A method and system for controlling production of parts by fusing digital twins

Abstract

This invention relates to the field of component production control technology, specifically to a component production control method and system integrating digital twins. The system achieves precise positioning and distribution of components by dividing a region image into grids, eliminating the problem of repeated display across grids. In subsequent staged evaluation processes, a staged evaluation route for the components is constructed, and the evaluation process is simulated using digital twin technology to obtain the corresponding component marker parameters at different stages. These marker parameters are then used to adjust the evaluation route for the next stage, thereby obtaining the evaluation results for each stage. This avoids the problem of repeated parameter collection during the evaluation process, ensuring a closed loop in component production control and improving anomaly response efficiency. Finally, the current production status of the component is determined based on the evaluation results of each stage, and the production equipment is adaptively adjusted accordingly.

Description

Technical Field

[0001] This invention relates to the field of parts production control technology, and more specifically, to a parts production control method and system that integrates digital twins. Background Technology

[0002] In the field of intelligent manufacturing, quality control and process optimization in parts production highly depend on real-time evaluation and feedback of production results. Existing mainstream production control methods typically follow a post-feedback paradigm of "perception-evaluation-control," which involves collecting post-production status data of parts (such as stacking posture, appearance defects, dimensional tolerances, etc.) through devices such as vision sensors and laser measurements, feeding this status data back to the central control system, and then adaptively adjusting the process parameters of the production line equipment (such as robotic arm trajectory, machine tool feed rate, spraying pressure, etc.) to improve quality in subsequent production.

[0003] However, as product customization deepens and production cycles accelerate, the inherent limitations of this traditional paradigm become increasingly apparent. The core issues lie in the serialization of the evaluation process and the siloing of data applications, specifically manifested as follows:

[0004] First, multi-step sequential evaluation leads to efficiency bottlenecks: existing evaluation processes are typically broken down into multiple single, sequentially executed subtasks. For example, 2D vision performs initial positioning and counting, then triggers a 3D sensor to perform a detailed scan of a specific area to obtain posture, and finally the data is transferred to another independent quality analysis module for defect determination. These steps proceed linearly according to a fixed process, with each step waiting for the previous step to complete and transmit data. During this time, a large amount of computing resources are idle, and delays in any link accumulate, resulting in an excessively long overall response cycle from status acquisition to control command generation, which cannot meet the real-time requirements of high-cycle dynamic production.

[0005] Secondly, data silos hinder information fusion and in-depth utilization: Data generated in different evaluation steps (such as texture information in 2D images, spatial geometric information in 3D point clouds, and semantic information in quality inspection results) are often processed by independent subsystems, lacking a unified data model for correlation and representation. For example, the 3D posture data of a part and the scratch detection results on its surface are separated at the data level, making it difficult for the control system to directly obtain deep correlation knowledge such as "what kind of defects are more likely to occur at a specific location under what posture." This results in massive amounts of high-value process data being used only to complete a single judgment, without uncovering the underlying process-quality correlation patterns, leading to low data utilization.

[0006] Therefore, there is an urgent need for a component production control method and system that can break down barriers in the evaluation process, achieve seamless data flow throughout the entire process, and integrate digital twins. Summary of the Invention

[0007] The purpose of this invention is to provide a component production control method and system that integrates digital twins to solve the problems mentioned in the background art.

[0008] To achieve the above objectives, one objective of this invention is to provide a component production control method integrating digital twins, comprising the following steps:

[0009] S1. Configure monitoring equipment and monitoring routes to collect images of the area where the parts are placed.

[0010] S2. Perform grid-based segmentation on the regional image and clearly locate and distribute each component;

[0011] S3. Construct a phased evaluation route for components, use digital twin technology to simulate the evaluation process, and obtain the corresponding component marker parameters at different stages;

[0012] S4. Extract the indicator parameters for each stage, use the indicator parameters to adjust the evaluation route for the next stage, adjust the monitoring equipment according to the evaluation route, and obtain the evaluation results for each stage.

[0013] S5. Based on the evaluation results of each stage, determine the current production status of the parts and make adaptive adjustments to the production equipment in combination with the production status.

[0014] Preferably, the monitoring device configured in S1 is a 3D structured light camera.

