Neural guidance rasterization-based twin manufacturing accessibility analysis method and system

Abstract

The invention provides a method and a system for analyzing accessibility of manufactured twin based on nerve guidance rasterization, and relates to the technical field of subtractive manufacturing, and the method comprises the following steps: obtaining a three-dimensional model and cutter parameters of a target manufactured twin; based on the tool parameters, discretization processing is carried out on the axial direction possibly adopted in the tool machining process, and a tool direction set is constructed; potential reachability guiding information in all cutter directions is predicted through a neural network, and a reachability guiding matrix is obtained; on the basis of the reachability guiding matrix, reachability verification is carried out on each tool direction in the tool direction set in sequence; after all cutter directions are verified, summarizing the accessibility states of all grid vertexes; according to the method, the final reachability result is directly output without depending on the neural network, the potential reachability of the model area in the cutter direction is predicted and guided by using the neural network, and final judgment is completed through reachability verification based on rasterization, so that analysis efficiency and industrial processing safety are both considered.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of subtractive manufacturing, in particular to a manufacturing twin-body reachability analysis method and system based on neural-guided rasterization. BACKGROUND

[0002] Manufacturing twin-body, as one of the core technologies of intelligent manufacturing, is a dynamic system that realizes process visualization, analysis and prediction by constructing digital mapping between physical space and virtual space, and fusing manufacturing environment, manufacturing resources and process data. In the construction and application of manufacturing twin-body, the transformation from design twin-body to manufacturing twin-body is a key link, and the tool reachability analysis is a basic prerequisite for determining the manufacturability of complex part design features in the transformation process. The analysis result directly determines whether the geometric features in the design twin-body can be transformed into effective machining features in the manufacturing twin-body, and is also an important basis for subsequent manufacturing twin-body process planning, tool path generation, process virtual reproduction and error prediction.

[0003] Tool reachability analysis in manufacturing twin-body refers to determining whether the specified position on the surface of the three-dimensional model of the design twin-body can be approached and machined by the tool in the allowed direction without geometric collision in the virtual space of the manufacturing twin-body, combined with the manufacturing resources and environmental constraints such as tool geometric size, tool axis posture, machine tool motion constraints, etc. The core is to realize the precise coupling of design intent and manufacturing feasibility in the manufacturing twin-body.

[0004] Existing tool reachability analysis methods are mainly based on precise geometric calculation. Such methods usually need to discretize the model in space, and perform collision detection or visibility judgment for each discrete sampling point in multiple candidate tool directions. Specifically, the spatial relationship between the tool and the model needs to be calculated repeatedly, and it is determined whether there is intersection or occlusion. Although the above method can give strict geometric judgment results in theory, the calculation process involves complex spatial queries, a large number of repeated geometric operations and high-dimensional parameter combinations. When the model sampling points are relatively dense, the overall calculation time will increase significantly, which is difficult to meet the efficiency requirements of interactive design and rapid process evaluation.

[0005] In recent years, some researches have tried to introduce deep learning methods to directly predict the tool accessibility distribution using neural networks, thereby avoiding a large number of geometric calculations in the running stage. These methods have improved the analysis speed to some extent, but since neural networks mainly rely on data-driven, it is difficult to strictly model complex geometric occlusion relationships, especially when there are small occlusions or multiple occlusion structures, which may misjudge the actual inaccessible region as an accessible region. The above misjudgment may cause the tool to collide with the model in actual machining, thereby causing equipment damage or processing accidents, so it is difficult to directly use in industrial application scenarios.

[0006] Therefore, under the conditions of complex three-dimensional models and multiple tool directions, the existing tool accessibility analysis cannot simultaneously consider the problems of computational efficiency and reliable judgment. SUMMARY

[0007] To solve the above problems, the present application proposes a manufacturing twin accessibility analysis method and system based on neural-guided rasterization, which does not rely on neural networks to directly output the final accessibility result, but uses neural networks to predict and guide the potential accessibility of the model region in the tool direction, and then completes the final judgment through accessibility verification based on GPU rasterization, thereby balancing analysis efficiency and industrial processing safety.

