Method and device for slice-integrated repair path planning based on three-dimensional assembly model

Abstract

The application relates to the technical field of high-end equipment maintenance, and provides a three-dimensional assembly body model slicing integrated maintenance path planning method and device to solve the problems of easy collision, poor accessibility, low efficiency and insufficient refinement in three-dimensional assembly body maintenance path planning, the method comprising the following steps: acquiring point cloud data of a three-dimensional assembly body model of equipment to be maintained; automatically identifying a maintenance area to be maintained by using a region expansion algorithm; generating a two-dimensional slice sequence containing the contour of the maintenance area to be maintained by using an adaptive layering algorithm to construct a maintenance accessibility score matrix; and realizing maintenance path planning by using a cross-layer ant colony optimization algorithm based on the accessibility score matrix as a path planning environment. The application can accurately identify the maintenance area, adaptively layer the maintenance constraints, and finally use the cross-layer ant colony optimization algorithm for global optimization, so that a three-dimensional maintenance path without collision, with high accessibility and smooth continuity, can be automatically generated in a refined and efficient manner.

Description

Technical Field

[0001] This invention belongs to the field of high-end equipment maintenance technology, specifically relating to a maintenance path planning method and device based on three-dimensional assembly model slice integration. Background Technology

[0002] In the field of intelligent manufacturing and equipment maintenance, maintenance path planning for complex assemblies (such as aero engines and industrial robots) has always been a key challenge.

[0003] First, traditional manual planning methods that rely on engineers' experience are difficult to guarantee accuracy and are inefficient. The space constraints of complex assemblies are prominent, affecting the accuracy and stability of refined path planning operations. For example, when the structure of equipment parts is complex and the assembly is dense, the movement path of maintenance tools is prone to interference with surrounding parts, making it difficult to achieve high-precision refined maintenance operations. For another example, in the repair of blades of equipment such as aero engines, the repair quality is prone to instability due to path overlap.

[0004] Secondly, existing intelligent maintenance path planning technologies have limitations. For example, Chinese invention application CN115374609A (publication date: November 22, 2022) proposes a maintenance path planning method for the complex internal environment of an aero-engine, mainly based on an improved... The algorithm introduces gravity constraints and hierarchical collision detection to improve the efficiency of path planning in confined spaces through dynamic sampling and collision detection. However, the complexity of directly searching in the three maintenance spaces using the improved algorithm is high, which makes it easy to get stuck in local dilemmas, makes it difficult to guarantee the global optimality of the path, and does not consider the accessibility of maintenance tools.

[0005] Therefore, there is an urgent need to propose a slice-integrated maintenance path planning method, which adaptively slices the three-dimensional assembly model into multiple two-dimensional spaces, and then integrates the maintenance path planning method through a global optimization algorithm to quickly generate collision-free and highly accessible maintenance paths, so as to realize refined and efficient operation of maintenance path design. Summary of the Invention

[0006] To effectively address the aforementioned problems in the existing technology, this invention provides a maintenance path planning method and apparatus based on a three-dimensional assembly model slice integration. The method adaptively slices the three-dimensional assembly model into multiple two-dimensional spaces, obtains the accessibility score matrix of each slice layer, and uses a cross-layer ant colony optimization algorithm for local path planning and global integration to quickly generate collision-free, highly accessible maintenance paths, facilitating the refinement and efficiency of maintenance path design.

[0007] This invention provides a maintenance path planning method based on slicing of a three-dimensional assembly model, comprising:

[0008] Step 110: Obtain the 3D assembly model of the equipment to be repaired and generate point cloud data;

[0009] Step 120: Using the 3D assembly model as input, the region expansion algorithm is used to automatically identify the area to be repaired and output the location of the area to be repaired.

[0010] Step 130: Adaptive layering algorithm is used to complete the adaptive layering and slicing process of the three-dimensional assembly model, generating multiple slice layers containing the outline of the area to be repaired in a two-dimensional slice sequence.

[0011] Step 140: Within each slice layer, the area to be repaired is discretized into regular grid cells. By calculating the comprehensive accessibility score of each grid cell, the accessibility score matrix of each slice layer is obtained.

[0012] Step 150: Using the accessibility score matrix as input, the cross-layer ant colony optimization algorithm is used to plan the maintenance path and generate the final maintenance path. The cross-layer ant colony optimization algorithm uses an ant colony algorithm based on the accessibility score matrix as the path planning environment and cross-layer connection point detection to find the optimal path on the slice layer and connect it with a smooth curve to form a continuous, collision-free three-dimensional maintenance path.

[0013] Furthermore, this invention provides a maintenance path planning device based on a three-dimensional assembly model slice integration, the device being used to implement the steps of the aforementioned method, the device comprising:

[0014] The first module is used to acquire the three-dimensional assembly model of the equipment to be repaired and generate point cloud data;

[0015] The second module is used to take the three-dimensional assembly model as input, use the region expansion algorithm to automatically identify the area to be repaired, and output the location of the area to be repaired.

[0016] The third module is used to perform adaptive layering and slicing processing on the three-dimensional assembly model using an adaptive layering algorithm, generating multiple slice layers containing a two-dimensional slice sequence of the area to be repaired.

[0017] The fourth module is used to discretize the area to be repaired into regular grid cells within each slice layer, and obtain the accessibility score matrix for each slice layer by calculating the comprehensive accessibility score of each grid cell.

[0018] The fifth module is used to take the accessibility score matrix as input and use the cross-layer ant colony optimization algorithm to realize maintenance path planning and generate the final maintenance path. The cross-layer ant colony optimization algorithm uses the ant colony algorithm based on the accessibility score matrix as the path planning environment and cross-layer connection point detection to find the optimal path on the slice layer and connect it with a smooth curve to form a continuous, collision-free three-dimensional maintenance path.

[0019] Compared with the prior art, the maintenance path planning method and device based on three-dimensional assembly model slice integration provided by the present invention has the following beneficial effects:

[0020] This invention, through adaptive slicing and incorporating tool attitude angles into the accessibility scoring, plans paths that guide maintenance tools to approach maintenance points at optimal angles, achieving precise operation. The accessibility scoring matrix pre-encodes collision risks and attitude constraints, upon which the cross-layer ant colony algorithm searches, fundamentally eliminating the generation of invalid and dangerous paths and ensuring high path safety and accessibility. The optimization objective directly includes path length and smoothness, ensuring that the generated path is not only the shortest or near-shortest but also exhibits smooth movement, reducing tool start-stop and jitter, thereby significantly shortening maintenance time, reducing tool wear, and achieving high efficiency in the maintenance process. The adaptive layering and intelligent optimization algorithms employed greatly reduce the search space and computation time, enabling rapid path planning for complex assemblies, meeting the requirements of modern industry for maintenance response speed, and demonstrating excellent application prospects. Attached Figure Description

[0021] 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 or examples of embodiments of the present invention.

[0022] Figure 1 This is a flowchart illustrating the steps of the integrated maintenance path planning method based on 3D assembly model slicing in the first embodiment of the present invention.

