A product reconstruction method and system based on three-dimensional scanning and CAD surface deviation distribution

Abstract

The application discloses a product reconstruction method and system based on three-dimensional scanning and CAD surface deviation distribution, and relates to the technical field of three-dimensional digital measurement. The method firstly carries out CAD pretreatment on each part to obtain a reference CAD model and a finite element model, and scans and acquires single-piece scanning point clouds. The point clouds are accurately aligned with the reference CAD model through single-piece registration, and then the projection point corresponding to each scanning point and the deviation vector thereof are calculated. After physical assembly, the part point clouds are positioned to a product coordinate system through product registration. Then, the surface nodes of the finite element model are classified based on the deviation vector. Finally, the corrected finite element model and the reference CAD model are repositioned to the product coordinate system, and a digital prototype and a high-fidelity finite element analysis model integrating actual geometric deviation and assembly information are output. The application significantly improves the convenience, precision of digital measurement and fidelity of finite element simulation.

Description

Technical Field

[0001] This invention relates to the field of three-dimensional digital measurement technology, specifically to a product reconstruction method and system based on three-dimensional scanning and CAD surface deviation distribution. Background Technology

[0002] In the process of industrial digital transformation, product reconstruction technology based on 3D scanning has become a core means of connecting physical products with digital models. However, existing technologies still have many key shortcomings, directly restricting their application effectiveness in high-precision scenarios. Traditional product reconstruction often adopts parametric fitting or point cloud meshing methods. Parametric fitting weakens the local geometric features of the product during the process, leading to distortion of the reconstructed model and failing to accurately reflect key information such as interference during actual assembly of parts. The fidelity of point cloud meshing is always limited by the quality and density of the original scanned point cloud, making it difficult to meet the needs of high-precision engineering analysis. For multi-part products, existing reconstruction methods either cannot effectively integrate the spatial position information after physical assembly, resulting in the digital model lacking the actual assembly correlation of the product; or they require additional installation of special positioning parts to assist in registration, which not only increases the pre-construction preparation cost of the reconstruction operation but also significantly reduces the applicability of the technology. At the same time, existing technologies lack a complete conversion path from scanned point clouds to high-precision finite element models. The generated finite element models do not incorporate deviation information during product processing and assembly, resulting in a large deviation between the subsequent finite element analysis results and the actual performance of the product. These issues collectively lead to insufficient geometric and assembly consistency between the reconstructed digital model and the physical product. This results in an inability to support accurate measurement of product geometric dimensions, and also makes it difficult to meet engineering requirements such as product performance analysis and process optimization, severely impacting the role of digital technology in improving production efficiency and product quality. Summary of the Invention

[0003] To address the shortcomings of existing technologies, the present invention aims to provide a product reconstruction method and system based on 3D scanning and CAD surface deviation distribution, which avoids data distortion caused by global point cloud processing, realizes accurate reverse reconstruction of multi-part products without dedicated tooling, and significantly improves the convenience, accuracy, and fidelity of digital measurement and finite element simulation.

[0004] To achieve the above objectives, the embodiments of this invention provide the following technical solutions:

[0005] This application provides a product reconstruction method based on 3D scanning and CAD surface deviation distribution, including the following steps: S1, obtaining the design geometric CAD model of each part, simplifying the design geometric CAD model by removing chamfers, fillets, and pin hole process features, and marking the processing datum on the simplified CAD model to obtain a datum CAD model for registration and deviation analysis, and discretizing the datum CAD model to generate a corresponding finite element model; S2, before the physical assembly of the product, using a 3D scanning device to scan each part one by one to obtain the single-part scan point cloud of each part; S3, for each part, comparing the single-part scan point cloud obtained in S2 with the single-part scan point cloud obtained in S1. S4. Register the single-part scan point cloud with the reference CAD model of the part to achieve precise alignment; S5. For each scan point in the single-part scan point cloud of each part, calculate the nearest point from the scan point to the surface of the corresponding part's reference CAD model, define the nearest point as the projection point, generate a vector from the projection point to its corresponding scan point, and define the vector as the deviation vector corresponding to the scan point; S6. Assemble each part into a complete product, scan the complete product using a 3D scanning device, and obtain the overall scan point cloud of the product including the processing reference of exposed parts; S7. For each part, compare the single-part scan point cloud obtained in S2 with the one obtained in S5. The overall scanned point cloud of the product is registered to determine the rigid body displacement transformation parameters required to transform the scanned point cloud of the single part from the part coordinate system to the overall product coordinate system; S7. For the finite element model of each part, the nodes on its outer surface are classified into face nodes and edge nodes. Nodes located on the face features of the finite element model but not on the edge features are face nodes, and nodes located on the edge features of the finite element model are edge nodes; S8. For each face node, a spherical search area is defined with the face node as the center. All projection points within the spherical search area are found, and the arithmetic mean of the deviation vectors corresponding to these projection points is calculated. This arithmetic mean is used as the face offset vector of the face node; S 9. For each edge node, establish a tubular search domain with the edge feature of the finite element model containing the edge node as the axis, find the projection points located in the tubular search domain, and generate the scan points corresponding to each projection point. Based on the found scan points, use the least squares method to fit the constituent surfaces on both sides of the edge feature to obtain two parameterized fitted surfaces. Calculate the spatial intersection of these two parameterized fitted surfaces, calculate the shortest distance vector from the edge node to the spatial intersection, and use the shortest distance vector as the edge offset vector of the edge node; S10. Based on the surface offset vector of each face node and the edge offset vector of each edge node, perform position offset on all outer surface nodes of the finite element model of each part;S11. Based on the rigid body displacement transformation parameters of each part, perform the same rigid body displacement transformation on both the precisely aligned reference CAD model of the part and the finite element model of the part with completed node offsets. After the displacement transformation, package the set of reference CAD models of all parts located in the overall product coordinate system, along with the projection point coordinates and deviation vector data associated with each reference CAD model, and output them together as a digital prototype of the product. Simultaneously, package the set of finite element models of all parts after the displacement transformation and output them as a finite element analysis model of the product.