[0015] Preferably, the method for gridding the region image in S2 includes the following steps:

[0016] S2.1 Determine the size of the unit grid based on the size of the area where the parts are placed;

[0017] S2.2 Map the image of the region after grid division to a two-dimensional coordinate system, and locate the coordinates of each unit grid using the two-dimensional coordinate system;

[0018] S2.3. Use the instance segmentation model to obtain the mask of each component, and use the binary pixel matrix in the instance segmentation model to extract the accurate outer contour of the component at different positions;

[0019] S2.4 Determine the distribution position of the parts in the two-dimensional coordinate system based on the outer contour of the parts.

[0020] Preferably, the method for clearly positioning and distributing each component in step S2 includes the following steps:

[0021] S2.10. Traverse all unit meshes and calculate the intersection area between the current component's outer contour and the current unit mesh;

[0022] S2.11. Based on the intersection area, locate the component in a single region and extract all unit meshes that intersect with the outer contour of the current component;

[0023] S2.12 Select the unit mesh with the largest intersection area, assign the component to the unit mesh, and mark the unit mesh as the sovereign mesh. All other unit meshes that intersect with the outer contour of the current component are marked as secondary weighted meshes.

[0024] S2.13. Traverse all unit grids. For each unit grid:

[0025] Only draw those components that are assigned to the current unit mesh;

[0026] Hide components that overlap with this unit's mesh but whose sovereignty belongs to another unit's mesh.

[0027] Preferably, the component phase evaluation route constructed in S3 includes component placement distribution, the number of components per unit grid, and component quality monitoring.

[0028] Preferably, the indicator parameter for the placement and distribution of the components is the coordinate marker and positional relationship of each component stack, and the indicator parameter for the number of components per unit grid is the difference rate.

[0029] Preferably, the evaluation method for the placement and distribution of the components includes the following steps:

[0030] S3.1, Mark the instance IDs of all components in different unit grids;

[0031] S3.2 Extract the 3D point cloud of the corresponding part in each unit grid from the 3D point cloud corresponding to the original part mask according to the instance ID, and mark it as a point set. Each point set represents a part in the unit grid.

[0032] S3.3. Merge the point sets of all components within the current unit grid to obtain the total point cloud set representing all components within the unit grid;

[0033] S3.4 Construct a spatial rectangular coordinate system and use a random sampling consistency plane fitting algorithm to segment the bearing plane from the total point cloud set;

[0034] S3.5 Calculate the vertical distance from the point with the largest Z coordinate value in the total point cloud set to the fitted bearing plane, and mark it as the stacking height of the component stack;

[0035] S3.6. For all 3D point clouds within the current unit grid, use a clustering algorithm based on Euclidean distance to process them, grouping spatially adjacent points into the same cluster and distant points into different clusters.

[0036] S3.7 Determine the number of clusters obtained by the clustering algorithm as the number of clustered parts in the current unit grid;

[0037] S3.8 Count the number of piled parts and the stacking height of the parts in each unit grid, and summarize and mark the parts placement distribution of the corresponding unit grid.

[0038] Preferably, the method for evaluating the number of components per unit grid includes the following steps:

[0039] S3.10 Count the number of individually placed components in a unit grid based on the component mask;

[0040] S3.11. Taking into account the diversity of component masks, count the individual components under different placement postures.

[0041] S3.12. Obtain the number of individual components with different placement orientations in a unit grid and mark them as the difference rate;

[0042] S3.13. When it is determined that there is a stack of parts based on the difference rate, the coordinate marks and positional relationships of each part stack are combined to construct the second-round image acquisition route of the 3D structured light camera, and targeted image acquisition is carried out on the part stacks of different unit grids.

[0043] S3.14. Identify and count the components based on the component masks in the stacked state, and count the number of components in each unit grid.

[0044] Preferably, the evaluation method for monitoring the quality of the components includes the following steps:

[0045] S3.20, Obtain the difference rate of each unit grid. ;

[0046] S3.21, Establish the difference rate threshold And perform difference rate calculation on each unit grid. Comparison:

[0047] Difference rate Below the difference rate threshold Remove units from the grid;

[0048] The difference rate is not lower than the difference rate threshold. The unit grid is labeled as the evaluation unit grid;

[0049] S3.22, Based on the difference rate The size is used to sort the grids of each evaluation unit in order of priority;

[0050] S3.23. Collect sample images of each evaluation unit grid in sequence according to the order of priority, and obtain the corresponding number of sample parts;

[0051] S3.24. Compare the component images with those in the standard database to assess the component anomaly rate in the current placement area.