[0023] Figure 2 This is a schematic diagram of the three-dimensional assembly model of the six-cylinder diesel engine used in the experiment of this invention;

[0024] Figure 3 This is a schematic diagram of the point cloud data of the three-dimensional assembly model of the six-cylinder diesel engine in the experiment of this invention.

[0025] Figure 4 This is a schematic diagram of the triangular mesh data of the three-dimensional assembly model of the six-cylinder diesel engine used in the experiment of this invention. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0027] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0028] In the first embodiment, as Figure 1 As shown, this invention proposes a maintenance path planning method based on slicing and integrating three-dimensional assembly models, including:

[0029] Step 110: Obtain the 3D assembly model of the equipment to be repaired and generate point cloud data;

[0030] Step 120: Input the point cloud data of the 3D assembly model, use the region expansion algorithm to automatically identify the area to be repaired, and output the location of the area to be repaired;

[0031] Step 130: Adaptive layering algorithm is used to complete the adaptive layering and slicing process of the three-dimensional assembly model, generating multiple slice layers containing the outline of the area to be repaired in a two-dimensional slice sequence.

[0032] Step 140: Within each slice layer, the area to be repaired is discretized into regular grid cells. By calculating the comprehensive accessibility score of each grid cell, the accessibility score matrix of each slice layer is obtained.

[0033] Step 150: Using the accessibility score matrix as input, the cross-layer ant colony optimization algorithm is used to plan the maintenance path and generate the final maintenance path. The cross-layer ant colony optimization algorithm uses an ant colony algorithm based on the accessibility score matrix as the path planning environment and cross-layer connection point detection to find the optimal path on the slice layer and connect it with a smooth curve to form a continuous, collision-free three-dimensional maintenance path.

[0034] In one embodiment, step 110 includes:

[0035] Data is imported using 3D MAX software to obtain a three-dimensional assembly model of the equipment to be repaired. The three-dimensional assembly model has a spatial surface structure composed of segmented continuous facets. Since the data structure of the three-dimensional assembly model is imported through 3D MAX software, it is usually automatically converted into polygon-based mesh data (mesh model) to obtain a corresponding three-dimensional mesh model. The facets are generally polygonal, which are converted into triangles through the meshing function of 3D MAX software to form triangular facets.

[0036] The 3D MAX software samples the mesh surface of the 3D assembly model through its scripting functions or specific tools. The sampling process is based on a preset surface sampling algorithm, which directly generates a large number of copies of seed points at specified locations on the model surface (e.g., all vertex positions or random areas of the 3D mesh model), constructs a dense set of 3D coordinate points, and converts the segmented continuous surface of the model into discrete point cloud data.

[0037] In one embodiment, the specific tool is the scatter modifier that comes with the 3D MAX software.

[0038] The point cloud data is exported as a standard point cloud file format for subsequent data processing and analysis.

[0039] Further, in step 120, the point cloud data of the 3D assembly model is input, and the automatic identification of the area to be repaired using the region expansion algorithm and the output of the location of the area to be repaired include:

[0040] Step 121: Input the point cloud data of the 3D assembly model; perform preprocessing of the point cloud data, including:

[0041] Gaussian filtering or statistical filtering is used for noise reduction to remove isolated noise points and outliers.

[0042] Voxel mesh downsampling is used to downsample point cloud data, which reduces data density and improves computational efficiency while maintaining the overall shape of the model;

[0043] Perform topology repair on the grid data to ensure the correct connectivity of the facets.

[0044] Step 122: Using the defect point as the seed point, expand the region based on the curvature change and the geometric features of the neighborhood to accurately segment the maintenance boundary.

[0045] First, one or more defect points are identified on the 3D assembly model through manual marking or automatic detection algorithms, and these defect points are used as seed points for the initial expansion of the region.

[0046] Set curvature threshold Similarity threshold for geometric features ;

[0047] Starting from the seed point, iteratively check the neighboring points of the seed point and perform the following region expansion operation:

[0048] Calculate the Gaussian curvature and mean curvature of the neighboring points of the seed point;

[0049] If the curvature difference of a neighboring point is less than the curvature threshold and the geometric feature similarity meets the similarity condition, then the neighboring point is included in the extended region; the curvature difference refers to the absolute difference between the Gaussian curvature or the average curvature of the current seed point and the new neighboring point.

[0050] Repeat the above process until no new neighboring points can be included in the expanded area, until the area expansion stops and the complete area to be repaired is obtained.

[0051] Within the area to be repaired, a boundary extraction algorithm is used to generate the repair boundary: the repair boundary is segmented by a boundary point determination formula; the boundary point determination formula means that if a point belongs to the area to be repaired, but there are other points in its neighborhood that do not belong to the area to be repaired, then this point is determined to be a boundary point; the set of all boundary points constitutes the final repair boundary.

[0052] Geometric feature similarity is an indicator that measures how similar two geometric features are. Geometric feature similarity includes: normal vector similarity and / or color similarity. Normal vector similarity is characterized by the dot product between two normal vectors. The higher the normal vector similarity, the closer the dot product is to 1. Color similarity is measured by calculating the Euclidean distance or absolute difference of the RGB values ​​of two points. The smaller the distance, the closer the colors are, and the higher the similarity.

[0053] The geometric feature similarity threshold is set based on the normal vector direction and / or color information; for the normal vector direction, a lower limit threshold for normal vector similarity is set; for color information, an upper limit threshold for color similarity is set.

[0054] The similarity conditions refer to: normal vector similarity being greater than the lower threshold of normal vector similarity, and / or color similarity being less than the upper threshold of color similarity.

[0055] Step 123: Mark and output the geometric coordinates and topological relationships of the area to be repaired.

[0056] Function extraction using 3D coordinates Obtain maintenance boundary vertices The coordinates of the boundary vertices are recorded. ;

[0057] Construct a topological connection diagram (such as an adjacency matrix) between the area to be repaired and the components of the assembly.

[0058] The intersection area between the region and its adjacent components is calculated using a topology analysis algorithm. Define topological relationship weights ( (total area), outputting a JSON or XML file containing the coordinate set and topological relationships.

[0059] The above steps employ a region expansion algorithm to automatically identify areas requiring repair, achieving automated and quantitative identification of these areas. This avoids subjective errors associated with manual labeling and provides a reliable input basis for subsequent planning. The region expansion algorithm is computationally efficient, quickly separating critical repair areas from complex assemblies and significantly reducing pre-processing time.

[0060] In step 130, an adaptive layering algorithm is used to perform adaptive layering and slicing processing on the 3D assembly model, generating multiple slice layers containing the outline of the area to be repaired in a 2D slice sequence, including:

[0061] Step 131: An adaptive layering algorithm is used to dynamically adjust the slice thickness based on maintenance process requirements and model geometric features; the maintenance process requirements include at least the thickness of each component of the equipment to be maintained. Specifically, this includes:

[0062] (1) The input parameters for initializing the algorithm shall include at least:

[0063] Repair process requirements: a set of threshold thicknesses for each component. ( For the first (Maximum permissible layer thickness threshold for each component);

[0064] Model geometric features:

[0065] Mesh data of a 3D assembly model, including vertex sets. and triangular facet set ;

[0066] Mesh data of a 3D assembly model: vertex set ;

[0067] Triangular facet set ;

[0068] Each vertex Included coordinate information , ;

[0069] Global layer thickness constraint: minimum layer thickness Maximum layer thickness .