[0008] According to some embodiments, the present application adopts the following technical solutions: The manufacturing twin accessibility analysis method based on neural-guided rasterization comprises: obtaining a three-dimensional model of a target manufacturing twin and tool parameters; based on the tool parameters, discretizing the axial directions that the tool may adopt in the machining process to construct a tool direction set; predicting the potential accessibility guidance information of all tool directions through a neural network to obtain an accessibility guidance matrix, which records the accessibility prediction value of all grid vertices of the three-dimensional model in each tool direction; based on the accessibility guidance matrix, using the principle of verifying the tool direction that can confirm a large number of accessible vertices first, sequentially verifying each tool direction in the tool direction set, and updating the accessibility state of the vertices in the grid vertex set that have not been confirmed to be accessible according to the verification result; after all tool direction verifications are completed, the accessibility states of all grid vertices are summarized and mapped back to the original three-dimensional model surface; wherein the accessibility verification is a geometric accessibility verification method based on rasterization rendering, which strictly judges whether the tool collides with the model through depth map and pooling operation, thereby verifying whether the grid vertex is accessible.

[0009] According to some embodiments, the present application adopts the technical solutions as follows: The manufacturing twin reachability analysis system based on neural-guided rasterization includes: The acquisition module is configured to acquire a three-dimensional model of a target manufacturing twin and tool parameters; The construction module is configured to discretize the axial direction that the tool can take in the machining process based on the tool parameters, and construct a tool direction set; The prediction module is configured to predict the potential reachability guidance information of all tool directions through a neural network, and obtain a reachability guidance matrix, which records the reachability prediction value of all grid vertices of the three-dimensional model under each tool direction; The verification module is configured to, based on the reachability guidance matrix, adopt the principle of verifying the tool direction that is most likely to confirm a large number of reachable vertices first, and sequentially verify each tool direction in the tool direction set, and update the reachability state of the vertices in the grid vertex set that have not been confirmed to be reachable according to the verification result; The summary module is configured to, after the verification of all tool directions is completed, summarize the reachability state of all grid vertices and map it back to the original three-dimensional model surface. The reachability verification is a geometric reachability verification method based on rasterization rendering, which strictly judges whether the tool and the model collide through a depth map and a pooling operation, so as to verify whether the grid vertex is reachable.

[0010] According to some embodiments, the present application adopts the technical solutions as follows: A computer program product includes a computer program, which, when executed by a processor, implements the manufacturing twin reachability analysis method based on neural-guided rasterization.

[0011] According to some embodiments, the present application adopts the technical solutions as follows: A non-transitory computer-readable storage medium for storing computer instructions, which, when executed by a processor, implements the manufacturing twin reachability analysis method based on neural-guided rasterization.

[0012] According to some embodiments, the present application adopts the technical solutions as follows: An electronic device includes a processor, a memory, and a computer program; wherein the processor is connected with the memory, and the computer program is stored in the memory; when the electronic device is running, the processor executes the computer program stored in the memory, so that the electronic device executes the manufacturing twin reachability analysis method based on neural-guided rasterization.

[0013] Compared with the prior art, the present application has the following beneficial effects: (1) The manufacturing twin accessibility analysis method based on the combination of neural guidance and rasterization geometric verification provided by the present application significantly improves the overall efficiency of complex three-dimensional model accessibility analysis under the premise of ensuring strict and reliable determination results. By introducing a neural network to predict the potential accessibility under different tool directions, only the prediction results are used as guidance information for the geometric verification sequence, and they are not directly involved in the final determination, so that a large number of truly accessible vertices can be confirmed in the tool direction at the front of the sequence. Once the vertex is confirmed to be accessible in any direction, the early stopping mechanism is triggered and the vertex is removed from the subsequent verification process, thereby effectively avoiding repeated execution of high-cost geometric calculations on the same vertex under multiple directions and multiple scales. Therefore, even if the number of model surface sampling points increases significantly, the computational overhead of the present application will not increase proportionally with the increase of sampling density, and the present application can maintain high running efficiency in complex model and high-precision analysis scenarios.