[0006] Further, S3 specifically includes: S31, uniformly sampling and generating a reference reference point cloud from the machining reference surface of the reference CAD model of the part, and using a point cloud registration algorithm based on three-dimensional feature descriptors to initially align the single-piece scan point cloud of the part with the reference reference point cloud to obtain a preliminary registration result; S32, using the geometric features of the machining reference marked in the reference CAD model of the part as constraints, using a least squares optimization algorithm to iteratively optimize the preliminary registration result so that the single-piece scan point cloud and the reference CAD model achieve a precise alignment.

[0007] Further, S6 specifically includes: S61, using a point cloud registration algorithm based on three-dimensional feature descriptors to initially align the single-piece scan point cloud with the overall product scan point cloud to obtain a preliminary alignment result; S62, using the ICP algorithm to iteratively optimize the preliminary alignment result to accurately align the single-piece scan point cloud with the overall product scan point cloud, and calculate the rigid body displacement change parameters.

[0008] Furthermore, the method also includes a step of performing digital measurement of geometric dimensions based on a digital prototype, specifically including: S121, determining the feature to be measured and at least two measurement endpoints on the reference CAD model in the digital prototype; S122, reading the theoretical values ​​of the design parameters of the geometric feature on the reference CAD model from the digital prototype; S123, searching for associated projection points in the neighborhood space for each measurement endpoint; S124, calculating the average value of the deviation vectors corresponding to all projection points associated with each measurement endpoint; S125, projecting the average value of each deviation vector onto the measurement direction of the geometric feature to obtain the dimension deviation components corresponding to each endpoint; S126, subtracting the sum of the dimension deviation components corresponding to all endpoints from the theoretical value of the design parameters to obtain the actual digital measurement value of the geometric feature.

[0009] Furthermore, in S9, if a node is identified as an edge node belonging to three different model edge features simultaneously, the following operations are performed on that node: S91, for each edge feature to which it belongs, calculate the shortest distance vector from the node to the intersection line of the parameterized fitted surface space corresponding to that edge feature, resulting in three shortest distance vectors; S92, calculate the average value of the three shortest distance vectors; S93, use the average value as the final edge offset vector of the node as the position offset in S10.

[0010] Accordingly, this application also provides a product reconstruction system based on 3D scanning and CAD surface deviation distribution. The system includes: a preprocessing module configured to execute S1 to generate a reference CAD model and perform finite element mesh generation; a scanning module configured to execute S2 and S5 to acquire the single-piece scan point cloud of the part and the overall scan point cloud of the product; a single-piece registration module configured to execute S3 to achieve high-precision alignment between the single-piece scan point cloud and the reference CAD model; a deviation calculation module configured to execute S4 to calculate the projection points and deviation vectors; a product registration module configured to execute S6 to achieve coordinate unification from the single-piece point cloud to the product coordinate system; a node offset module configured to execute S7, S8, S9 and S10 to complete the classification of finite element model nodes and position correction based on the deviation vector; and an output module configured to execute S11 to reposition the reference CAD model and the initial finite element model, and integrate and output a digital prototype and a finite element analysis model.

[0011] Furthermore, the single-piece registration module includes: a first coarse registration unit configured to execute S31, extracting a reference reference point cloud from the reference CAD model and performing initial alignment using an algorithm based on three-dimensional feature descriptors; and a first fine registration unit configured to execute S32, performing least-squares optimization based on machining reference constraints.

[0012] Furthermore, the product registration module includes: a second coarse registration unit configured to execute S61, which uses an algorithm based on three-dimensional feature descriptors to achieve initial alignment between the single-piece point cloud and the overall point cloud; and a second fine registration unit configured to execute S62, which uses the ICP algorithm to achieve precise alignment and solve displacement parameters.

[0013] Furthermore, the node offset module includes: a node classification unit configured to execute S7, classifying the nodes on the outer surface of the finite element model into face nodes and edge nodes; a face node processing unit configured to execute S8, calculating the average spherical neighborhood deviation of each face node to generate a face offset vector; an edge node processing unit configured to execute S9, fitting the edge feature surface and calculating the intersection line to generate an edge offset vector; and an offset execution unit configured to execute S10, performing position offset on the nodes based on the face offset vector and the edge offset vector.

[0014] Furthermore, the deviation calculation module includes: a projection point calculation unit configured to calculate the nearest point from the scan point to the surface of the reference CAD model to obtain the projection point; and a deviation vector generation unit configured to generate a deviation vector pointing from the projection point to the corresponding scan point.

