Multi-variety assembly production line reconfigurable switching method and system based on digital twinning

Abstract

The invention provides a multi-variety assembly production line reconfigurable switching method and system based on digital twinning. The method comprises the steps that the dynamic mapping relation between the variety of a product to be produced and assembly steps is achieved through a reconfigurable logic tree model; feature parameters of the to-be-produced product are obtained through visual identification or order data analysis; importing the reconfigurable logic tree model into a digital twin system, fusing actual production data to dynamically reconstruct an assembly model, and optimizing an assembly path through simulation rehearsal; automatically generating an executable assembly control program according to the rehearsal result; and issuing the executable assembly control program to a production field to execute actual switching, and feeding back closed-loop adjustment through real-time data. According to the invention, the production line switching efficiency and the flexibility level can be improved, and the human resource consumption is reduced.

Description

Technical Field

[0001] The present invention relates to the field of intelligent manufacturing technology, and in particular to a reconfigurable switching method and system for multi-variety assembly production lines based on digital twins. Background Art

[0002] With the rapid development of intelligent manufacturing, the current manufacturing industry is transforming towards a multi-variety, small-batch, and customized production model, while traditional assembly production lines still face three core problems: (1) Low process switching efficiency. Traditional production lines rely on manual adjustment of process parameters and equipment configuration. Switching between different products requires several months of tooling and fixture replacement, process path verification, and program debugging, which seriously affects flexible production capabilities; (2) Process planning and execution are separated. Existing digital twin technologies mostly focus on optimizing a single link. It is difficult to directly generate executable control code from virtual simulation results, and manual secondary programming is still required; (3) Insufficient closed-loop control capabilities. Most digital twin systems only achieve physical-virtual one-way mapping and lack a dynamic feedback mechanism based on real-time data. Assembly deviations that occur after production line switching cannot be corrected in a timely manner, resulting in quality fluctuations.

[0003] Patent document CN117075543A discloses a virtual production line planning and adaptive scheduling method based on digital twins, including: establishing a static digital model of the production line based on a semantic unified description model for each element of the production line manufacturing resources; planning the production line according to the production requirements of the production line; guiding the production line planning in the real environment based on the static digital model of the production line planned in the virtual environment to enable it to have the conditions for formal operation; giving the static digital model of the production line a real-time state based on real-time data collected at the operation site of the production line manufacturing resources to form a digital twin model of the production line; making corresponding adjustments to the production instructions for production disturbances that occur during the operation of the production line and ensuring the stable operation of the production line; conveying the scheduling production instructions to the physical production line, controlling the manufacturing resources in the physical production line, completing the production tasks, and achieving the purpose of rapid switching and rapid response of the production line.

[0004] However, the method described in patent document CN117075543A mainly focuses on production line planning and scheduling, that is, how to reconstruct and assemble the production line to complete the order under the premise of meeting the high-prerequisite orders, without involving the change and reconstruction of processes and products. The patent document still relies on manual adjustment of process parameters, fixtures and equipment configurations, and requires multiple trial and error verifications to complete the production line reconstruction, resulting in a switching cycle of several months, which cannot meet the needs of flexible production.

[0005] Therefore, the market is in urgent need of a reconfigurable intelligent switching method and system for multi-variety assembly production lines based on digital twins that can improve production line switching efficiency and flexibility and reduce human resource consumption. Summary of the invention

[0006] Aiming at the defects in the prior art, the purpose of the present invention is to provide a reconfigurable switching method and system for multi-variety assembly production lines based on digital twins.

[0007] A reconfigurable intelligent switching method for multi-variety assembly production lines based on digital twins provided by the present invention includes: Step S1: Implement the dynamic mapping relationship between the product variety to be produced and the assembly process steps through a reconfigurable logic tree model; Step S2: Obtain the characteristic parameters of the product to be produced through visual recognition or order data parsing; Step S3: Import the reconfigurable logic tree model into the digital twin system, dynamically reconstruct the assembly model by integrating actual production data, and optimize the assembly path through simulation preview; Step S4: Automatically generate an executable assembly control program according to the preview result; Step S5: Send the executable assembly control program to the production site to execute the actual switching, and perform closed-loop adjustment through real-time data feedback.