[0014] (2) The present application always maintains strict conservative characteristics throughout the analysis process, which can avoid misjudging the actual inaccessible region as an accessible region while maintaining high determination accuracy. Specifically, the neural network prediction result is only used as prior guidance information, and all final accessibility determinations are based on the geometric verification process of rasterization rendering. The geometric verification process conservatively detects potential tool-model collisions based on depth map minimum pooling and tool geometry parameters, and compensates for rasterization discretization errors through boundary expansion and other measures, so that any possible collision situation will not be missed, and the present application has high engineering safety and industrial applicability.

[0015] (3) Since the accessibility determination core of the present application is based on general geometric rendering and parameterized tool models, it does not depend on specific model structures or fixed scale features, and has good adaptability to three-dimensional models of different complexity and tools of different sizes, shapes and parameter configurations. Even if the model size changes or the tool parameters are adjusted, the present application can still maintain stable and reliable analysis performance, has good universality and engineering robustness, and is suitable for popularization and use in actual application scenarios such as numerical control subtractive manufacturing. BRIEF DESCRIPTION OF DRAWINGS

[0016] The accompanying drawings, which form a part of this description, are included to provide further understanding of the application, and are incorporated in and constitute a part of this application. The schematic embodiments of the present application and their descriptions are used to explain the present application, and do not constitute an improper limitation on the present application.

[0017] Figure 1 The method flowchart of Example 1.

[0018] Figure 2 The three-dimensional model example diagram of the target manufacturing twin of Example 1.

[0019] Figure 3 Figure 1 is an example of tool parameters for Example 1.

[0020] Figure 4 Figure 2 is an example of tool orientation for Example 1.

[0021] Figure 5 Figure 3 is a neural network architecture diagram based on O-CNN for Example 1.

[0022] Figure 6 Figure 4 is an example of tool orientations to be verified for Example 1.

[0023] Figure 7 Figure 5 is a depth map for Example 1.

[0024] Figure 8 Figure 6 is a tool center depth map for Example 1.

[0025] Figure 9 Figure 7 is a pooling depth result map for Example 1, in which (a) and (b) are small radius window and large radius window, respectively.

[0026] Figure 10 Figure 8 is a reachability determination result map for Example 1 under a certain tool orientation.

[0027] Figure 11 Figure 9 is a final reachability analysis three-dimensional result for Example 1. DETAILED DESCRIPTION

[0028] The present application will be further described below in conjunction with the accompanying drawings and examples.

[0029] It should be noted that the following detailed description is exemplary in nature and is intended to provide further description of the application. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0030] It is to be understood that the singular forms “a,” “an,” and “the” include plural referents unless the context clearly dictates otherwise. It should also be noted that, as used herein, the term “and / or” includes any and all combinations of one or more of the associated listed items. In addition, it should be noted that, as used herein, the terms “the” and / or “a,” and “one or more of the” preceding a list of items should be interpreted as including any and all combinations of one or more of the associated listed items.

[0031] Example 1 In one embodiment of the present application, a neural-guided rasterization-based manufacturing twin reachability analysis method is provided, mainly for subtractive manufacturing processes, which analyzes the three-dimensional model input by the user, and under the premise of ensuring that the determination result is strictly conservative, efficiently determines the reachability distribution of each position on the model surface under the condition of multiple tool orientations, such asFigure 1 The specific steps are shown as follows: Step S1: Obtain a three-dimensional model of a target manufacturing twin and tool parameters.

[0032] The three-dimensional model is approximated and stored in a discrete data structure, such as a triangular mesh, a point cloud, or an SDF (signed distance function) representation method.