[0015] The beneficial effects of this invention are as follows: By adopting a phased registration logic of "single-part registration - product registration," it ensures accurate alignment between individual parts and the CAD model, and achieves precise repositioning of multiple parts in the product coordinate system. This effectively integrates the assembly information of the actual product and solves the problem of lack of assembly realism in multi-part reconstruction. By differentiating between face nodes and edge nodes for differential deviation processing, it avoids local distortion caused by global deviation processing, significantly improving the fidelity of the reconstructed model to the local geometric features of the product, making the digital model closer to the actual state. It simultaneously outputs a composite digital prototype containing the CAD model, projection points, and deviation vectors, as well as an optimized finite element model, which not only meets the needs of digital measurement of product geometric dimensions but also provides a high-fidelity carrier for product performance analysis, realizing "one-time reconstruction, multiple applications." It provides a complete conversion path from scanned point cloud to finite element model, and the generated finite element model incorporates product processing and assembly deviation information, significantly improving the accuracy of subsequent finite element analysis. It eliminates the need for dedicated positioning parts, completing multi-part registration through the product's exposed processing datum, broadening the applicability of the technology and reducing the pre-construction preparation costs. Attached Figure Description

[0016] Figure 1 A flowchart illustrating a product reconstruction method based on 3D scanning and CAD surface deviation distribution, provided for an embodiment of this application;

[0017] Figure 2 This is a schematic diagram of a product reconstruction system based on 3D scanning and CAD surface deviation distribution, provided as an embodiment of this application. Detailed Implementation

[0018] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the scope of the present invention.

[0019] In this invention, the terms "system" and "network" are used interchangeably. "Multiple" refers to two or more; therefore, in this invention, "multiple" can also be understood as "at least two." "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. Additionally, the character " / ", unless otherwise specified, generally indicates that the preceding and following related objects have an "or" relationship. Furthermore, it should be understood that in the description of this invention, terms such as "first" and "second" are used only for descriptive purposes and should not be construed as indicating or implying relative importance or order.

[0020] In existing product reconstruction technologies, parameter fitting methods weaken local geometric features of the product during the fitting process, leading to model distortion and failing to accurately reflect key information such as interference during actual assembly of parts. The fidelity of point cloud meshing technology is always limited by the original point cloud, making it difficult to achieve high-precision reconstruction. For multi-part products, existing reconstruction methods either cannot integrate physical assembly information or require the installation of special positioning parts, resulting in a narrow scope of application. Furthermore, there is a lack of specific and feasible solutions for converting scanned point clouds into finite element models, which cannot provide high-fidelity models for subsequent finite element analysis and seriously affects the accuracy of product inspection and performance analysis.

[0021] like Figure 1As shown, this application provides a product reconstruction method based on 3D scanning and CAD surface deviation distribution, including the following steps: S1, obtaining the design geometric CAD model of each part, simplifying the design geometric CAD model by removing chamfers, fillets, and pin hole process features, and marking the processing datum on the simplified CAD model to obtain a datum CAD model for registration and deviation analysis, and discretizing the datum CAD model to generate the corresponding finite element model; S2, before the physical assembly of the product, using a 3D scanning device to scan each part one by one to obtain the single-part scanning point cloud of each part; S3, for each part, comparing the single-part scanning point cloud obtained in S2 with the point cloud obtained in S1. S4. Register the obtained reference CAD model of the part to achieve precise alignment between the single-part scan point cloud and the reference CAD model; S5. For each scan point in the single-part scan point cloud of each part, calculate the nearest point from the scan point to the surface of the corresponding part's reference CAD model, define the nearest point as the projection point, generate a vector from the projection point to its corresponding scan point, and define the vector as the deviation vector corresponding to the scan point; S6. Assemble each part into a complete product, scan the complete product using a 3D scanning device, and obtain the overall scan point cloud of the product including the processing reference of exposed parts; S7. For each part, register the single-part scan point cloud obtained in S2 with the one obtained in S5. The obtained overall scanned point cloud of the product is registered to determine the rigid body displacement transformation parameters required to transform the scanned point cloud of the single part from the part coordinate system to the overall product coordinate system; S7. For the finite element model of each part, the nodes on its outer surface are classified into face nodes and edge nodes. Among them, the nodes located on the face features of the finite element model but not at the edge features are face nodes, and the nodes located on the edge features of the finite element model are edge nodes; S8. For each face node, a spherical search area is defined with the face node as the center. All projection points located in the spherical search area are found. The arithmetic mean of the deviation vectors corresponding to these projection points is calculated, and the arithmetic mean is used as the face offset vector of the face node; S9. For each edge node, establish a tubular search domain with the edge feature of the finite element model containing the edge node as the axis, find the projection points located in the tubular search domain, and generate the scan points corresponding to each projection point. Based on the found scan points, use the least squares method to fit the constituent surfaces on both sides of the edge feature to obtain two parameterized fitted surfaces. Calculate the spatial intersection of these two parameterized fitted surfaces, calculate the shortest distance vector from the edge node to the spatial intersection, and use the shortest distance vector as the edge offset vector of the edge node. S10. Based on the surface offset vector of each face node and the edge offset vector of each edge node, perform position offset on all outer surface nodes of the finite element model of each part.S11. Based on the rigid body displacement transformation parameters of each part, perform the same rigid body displacement transformation on both the precisely aligned reference CAD model of the part and the finite element model of the part with completed node offsets. After the displacement transformation, package the set of reference CAD models of all parts located in the overall product coordinate system, along with the projection point coordinates and deviation vector data associated with each reference CAD model, and output them together as a digital prototype of the product. Simultaneously, package the set of finite element models of all parts after the displacement transformation and output them as a finite element analysis model of the product.