[0070] (2) Determine the slicing direction and model height range, including:

[0071] Select the 3D assembly model globally The positive axis direction is the slicing direction, which is also the repair direction of the maintenance.

[0072] Define the model height range along the slicing direction as follows: , This represents the minimum slice height. This represents the maximum slice height; the initial slice height is... .

[0073] (3) Dynamic calculation of layer thickness, including:

[0074] Set constraints on the thickness of each component: for the current height Determine the component to be represented by the cross-section of the model at that location. Obtain the corresponding maximum layer thickness threshold. ;

[0075] Extract model geometric features: Calculate current height The normal vector of all triangular facets By using the normal vector and the slice direction ( (axis) included angle Characterizing surface tilt:

[0076] ;

[0077] Dynamic adjustment of layer thickness is performed using the following adjustment formula:

[0078] ;

[0079] in, It is a component Thickness threshold, The thickness decreases as the inclination increases, thus reducing the layer thickness in the inclined region.

[0080] (4) Update the slice height using the following update formula:

[0081] .

[0082] Repeat step (3) until the result in (4) is obtained. satisfy .

[0083] Step 132, reducing the layer thickness in areas with large curvature changes to improve repair accuracy, including:

[0084] (1) Perform curvature calculation, including:

[0085] For each vertex of the triangular facets on the model surface, calculate... Gaussian curvature and mean curvature Through principal curvature Characterizing the magnitude of curvature:

[0086] .

[0087] Define curvature threshold (e.g., preset values ​​for the curvature region at the blade edge), if , marked as high curvature region.

[0088] (2) Perform layer thickness correction in high curvature regions, including:

[0089] For regions with high curvature, a curvature correction coefficient is introduced. The layer thickness formula is adjusted to:

[0090] ;

[0091] in, The layer thickness is for high curvature regions and decreases as curvature increases. This is the curvature sensitivity parameter.

[0092] In one embodiment, take .

[0093] (3) Transition processing of maintenance boundaries, including: using linear interpolation to smooth the layer thickness change in the transition zone between high curvature areas and flat areas to avoid abrupt changes in layer thickness:

[0094] ;

[0095] in, For the transition curvature region layer thickness, For regions with flat curvature, the layer thickness is [not specified]. For transition distance weights, .

[0096] Step 133, increasing the layer thickness in flat areas to improve planning efficiency, includes:

[0097] (1) Flat region determination: by calculating the variance of the normal vector of the current region. ,like , It is determined to be a flat region by setting a flatness threshold; or by using the average curvature. Directly determine, It is a preset flat curvature threshold.

[0098] (2) Optimize the layer thickness in flat areas: while meeting process accuracy requirements (such as surface roughness) Under the premise of ), maximize the layer thickness in the flat curvature region:

[0099] ;

[0100] in, For the preset surface roughness, These are constants related to materials and processes, such as those for metals. .

[0101] (3) Verify the efficiency improvement

[0102] The reduction in the total number of slices is calculated based on the proportion of flat areas:

[0103] .

[0104] Step 134: Generate a set of two-dimensional slices containing the outline of the repair area, including:

[0105] (1) Define the maintenance area mask

[0106] Input the 3D mask of the maintenance area (Obtained through user annotation or defect detection algorithms), slicing is performed only on the area within the mask.

[0107] (2) Extract the cross-sectional profile

[0108] Find the intersection line between the triangular facet and the slice plane at a specified current height. With coordinates The points on the map are used to traverse the 3D mask of the maintenance area. For all triangular faces, determine if they are aligned with the slice plane set at the current height. intersect:

[0109] If the vertex of the face Coordinates are all greater than / less than No intersections;

[0110] Otherwise, calculate the intersection point using the line segment-plane intersection formula. and This yields a discrete set of intersection points.

[0111] (3) Construct the outline polygon

[0112] The set of all intersection points is connected into a closed contour polygon using a spatial sorting algorithm (such as polar coordinate sorting).

[0113] Calculate the centroid of each intersection point in the intersection set. For each intersection point Calculate polar angle ;

[0114] according to Sort the intersection points from smallest to largest, and connect the endpoints sequentially to form a closed contour polygon. .

[0115] (4) Perform contour optimization and output

[0116] Redundant point removal: Delete collinear points using the cross product of vectors. Determine the collinear points;

[0117] Output format: Save the outline polygon as a 2D coordinate sequence: , as input for maintenance path planning.

[0118] The above steps utilize an adaptive layering algorithm to dynamically combine maintenance process requirements with model geometric features. Layer thickness is reduced in areas of high curvature to improve accuracy, while layer thickness is increased in flat areas to enhance efficiency. Redundant computations are reduced through contour optimization and mask focusing. Specifically, dynamic layer thickness adjustment and curvature adaptation improve the accuracy and safety of path planning, helping to address the issue of collisions in maintenance paths. Slice direction optimization and contour continuity constraints based on transition processing enhance the accessibility of tool paths. Layer thickness optimization in flat areas and slice count control accelerate the path planning process. Integrating curvature sensitivity and process parameters improves the refinement of high-precision maintenance path planning.

[0119] Further, step 140 includes:

[0120] Step 141: Within each slice layer, the maintenance space is discretized into grid cells, and obstacle grids are marked.

[0121] (1) Determine the maintenance boundary of the space

[0122] Input the 2D projection area of ​​the 3D repair space onto the current slice layer, and define the repair boundary coordinate range as follows. and ,in, The two-dimensional coordinate axes of the slicing plane, yes Minimum value in the axial direction yes Maximum value in the axial direction yes Minimum value in the axial direction yes Maximum value in the axial direction. Set the grid resolution according to the accuracy requirements of the maintenance task. (Unit: mm), Calculate the number of grid rows and the number of grid columns .

[0123] (2) Perform grid cell encoding and coordinate mapping

[0124] Two-dimensional index encoding is performed on the grid cells, the first... Line 1 The center coordinates of the column cell are:

[0125] ;

[0126] in, ;

[0127] Create an index With physical coordinates A mapping table stores the maintenance boundary coordinates of each unit. .

[0128] (3) Mark the obstacle grid

[0129] Input the 2D contours of obstacles within the slice layer (such as the projected polygons of equipment casings or fixed structures), and use the ray casting method to determine the center of each mesh cell. Is it located inside an obstacle? If it is determined to be an obstacle, then mark the unit as unreachable (accessibility score). Otherwise, mark it as a unit to be evaluated.

[0130] Step 142: Calculate the comprehensive accessibility score for each mesh cell, including physical constraints based on the maintenance tool (e.g., robotic arm) (e.g., tool size, joint range of motion), and calculate the accessibility score for each mesh cell. :

[0131] ;

[0132] in:

[0133] : Distance from the center point of the tool to the repair point (normalized value);

[0134] : Distance from the tool to the nearest obstacle (normalized value);

[0135] : The angle between the tool axis and the normal of the maintenance point (measures the accessibility of the posture);

[0136] , , These are the weighting coefficients, and Adjustments will be made based on the actual maintenance process.