[0052] The second objective of this invention is to provide a system for implementing a component production control method that integrates digital twins, including an equipment end, a communication end, and a processing end;

[0053] The device is equipped with a 3D structured light camera to acquire images of the parts placement area. The control equipment adjusts the position of the 3D structured light camera to perform directional image acquisition. It is also equipped with production equipment, which can be adjusted to adapt to the current parts stacking state.

[0054] The communication terminal is used to establish a network channel between the device and the processing terminal. It is used to upload the regional images acquired by the device to the processing terminal, use the edge computing module on the processing terminal to process the regional images, construct a phased evaluation route for the components, perform digital twin simulation evaluation, extract the marker parameters of each stage, use the marker parameters to adjust the evaluation route of the next stage, obtain the evaluation results of each stage, and feed them back to the device through the network channel. The control device on the device adjusts the position and route planning of the 3D structured light camera according to the evaluation results and evaluation route, and executes the evaluation route and targeted acquisition work.

[0055] The component production control method and system integrating digital twins of this invention have the following beneficial effects: This invention acquires images of the component placement area by configuring monitoring equipment and routes, and then segments the obtained area images into a grid, achieving clear positioning and refined distribution management of the components. Based on this, a phased evaluation route based on digital twin technology is constructed. The evaluation process is simulated in virtual space to obtain the component's marker parameters at different stages. The extracted marker parameters are then used to dynamically adjust subsequent evaluation routes, guiding the monitoring equipment to adaptively adjust and obtain evaluation results at each stage. By integrating evaluation data from each stage, the current component production status is accurately located, and the production equipment is adaptively adjusted accordingly. This method, through the combination of grid segmentation and digital twin technology, significantly improves the accuracy and efficiency of component status monitoring, achieves dynamic optimization and closed-loop feedback in the evaluation process, effectively reduces physical losses and the need for human intervention in on-site inspection, provides reliable data support for the intelligent control of production equipment, and thus improves the overall automation level and operational stability of the production process. Attached Figure Description

[0056] Figure 1This is a flowchart of the component production control method integrating digital twins according to the present invention;

[0057] Figure 2 This is a flowchart of the gridded segmentation of a region image according to the present invention;

[0058] Figure 3 This is a schematic diagram of the component sovereignty grid division process of the present invention;

[0059] Figure 4 This is a demonstration diagram of a component sovereignty grid division according to the present invention;

[0060] Figure 5 This is a demonstration diagram of another component sovereignty grid division according to the present invention;

[0061] Figure 6 This is a flowchart for calculating the number of stacked components in this invention;

[0062] Figure 7 This is a flowchart illustrating the clear positioning and distribution of various components according to the present invention.

[0063] Figure 8 This is a flowchart of the evaluation method for the placement and distribution of components according to the present invention;

[0064] Figure 9 This is a flowchart of the method for evaluating the number of components per unit grid according to the present invention;

[0065] Figure 10 This is a flowchart of the component quality monitoring and evaluation method of the present invention;

[0066] Figure 11 This is a structural block diagram of the component production control system of the present invention. Detailed Implementation

[0067] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0068] Please see Figure 1 As shown, one of the objectives of this invention is to provide a component production control method integrating digital twins, comprising the following steps:

[0069] S1. Configure monitoring equipment and monitoring routes to collect images of the area where the parts are placed.

[0070] S2. Perform grid-based segmentation on the regional image and clearly locate and distribute each component;

[0071] S3. Construct a phased evaluation route for components, use digital twin technology to simulate the evaluation process, and obtain the corresponding component marker parameters at different stages;

[0072] S4. Extract the indicator parameters for each stage, use the indicator parameters to adjust the evaluation route for the next stage, adjust the monitoring equipment according to the evaluation route, and obtain the evaluation results for each stage.

[0073] S5. Based on the evaluation results of each stage, determine the current production status of the parts and make adaptive adjustments to the production equipment in combination with the production status.

[0074] The details are as follows:

[0075] Since the distribution and quality inspection results of components after production can reflect the current production status and serve as a reference for adjusting the next stage of component production, a production assessment is required after the first round of component production is completed.