[0008] Preferably, the reconfigurable logic tree model includes a process knowledge graph, where the nodes in the process knowledge graph represent the feature vectors of assembly process steps, and the edge weights represent the process constraint relationships; The feature vectors of the assembly process steps include torque range and positioning accuracy; The reconfigurable logic tree model is established based on the structure of the assembled product. In the reconfigurable logic tree, the process elements of different products are logically associated with the corresponding assembly sequences, tooling fixtures, and inspection procedures information of the products through semantic mapping.

[0009] Preferably, the reconfigurable logic tree model can dynamically optimize the topological connection weights through the gradient descent algorithm based on real-time working condition data, update the topological relationship of the logic tree model, and realize the online reconstruction of the assembly sequence and process elements through semantic mapping.

[0010] Preferably, the characteristic parameters include a multi-source data fusion method of workpiece geometric features, order batch information, and production beat requirements.

[0011] Preferably, the assembly model refers to the assembly line model after switching products in the digital twin virtual platform; The digital twin system is built through five-dimensional modeling based on the Unity3D engine. The five dimensions include geometry, motion, data, communication, and control, and OPC UA communication is used to achieve unified communication of industrial equipment.

[0012] Preferably, the reconfigurable logic tree model is data-bound to the physical engine of the digital twin system through an API interface to realize the dynamic association of process parameters and three-dimensional models.

[0013] Preferably, the executable control program includes robot code and PLC code; The robot code and PLC code are modular codes, which decompose the robot / PLC code into fixed logic segments and variable parameter slots.

[0014] Preferably, the real-time data feedback closed-loop adjustment is based on the on-site assembly process data. Through the digital twin system, the virtual environment is kept consistent with the real environment. In the digital twin system, the motion processes of the robot and PLC are adjusted, and the generated code is sent to the site to ensure closed-loop control. Preferably, the digital twin rehearsal is to pre-assemble the product in the virtual environment, simulate the assembly process of the actual environment, and timely adjust the incorrect parts of the assembly.

[0015] A reconfigurable intelligent switching system for multi-variety assembly production lines based on digital twins provided by the present invention includes: Logic processing layer: realizing the dynamic mapping relationship between the product variety to be produced and the assembly steps through the reconfigurable logic tree model; Data acquisition layer: obtaining the characteristic parameters of the product to be produced through visual recognition or order data parsing; Digital twin layer: importing the reconfigurable logic tree model into the digital twin system, dynamically reconstructing the assembly model by integrating the actual production data, and optimizing the assembly path through simulation rehearsal; Execution layer: automatically generating an executable assembly control program according to the rehearsal result; Feedback layer: sending the executable assembly control program to the production site to execute the actual switching, and performing closed-loop adjustment through real-time data feedback.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. Through the deep integration of the reconfigurable logic tree and the digital twin system, the present invention realizes the full-automatic switching from product feature recognition (such as visual / order data parsing), process dynamic reconstruction to control program generation, compresses the switching cycle of the traditional production line from several months to within several weeks, and improves the switching efficiency of the production line.

[0017] 2. By adopting five-dimensional modeling (geometry - motion - data - communication - control) and a closed-loop feedback mechanism, the present invention detects more than 90% of the assembly path conflicts through the physical engine in the digital twin virtual rehearsal stage, and dynamically corrects the equipment trajectory through the real-time data of the sensor in the on-site execution stage, greatly improving the assembly accuracy.

[0018] 3. Through the modular code generation technology, the present invention decomposes robot / PLC instructions into fixed logic segments and variable parameter slots, and automatically fills in key parameters such as position coordinates and speed through the digital twin rehearsal data, realizing the automatic generation of code, reducing the consumption of human resources, improving the efficiency of code writing, enhancing the flexibility level of the production line, and achieving a reconfigurable production line. Description of the Drawings

[0019] By reading the detailed description of the non-restrictive embodiments with reference to the following drawings, other features, objectives, and advantages of the present invention will become more apparent: Figure 1 It is a schematic flowchart of the working method of the present invention; Figure 2 It is a system architecture diagram of the present invention; Figure 3 It is a schematic diagram of the reconfigurable logic tree of the present invention; Figure 4 It is a schematic diagram of the modular code of the robot and PLC of the present invention. Detailed Embodiments