[0137] Step 143: Obtain the accessibility score matrix of each slice layer as input for subsequent maintenance path planning.

[0138] Step 140 above structures the problem through spatial discretization, quantifies the physical constraints of the maintenance process (tool size, attitude, obstacle avoidance) into computable data through multi-factor comprehensive accessibility scoring, and finally outputs a guiding accessibility scoring matrix, which lays a solid foundation for subsequent path planning algorithms, enabling them to efficiently generate refined maintenance paths with high accessibility, low collision risk and in line with process requirements.

[0139] Specifically, in step 150, the cross-layer ant colony optimization algorithm includes:

[0140] Step 151, perform local path generation:

[0141] (1) Initialize parameters

[0142] Set basic parameters such as ant colony size, pheromone evaporation factor, and heuristic function weight;

[0143] Input the reachability score matrix of the current slice layer as the path planning environment.

[0144] (2) Path construction

[0145] Each ant starts from the beginning of the maintenance area and selects the next grid cell based on pheromone concentration and accessibility score:

[0146] Transition probability formula:

[0147] ;

[0148] in, For path pheromone concentration, ( For the unit in the accessibility rating matrix (rating score) , These are the weighting coefficients for pheromones and heuristic functions, respectively.

[0149] Ants record path nodes as they move, avoiding unreachable areas (grid cells with scores below a threshold).

[0150] (3) Calculate the fitness function

[0151] Calculate the fitness value for the path generated by each ant:

[0152] ;

[0153] in:

[0154] The path length is the sum of the Euclidean distances between grid cells.

[0155] The average reachability score for the path (the mean of the scores of the path coverage grid cells);

[0156] Path smoothness (the rate of change of the angle between the lines connecting adjacent nodes; the smaller the value, the smoother the path).

[0157] , , This is a weighting coefficient, which is adjusted according to the maintenance process requirements.

[0158] (4) Update pheromones

[0159] Local update: After the ant completes its path, it updates the grid cells along the path according to the formula. Volatile pheromones ( As a volatile factor, (for initial pheromones).

[0160] Global Update: The optimal fitness path gains additional pheromones. ( For pheromone constants, (This represents the fitness value of the optimal path).

[0161] (5) Terminate iteration

[0162] When the maximum number of iterations or the path fitness value converges, the optimal local path of the current slice layer is output.

[0163] Step 152, perform global path integration, including:

[0164] (1) Cross-layer connection point detection

[0165] For adjacent slice layers (such as the first slice layer) Layer and First The optimal path of the layer is obtained, and the endpoints and intermediate feature points (such as curvature extrema) of the path are extracted.

[0166] Calculate the Euclidean distance between the two path points in three-dimensional space, and filter point pairs with a distance less than a threshold as candidate connection points.

[0167] (2) Smooth curve connection

[0168] The candidate connection points are fitted using B-spline curves or Bézier curves to ensure that the curves satisfy the following conditions:

[0169] Continuity: The position of the curve at the connection point and the first derivative (tangent direction) are continuous;

[0170] Collision-free: Discretely sample the curve and check whether the sampled points interfere with the assembly model (verified by a 3D model collision detection algorithm).

[0171] (3) Path optimization

[0172] If there is interference in the connecting curves, adjust the position of the connecting points or the curve control points, and refit until the collision-free condition is met; calculate the total length and smoothness of the integrated 3D path, and retain the optimal connecting scheme.

[0173] Step 153: Perform final optimization and output to generate a three-dimensional maintenance path.

[0174] (1) Path splicing

[0175] Connect the local paths of all slice layers sequentially using cross-layer smooth curves to form a preliminary three-dimensional path.

[0176] (2) Global collision-free verification

[0177] Perform global interference detection on the 3D path: Discretize the path into dense sampling points (e.g., one point every 0.1 mm); for each sampling point, calculate the minimum distance between it and obstacles in the assembly model. If the distance is less than the tool radius, it is determined to be a collision point.

[0178] (3) Path correction

[0179] Local path replanning is adopted for the collision point area: the ant colony algorithm is called again to generate obstacle bypass sub-paths with 5 nodes before and after the collision point as the starting point and the ending point; the sub-paths are connected to the original path with smooth curves to ensure overall continuity.

[0180] (4) Output results

[0181] Generate a continuous 3D path from the start point to the end point of the maintenance, and output the 3D coordinate sequence of the path points: , , ..., This serves as the command for generating maintenance path connections.

[0182] The specific process of using the cross-layer ant colony optimization algorithm to plan maintenance paths and generate the final maintenance path is shown in Table 1 as the pseudocode of the cross-layer ant colony algorithm for optimizing maintenance paths.

[0183] Table 1. Pseudocode of the cross-layer ant colony algorithm for optimizing maintenance path planning.

[0184]

[0185] The aforementioned cross-layer ant colony optimization algorithm combines local ant colony optimization with cross-layer connection point detection. It fully utilizes the local optimization capability of ant colony optimization and combines it with cross-layer detection to achieve global optimization of continuous cross-layer paths. The fitness function used comprehensively optimizes path length, safety and accessibility, and operational smoothness. The generated path is not only safe but also efficient and stable. The entire planning process requires no manual intervention, has a high degree of automation, and can quickly respond to different maintenance scenarios and models.

[0186] The effectiveness of the method was also verified experimentally. For example... Figure 2As shown, the experiment used the third piston ring disassembly path planning of a three-dimensional assembly model of a six-cylinder diesel engine. Details are as follows:

[0187] (i) Acquisition of 3D assembly data and identification of areas to be repaired

[0188] (1) Acquisition of 3D assembly model and recognition of region expansion

[0189] Obtain the 3D assembly model of the six-cylinder diesel engine (in STL or PLY format) into 3D MAX software and construct a 3D solid model that the software can recognize. For the third piston ring (located in the second ring groove at the top of the piston of the third cylinder on the left side of the engine), manually mark the wear defect points on the piston ring surface as seed points and start the region expansion algorithm.

[0190] (2) Point cloud / mesh data input and preprocessing

[0191] How to obtain point cloud data of a 3D assembly model in 3ds Max software? Figure 3 As shown, the triangular mesh data is as follows Figure 4 As shown, Figure 3 and Figure 4 The data includes point cloud data and mesh data for components such as pistons, piston rings, connecting rods, and cylinder blocks, as well as vertex sets (3D coordinates). The point cloud data / mesh data is denoised using a set of triangular facets (each facet contains 3 vertex indices). The denoising process is then applied to the point cloud / mesh data (removing points with a distance of 3 from the mean). (Except for abnormal points outside the defect), ensure that the coordinate accuracy error of the defect point is ≤0.02mm.