[0076] This solution uses image monitoring to acquire images of the component placement area. Specifically, it involves configuring monitoring equipment and a monitoring route to acquire images of the component placement area. The monitoring equipment uses a 3D structured light camera to capture images of the component placement area and obtain an overall regional image.

[0077] After completing the regional image acquisition, the differences in the scattered state, quality, stacking state, and neatness of components at different locations within the regional images are also important indicators of production results. Therefore, to improve the subsequent evaluation effect, it is necessary to perform fine processing on the regional images, namely, to divide the regional images into grids and to clearly locate and distribute each component. Figures 2 to 3 As shown, the specific segmentation method is as follows:

[0078] First, determine the size of the unit grid. This depends on the specific scene and the average size of the parts. For example, if the placement area is 2000×2000 pixels and the average size of the parts is 50×50 pixels, then it is most appropriate to divide the unit grid into 200×200 pixel units. The unit grid should not be too small (otherwise almost every part will cross the grid) or too large (losing the meaning of division).

[0079] After completing the mesh generation, as follows Figure 7As shown, the entire region image is mapped to a two-dimensional coordinate system. Coordinate positioning of each unit grid is performed using this system. An instance segmentation model is used to obtain the masks of each component. The binary pixel matrix in the instance segmentation model is then used to extract the precise outer contours of the components at different locations. The distribution position of the components in the two-dimensional coordinate system is determined based on their outer contours. After gridding, some components span multiple unit grids, resulting in unclear distribution positions and potentially affecting subsequent evaluation. Therefore, to address this issue, precise positioning of each component is required. This is achieved by traversing all unit grids and calculating the intersection area between the current component's outer contour and the current unit grid. This solution uses a computational geometry library to calculate this intersection area. Both the current component's outer contour and the current unit grid are converted to Polygon objects. The `.intersection()` method is called to obtain the intersection region, and the `.area` property is used to obtain the intersection area. After obtaining the intersection area between each component's outer contour and its corresponding unit grid, area arbitration logic is used to locate the component in a single region: First, all unit grids intersecting with the current component are extracted, such as... Figure 4 As shown, the region image is divided into 3×3 unit grids. Each unit grid is located using coordinates in a two-dimensional coordinate system. The region image includes three components: component I, component II, and component III. The grid coordinates overlapping with component I are (2,2), (2,3), (3,3), and (3,2), respectively; the grid coordinates overlapping with component II are (3,1) and (3,2), respectively; and the grid coordinates overlapping with component III are (1,1) and (1,2), respectively. For each component, the intersection area of ​​all unit grids intersecting with that component is compared. The unit grid with the largest intersection area is selected, and the component is assigned to that unit grid, which is then marked as the sovereign grid of that component. The remaining intersecting unit grids are marked as sub-suppressive grids of that component. Figure 5 As shown, component I has the largest intersection area with the unit grid (2,3). This unit grid (2,3) will be marked as the sovereign grid of the current component I. The remaining unit grids (2,2), (3,3), and (3,2) are all marked as sub-sovereign grids. Only the sovereign grid has the right to display the current component. Traverse all unit grids. For each unit grid:

[0080] Only draw those components that are assigned to the current unit mesh;

[0081] Do not draw (hide) components that overlap with this unit grid but whose sovereignty belongs to other unit grids.

[0082] like Figure 5This results in a clearly defined view where each component is displayed fully or partially within its own sovereign grid, eliminating the problem of duplicate displays across grids.

[0083] Furthermore, after completing the component positioning work, the next step is to evaluate the component status. This solution constructs a phased evaluation route for the components and uses digital twin technology to simulate the evaluation process to obtain the corresponding component marker parameters at different stages. The phased evaluation route includes the component placement distribution, the number of components per unit grid, and component quality monitoring.

[0084] The evaluation process for the placement and distribution of components is as follows:

[0085] like Figure 6 as well as Figure 8 As shown, the component placement distribution is used to evaluate the placement status of components in different areas, including the stacking height of component piles and the stacking status (number of components in a pile).

[0086] First, after completing the mesh segmentation, each component is only displayed on the sovereign mesh. The instance IDs of all components in different unit meshes are marked. Based on the instance IDs, the 3D point clouds of the corresponding components in each unit mesh are extracted from the 3D point clouds corresponding to the original component masks and marked as point sets. Each point set represents one (or part: partially hidden) component in the unit mesh.