[0020] The present invention will be described in detail below with reference to specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but do not limit the present invention in any form. It should be noted that those of ordinary skill in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

[0021] The present invention realizes the online reconstruction of product structure and process elements by establishing a reconfigurable logic tree dynamic mapping mechanism, constructing a logic tree model containing a process knowledge graph based on semantic mapping, using assembly process step feature vectors (such as torque range, positioning accuracy) for nodes, and dynamically optimizing the edge weights through the gradient descent algorithm; digital twin full-element simulation, integrating real-time device data through the OPC UA protocol, and realizing geometric-motion-control parameter linkage simulation in the Unity3D environment, and more than 90% of assembly path conflicts can be found in the rehearsal stage; automatic code generation, establishing a modular code architecture, decomposing robot / PLC instructions into fixed logic segments and variable parameter slots, and automatically filling in parameters such as position coordinates and motion speed through the digital twin rehearsal data, realizing automatic code generation and distribution. Through the above technologies, the present invention can compress the switching time of multi-variety assembly production lines from the industry average of several months to within a few weeks, reduce the manual participation link by 83%, improve the switching efficiency and flexibility level of the production line, and reduce the consumption of human resources.

[0022] Embodiment 1 According to a reconfigurable intelligent switching method for multi-variety assembly production lines based on digital twins provided by the present invention, data interaction between each step is realized through an industrial Internet of Things platform, such asFigure 1 As shown in the figure, it includes the following steps: Step S1: Implement the dynamic mapping relationship between the product varieties to be produced and the assembly process steps through a reconfigurable logic tree model. The reconfigurable logic tree model contains a process knowledge graph, in which, as Figure 3 shown in the figure, the nodes represent the assembly process step feature vectors, and the edge weights represent the process constraint relationships. The assembly process step feature vectors include torque range and positioning accuracy. The reconfigurable logic tree model can dynamically optimize the topological connection weights based on real-time working condition data through the gradient descent algorithm, update the topological relationship of the logic tree model, and realize the online reconstruction of the assembly sequence and process elements through semantic mapping. The reconfigurable logic tree model is established based on the structure of the assembled product. In the reconfigurable logic tree, the process elements of different products are logically associated with their assembly sequences, tooling fixtures, inspection procedures, etc. through semantic mapping.

[0023] Step S2: Obtain the characteristic parameters of the product to be produced through visual recognition or order data parsing. The characteristic parameters include the multi-source data fusion method of workpiece geometric features, order batch information, and production rhythm requirements. Among them, the workpiece geometric features are collected through image recognition by an industrial camera, the order batch information is read through RFID, and the multi-source data fusion method of production rhythm requirements is obtained through the MES system.

[0024] Step S3: Import the reconfigurable logic tree model into the digital twin system, dynamically reconstruct the assembly model by fusing the actual production data, and optimize the assembly path through simulation preview. The assembly model refers to the assembly line model after switching products in the digital twin virtual platform. That is, originally in the digital twin platform, there was a set of conventional assembly line models. After changing according to the actual product, the assembly line model reconstructs the original conventional assembly line model according to the digital model of the replaced product. Among them, the reconstruction content mainly focuses on the import of the product digital model, that is, importing the digital model of the new product to replace the original conventional product digital model to realize the reconstruction of the assembly line model.

[0025] The digital twin system is built based on the Unity3D engine through five-dimensional modeling. The five dimensions include geometry, motion, data, communication, and control. OPC UA communication is used to achieve unified communication of industrial equipment. Specifically, it includes the following: 1. Model building: Draw the corresponding digital model based on the physical entity to ensure the consistency of its basic dimensions with the physical entity. 2. Kinematics building: Based on the kinematics of the physical entity, build the kinematics of the digital model in the digital twin platform (Unity). In the Unity platform, the kinematics is generally a parent-child relationship, that is, the multi-axis follow-up can be achieved through the parent-child relationship between axes, while the rotation and movement of a single axis can be controlled by the digital model through the twin transmission data for code implementation. 3. Data acquisition: Collect relevant data of the built digital twin system through measures such as sensors and the robot body. 4. Data communication: Select a suitable communication protocol (such as TCP / IP) for data communication of the collected data. The data collected from the field is communicated, and the transmission frequency is set and transmitted to the upper computer (digital twin platform). The upper computer receives and analyzes the data. 5. Control: Based on the transmitted twin data and the kinematic relationship of different components, control the digital twins of different components through C# code, and refresh the motion state of the components in real time through high-frequency data transmission to achieve the digital twin effect.