[0033] In this embodiment, the three-dimensional model is represented by a triangular mesh, as shown in Figure 2 The three-dimensional model is represented by a triangular mesh, as shown in Figure 3 The tool parameters include tool ball radius, tool top radius, tool total length, and tool rod length, etc. The input triangular mesh model needs to meet two key conditions: water tightness and flow pattern: 1) The model surface should have no holes and no self-intersection phenomenon to ensure the integrity of the geometric structure to meet the requirements of subtractive manufacturing process; 2) The model must have flow pattern topology characteristics, that is, the adjacency relationship of all triangular facets meets the standard topology connection rules, excluding boundary points, non-closed areas and other abnormal geometric shapes.

[0034] The above constraints ensure the effective compatibility of the model and the manufacturing process, and provide a reliable geometric basis for subsequent reachability analysis.

[0035] Step S2: Discretize the axial direction that the tool may take during processing based on the tool parameters, and construct a tool direction set.

[0036] Further, the construction of the tool direction set includes: Since the tool is processed from above the workpiece, the tool direction is uniformly sampled on the unit upper hemisphere with a radius of 1, a sphere center at the origin, and z≥0, to construct a finite number of candidate tool direction sets, which are discretized to handle the axial direction that the tool may take during processing, for simulating the possible entry direction of the tool during multi-axis NC machining.

[0037] Specifically, to facilitate later projection, the tool direction is imagined as a virtual camera, and the camera's shooting angle is consistent with the tool direction, as shown in Figure 4Each direction represents a fixed pose of the tool axis in the model coordinate system, which is used to simulate the possible entry direction of the tool in the multi-axis NC machining process. By constructing a discrete set of tool directions, the continuous direction space can be converted into a finite verification problem, providing a clear search space for subsequent reachability analysis and avoiding the problem of uncontrollable calculation due to the continuity of the direction.

[0038] Step S3: predicting the potential reachability guidance information of all tool directions through the neural network to obtain a reachability guidance matrix, which records the reachability prediction value of all grid vertices of the three-dimensional model under each tool direction.

[0039] Further, the step of predicting the potential reachability guidance information of all tool directions through the neural network comprises: The three-dimensional model is represented as an octree structure, and the triangular mesh model of the octree structure is input into the trained neural network together with the tool parameters for inference. The binary potential reachability prediction value of each grid vertex under each tool direction is output, and a reachability guidance matrix is constructed. wherein, is the number of grid vertices, is the number of tool directions in the tool direction set, and the matrix element represents the vertex has potential reachability under the tool direction represents that it is predicted to be potentially unreachable.

[0040] Specifically, the neural network adopts an Unet architecture based on O-CNN as shown in Figure 5 , which is used to predict the potential reachability distribution of each vertex in the model under different tool directions. The binary (0 / 1) reachability prediction is performed on all grid vertices under each candidate tool direction. Figure 5 The red area on the three-dimensional model inrepresents the most probable direction visualization, thereby constructing a vertex-direction reachability guidance matrix. This matrix only serves as guidance information in the subsequent geometric reachability verification process and is not directly used as the final tool reachability judgment basis. Its role is to provide prior reference for the subsequent verification stage, thereby reducing unnecessary geometric calculations.

[0041] The present application randomly extracts 11000 triangular mesh models from a public dataset ABC (A Big CAD Model Dataset For Geometric Deep Learning), calculates whether the sampling points are reachable on each unit hemispherical sampling direction based on traditional geometric methods, and constructs the dataset. Among them, after scale normalization processing, 8800 models are used for training the neural network, 2000 models constitute the test set, and the remaining 200 models are used for quantitative evaluation of the calculation efficiency and accuracy outside the gallery results, and the training process adopts the cross-entropy loss function.

[0042] Step S4: Based on the reachability guide matrix, the principle of verifying the tool direction that is most likely to confirm a large number of reachable vertices is adopted, and the reachability verification of each tool direction in the tool direction set is carried out in turn, and the reachability state of the vertices in the current grid vertex set that has not been confirmed to be reachable is updated according to the verification result.