[0087] The stacking height is then calculated by merging the points of all components within the current unit grid to obtain a total point cloud representing all components within that unit grid. To calculate the relative stacking height, after constructing a spatial Cartesian coordinate system, the supporting plane (e.g., pallet surface, conveyor belt surface, or ground) needs to be identified. A Random Sample Consensus (RANSAC) plane fitting algorithm is used to segment the largest plane from the total point cloud; this largest plane is considered the supporting plane. The perpendicular distance from the point with the largest Z-coordinate value in the total point cloud to the fitted supporting plane is calculated and marked as the stacking height of the component stack within that unit grid. Following these steps, the stacking height of each component stack in each unit grid is calculated and marked, identifying the component stack with the highest stacking height in each unit grid.

[0088] After calculating the stacking height, the stacking state, i.e., the number of stacked components, is calculated. This invention uses clustering counting based on 3D point clouds. In actual stacking scenarios, instance segmentation may fail due to severe occlusion, for example, identifying multiple closely spaced components as one. In such cases, the spatial continuity of 3D point clouds needs to be used for supplementary judgment.

[0089] A clustering algorithm based on Euclidean distance is used on all 3D point clouds within the current unit grid to group points that are close to each other into the same cluster, while points that are far apart into different clusters, resulting in a certain number of clusters. The number of clusters obtained by the clustering algorithm is then determined as the number of clustered parts in the current unit grid.

[0090] Finally, by combining the spatial rectangular coordinate system, the highest point of each stack of parts is marked with coordinates, and the positional relationship between each stack of parts in different unit grids is obtained. The stacking height, the number of stacked parts, and the positional relationship between the stacked parts are bound to the spatial positional information to form a complete grid state unit. The grid state units of all unit grids constitute a quantitative digital twin description of the part placement area, that is, the part placement distribution.

[0091] After completing the component placement and distribution assessment, to facilitate the subsequent assessment of the number of components per unit grid, it is necessary to extract the marker parameters from the component placement and distribution assessment process for planning the next stage of the assessment route. In this scheme, the coordinate markers and positional relationships of each component stack during the component placement and distribution assessment process are used as marker parameters. For example... Figure 9 As shown, the biggest obstacle during the counting process is that some parts cannot be collected due to accumulation or shooting angle issues. To avoid these problems, when assessing the number of parts:

[0092] This solution employs a precise 3D point cloud analysis method. First, based on the instance segmentation mask, the number of individually placed components (unstacked components) within a unit grid is counted. Then, combining the component orientation (upright, sideways, tilted, etc.) reflected by the component mask, the number of individual components in different placement states is counted, and the number of individual components with differing placement postures within the unit grid is recorded as the difference rate. After completing the counting of individually placed components, a second-round image acquisition route for the 3D structured light camera is constructed based on the coordinate markers and positional relationships of each component stack. Targeted image acquisition is then performed on component stacks in different unit grids. Specifically, the 3D structured light camera's displacement path is located based on the coordinate markers of the component stacks, the coordinates of the highest point of each component stack are located for scanning and imaging, and the component masks in the stacked state are used for identification and counting. The number of components within each unit grid is then counted, completing the component counting process.

[0093] In order to ensure the diversity of assessments during the quality assessment process, such as Figure 10 As shown, the difference rate of each unit grid needs to be used as a flag parameter. During quality assessment, the difference rate of each unit grid is obtained. Establish a difference rate threshold and calculate the difference rate for each unit grid. Comparison;

[0094] Difference rate Below the difference rate threshold Unit grids that are not included in the quality assessment will be discarded.

[0095] Difference rate Not lower than the difference rate threshold The unit grid is labeled with the evaluation unit grid, and the difference rate of each evaluation unit grid is combined. The components are sorted by size, and sample images are collected for each evaluation unit grid in sequence according to the sorting order. The corresponding number of sample components are obtained (determined by the number of components in the current placement area, and the two are positively correlated). The components are compared with the component images in the standard database to evaluate the component anomaly rate in the current placement area.