[0026] The reconfigurable logic tree model is data-bound to the physical engine of the digital twin system through the API interface to achieve the dynamic association of process parameters and the three-dimensional model. The digital twin preview is to pre-assemble the product in a virtual environment, simulate the assembly process of the actual environment, and timely adjust the incorrect parts of the assembly to ensure the assembly success rate and assembly accuracy.

[0027] Step S4: Automatically generate an executable assembly control program according to the preview result. The executable control program includes robot code and PLC (programmable logic controller) code. The robot code and PLC code are modular codes, and the robot / PLC code is decomposed into fixed logic segments and variable parameter slots. Taking the robot as an example, the commonly used instructions are integrated into program modules, such as the robot moving to the target point, the robot grasping, and the robot releasing. Different modular programs set start and stop signals and can wait for the modular program start signal to loop and execute at the robot end. Such as Figure 4As shown, the same assembly process is unified into a section of code. The fixed logic section code remains unchanged. Through the recognition in step S2 and the digital twin assembly preview in step S3, parameters such as position coordinates and movement speeds are automatically filled, realizing the automatic generation and distribution of the code. Specifically, taking the module of the robot moving to the target point as an example, the part involving specific movement data of this module is set as a variable. In the digital twin platform, the twin model robot moves to the specified point, and the angles moved by each axis are known. This data can be communicated and transmitted into the variable of the modular program. At the same time, a start signal is given, and the main program that loops and executes on the robot side receives the start signal, executes the modular program of moving to the target point, and drives the physical robot side to move to the same target point. Taking the target point as an example, in the model of the digital twin platform, each point that the robot moves to has a relative position relationship with the robot body. That is, there is a coordinate system conversion relationship between the target point in the model and the target point in the actual assembly line. Through the pre-calibrated transformation matrix in the early stage, when the robot gives the target point to move to in the model, after coordinate conversion, the target point that the robot actually needs to move to in reality is obtained. Taking this point as a variable, as described in 1, a start signal is given to make the robot side execute the program, reach the target point, and at the same time ensure that the position is the same as the target point on the model side.

[0028] Step S5: Send the executable assembly control program to the production site to execute the actual switch, and perform closed-loop adjustment through real-time data feedback. The real-time data feedback closed-loop adjustment is based on the on-site assembly process data. Through the digital twin system, the virtual environment is kept consistent with the real environment. In the digital twin system, the movement processes of the robot and the PLC are adjusted, and the generated code is sent to the site to ensure closed-loop control. Furthermore, the present invention is specifically described as follows: The reconfigurable logic tree is a directed graph structure, where each node represents an assembly step, the connection lines (edges) between the nodes represent the sequential dependency relationships between the steps, and the weight values on the edges represent the strength or constraint degree of the dependency relationships. Each assembly step node can specifically include the following characteristic parameters: Torque range parameter: Record the minimum and maximum torque values required when this step is executed. For example, a certain tightening step requires a torque range of 2.5 - 3.0 Nm; Positioning accuracy parameter: Record the position tolerance required for this step. For example, ±0.05 mm; Execution time parameter: Record the standard completion time of this step, such as 3.5 seconds; The above node parameters are stored in numerical form to form the characteristic vector of the node. For example, the characteristic vector of a bolt tightening step may be [2.5 - 3.0 Nm, ±0.05 mm, 3.5 s].

[0029] First, when the product to be produced enters the production line, the geometric features of the workpiece are first recognized by an industrial camera, and the order data is parsed to obtain product parameters, realizing the fusion of multi-source data.