[0192] (3) Area expansion and maintenance boundary division

[0193] Centered on the seed point, calculate the curvature change of neighboring vertices (Gaussian curvature threshold set to 0.05mm). - ¹) Based on the normal vector feature (the angle between the normal vectors of neighboring points is ≤15°), the region is expanded through an eight-neighbor search. The expansion stops when the expanded region includes the complete circumferential contour of the third piston ring (diameter 100mm, width 3mm) and a 5mm range above and below the ring groove, generating a 3D bounding box for the maintenance boundary (coordinate range: x=450-550mm, y=320-330mm, z=180-190mm).

[0194] (4) Geometric coordinates and topological relation marking

[0195] Mark the key geometric coordinates of the maintenance area: piston ring center (500, 325, 185) mm, upper edge of ring groove z=187 mm, lower edge z=184 mm; the topological relationship includes the nesting relationship between piston ring and piston (clearance fit, clearance 0.05 mm), and the radial clearance between piston ring and cylinder liner (0.1 mm), stored as an XML format topology tree structure.

[0196] (ii) Adaptive layered slicing processing

[0197] (1) Initialization of parameters and dynamic layer thickness calculation of adaptive layering algorithm

[0198] Input parameters: Repair process requirements (piston ring thickness 3mm, layer thickness threshold range 0.2-1mm, ring groove accuracy requirement ±0.03mm); Model geometric features (12540 vertices of triangular mesh, 25080 faces); Global layer thickness constraints (minimum 0.2mm, maximum 1mm).

[0199] Slicing direction and height range: The maintenance area height range is defined as 180-190mm along the piston axis (Z-axis), and the initial slice height is z=180mm.

[0200] Dynamic layer thickness calculation:

[0201] Component thickness constraints: The current height z=184-187mm is the piston ring body, with a corresponding thickness threshold of 0.3mm; z=180-184mm and 187-190mm are the ring groove transition areas, with a threshold of 0.5mm.

[0202] Geometric feature extraction: Calculate the angle between the normal vectors of the slicing plane and the triangular facet. , A slope of ≤30° is considered a steep area. >60° is considered a flat area.

[0203] Dynamic adjustment of layer thickness: Layer thickness in steep areas = 0.3mm + 0.2mm × The thickness of the flat area is min(1mm, 0.5mm × 100mm). ).

[0204] (2) Thickness correction in high curvature region (piston ring edge)

[0205] Curvature calculation: Calculate the Gaussian curvature (0.12) at the apex of the piston ring edge. ) and mean curvature (0.08) Principal curvature > 0.1 Set the curvature threshold to 0.08. .

[0206] Layer thickness correction: A curvature correction coefficient is introduced for the high curvature area at the edge (circumferential width 5mm). =1 / (1+0.5×principal curvature), corrected layer thickness =0.3mm× ≈0.2mm.

[0207] Boundary transition processing: The transition zone (3mm wide) uses linear interpolation, and the layer thickness smoothly transitions from 0.2mm to 0.3mm to avoid path jitter caused by abrupt changes in layer thickness.

[0208] (3) Optimization of layer thickness in flat areas (piston top surface)

[0209] Flat region determination: Average curvature of piston top surface (z=190mm) < 0.01 If the variance of the normal vector is less than 0.05, it is determined to be a flat region.

[0210] Layer thickness optimization: while meeting surface roughness requirements Under the requirement, the layer thickness is set to 1mm (maximum allowable value), which improves efficiency by 300% in steeper areas.

[0211] Efficiency verification: The number of slices in the flat area was reduced from 30 layers with a thickness of 0.3mm to 10 layers, and the slicing time was shortened from 2.5 minutes to 0.8 minutes.

[0212] (4) Generation of two-dimensional slices of the maintenance area

[0213] Maintenance area mask definition: Input the 3D mask of the third piston ring (coordinate range) , , Only slices the area within the mask.

[0214] Section contour extraction: Traverse the height of each slice (e.g.) For a triangular facet, the endpoints of the intersection line are calculated using the line segment-plane intersection formula. For example, the vertex coordinates of a certain triangular facet are (499,325,184.8), (501,325,184.8), and (500,326,185.2). The planes intersect, and the endpoints of the intersection line are calculated to be (499.5, 325, 185) and (500.5, 325, 185).

[0215] Contour polygon construction: Calculate the centroid of the intersection endpoints (500, 325, 185), sort the endpoints by polar angle, and connect them to form a closed contour polygon (piston ring cross-section contour, diameter 100mm).

[0216] Contour optimization and output: Eliminating collinear points (vector cross product < 0.01) (points), output two-dimensional coordinate sequence As input for path planning.

[0217] (III) Construction of Maintenance Accessibility Scoring Matrix (Accessibility Map)

[0218] (1) Grid cell discretization

[0219] Spatial boundary determination: current slice layer (e.g.) The two-dimensional projection area is , The grid resolution is set to 0.5mm, the number of grid rows = (550-450) / 0.5 = 200, the number of columns = (330-320) / 0.5 = 20, and the total number of cells is 4000.

[0220] Mesh encoding and coordinate mapping: Part 1 Line 1 The coordinates of the column cell center are (450 + 0.5). 320+0.5 (185) mm, create an index ( , A mapping table between physical coordinates and coordinates.

[0221] Obstacle mesh marking: Input the inner wall of the annular groove ( , The two-dimensional contour of a grid cell is used to determine whether the center of each grid cell is within an obstacle. For example, a cell ( =100, =10) The center (500, 325, 185) is located inside the annular groove and is marked as unreachable. );unit( =80, =10) The center (490,325,185) is free space and is marked as to be evaluated.

[0222] (2) Accessibility score calculation

[0223] Based on the physical constraints of the disassembly tool (an 8mm diameter elastic expansion clamp), an accessibility score is calculated for each mesh cell. :

[0224] Distance between tool and obstacle (Tool radius) (Fully accessible);

[0225] (Partially accessible);

[0226] (Unreachable).

[0227] For example, the outer grid unit of the piston ring ( =70, =10) The distance between the center (485,325,185) and the inner wall of the cylinder liner is 5mm. .

[0228] (3) Accessibility rating matrix output

[0229] Generate a 200×20 two-dimensional rating matrix, where the matrix element values ​​are the values ​​of the corresponding grid cells. The value serves as the environmental input for the ant colony algorithm's path planning.

[0230] (iv) Cross-layer ant colony optimization path planning

[0231] (1) Local path generation

[0232] Parameter initialization: ant colony size 50, pheromone evaporation factor 0.1, heuristic function weights (reachability score weight 0.7, distance weight 0.3).

[0233] Path construction: Ants start from the repair starting point ( , , Starting from the initial position of the tool, in the current slice layer (e.g., ... Select the next grid cell: Preferred selection Furthermore, avoid units with high pheromone concentrations. The annular groove region is used to generate an annular disassembly path profile.

[0234] Fitness function: Path fitness value = (total path length × 0.3) + (number of unreachable units × 10), with the optimal path having the smallest fitness value.

[0235] Pheromones are updated in two ways: Local update (the pheromone content of each unit on the path is reduced by 10%) and Global update (the pheromone content of the optimal path is increased by 20%).