[0043] Among them, the reachability verification is a geometric reachability verification method based on rasterization rendering, which strictly judges whether the tool and the model collide through depth map and pooling operation, so as to verify whether the grid vertex is reachable.

[0044] Specifically, the feasibility of all grid vertices in the grid vertex set is confirmed by adopting multi-round reachability verification. Once a vertex is confirmed to be reachable in any tool direction, the vertex is removed from the current grid vertex set that has not been confirmed to be reachable.

[0045] In each round of reachability verification, the tool direction set is sorted based on the neural network prediction result to optimize the calculation order of subsequent geometric reachability verification. Specifically, a greedy sorting strategy based on coverage is adopted, and in each round of reachability verification process, the tool direction that can cover the most number of vertices that have not been confirmed to be reachable is preferentially selected as the tool direction to be verified in the current round. The tool direction selected in the first round satisfies:

[0046] Among them, represents the current grid vertex set that has not been confirmed to be reachable, represents the set of tool directions that have been verified. Through the above sorting method, in the subsequent geometric verification stage based on rasterization, the tool direction that is more likely to confirm a large number of reachable vertices can be preferentially verified, so as to improve the triggering probability of the early stopping mechanism, and effectively reduce the overall calculation overhead on the premise of ensuring the conservativeness of the final judgment result.

[0047] For the first ​The tool direction is selected As shown in Figure 6 , taking the downward tool direction as an example, the geometric accessibility verification based on rasterization rendering is performed on the grid vertex which has not been confirmed to be accessible. The verification method utilizes the axial symmetry characteristics of the tool geometry, converts the tool-model collision detection problem in three-dimensional space into a two-dimensional parallel computing process based on depth map, thereby significantly reducing the calculation cost under the premise of ensuring that the determination result is strictly conservative. Then the accessibility verification of the wheel is specifically: Step S4-1: for the current tool direction to be verified , direction normalization is performed, the normalized tool direction is set as the camera observation direction, and the rotation matrix from the world coordinate system to the camera coordinate system is calculated , through the rotation transformation, the vertices and their normal vectors of the model are uniformly transformed to the camera coordinate system:

[0048] wherein, V represents the vertex set of the three-dimensional model, V represents the vertex coordinates of the model in the world coordinate system, V represents the vertex coordinates of the vertex transformed to the camera coordinate system, N represents the vertex normal vector of the model in the world coordinate system, N represents the vertex normal vector of the vertex transformed to the camera coordinate system. According to the bounding box of the rotated model, the orthogonal projection camera parameters are adaptively set, so that the entire model falls into the fixed resolution viewport range.

[0049] Step S4-2: in the current tool direction , the rasterization rendering based on GPU is adopted to perform orthogonal projection on the rotated three-dimensional model, and the corresponding depth map is generated.

[0050] The depth map is a two-dimensional image used to record three-dimensional space information. It is very similar to a photo, but the color in the photo does not record the brightness or texture of the object, but the distance of the object in the scene from the observer.

[0051] In this embodiment, the rasterization rendering based on GPU is adopted, the virtual camera is aligned to the rotated three-dimensional model, the orthogonal projection matrix is used to project the three-dimensional model onto a two-dimensional plane in parallel, the accurate size of the object is preserved, and the perspective effect of "big near and small far" is eliminated, and the depth map as shown in Figure 7 is obtained. In the visualization image of the depth map, generally, the lighter the color, the farther the distance, and the darker the color, the closer the distance.

[0052] Each pixel in the depth map records the depth value of the frontmost visible surface under the current tool orientation. Pixels not covered by any geometry are assigned an infinite depth. To ensure the conservativeness of the analysis result, a local minimum value expansion operation is performed on the model boundary region after the depth map is generated to compensate for the discrete error possibly caused by the limited resolution.