[0030] Then, based on the reconfigurable logic tree model, the product feature parameters are dynamically mapped to the assembly steps. Through semantic association, an initial assembly sequence is generated, and real-time working condition data is imported into the digital twin environment. Combining with the five-dimensional virtual model constructed by the Unity3D engine, the assembly path simulation is pre-run, and the tooling layout and robot motion trajectory are dynamically optimized. The dynamic mapping of the product feature parameters to the assembly steps based on the reconfigurable logic tree model includes the following steps: Step 1: Obtain the set of feature parameters of the new product, including: product geometric dimensions (length, width, height, etc.), connection methods (bolt connection, etc.), and assembly specification requirements (torque requirements, precision requirements, etc.).

[0031] Step 2: Use the vector similarity method to calculate the matching degree between the features of the new product and each process step node in the logic tree. Step 2 includes: performing dimension unification processing on each product feature and process step feature so that they can be numerically compared; then calculating their similarity, with the numerical range between 0 and 1, and the larger the value, the higher the matching degree; then set a matching threshold (for example, 0.75). When the similarity exceeds this threshold, it is considered that this product feature can establish a mapping relationship with this process step node. For example, the matching degree between the bolt connection feature of a new product and the existing "6mm bolt tightening process step" is 0.85, which is higher than the set threshold, so a mapping relationship is established.

[0032] Step 3: Based on the established mapping relationship between the product features and the process steps, all the process step nodes that have established a mapping relationship with the features of the new product are screened out from the reconfigurable logic tree. The process step nodes form a candidate process step set required for the assembly of the new product, and the original dependencies between the nodes (i.e., the edges and weights connecting them) are retained to form a subgraph structure. Topological sorting is performed on the subgraph to determine the process step execution order, that is, the initial assembly sequence. In addition, various assembly constraints need to be considered during the sorting process, including pre-order constraints and parallel constraints. Among them, the pre-order constraint means that some process steps must be completed before other process steps, and the parallel constraint means that some process steps can be executed in parallel. For example, for a simple assembled product, after screening, there may be five process steps: ["install the base", "install the bracket", "fix the bolt", "install the panel", "inspect"], and through topological sorting, the execution order may be ["install the base", "install the bracket", "fix the bolt", "install the panel", "inspect"].

[0033] Step 4: Import the precise 3D model of the assembly line in the Unity3D platform, including all equipment, workpieces, and tooling fixtures. Set the correct kinematic parameters for each robot, conveyor belt, and other equipment to ensure that their movements in the virtual environment are consistent with the actual situation. Set collision volumes and physical properties for all objects for collision detection. Plan and calculate the movement paths of the equipment for each process step. At the same time, perform multi-objective optimization on the initially planned paths, and the following factors can be considered: objectives such as time minimization and accuracy maximization. For example, there may be multiple options for the path of the robot to pick up a workpiece from the workpiece rack and move it to the assembly position. The system will test these paths in the virtual environment and calculate various indicators. For example, Path 1 takes 4.2 seconds and Path 2 takes 4.5 seconds. Finally, the best path will be selected according to the objectives.

[0034] Next, after verification through simulation, disassemble the assembly sequence into fixed logic code segments and variable parameter slots for the robot. Automatically fill in simulation optimization parameters such as coordinates and speeds, and generate executable control programs for the robot and PLC. Send them to the PLC and robot controller through the OPC UA protocol. Specifically, it includes the following: The movement module controls the equipment to move from the current position to the target position; the grasping module controls the manipulator or gripper to grasp the workpiece; the releasing module controls the manipulator or gripper to release the workpiece; the waiting module waits for a specific signal or waits for a specific time. For example, a fixed logic segment of a robot movement module may be the overall logic to control the robot to start smoothly, move along a specific path, and decelerate and stop, while the variable parameter slots include parameters such as target position coordinates, movement speed, and accuracy area that need to be customized for specific process steps.