[0236] Iteration Termination: After 50 iterations, the path converges, and the optimal local path of the current layer is output (e.g., ...). Layer path: from (550, 325) clockwise around the piston ring to (548, 325), length 314mm.

[0237] (2) Global path integration (cross-layer connections and smoothing)

[0238] Cross-layer connection point detection: Extracting adjacent layers ( and For each path endpoint, calculate the three-dimensional Euclidean distance and filter point pairs with a distance <1mm (such as (548,325,187) and (547.5,325,186)) as candidate connection points.

[0239] Smooth curve connection: A third-order Bézier curve is used to fit the connection points, with control points set at (548,325,187), (547.8,325,186.7), (547.6,325,186.3), and (547.5,325,186) to ensure positional continuity. , ), Tangent direction is continuous (first derivative angle < 5°).

[0240] Path optimization: Sampling and detection of the connection curve revealed that (547.7,325,186.5) is 0.8mm away from the top of the piston (tool radius 4mm), with no collision, so this connection scheme is retained.

[0241] (3) 3D path generation (final optimization and output)

[0242] Path splicing: Connecting the local paths of each slice layer using Bézier curves to form a path from... arrive The spiral descent three-dimensional path.

[0243] Global collision-free verification: A path point is sampled every 0.1mm, and the minimum distance to the obstacle is calculated. For example, if the path point (500,325,185.5) is 5mm away from the upper edge of the groove (4mm, tool radius), there is no collision; if the point (499,325,184.2) is 0.3mm away from the lower edge of the groove (4mm, collision point), it is determined to be a collision point.

[0244] Path correction: Using the 5 points before the collision point (499.5, 325, 184.7) and the 5 points after the collision point (498.5, 325, 183.7) as the starting and ending points, replan the path around the barrier (offset outward by 0.5mm), and connect them with a smooth curve to ensure that the distance between the corrected path and the ring groove is ≥4mm.

[0245] Output: A continuous 3D path coordinate sequence (3142 points in total) is generated. A partial list of points from the continuous 3D path coordinate sequence is shown below:

[0246] (550.0, 325.0, 190.0) (549.8,325.1,189.9) (547.5, 325.0, 184.0)

[0247] A maintenance path is generated by connecting all consecutive point sequences in sequence.

[0248] Analysis of the experimental results further illustrates the maintenance path planning method based on 3D assembly model slicing integration provided by this invention. By designing an adaptive layered slicing algorithm, the layer thickness is dynamically adjusted according to the model's geometric features (curvature, tilt) and maintenance process requirements (component thickness, surface roughness). This ensures that in areas with high curvature such as blade edges, precise contour details are captured by reducing the layer thickness; while in flat areas, the layer thickness is increased to improve efficiency, generating a set of 2D slice sequences, providing a geometric basis that is both faithful and efficient for path planning. Furthermore, by innovatively introducing the angle between the tool axis and the maintenance point normal in the accessibility scoring, collision risks and attitude constraints are pre-encoded into each grid cell. This ensures that the planned path not only reaches the target point but also guides the maintenance tool to approach the maintenance point at the optimal operating angle, achieving precise operation for processes such as welding and spraying that require specific attitudes.

[0249] Furthermore, by discretizing the maintenance space and calculating the comprehensive accessibility score matrix as the search basis for the cross-layer ant colony algorithm, the algorithm can naturally avoid obstacles and prioritize areas with high tool orientation accessibility during optimization. This mechanism greatly reduces the computational complexity and search time of the algorithm, and fundamentally eliminates the generation of invalid, unreachable, and dangerous paths that interfere with obstacles, ensuring the absolute safety and reliable accessibility of the path from the source. When using the cross-layer ant colony algorithm for maintenance path search, the optimization objective function directly integrates path length and smoothness, making the generated path not only the shortest or near shortest in space, but also possessing excellent kinematic characteristics, with smooth and continuous motion. This significantly reduces the start-stop, jitter, and sharp turns of maintenance tools during operation, thereby shortening the total time of a single maintenance task.

[0250] Furthermore, in one embodiment, the present invention provides a maintenance path planning device based on a three-dimensional assembly model slice integration, the device being used to implement the steps of the method described in the first embodiment, the device comprising:

[0251] The first module is used in step 110 to obtain the three-dimensional assembly model of the equipment to be repaired and generate point cloud data.

[0252] The second module is used to input the point cloud data of the 3D assembly model, automatically identify the area to be repaired using the region expansion algorithm, and output the location of the area to be repaired.

[0253] The third module is used to perform adaptive layering and slicing processing on the three-dimensional assembly model using an adaptive layering algorithm, generating multiple slice layers containing a two-dimensional slice sequence of the area to be repaired.

[0254] The fourth module is used to discretize the area to be repaired into regular grid cells within each slice layer, and obtain the accessibility score matrix for each slice layer by calculating the comprehensive accessibility score of each grid cell.

[0255] The fifth module is used to take the accessibility score matrix as input and use the cross-layer ant colony optimization algorithm to realize maintenance path planning and generate the final maintenance path. The cross-layer ant colony optimization algorithm uses the ant colony algorithm based on the accessibility score matrix as the path planning environment and cross-layer connection point detection to find the optimal path on the slice layer and connect it with a smooth curve to form a continuous, collision-free three-dimensional maintenance path.

[0256] In one embodiment, the present invention provides a computer device, which may be a server, comprising a processor, a memory, a network interface, and a database connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores maintenance path planning data based on a 3D assembly model slice integrated approach. The network interface communicates with external terminals via a network connection. When the computer program is executed by the processor, it implements the 3D assembly model slice integrated maintenance path planning method.

[0257] Those skilled in the art will understand that the description of the device technical features in the above embodiments does not constitute a limitation on all devices to which the present invention is applied. Specific devices may include more or fewer components, or combinations of certain components, or different component arrangements.

[0258] In another embodiment, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the integrated maintenance path planning method based on three-dimensional assembly model slices provided in any of the above embodiments.

[0259] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0260] Matters not covered in this invention are common knowledge.

[0261] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0262] The embodiments described above are merely examples of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention.