[0096] Finally, the evaluation results of each evaluation stage are obtained, the current production status of the parts is located, the production equipment is adaptively adjusted based on the production status, abnormal production status is evaluated based on historical evaluation data, and the production equipment parameters are adjusted based on the current placement and distribution of parts, the number of parts per unit grid, and the results of parts quality monitoring.

[0097] This invention achieves clear positioning and distribution of components by dividing the regional image into grids, resulting in a clearly defined view and eliminating the problem of repeated display across grids. Simultaneously, during subsequent stage evaluations, a staged evaluation route for each component is constructed, and digital twin technology is used to simulate the evaluation process, obtaining the corresponding component marker parameters at different stages. These marker parameters are then used to adjust the evaluation route for the next stage, ensuring the effectiveness of the evaluation route, avoiding the problem of repeated parameter collection during the evaluation process, guaranteeing a closed-loop system for component production control, and improving anomaly response efficiency.

[0098] Please see Figure 11 The second objective of this invention is to provide a system for implementing the integrated digital twin component production control method, comprising an equipment end, a communication end, and a processing end;

[0099] The equipment is equipped with a 3D structured light camera to acquire images of the parts placement area, and the position of the 3D structured light camera is adjusted in conjunction with the control equipment to achieve directional image acquisition. It is also equipped with production equipment, and the parameters of the production equipment are adjusted to adapt to the current stacking state of the parts.

[0100] The communication terminal establishes a network channel (5G network and IoT) between the device and the processing terminal, uploading image data acquired by the device to the processing terminal. The processing terminal utilizes its edge computing module to process the image data. This edge computing module includes an FPGA (Field-Programmable Gate Array, primarily used for 3D vision data acquisition and preprocessing), a GPU (Graphics Processing Unit, primarily used for instance segmentation model inference, point cloud clustering, and height calculation), and a CPU (Central Processing Unit, primarily used for mesh generation and area arbitration). A phased evaluation route for each component is then constructed, and digital twin simulation evaluation is performed, extracting key parameters for each stage. These parameters are used to adjust the evaluation route for the next stage, obtaining the evaluation results for each stage, and finally feeding them back to the device via the network channel. The control equipment on the device terminal adjusts the position and route planning of the 3D structured light camera based on the evaluation results and the evaluation route, executing the evaluation route and performing targeted data acquisition.

[0101] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method of controlling production of a component by fusion of digital twins, characterized by, Includes the following steps: S1. Configure monitoring equipment and monitoring routes to collect images of the area where the parts are placed. S2. Perform grid-based segmentation on the regional image and clearly locate and distribute each component; The gridding of the region image includes: determining the size of the unit grid according to the size of the component placement area, obtaining the mask of each component using the instance segmentation model, and extracting the precise outer contour of the component at different positions using the binary pixel matrix in the instance segmentation model. The process of clearly defining the location and distribution of each component includes: S2.

10. Traverse all unit meshes and calculate the intersection area between the current component's outer contour and the current unit mesh; S2.

11. Based on the intersection area, locate the component in a single region and extract all unit meshes that intersect with the outer contour of the current component; S2.12 Select the unit mesh with the largest intersection area, assign the component to the unit mesh, and mark the unit mesh as the sovereign mesh. All other unit meshes that intersect with the outer contour of the current component are marked as secondary weighted meshes. S2.

13. Traverse all unit meshes. For each unit mesh: only draw those parts that are assigned to the current unit mesh, and hide parts that overlap with the current unit mesh but whose ownership belongs to other unit meshes. S3. Construct a phased evaluation route for components, use digital twin technology to simulate the evaluation process, and obtain the corresponding component marker parameters at different stages; The phased evaluation route for constructing components includes component placement and distribution, the number of components per unit grid, and component quality monitoring; The indicator parameters for the placement and distribution of the components are the coordinate markers and positional relationships of each component stack, and the indicator parameter for the number of components per unit grid is the difference rate. The difference rate is as follows: First, the number of individually placed parts in a unit grid is counted according to the part mask. Then, the orientation of the parts reflected by the part mask is combined with the number of individually placed parts in different placement states. The number of individually placed parts with different placement postures in the unit grid is obtained and marked as the difference rate. S4. Extract the indicator parameters for each stage, use the indicator parameters to adjust the evaluation route for the next stage, adjust the monitoring equipment according to the evaluation route, and obtain the evaluation results for each stage. S5. Based on the evaluation results of each stage, determine the current production status of the parts and make adaptive adjustments to the production equipment in combination with the production status.