[0022] In another possible embodiment, CAD preprocessing is first performed, inputting the CAD models of each part of the product one by one, removing process features such as chamfers, fillets, and pin holes that are not of concern to geometric measurement and finite element analysis, marking the machining datum of each part, and then discretizing the CAD model of the part into a finite element model based on the accuracy requirements of finite element analysis and the available computing resources; before product assembly, each part is scanned one by one using a 3D scanning device to obtain the scanned point cloud of each part; then, single-part registration is performed, first uniformly extracting spatial point clouds on the datum surface of the CAD model, and then using the ISS algorithm to register the scanned point clouds of each part with the extracted point clouds, so that the scanned point clouds are after rigid body displacement. After roughly aligning with the CAD model, the least squares method is used to register the machining datum and scanned point cloud of the CAD model. Keeping the CAD model fixed, the scanned point cloud is precisely aligned with the CAD model after rigid body displacement. Next, projection and deviation extraction are performed. The coordinates of each scanned point to its nearest point on the precisely aligned CAD model surface in the part's CAD coordinate system are calculated and exported, and this point is defined as the projection point. Simultaneously, the vector pointing from each projection point to its corresponding scanned point is calculated and exported, and this vector is defined as the deviation vector. After completing the above steps, the physical product is assembled, and the exposed machining datum of the product is scanned to form the product's scanned point cloud. Then, product registration is performed, first using I... The SS algorithm registers the individual scanned point clouds of each part with the scanned point cloud of the assembled product, keeping the product scanned point cloud fixed so that the individual scanned point clouds are roughly aligned with the product scanned point cloud after rigid body displacement. Then, the ICP algorithm is used for precise registration to ensure complete alignment between the individual scanned point clouds and the product scanned point cloud. Subsequently, the mesh nodes on the outer surface of the finite element model of each part are classified: nodes located on the surface features of the finite element model are defined as surface nodes, and nodes located on the edge features are defined as edge nodes. During the averaging deviation operation, the CAD model of each part after individual registration is replaced in situ with the finite element model, and the projection within the spherical neighborhood of each surface node is searched. For shadow points, calculate and derive the average value of the deviation vectors corresponding to all projection points in the neighborhood until all face nodes obtain the corresponding average deviation vectors; in the intersection line calculation stage, determine the composition surface type of the edge where each edge node is located based on the CAD model, establish a tubular search domain with the edge as the axis, fit the parameterized composition surface using the least squares method based on the scan points corresponding to the projection points in the search domain, and then calculate and derive the intersection line of the two parameterized fitted composition surfaces until each edge of the finite element model obtains a corresponding parameterized intersection line; when offsetting nodes, offset all face nodes according to their corresponding average deviation vectors, and offset edge nodes along the shortest distance to the parameterized intersection line corresponding to the edge where the node is located.Finally, model repositioning is performed. Based on the rigid body displacement experienced by the single-part scanned point cloud during the product registration stage, equal rigid body displacements are applied to its CAD model and the finite element model after offsetting surface nodes. This ensures precise alignment between the part's CAD model and finite element model and the product scanned data. The coordinates of the part's projection points in the product coordinate system after repositioning and their deviation vectors are calculated and recorded. The repositioned CAD model set, the corresponding projection point coordinates and deviation vectors of each CAD model are packaged and output as a digital prototype. Simultaneously, the repositioned finite element model set is packaged and output as the product's finite element model.

[0023] By adopting a phased registration logic of "single-part registration - product registration," the system ensures precise alignment between individual parts and the CAD model while achieving accurate repositioning of multiple parts in the product coordinate system. This effectively integrates the assembly information of the actual product, solving the problem of lack of assembly realism in multi-part reconstruction. By differentiating between face nodes and edge nodes for differential deviation processing, the system avoids local distortion caused by global deviation processing, significantly improving the fidelity of the reconstructed model to the local geometric features of the product, making the digital model closer to the actual product. The system simultaneously outputs a composite digital prototype containing the CAD model, projection points, and deviation vectors, as well as an optimized finite element model. This not only meets the needs of digital measurement of product geometric dimensions but also provides a high-fidelity carrier for product performance analysis, achieving "one-time reconstruction, multiple applications." The system provides a complete conversion path from scanned point clouds to finite element models, and the generated finite element models incorporate product processing and assembly deviation information, significantly improving the accuracy of subsequent finite element analysis. The system eliminates the need for dedicated positioning parts, completing multi-part registration through the product's exposed processing datum, broadening the applicability of the technology and reducing the pre-construction preparation costs.