[0053] Step S4-3: After obtaining the depth map, first filter out the non-negative triangle face IDs from the pixel-triangle face mapping table output by the rasterization, to obtain the set of visible triangle faces under the current view angle, then flatten the vertex IDs of all the visible triangle faces and take the unique values to obtain the initial set of visible vertices , and project them to the screen space.

[0054] For each visible vertex , offset the tool cutting region radius in the direction of its rotated unit normal vector to construct the corresponding tool center position:

[0055] Subsequently, project the tool center to the screen space to obtain its pixel coordinates and depth value, and represent it with a tool center depth map as shown in Figure 8 , where each point is a tool center constructed by a visible vertex. Figure 8

[0056] Step S4-4: Take the projected tool center position as the center of a circle, and construct a circular detection window corresponding to the tool geometry on the depth map .

[0057] Specifically, as shown in Figure 9 , according to the structural characteristics of the tool composed of cylindrical segments of different radii, set a small radius window as shown in Figure 9 (a) and a large radius window as shown in Figure 9 (b) respectively, and perform a minimum value pooling operation on the depth map in the corresponding window, which can be expressed by the formula:

[0058]

[0059] wherein, represents the circular pixel region with the tool center projection position as the center and the radius , represents the pixel, represents the minimum depth value in the small radius pooling window, ​represents the minimum depth value in the large radius pooling window.

[0060] The present embodiment takes a tool composed of two cylindrical segments as an example, but a tool composed of non-cylindrical segments can be solved by modifying the pooling window, and the determination of a tool composed of more segments can be solved by slightly modifying the determination condition and adding a pooling window. The solving idea is to determine by depth difference.

[0061] Step S4-5: Determine whether the tool collides with the model in the current direction by comparing the difference between the depth of the tool center and the minimum depth value in each circular window, and obtain Figure 10 The reachability determination result shown in FIG. 6, Figure 10 The medium red color is reachable, and the gray color is unreachable.

[0062] Specifically, the depth difference is calculated as follows:

[0063]

[0064] wherein, represents the depth value of the tool center, and if exceeds the height of the corresponding cylindrical segment, it is determined that the vertex is unreachable in the current direction of the tool; otherwise, it is determined that the vertex is reachable in the direction.

[0065] After completing the raster-based reachability verification in the current round of tool direction , the vertex set that is still not confirmed to be reachable is updated, and early stop determination is performed to avoid unnecessary subsequent geometric calculations.

[0066] Early stop determination refers to the monotonicity feature of tool reachability determination, that is, once a vertex is strictly confirmed to be reachable in any direction of the tool, there is no need to continue to verify the vertex in other directions. All vertices that have been confirmed to be reachable will be marked and excluded from the geometric verification process in subsequent tool directions, while vertices that have not been confirmed to be reachable will continue to participate in the verification process in subsequent directions.

[0067] Step S5: After all tool direction verifications are completed, the reachability status of all grid vertices is summarized and mapped back to the original three-dimensional model surface.

[0068] After all tool direction verifications are completed, the vertex-level reachability determination result is summarized, and the final reachability label is mapped back to the original three-dimensional model surface. As shown in FIG. 6, Figure 11 According to the final determination state of the vertex, the model surface is divided into reachable and unreachable regions, and further output in the form of color annotation, region segmentation or data file for subsequent numerical control machining process evaluation, tool path planning or manufacturing feasibility analysis.