[0035] The specific process of extracting parameters from the digital twin simulation and filling them into the code template is as follows: Identify the parameter types required for each module. For example, the movement module requires position, speed, and accuracy parameters. Extract the corresponding parameters from the simulation results. For example, the target position coordinates where the robot moves in the virtual environment. Convert the coordinates in the digital twin environment to the coordinates in the actual physical environment. The conversion is carried out through a pre-calibrated conversion matrix to ensure that the positions in the virtual environment can be accurately mapped to the physical world. For example, the point (100, 200, 50) in the virtual environment may need to be converted to (105, 198, 52) in the actual environment. Then fill the extracted and converted parameter values into the predefined code template. And send an execution instruction, and the robot or PLC can start executing this modular code. For example, for a certain assembly process step that requires the robot to move to the workpiece position and install components, the system will extract the accurate coordinates (X = 320.5, Y = 150.2, Z = 45.7) of this position and the best movement speed (80 mm / s) from the digital twin simulation, and then fill these parameters into the code template of the movement module to generate an executable robot instruction.

[0036] Meanwhile, during the actual assembly process, the industrial Internet of Things platform collects data such as the device positioning accuracy in real time and feeds it back to the digital twin system. The weights of the logic tree are dynamically corrected through the gradient descent algorithm. If an assembly deviation is detected, the virtual model is reconstructed and a compensation instruction is generated to achieve closed-loop adjustment of the assembly path, and finally the adaptive switching and precise assembly of multi-variety production lines are completed. Specifically, a multi-objective loss function is used to evaluate the effect of the current assembly plan and serve as the basis for weight optimization. The loss function can include time loss: the difference between the actual assembly time and the target time. For example, the actual time is 23.5 seconds, the target time is 20 seconds, and the difference is 3.5 seconds; accuracy loss: the deviation between the actual assembly accuracy and the required accuracy. For example, the required accuracy is ±0.05 mm, and the actual accuracy reaches ±0.08 mm, with a deviation of 0.03 mm. The calculation formula is as follows: Total loss = 0.4 × Time loss + 0.6 × Accuracy loss Before calculating each loss term, normalization processing is performed to make their dimensions consistent. For example, the time loss can be expressed as: (Actual time - Target time) / Target time.

[0037] The system dynamically adjusts the weights of each edge in the logic tree by using the gradient descent algorithm. For each edge in the logic tree (representing the dependency relationship between process steps), calculate the impact of its weight change on the total loss function, and use the numerical differentiation method to calculate the gradient: slightly change the weight of a certain edge and observe the change in the loss function to calculate the approximate gradient. For example, if the weight of edge e1 is increased from 0.75 to 0.751, and the loss function decreases from 0.35 to 0.348, the gradient of this weight is approximately -2. After each batch of product assemblies is completed, collect the actual performance data, calculate the total loss of the current assembly plan, calculate the gradients for all edge weights and update them, and use the updated weights to regenerate the assembly plan. Repeat the above process until the loss function converges or reaches the preset number of iterations.

[0038] Suppose that during the assembly process, it is found that the dependency strength between the two steps of "installation bracket" and "fixing bolt" is insufficient, resulting in the bracket being unstable before the bolt is tightened, affecting the accuracy. Then calculate the impact of increasing the weight of this dependency on the loss function. If increasing the weight can reduce the loss, then increase the weight value of this edge, for example, from 0.6 to 0.8, so that in subsequent assembly, the system will ensure that the bracket is more stable before tightening the bolt. Deviations during the assembly process are monitored in real time through a variety of sensors. For example, for a certain bearing installation step: Expected position: hole center (100.00, 200.00, 30.00) mm, Actual position: detected by the sensor as (100.12, 199.95, 29.97) mm. Calculate the Euclidean distance deviation: 0.14 mm. Compare with the position deviation threshold (±0.10 mm), and find that it exceeds the threshold, so adjustment is required. Based on the detected deviation, calculate the position compensation amount: (-0.12, 0.05, 0.03) mm. Generate a compensation instruction and perform a closed-loop adjustment. After compensation, detect the position again to confirm that the bearing is installed in place.

[0039] The present invention aims to solve the problem of how to quickly realize the reconfiguration of the production line when the product type is changed and the process is changed in a single assembly line. This reconfiguration mainly describes the reconfiguration in code and products.

[0040] The present invention realizes the dynamic semantic mapping between product features and assembly steps through a reconfigurable logic tree model, automatically extracts process parameters by combining visual recognition and order data parsing, and simulates and optimizes the assembly path in a digital twin environment. A control program can be generated without manual intervention, compressing the switching time from several months to within several weeks, and significantly improving the efficiency.