[0024] In existing technologies, the coarse registration method for single-piece registration is singular and fixed, relying solely on a specific algorithm to achieve approximate alignment between the scanned point cloud and the CAD model. When encountering situations such as insufficient accuracy of the scanning equipment, complex geometric structure of the part, or a lot of noise in the scanned point cloud, a single algorithm is difficult to achieve the ideal initial alignment effect. This not only increases the number of iterations and computational workload of subsequent fine registration, but may also lead to difficulties in fine registration convergence, affecting the overall registration accuracy and efficiency.

[0025] In the embodiments of this application, S3 specifically includes: S31, uniformly sampling and generating a reference reference point cloud from the machining reference surface of the reference CAD model of the part, and using a point cloud registration algorithm based on three-dimensional feature descriptors to initially align the single-piece scan point cloud of the part with the reference reference point cloud to obtain a preliminary registration result; S32, using the geometric features of the machining reference marked in the reference CAD model of the part as constraints, using a least squares optimization algorithm to iteratively optimize the preliminary registration result so that the single-piece scan point cloud and the reference CAD model achieve a precise alignment state.

[0026] In another possible embodiment, during the single-part registration stage, when it is necessary to achieve approximate alignment between the scanned point cloud and the CAD model, if the coarse registration method based on the ISS algorithm cannot adapt to the current scenario, other coarse registration algorithms based on key points and descriptors can be selected according to the quality of the scanned point cloud, the geometric complexity of the part, and the accuracy parameters of the scanning equipment. The initial alignment between the scanned point cloud and the extracted point cloud can be achieved through the feature extraction and matching logic of the algorithm itself. If the scanned point cloud has severe noise or the part structure is extremely special, causing various automatic coarse registration algorithms to fail to achieve the desired effect, the rigid body displacement of the scanned point cloud can be adjusted manually to adjust the scanned point cloud and the CAD model to a state where they visually roughly overlap. After the initial alignment is completed, fine registration is performed according to the least squares method to ensure the precise alignment between the scanned point cloud and the CAD model.

[0027] This invention breaks through the limitation of a single coarse registration method, providing diverse options that can adapt to different scanning equipment precision, part complexity, and scan point cloud quality scenarios, significantly improving the applicability and flexibility of the technology. By selecting the optimal coarse registration method for different scenarios, high alignment accuracy can be guaranteed in the initial alignment stage, reducing the optimization difficulty and computational workload of subsequent fine registration, shortening the overall registration time, and improving reconstruction efficiency. The supplementary manual control method enables initial alignment to be completed even in extreme scenarios where the scan point cloud quality is extremely poor or the part structure is special and the algorithm cannot work effectively, ensuring the continuity of the reconstruction process and avoiding the entire project from stalling due to registration problems.

[0028] The existing coarse registration method in the product registration stage is relatively fixed, relying on a single algorithm to achieve the initial alignment of the single part scan point cloud with the product scan point cloud. However, assembled products often have problems such as part occlusion and uneven distribution of scan point clouds. When faced with these complex situations, the single algorithm is prone to feature matching errors, resulting in low initial alignment accuracy. This increases the convergence difficulty of the ICP algorithm for fine registration, prolongs the registration time, and may even affect the final product-level registration accuracy.

[0029] In the embodiments of this application, S6 specifically includes: S61, using a point cloud registration algorithm based on three-dimensional feature descriptors to initially align the single-piece scan point cloud with the overall product scan point cloud to obtain a preliminary alignment result; S62, using the ICP algorithm to iteratively optimize the preliminary alignment result to accurately align the single-piece scan point cloud with the overall product scan point cloud, and calculate the rigid body displacement change parameters.

[0030] In another possible embodiment, during the product registration stage, when it is necessary to achieve approximate alignment between the single-piece scanned point cloud and the product scanned point cloud, if the ISS algorithm cannot handle issues such as occlusion or uneven point cloud distribution in the product scanned point cloud, other coarse registration algorithms based on key points and descriptors can be selected. Through the precise extraction and matching of feature points by the algorithm, the initial alignment between the single-piece scanned point cloud and the product scanned point cloud can be achieved. If the automatic coarse registration algorithm is difficult to function due to severe occlusion or excessive noise in the product scanned point cloud, the rigid body displacement of the single-piece scanned point cloud can be adjusted manually to make the single-piece scanned point cloud and the product scanned point cloud visually roughly aligned. After the initial alignment is completed, the ICP algorithm is used for fine registration. Through iterative optimization, the two are precisely aligned, and the corresponding rigid body displacement is recorded.

[0031] By expanding the range of product-level coarse registration methods, the coarse registration strategy can be flexibly adjusted according to the scanned point cloud state of the assembled product (such as occlusion degree and point cloud density), improving the success rate and accuracy of coarse registration in complex assembly scenarios. The optimized coarse registration effect can reduce the number of iterations in the fine registration process of the ICP algorithm, accelerate the convergence speed of fine registration, shorten the overall product registration time, and improve reconstruction efficiency. The addition of manual control mode enables initial alignment to be completed even in extreme cases where the product scanned point cloud is severely occluded or the point cloud quality is extremely poor, ensuring the smooth progress of the multi-part product registration process and avoiding the inability to build a complete digital image of the product due to registration failure.