2. The fused digital twin of part production control method of claim 1, wherein, The monitoring device configured in S1 is a 3D structured light camera.

3. The fused digital twin of part production control method of claim 2, wherein, The method for gridding the region image in S2 includes the following steps: S2.1 Map the image of the region after grid division to a two-dimensional coordinate system, and locate the coordinates of each unit grid using the two-dimensional coordinate system; S2.2 Determine the distribution position of the parts in the two-dimensional coordinate system based on the outer contour of the parts.

4. The fused digital twin of part production control method of claim 3, wherein, The evaluation method for the placement and distribution of the components includes the following steps: S3.1, Mark the instance IDs of all components in different unit grids; S3.2 Extract the 3D point cloud of the corresponding part in each unit grid from the 3D point cloud corresponding to the original part mask according to the instance ID, and mark it as a point set. Each point set represents a part in the unit grid. S3.

3. Merge the point sets of all components within the current unit grid to obtain the total point cloud set representing all components within the unit grid; S3.4 Construct a spatial rectangular coordinate system and use a random sampling consistency plane fitting algorithm to segment the bearing plane from the total point cloud set; S3.5 Calculate the vertical distance from the point with the largest Z coordinate value in the total point cloud set to the fitted bearing plane, and mark it as the stacking height of the component stack; S3.

6. For all 3D point clouds within the current unit grid, use a clustering algorithm based on Euclidean distance to process them, grouping spatially adjacent points into the same cluster and distant points into different clusters. S3.7 Determine the number of clusters obtained by the clustering algorithm as the number of clustered parts in the current unit grid; S3.8 Count the number of piled parts and the stacking height of the parts in each unit grid, and summarize and mark the parts placement distribution of the corresponding unit grid.

5. The fused digital twin of part production control method of claim 3, wherein, The method for evaluating the number of components per unit grid includes the following steps: S3.10 Count the number of individually placed components in a unit grid based on the component mask; S3.

11. Taking into account the diversity of component masks, count the individual components under different placement postures. S3.

12. Obtain the number of individual components with different placement orientations in a unit grid and mark them as the difference rate; S3.

13. When it is determined that there is a stack of parts based on the difference rate, the coordinate marks and positional relationships of each part stack are combined to construct the second-round image acquisition route of the 3D structured light camera, and targeted image acquisition is carried out on the part stacks of different unit grids. S3.

14. Identify and count the components based on the component masks in the stacked state, and count the number of components in each unit grid.

6. The fused digital twin of part production control method of claim 3, wherein, The evaluation method for monitoring the quality of the components includes the following steps: S3.20, acquiring the difference rate of each unit grid ; S3.21, Establishing a difference rate threshold and the difference rate of each unit grid Comparison: differential rate below the differential rate threshold units grid are culled; marking a unit grid having a difference rate not less than a difference rate threshold as an evaluation unit grid the unit grid having a difference rate not less than a difference rate threshold as an evaluation unit grid S3.22, according to the difference rate Size ranks the individual assessment unit grids in order. S3.

23. Collect sample images of each evaluation unit grid in sequence according to the order of priority, and obtain the corresponding number of sample parts; S3.

24. Compare the component images with those in the standard database to assess the component anomaly rate in the current placement area.

7. A system for implementing the method of claim 1 for controlling production of a component by fusion of digital twins, characterized in that This includes the device side, the communication side, and the processing side; The device is equipped with a 3D structured light camera to acquire images of the parts placement area. The control equipment adjusts the position of the 3D structured light camera to perform directional image acquisition. It is also equipped with production equipment, which can be adjusted to adapt to the current parts stacking state. The communication end is used for building a network channel between the device end and the processing end, for uploading the area image collected by the device end to the processing end, processing the area image by using an edge computing module carried by the processing end, constructing a phased evaluation route of the parts, performing digital twin simulation evaluation, extracting a mark parameter of each phase, using the mark parameter to regulate and control the evaluation route of the next phase, obtaining an evaluation result of each phase, and feeding back to the device end through the network channel. The control device of the device end adjusts the position and route of the 3D structured light camera according to the evaluation result and the evaluation route, executes the evaluation route, and performs targeted collection.

Images