[0041] Embodiment 2 The present invention also provides a reconfigurable intelligent switching system for a multi-variety assembly production line based on digital twin. The reconfigurable intelligent switching system for a multi-variety assembly production line based on digital twin can be realized by executing the process steps of the reconfigurable intelligent switching method for a multi-variety assembly production line based on digital twin. That is, those skilled in the art can understand the reconfigurable intelligent switching method for a multi-variety assembly production line based on digital twin as a preferred implementation manner of the reconfigurable intelligent switching system for a multi-variety assembly production line based on digital twin.

[0042] According to a reconfigurable intelligent switching system for a multi-variety assembly production line based on digital twin provided by the present invention, as Figure 2 shown, it includes: Logical processing layer: It realizes the dynamic mapping relationship between the product varieties to be produced and the assembly process steps through a reconfigurable logic tree model; the reconfigurable logic tree model includes a process knowledge graph, where the nodes in the process knowledge graph represent the assembly process step feature vectors, and the edge weights represent the process constraint relationships; the assembly process step feature vectors include torque range and positioning accuracy; the reconfigurable logic tree model is established based on the assembly product structure, and in the reconfigurable logic tree, the process elements of different products are logically associated with the corresponding assembly sequences, tooling fixtures, and inspection procedures information through semantic mapping. The reconfigurable logic tree model can dynamically optimize the topological connection weights based on real-time working condition data through the gradient descent algorithm, update the topological relationship of the logic tree model, and realize the online reconstruction of the assembly sequence and process elements through semantic mapping. The reconfigurable logic tree model is data-bound to the physical engine of the digital twin system through the API interface to realize the dynamic association between process parameters and the 3D model.

[0043] Data acquisition layer: It obtains the characteristic parameters of the product to be produced through visual recognition or order data parsing. The characteristic parameters include a multi-source data fusion method of workpiece geometric features, order batch information, and production beat requirements.

[0044] Digital twin layer: The reconfigurable logic tree model is imported into the digital twin system, and the assembly model is dynamically reconstructed by integrating the actual production data, and the assembly path is optimized through simulation preview. The assembly model refers to the assembly line model after switching products in the digital twin virtual platform. The digital twin system is built through five-dimensional modeling based on the Unity3D engine, and the five dimensions include geometry, motion, data, communication, and control, and the OPC UA communication is used to realize the unified communication of industrial equipment.

[0045] Execution layer: Automatically generates an executable assembly control program according to the preview result. The executable control program includes robot code and PLC code. The robot code and PLC code are modular codes, and the robot / PLC code is decomposed into fixed logic segments and variable parameter slots.

[0046] Feedback layer: The executable assembly control program is sent to the production site for actual switching, and closed-loop adjustment is performed through real-time data feedback. The real-time data feedback closed-loop adjustment is based on the on-site assembly process data, and the digital twin system is used to keep the virtual environment consistent with the real environment. In the digital twin system, the motion process of the robot and PLC is adjusted, and the generated code is sent to the site to ensure closed-loop control. The digital twin preview is to perform pre-assembly of the product in the virtual environment, simulate the assembly process of the actual environment, and timely adjust the incorrect parts of the assembly.

[0047] Those skilled in the art know that in addition to implementing the system and its various devices, modules, and units provided by the present invention in the form of pure computer-readable program code, the method steps can be logically programmed to enable the system and its various devices, modules, and units provided by the present invention to be implemented in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers, etc., to achieve the same functions. Therefore, the system and its various devices, modules, and units provided by the present invention can be considered as a kind of hardware component, and the devices, modules, and units included therein for implementing various functions can also be regarded as the structures within the hardware component; the devices, modules, and units for implementing various functions can also be regarded as either software modules for implementing the method or structures within the hardware component.

[0048] The specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art can make various changes or modifications within the scope of the claims, which does not affect the essence of the present invention. Without conflict, the embodiments of the present application and the features in the embodiments can be combined with each other arbitrarily.