Claims

1. A method for slice-integrated repair path planning based on a three-dimensional assembly model, characterized in that, The method comprises the following steps: Step 110, obtaining a three-dimensional assembly model of the equipment to be repaired and generating point cloud data; Step 120, using a region growing algorithm to automatically identify the repair area by taking the three-dimensional assembly model as input, and outputting the position of the repair area; Step 130, using an adaptive layering algorithm to complete adaptive layering and slicing of the three-dimensional assembly model, and generating a plurality of slice layers of a two-dimensional slice sequence containing the contour of the repair area; Step 140, in each slice layer, the repair area is discretized into regular grid cells, and the accessibility score matrix of each slice layer is obtained by calculating the comprehensive accessibility score of each grid cell; comprising: Step 141, in each slice layer, the repair space is discretized into grid cells, and the obstacle grid is marked, comprising: inputting the two-dimensional projection area of the three-dimensional repair space in the current slice layer to define the repair boundary coordinate range; setting the grid resolution according to the repair task accuracy requirement, calculating the number of grid rows and the number of grid columns; two-dimensional index coding is performed on the grid cells to obtain the center coordinates of the grid cells; a mapping relationship table of grid index and physical coordinates is established to store the repair boundary coordinates of each cell; inputting the two-dimensional contour of the obstacle in the slice layer, and using the ray method to judge whether the center coordinates of each grid cell are located inside the obstacle; if it is judged that it is located inside the obstacle, the grid cell is marked as unreachable, and the accessibility score is zero, otherwise, the grid cell is marked as an evaluation unit; Step 142, calculating the comprehensive accessibility score of each grid cell, comprising calculating the accessibility score of each grid cell based on the physical constraints of the repair tool, which is given by the following formula: ； wherein, is the normalized distance from the tool center point to the maintenance point, is the normalized distance from the tool to the nearest obstacle, is the angle between the tool axis and the normal of the maintenance point, used to measure the pose reachability; , , is the reachability score weight coefficient, and satisfies ; Step 143, obtaining the accessibility score matrix of each slice layer as the input of the subsequent repair path planning; Step 150, taking the accessibility score matrix as input, and using a cross-layer ant colony optimization algorithm to realize repair path planning to generate a final repair path; the cross-layer ant colony optimization algorithm finds the optimal path on the slice layer and connects through a smooth curve to form a continuous and collision-free three-dimensional repair path through an ant colony algorithm based on the accessibility score matrix as the path planning environment and cross-layer connection point detection.

2. The method of claim 1, wherein, The step 110 comprises: using 3D MAX software to obtain the three-dimensional assembly model of the repair equipment; the 3D MAX software samples the grid surface of the three-dimensional assembly model through the script function or the scatter modifier, generates a plurality of copies of the seed points at the vertex positions or random areas on the surface of the three-dimensional assembly model, and converts the surface of the three-dimensional assembly model into discrete point cloud data; the point cloud data is exported as a standard point cloud file format.

3. The method of claim 2, wherein, The step 120 comprises: Step 121, inputting the point cloud data to preprocess the point cloud data: using Gaussian filtering or statistical filtering for denoising; using voxel grid downsampling for point cloud data downsampling; topology repair is performed on the grid data to ensure that the connection relationship of the patches is correct; Step 122, taking the defect point as a seed point, region growing is performed according to the curvature change and the geometric features of the neighborhood to accurately segment the repair boundary: Determine one or more defect points on the three-dimensional assembly model by manual marking or automatic detection algorithm, and expand the initial seed point as a region with the defect points as the region; Setting a curvature threshold and a geometric feature similarity threshold ; Take the seed point as the starting point, iteratively check the neighborhood points of the seed point, and perform the following region expansion operation: Calculate the Gaussian curvature or average curvature of the neighborhood points of the seed point; If the curvature difference of a neighborhood point is less than the curvature threshold, and the geometric feature similarity meets the similarity condition, the neighborhood point is included in the expanded region; the curvature difference refers to the absolute difference of the Gaussian curvature or average curvature of the current seed point and the new neighborhood point; Repeat the above process until there is no new neighborhood point that can be included in the expanded region, until the region expansion stops, and obtain the complete repair region; In the repair region, a boundary extraction algorithm is used to generate a repair boundary; Step 123, mark and output the geometric coordinates and topological relationship of the repair region position: By a three-dimensional coordinate extraction function , the coordinates of the repair boundary vertex are acquired and recorded ; Construct a topological connection graph of the repair region and the components of the assembly; A topological analysis algorithm is used to calculate the intersection area of a region and an adjacent component , defining a topological relationship weight wherein is the total area; Output a JSON or XML file containing the coordinate set and topological relationship.

4. The method of claim 3, wherein, The boundary extraction algorithm segments the repair boundary through boundary point determination; The boundary point determination includes: if a point of the three-dimensional assembly model belongs to the repair region, but there are other points in the neighborhood of the point that do not belong to the repair region, then the point belonging to the repair region is determined as a boundary point; the set of all boundary points constitutes the final repair boundary; The geometric feature similarity includes: normal vector similarity and / or color similarity; The normal vector similarity is characterized by the dot product value between two normal vectors; The color similarity is measured by calculating the Euclidean distance or absolute difference of the RGB values of two points; The geometric feature similarity threshold is set based on the normal vector direction and / or color information: set the lower limit threshold of the normal vector similarity based on the normal vector direction; set the upper limit threshold of the color similarity based on the color information; The similarity condition includes: the normal vector similarity is greater than the lower limit threshold of the normal vector similarity, and / or the color similarity is less than the upper limit threshold of the color similarity.

5. The method of claim 4, wherein, The step 130 includes: Step 131, use an adaptive layering algorithm to dynamically adjust the slice thickness according to the repair process requirements and model geometric features; The repair process requirements at least include the thickness of each component of the repair equipment; The adaptive layering algorithm includes: First step, initialize the input parameters of the algorithm: Threshold set of thicknesses for each component part required by the repair process wherein, is the maximum layer thickness threshold allowed for the th component part. Model geometric features: Grid data of a three-dimensional assembly model, comprising a set of vertices and a set of triangular facets ; Grid data of a three-dimensional assembly model: a set of vertices ; Triangular facet set ; each vertex coordinate information , ; Global layer thickness constraint: minimum layer thickness , maximum layer thickness ; Second step, determine the slice direction and model height range, including: Selecting a three-dimensional assembly model globally The axial positive direction is the slicing direction; The model height range is determined along the slice direction as , The minimum value of the slice height is The maximum value of the slice height is; and the initial slice height is ; Third step, dynamically calculate the layer thickness, including: Set constraints on the thickness of each component: for the current height Determine the component to be represented by the cross-section of the model at that location. Obtain the corresponding maximum layer thickness threshold. ; Extracting model geometric features: calculating the current height the normal vector of all the triangular facets by the angle between the normal vector and the tangent direction characterizing the surface inclination; Perform dynamic adjustment of the layer thickness using the following adjustment formula: ； In a fourth step, the slice height is updated according to the following formula: ; Step 5. Repeat the step of dynamically calculating the layer thickness and updating the slice height until the slice height is satisfied , the algorithm terminates. Step 132, reduce the layer thickness in the area with large curvature change to improve the repair accuracy, including: For each triangle of the model surface, the principal curvatures are computed The Gaussian curvature and the mean curvature of each vertex are computed ； Setting a curvature threshold , if , then mark as high curvature region; Perform layer thickness correction for high curvature areas, and adjust the layer thickness formula for high curvature areas to: ； wherein, is the layer thickness for high curvature regions, is a curvature correction factor, is a curvature sensitivity parameter; Perform transition processing on the repair boundary, including: for the transition zone between high curvature areas and flat areas, use linear interpolation to smooth the layer thickness change to avoid sudden changes in layer thickness: ； wherein, is a transition curvature region layer thickness, is a flat curvature region layer thickness, is a transition distance weight, .

