fixture, the positioning clamping mechanism, and the cap removal execution mechanism assembly.

6. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored program, wherein, when the program is executed, it controls the device on which the computer-readable storage medium is located to perform the method according to any one of claims 1 to 4.

7. A computer device, characterized in that, include: Memory and processor The memory stores computer programs; The processor is configured to execute a computer program stored in the memory, wherein when the computer program is executed, the processor performs the method according to any one of claims 1 to 4.

8. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 4.

Images