Claims

1. A reconfigurable intelligent switching method for multi-variety assembly production lines based on digital twins, characterized in that Including: Step S1: Implement the dynamic mapping relationship between the product varieties to be produced and the assembly steps through a reconfigurable logic tree model; Step S2: Obtain the characteristic parameters of the product to be produced through visual recognition or order data parsing; Step S3: Import the reconfigurable logic tree model into the digital twin system, dynamically reconstruct the assembly model by integrating actual production data, and optimize the assembly path through simulation rehearsal; Step S4: Automatically generate an executable assembly control program according to the rehearsal result; Step S5: Send the executable assembly control program to the production site for actual switching, and perform closed-loop adjustment through real-time data feedback.

2. The method for reconfigurable intelligent switching of a multi-variety assembly production line based on digital twin according to claim 1, wherein The reconfigurable logic tree model includes a process knowledge graph, where the nodes in the process knowledge graph represent the feature vectors of the assembly steps, and the edge weights represent the process constraint relationships; The feature vectors of the assembly steps include torque range and positioning accuracy; The reconfigurable logic tree model is established based on the structure of the assembled product. In the reconfigurable logic tree, the process elements of different products are logically associated with the corresponding assembly sequences, tooling fixtures, and inspection procedures information through semantic mapping.

3. The reconfigurable intelligent switching method for multi-variety assembly production lines based on digital twins according to claim 1, characterized in that, The reconfigurable logic tree model can dynamically optimize the topological connection weights through the gradient descent algorithm based on real-time working condition data, update the topological relationship of the logic tree model, and realize the online reconstruction of the assembly sequence and process elements through semantic mapping.

4. The method for reconfigurable intelligent switching of a multi-variety assembly production line based on digital twins according to claim 1, wherein The characteristic parameters include a multi-source data fusion method of workpiece geometric features, order batch information, and production beat requirements.

5. The method for reconfigurable intelligent switching of a multi-variety assembly production line based on digital twins according to claim 1, wherein The assembly model refers to the assembly line model after switching products in the digital twin virtual platform; The digital twin system is built through five-dimensional modeling based on the Unity3D engine. The five dimensions include geometry, motion, data, communication, and control, and OPC UA communication is used to achieve unified communication of industrial equipment.

6. The method for reconfigurable intelligent switching of a multi-variety assembly production line based on digital twins according to claim 1, wherein The reconfigurable logic tree model is data-bound to the physical engine of the digital twin system through an API interface to realize the dynamic association between process parameters and the three-dimensional model.

7. The method for reconfigurable intelligent switching of a multi-variety assembly production line based on digital twins according to claim 1, wherein The executable assembly control program includes robot code and PLC code; The robot code and PLC code are modular codes, and the robot / PLC code is decomposed into fixed logic segments and variable parameter slots.

8. The method for reconfigurable intelligent switching of a multi-variety assembly production line based on digital twin according to claim 1, wherein The real-time data feedback closed-loop adjustment is based on the on-site assembly process data. Through the digital twin system, the virtual environment is kept consistent with the real environment. The motion processes of the robot and PLC are adjusted in the digital twin system, and the generated code is sent to the site to ensure closed-loop control.

9. The method for reconfigurable intelligent switching of a multi-variety assembly production line based on digital twins according to claim 1, characterized in that Digital twin rehearsal is to perform pre-assembly of the product in the virtual environment, simulate the assembly process of the actual environment, and timely adjust the incorrect parts of the assembly.

10. A reconfigurable intelligent switching system for multi-variety assembly production lines based on digital twins, characterized in that, Including: Logic processing layer: Implement the dynamic mapping relationship between the product varieties to be produced and the assembly steps through a reconfigurable logic tree model; Data acquisition layer: Obtain the characteristic parameters of the product to be produced through visual recognition or order data parsing; Digital twin layer: Import the reconfigurable logic tree model into the digital twin system, dynamically reconstruct the assembly model by integrating actual production data, and optimize the assembly path through simulation rehearsal; Execution layer: Automatically generate an executable assembly control program according to the rehearsal result; Feedback layer: The executable assembly control program is sent to the production site for actual switching, and closed-loop adjustment is performed through real-time data feedback.

Images