[0069] Embodiment 2 In an embodiment of the present application, a manufacturing twin reachability analysis system based on neural-guided rasterization is provided, comprising: An acquisition module configured to acquire a three-dimensional model of a target manufacturing twin and tool parameters; A construction module configured to discretize the axial direction that the tool can take in the machining process based on the tool parameters, and construct a tool direction set; A prediction module configured to predict the potential reachability guidance information of all tool directions through a neural network, and obtain a reachability guidance matrix, which records the reachability prediction value of all grid vertices of the three-dimensional model under each tool direction; A verification module configured to verify the reachability of each tool direction in the tool direction set in turn based on the reachability guidance matrix, and update the reachability state of the vertices in the grid vertex set that have not been confirmed to be reachable according to the verification result; A summary module configured to summarize the reachability state of all grid vertices and map it back to the original three-dimensional model surface after the verification of all tool directions is completed. Wherein, the reachability verification is a geometric reachability verification method based on rasterization rendering, which strictly judges whether the tool and the model collide through depth map and pooling operation, so as to verify whether the grid vertex is reachable.

[0070] Embodiment 3 In an embodiment of the present application, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the manufacturing twin reachability analysis method based on neural-guided rasterization.

[0071] Embodiment 4 In an embodiment of the present application, a non-transitory computer readable storage medium is provided for storing computer instructions, which, when executed by a processor, implement the manufacturing twin reachability analysis method based on neural-guided rasterization.

[0072] Embodiment 5 In an embodiment of the present application, an electronic device is provided, comprising a processor, a memory and a computer program; wherein the processor is connected with the memory, and the computer program is stored in the memory; when the electronic device is running, the processor executes the computer program stored in the memory, so that the electronic device executes the manufacturing twin reachability analysis method based on neural-guided rasterization.

[0073] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flow or blocks Figure 1 means for functionally implementing the steps in one or more flow or blocks.

[0074] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flow or blocks Figure 1 steps for functionally implementing the steps in one or more flow or blocks.

[0075] The above description is only specific implementation of the present application, and is not intended to limit the protection scope of the present application. Any modification or change made by those skilled in the art without creative effort based on the technical solution of the present application shall fall within the protection scope of the present application.

Claims

1. A method for accessibility analysis of fabricated twins based on neural-guided rasterization, characterized in that, include: Obtain the 3D model of the target manufacturing twin and the tool parameters; Based on the tool parameters, the axial directions that the tool may use during the machining process are discretized to construct a set of tool directions; The potential reachability guidance information for all tool directions is predicted by a neural network to obtain an reachability guidance matrix. The reachability guidance matrix records the reachability prediction values ​​of all mesh vertices of the 3D model under each tool direction. Based on the reachability guidance matrix, the principle of prioritizing the verification of tool directions that are most likely to confirm a large number of reachable vertices is adopted. The reachability of each tool direction in the tool direction set is verified in turn. Based on the verification results, the reachability status of vertices in the set of mesh vertices that have not yet been confirmed to be reachable is updated. After all tool orientations have been verified, the reachability status of all mesh vertices is summarized and mapped back to the original 3D model surface. The reachability verification adopts a geometric reachability verification method based on rasterization rendering. It strictly judges whether a collision occurs between the tool and the model through depth map and pooling operation, thereby verifying whether the mesh vertices are reachable.

2. The method for accessibility analysis of fabricated twins based on neural-guided rasterization as described in claim 1, characterized in that, The set of tool directions specifically includes: The tool direction is uniformly sampled on the unit upper hemisphere to construct a finite set of candidate tool directions. The axial direction that the tool may take during the machining process is discretized to simulate the entry direction that the tool may take during multi-axis CNC machining.

3. The method for accessibility analysis of fabricated twins based on neural-guided rasterization as described in claim 1, characterized in that, The method of predicting potential reachability guidance information for all tool directions using a neural network specifically includes: The 3D model is encoded into a 3D representation that can be processed by a neural network, and input along with tool parameters into the trained neural network for inference. The binary potential reachability prediction value of each grid vertex under each tool direction is output, and an reachability guidance matrix is ​​constructed. ,in, For grid vertex data, The number of tool directions in the tool direction set, matrix elements Represents vertices In the direction of the tool It has potential accessibility. This indicates that the prediction is potentially unreachable.