[0032] Traditional digital prototypes rely solely on the design parameters of the CAD model for dimensional measurement, completely ignoring the geometric deviations generated during product manufacturing. This results in significant discrepancies between the measurement results and the actual product dimensions, failing to reflect the product's true geometric state and consequently affecting the accuracy of subsequent work such as process adjustments and assembly optimization based on the measurement results.

[0033] In embodiments of this application, the method further includes a step of performing digital measurement of geometric dimensions based on a digital prototype, specifically including: S121, determining the feature to be measured and at least two measurement endpoints on the reference CAD model in the digital prototype; S122, reading the theoretical values ​​of the design parameters of the geometric feature on the reference CAD model from the digital prototype; S123, searching for associated projection points in the neighborhood space for each measurement endpoint; S124, calculating the average value of the deviation vectors corresponding to all projection points associated with each measurement endpoint; S125, projecting the average value of each deviation vector onto the measurement direction of the geometric feature to obtain the dimension deviation components corresponding to each endpoint; S126, subtracting the sum of the dimension deviation components corresponding to all endpoints from the theoretical value of the design parameters to obtain the actual digital measurement value of the geometric feature.

[0034] In another possible embodiment, when measuring the product dimension of interest in a composite digital prototype, the theoretical value of the design parameter corresponding to the dimension of interest is first read directly from the CAD model contained in the digital prototype. Then, the two measurement endpoints of the dimension are determined, and the corresponding projection points are searched in the vicinity of each measurement endpoint. The deviation vectors corresponding to these projection points are averaged to obtain the average value of the deviation vectors near each measurement endpoint. Then, the average value is decomposed along the measurement direction to obtain the deviation component of each endpoint in the measurement direction. Finally, the theoretical value of the design parameter of the CAD model and the deviation components of the two endpoints are superimposed according to the corresponding logic to obtain the actual digital measurement result of the dimension of interest.

[0035] By organically combining the design values ​​of the CAD model with the actual scanning deviation vector, the measurement results can truly reflect the geometric dimensions of the actual product, completely solving the shortcomings of traditional measurement that relies solely on theoretical values, and significantly improving the accuracy and reliability of digital measurement. The measurement process does not require physical measuring tools and can be completed using only a digital prototype. Furthermore, the averaging of the deviation vector can reduce the impact of single-point deviations on the measurement results, further improving measurement accuracy. The measurement logic is simple and clear, and can be automated through programming, reducing manual operation costs and measurement errors. At the same time, it can quickly obtain measurement results for any dimension of interest, improving measurement efficiency and providing timely data support for production process adjustments.

[0036] Existing technologies, when processing the offset of edge nodes in finite element models, do not consider the special case where some edge nodes belong to multiple edges simultaneously. They only perform offset processing based on the intersection line of a single edge, resulting in the offset position of such nodes not matching the actual geometric features of the product. This affects the overall fidelity of the finite element model and reduces the accuracy of subsequent finite element analysis results.

[0037] In the embodiments of this application, in S9, if a node is identified as an edge node belonging to three different model edge features simultaneously, the following operations are performed on the node: S91, for each edge feature to which it belongs, calculate the shortest distance vector from the node to the intersection line of the parameterized fitted surface space corresponding to the edge feature, and obtain three shortest distance vectors in total; S92, calculate the average value of the three shortest distance vectors; S93, use the average value of the vectors as the final edge offset vector of the node as the position offset in S10.

[0038] In another possible embodiment, during the node offset stage, when it is determined that a node on one side belongs to all three sides, the shortest vector from the node to the parameterized intersection line corresponding to each of the three sides is calculated first to ensure that each shortest vector can accurately reflect the distance and direction from the node to the corresponding intersection line. Then, the three shortest vectors are averaged to obtain an average vector that integrates the influence of the geometric features of the three sides. Finally, the node on one side is offset according to the average vector so that the final position of the node can fit the actual geometric state of the product.

[0039] A special handling scheme is proposed for the special case of "multi-edge shared nodes" to ensure that the offset position of such nodes can fit the actual geometric state of the product, thus solving the model distortion problem caused by traditional processing methods. By calculating the shortest vector from the node to each intersecting line and taking the average value, the influence of the geometric characteristics of each edge can be taken into account, making the node offset more reasonable and accurate, and further improving the fidelity of the finite element model. The optimized edge node offset processing logic enables the finite element model to more accurately reflect the structural characteristics of the product, providing a more reliable model foundation for subsequent finite element analysis and improving the credibility of the analysis results.

[0040] like Figure 2 The present application also provides a product reconstruction system based on 3D scanning and CAD surface deviation distribution. The system includes: a preprocessing module configured to execute S1 to generate a reference CAD model and perform finite element mesh generation; a scanning module configured to execute S2 and S5 to acquire single-piece scan point clouds of parts and overall scan point clouds of the product; a single-piece registration module configured to execute S3 to achieve high-precision alignment between the single-piece scan point cloud and the reference CAD model; a deviation calculation module configured to execute S4 to calculate projection points and deviation vectors; a product registration module configured to execute S6 to achieve coordinate unification from the single-piece point cloud to the product coordinate system; a node offset module configured to execute S7, S8, S9 and S10 to complete the classification of finite element model nodes and position correction based on deviation vectors; and an output module configured to execute S11 to reposition the reference CAD model and the initial finite element model, and integrate and output a digital prototype and a finite element analysis model.