6. The method of claim 5, wherein, The step 130 also includes: Step 133, increase the layer thickness in the flat area to improve the planning efficiency, including: Flat region determination: by calculating the normal vector variance of the current region , if , is a preset flat threshold, the region is determined as a flat region; or by averaging the curvature directly determine is a preset flat curvature threshold Optimize the layer thickness of the flat area: Maximize the layer thickness of the flat curvature region under the premise of meeting the precision requirements of the repair process: ； wherein, is a pre-set surface roughness, is a constant related to the material, process; The efficiency improvement is verified by the total slice number reduction amount calculated by the flat area proportion, which is given by the following formula: ; Step 134, generate a set of two-dimensional slices containing the repair area contour, including: Input the three-dimensional mask of the repair area; Set the slice plane at the specified height, calculate the intersection line of the slice plane and the triangular patches within the three-dimensional mask of the repair area by traversing all triangular patches; when the triangular patch intersects with the slice plane, calculate the two intersection points by the line-segment-plane intersection formula to obtain a discrete set of intersection points; Connect the intersection point set to form a closed contour polygon through a spatial sorting algorithm, including: calculate the centroid of the intersection point set; calculate the polar angle of each intersection point with the centroid as the reference point; sort and connect the intersection points in order to form a closed contour polygon; Optimize the contour polygon by judging and deleting collinear points through the cross product of adjacent edge vectors to remove redundant vertices, and output the two-dimensional coordinate sequence as the input of the repair path planning.

7. The method of claim 1, wherein, In step 150, the cross-layer ant colony optimization algorithm includes: Step 151, generate a local path, including: Initialize parameters: Set the basic parameters such as the size of the ant colony, the information volatility factor, and the heuristic function weight; Input the accessibility score matrix of the current slice layer as the path planning environment; Path construction: Each ant starts from the starting point of the repair area, transfers to the next grid unit based on the information concentration and accessibility score, and uses the following probability formula to select the grid unit for transfer: ； wherein, is the pheromone concentration, of the path , is the score value of the grid cell in the reachability score matrix, , are the weight coefficients of the pheromone and heuristic function, respectively. The ant records the path nodes during the movement according to the probability formula, avoiding the inaccessible area, which refers to the area where the grid unit score is below the threshold; Calculate the fitness value of the path generated by each ant: ； wherein, is a path length, is a path average reachability score, is a path smoothness; , , is an adaption weight coefficient; After the ants complete the path, the pheromone is locally updated, including the grid cells on the path according to the formula volatile pheromones, wherein is the volatile factor, is the initial pheromone; Update the information globally, including adding extra information to the path with the optimal fitness value according to the following formula: ； wherein, is a pheromone constant, is the optimal path fitness value; When the maximum number of iterations is reached or the path fitness value converges, output the optimal local path of the current slice layer.

8. The method of claim 7, wherein, The cross-layer ant colony optimization algorithm also includes: Step 152, global path integration, including: Use cross-layer connection point detection to extract the path endpoints and intermediate feature points for the optimal paths of two adjacent slice layers; for all point pairs formed by one point on each of the two adjacent slice layers, filter the point pairs with a Euclidean distance less than the threshold as candidate connection points; For the candidate connection points, use B-spline or Bezier curve for smooth curve connection and fitting as a connection curve; the connection curve satisfies the continuity and non-collision conditions: the first-order derivative is continuous at the connection point; perform discrete sampling on the connection curve and check whether the sampling points interfere with the three-dimensional assembly model; If the sampling points interfere with the three-dimensional assembly model, adjust the candidate connection points and re-fit until the non-collision condition is met; Integrate all the connection curves of adjacent slice layers through cross-layer connection point detection and smooth curve connection to form a preliminary three-dimensional repair path. Step 153, final optimization and output, generate a three-dimensional repair path, including: Input the preliminary three-dimensional repair path as the original path; Global interference detection on the original path: Discretize the original path into dense sampling points at a predetermined step length; For each sampling point, the minimum distance to the obstacles in the three-dimensional assembly model is calculated, and if the distance is less than the tool radius, it is determined as a collision point; The local path is re-planned for the maintenance area where the collision point is located: The five nodes before and after the collision point are taken as the starting point and the end point, and the ant colony algorithm is called to generate the obstacle-avoiding sub-path; The sub-path and the original path are connected by a smooth curve to ensure the overall continuity, and the re-planned three-dimensional maintenance path is obtained; The re-planned three-dimensional maintenance path is taken as the new original path; The above steps are repeated until the total length and smoothness of the re-planned three-dimensional maintenance path tend to be stable, the cycle is terminated, and the re-planned three-dimensional maintenance path is output as the output.

9. A slice-integrated repair path planning device based on a three-dimensional assembly model, characterized by, The device is used to implement the steps of the method of claim 1, and the device comprises the following modules: A first module for obtaining a three-dimensional assembly model of a device to be maintained and generating point cloud data; A second module for inputting the three-dimensional assembly model and automatically identifying the maintenance area using a region growing algorithm to output the location of the maintenance area; A third module for performing adaptive layering and slicing of the three-dimensional assembly model using an adaptive layering algorithm to generate a plurality of slice layers of a two-dimensional slice sequence containing the outline of the maintenance area; A fourth module for discretizing the maintenance area into regular grid cells in each slice layer and obtaining an accessibility score matrix for each slice layer by calculating the comprehensive accessibility score of each grid cell; comprising: A first submodule for discretizing the maintenance space into grid cells and marking obstacle grids in each slice layer, comprising: Inputting the two-dimensional projection area of the three-dimensional maintenance space in the current slice layer to define the maintenance boundary coordinate range; Setting the grid resolution according to the maintenance task accuracy requirement and calculating the number of grid rows and columns; Indexing and encoding the grid cells in two dimensions to obtain the center coordinates of the grid cells; Establishing a mapping relationship table between the grid index and the physical coordinates to store the maintenance boundary coordinates of each cell; Inputting the two-dimensional outline of the obstacles in the slice layer and using the ray method to determine whether the center coordinates of each grid cell are inside the obstacles; if it is determined that the center coordinates are inside the obstacles, the grid cell is marked as unreachable and the accessibility score is zero, otherwise, the grid cell is marked as a to-be-evaluated cell; A second submodule for calculating the comprehensive accessibility score of each grid cell, comprising calculating the accessibility score of each grid cell based on the physical constraints of the maintenance tool, which is given by the following formula: ； wherein, is the normalized distance from the tool center point to the maintenance point, is the normalized distance from the tool to the nearest obstacle, is the angle between the tool axis and the normal of the maintenance point, used to measure the pose reachability; , , is the reachability score weight coefficient, and satisfies ; A third submodule for obtaining the accessibility score matrix of each slice layer as input for subsequent maintenance path planning; A fifth module for inputting the accessibility score matrix and implementing maintenance path planning using a cross-layer ant colony optimization algorithm to generate the final maintenance path; the cross-layer ant colony optimization algorithm finds the optimal path on the slice layer and connects them through a smooth curve to form a continuous and collision-free three-dimensional maintenance path by using an ant colony algorithm based on the accessibility score matrix as the path planning environment and cross-layer connection point detection.

Images