4. The method for accessibility analysis of fabricated twins based on neural-guided rasterization as described in claim 1, characterized in that, The principle of prioritizing the verification of tool directions most likely to confirm a large number of reachable vertices specifically includes: A greedy sorting strategy based on coverage is adopted. In each round of verification, the tool direction that covers the largest number of currently unconfirmed reachable vertices is selected from the set of remaining unverified tool directions as the current verification object, which can be expressed by the formula: in, This represents the set of mesh vertices that have not yet been confirmed as reachable. For tool direction set, This represents the set of verified tool orientations. This represents the set of remaining unverified tool directions.

5. The method for accessibility analysis of fabricated twins based on neural-guided rasterization as described in claim 1, characterized in that, The geometric reachability verification method based on rasterization rendering specifically includes: The current tool direction is normalized, and the normalized tool direction is set as the camera viewing direction. The world coordinate system is aligned with the camera coordinate system through rotation transformation, and the model vertices and normal vectors are transformed to the camera coordinate system. Generate depth maps under orthographic projection using GPU rasterization rendering; For the visible vertices in the model, offset the radius of the tool cutting area along its normal vector direction to construct the tool center position and project it onto the screen space; On the depth map, a circular detection window corresponding to the tool geometry is constructed with the tool center projection position as the center. The minimum depth within the window is calculated by minimum pooling and compared with the tool center depth to determine whether the vertex is reachable in the current tool direction.

6. The method for accessibility analysis of fabricated twins based on neural-guided rasterization as described in claim 5, characterized in that, The construction of the circular detection window corresponding to the tool geometry specifically includes: Based on the structural characteristics of the cutting tool, which consists of cylindrical segments of different radii, small radius windows are set respectively. and large radius window These correspond to different geometric segments of the cutting tool; Calculate the difference between the tool center depth and the minimum depth value within each window. and ,like or If the height exceeds the corresponding cylindrical segment height, the vertex is determined to be unreachable in the current tool direction; otherwise, it is determined to be reachable.

7. The method for accessibility analysis of fabricated twins based on neural-guided rasterization as described in claim 1, characterized in that, The process of sequentially verifying the reachability of each tool direction in the tool direction set also includes an early stop determination step: For any mesh vertex that has been confirmed to be reachable, it is immediately marked and removed from the subsequent geometric verification process in the tool direction, and will no longer participate in the subsequent verification, so as to reduce unnecessary geometric calculations by taking advantage of the monotonicity of tool reachability determination.

8. A manufacturing twin accessibility analysis system based on neural-guided rasterization, characterized in that, include: The acquisition module is configured to acquire the 3D model of the target manufacturing twin and tool parameters; The construction module is configured to: discretize the axial directions that the tool may use during the machining process based on the tool parameters, and construct a set of tool directions; The prediction module is configured to: predict potential reachability guidance information for all tool directions through a neural network to obtain an reachability guidance matrix, wherein the reachability guidance matrix records the reachability prediction values ​​of all mesh vertices of the 3D model under each tool direction; The verification module is configured to: based on the reachability guidance matrix, adopt the principle of prioritizing the verification of tool directions that are most likely to confirm a large number of reachable vertices, sequentially verify the reachability of each tool direction in the tool direction set, and update the reachability status of vertices in the set of mesh vertices that have not yet been confirmed as reachable based on the verification results; The summary module is configured to: after all tool orientation verifications are completed, summarize the reachability status of all mesh vertices and map it back to the original 3D model surface; The reachability verification adopts a geometric reachability verification method based on rasterization rendering. It strictly judges whether a collision occurs between the tool and the model through depth map and pooling operation, thereby verifying whether the mesh vertices are reachable.

9. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium is used to store computer instructions, which, when executed by a processor, implement the neural-guided rasterization-based manufacturing twin accessibility analysis method as described in any one of claims 1-7.

10. An electronic device, characterized in that, include: The device includes a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to enable the electronic device to perform the method for analyzing the accessibility of a manufacturing twin based on neural-guided rasterization as described in any one of claims 1-7.

Images