[0041] In the embodiments of this application, the single-piece registration module includes: a first coarse registration unit configured to execute S31, extract a reference reference point cloud from the reference CAD model, and perform initial alignment using an algorithm based on three-dimensional feature descriptors; and a first fine registration unit configured to execute S32, perform least squares optimization based on machining reference constraints.

[0042] In the embodiments of this application, the product registration module includes: a second coarse registration unit, configured to execute S61, using an algorithm based on three-dimensional feature descriptors to achieve initial alignment between the single-piece point cloud and the overall point cloud; and a second fine registration unit, configured to execute S62, using the ICP algorithm to achieve precise alignment and calculate displacement parameters.

[0043] In the embodiments of this application, the node offset module includes: a node classification unit configured to execute S7, classifying the nodes on the outer surface of the finite element model into face nodes and edge nodes; a face node processing unit configured to execute S8, calculating the average spherical neighborhood deviation of each face node to generate a face offset vector; an edge node processing unit configured to execute S9, fitting the edge feature surface and calculating the intersection line to generate an edge offset vector; and an offset execution unit configured to execute S10, performing position offset on the nodes according to the face offset vector and the edge offset vector.

[0044] In an embodiment of this application, the deviation calculation module includes: a projection point calculation unit configured to calculate the nearest point from the scan point to the surface of the reference CAD model to obtain the projection point; and a deviation vector generation unit configured to generate a deviation vector pointing from the projection point to the corresponding scan point.

[0045] The optional embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the embodiments of the present invention are not limited to the specific details in the above embodiments. Within the scope of the technical concept of the embodiments of the present invention, various simple modifications can be made to the technical solutions of the embodiments of the present invention, and these simple modifications all fall within the protection scope of the embodiments of the present invention.

[0046] It should also be noted that the various specific technical features described in the above embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, the embodiments of the present invention will not describe the various possible combinations separately.

[0047] Furthermore, various different implementations of the present invention can be combined arbitrarily, as long as they do not violate the spirit of the present invention, they should also be regarded as the content disclosed in the present invention.

Claims

1. A product reconstruction method based on 3D scanning and CAD surface deviation distribution, characterized in that, Includes the following steps: S1. Obtain the design geometric CAD model of each part, simplify the design geometric CAD model, remove chamfer, fillet and pin hole process features, and mark the machining datum on the simplified CAD model to obtain the datum CAD model for registration and deviation analysis. At the same time, discretize the datum CAD model to generate the corresponding finite element model. S2. Before the physical assembly of the product, use a 3D scanning device to scan each part one by one to obtain the single-piece scan point cloud of each part. S3. For each part, register the single-part scan point cloud obtained in S2 with the reference CAD model of the part obtained in S1, so that the single-part scan point cloud and the reference CAD model are accurately aligned. S4. For each scan point in the single-piece scan point cloud of each part, calculate the nearest point from the scan point to the surface of the corresponding part's reference CAD model, define the nearest point as the projection point, generate a vector from the projection point to its corresponding scan point, and define the vector as the deviation vector corresponding to the scan point. S5. Assemble each part into a complete product, and use a 3D scanning device to scan the complete product to obtain the overall scan point cloud of the product, including the processing reference of the exposed parts. S6. For each part, register the single-part scan point cloud obtained in S2 with the overall product scan point cloud obtained in S5, and determine the rigid body displacement transformation parameters required to transform the single-part scan point cloud from the part coordinate system to the overall product coordinate system. S7. For the finite element model of each part, classify the nodes on its outer surface into face nodes and edge nodes. Among them, the nodes located on the face features of the finite element model but not on the edge features are face nodes, and the nodes located on the edge features of the finite element model are edge nodes. S8. For each face node, define a spherical search area centered on the face node, find all projection points within the spherical search area, calculate the arithmetic mean of the deviation vectors corresponding to each of these projection points, and use the arithmetic mean as the face offset vector of the face node. S9. For each edge node, a tubular search domain is established with the edge feature of the finite element model where the edge node is located as the axis. The projection points located in the tubular search domain are found, and the scan points corresponding to each projection point are generated. Based on the scan points found, the least squares method is used to fit the constituent surfaces on both sides of the edge feature respectively to obtain two parameterized fitting surfaces. The spatial intersection of the two parameterized fitting surfaces is calculated, and the shortest distance vector from the edge node to the spatial intersection is calculated. The shortest distance vector is used as the edge offset vector of the edge node. S10. Based on the face offset vector of each face node and the edge offset vector of each edge node, perform position offset on all outer surface nodes of the finite element model of each part. S11. Based on the rigid body displacement transformation parameters of each part, perform the same rigid body displacement transformation on the precisely aligned reference CAD model of the part and the finite element model of the part with completed node offset. After the displacement transformation, the set of reference CAD models of all parts located in the overall coordinate system of the product, as well as the projection point coordinates and deviation vector data associated with each reference CAD model, are packaged and output as the digital prototype of the product. At the same time, the set of finite element models of all parts after the displacement transformation is packaged and output as the finite element analysis model of the product.

2. The product reconstruction method based on 3D scanning and CAD surface deviation distribution according to claim 1, characterized in that, S3 specifically includes: S31. From the machining reference surface of the reference CAD model of the part, uniformly sample to generate a reference reference point cloud. Use a point cloud registration algorithm based on three-dimensional feature descriptors to initially align the single-piece scan point cloud of the part with the reference reference point cloud to obtain a preliminary registration result. S32. Using the geometric features of the machining datum marked in the reference CAD model of the part as constraints, the least squares optimization algorithm is used to iteratively optimize the preliminary registration results so that the single-part scan point cloud and the reference CAD model are accurately aligned.

3. The product reconstruction method based on 3D scanning and CAD surface deviation distribution according to claim 1, characterized in that, S6 specifically includes: S61. Using a point cloud registration algorithm based on three-dimensional feature descriptors, the single-piece scanned point cloud is initially aligned with the overall product scanned point cloud to obtain preliminary alignment results. S62. The ICP algorithm is used to iteratively optimize the preliminary alignment results, so that the single-piece scan point cloud is accurately aligned with the overall scan point cloud of the product, and the rigid body displacement change parameters are calculated.

4. The product reconstruction method based on 3D scanning and CAD surface deviation distribution according to claim 1, characterized in that, The method also includes a step of digitally measuring geometric dimensions based on a digital prototype, specifically including: S121. Determine the characteristics of the set to be measured and at least two measurement endpoints on the reference CAD model in the digital prototype; S122. Read the theoretical values ​​of the design parameters of the geometric features on the reference CAD model from the digital prototype; S123. For each measurement endpoint, search for associated projection points in its neighborhood space; S124. Calculate the average value of the deviation vector corresponding to all projection points associated with each measurement endpoint; S125. Project the average value of each deviation vector onto the measurement direction of the geometric feature to obtain the dimensional deviation components corresponding to each endpoint. S126. Subtract the sum of the dimensional deviation components corresponding to all endpoints from the theoretical value of the design parameters to obtain the actual dimensional measurement value of the geometric feature.

5. The product reconstruction method based on 3D scanning and CAD surface deviation distribution according to claim 1, characterized in that, In S9, if a node is identified as an edge node belonging to three different model edge features simultaneously, then the following operations are performed for that node: S91. For each edge feature to which it belongs, calculate the shortest distance vector from the node to the parameterized fitted surface space intersection line corresponding to the edge feature, and obtain three shortest distance vectors in total. S92. Calculate the average of the three shortest distance vectors; S93. Use the average value of the vector as the final edge offset vector of the node as the position offset in S10.

6. A product reconstruction system based on 3D scanning and CAD surface deviation distribution, used to implement the product reconstruction method based on 3D scanning and CAD surface deviation distribution as described in any one of claims 1-5, characterized in that, The system includes: The preprocessing module is configured to execute S1 to complete the generation of the baseline CAD model and the generation of the finite element mesh. The scanning module is configured to execute S2 and S5 to acquire the single-piece scan point cloud of the part and the overall scan point cloud of the product. The single-piece registration module is configured to execute S3 to achieve high-precision alignment between the scanned point cloud of a single piece and the benchmark CAD model. The deviation calculation module is configured to execute S4 to calculate the projection points and deviation vectors. The product registration module is configured to execute S6 to achieve coordinate unification from the point cloud of a single item to the product coordinate system; The node offset module is configured to execute S7, S8, S9 and S10 to complete the classification of nodes in the finite element model and the position correction based on the deviation vector. The output module is configured to execute S11 to reposition the baseline CAD model and the initial finite element model, and integrate and output the digital prototype and finite element analysis model.

7. The product reconstruction system based on 3D scanning and CAD surface deviation distribution according to claim 6, characterized in that, The single-piece registration module includes: The first coarse registration unit is configured to execute S31, extract the reference reference point cloud from the reference CAD model, and perform initial alignment using an algorithm based on 3D feature descriptors; The first fine registration unit is configured to execute S32, which performs least squares optimization based on machining datum constraints.

8. The product reconstruction system based on 3D scanning and CAD surface deviation distribution according to claim 6, characterized in that, The product registration module includes: The second coarse registration unit is configured to execute S61, which uses an algorithm based on three-dimensional feature descriptors to achieve initial alignment between the single point cloud and the overall point cloud. The second fine registration unit is configured to execute S62, which uses the ICP algorithm to achieve accurate alignment and solve displacement parameters.

9. The product reconstruction system based on 3D scanning and CAD surface deviation distribution according to claim 6, characterized in that, The node offset module includes: The node classification unit is configured to execute S7 to classify the nodes on the outer surface of the finite element model into face nodes and edge nodes. The face node processing unit is configured to execute S8 to calculate the mean spherical neighborhood deviation of each face node to generate a face offset vector. The edge node processing unit is configured to execute S9, fit the edge feature surface and calculate the intersection to generate the edge offset vector; The offset execution unit is configured to execute S10, which performs position offset on the node based on the face offset vector and the edge offset vector.

10. The product reconstruction system based on 3D scanning and CAD surface deviation distribution according to claim 6, characterized in that, The deviation calculation module includes: The projection point calculation unit is configured to calculate the nearest point from the scan point to the surface of the reference CAD model to obtain the projection point; The deviation vector generation unit is configured to generate a deviation vector pointing from the projection point to the corresponding scan point.

Images