Skin stretching device adaptive iterative optimization control system based on digital twinning

Abstract

This invention provides an adaptive iterative optimization control system for skin forming equipment based on digital twins. It establishes a closed-loop architecture encompassing detection, simulation, and control, and utilizes digital twin technology for simulation and inverse optimization calculation of process parameters. Compared to existing forward trial-and-error and open-loop control methods, this significantly improves the efficiency and accuracy of parameter optimization. The system also establishes a process knowledge base based on a self-learning mechanism, continuously acquiring and accumulating experience from each production run, and continuously improving the accuracy of the digital twin model and the efficiency of the optimization algorithm. This system has strong applicability and can meet the high production requirements of large and complex skins.

Description

Technical Field

[0001] This invention belongs to the field of aircraft skin stretching equipment control technology, specifically relating to an adaptive iterative optimization control system for skin stretching equipment based on digital twins. Background Technology

[0002] Large skin structures used in the aerospace field are typically manufactured using a stretch forming process, which places extremely high demands on the precision of the process parameters of the stretch forming equipment. Current technologies for setting and optimizing the process parameters of stretch forming equipment mainly rely on three methods: offline finite element simulation, searching experience databases, and multiple rounds of trial and error correction. However, these methods suffer from several problems, including parameters deviating from actual materials and operating conditions, and limited improvement in product accuracy and consistency. Furthermore, the boundary conditions and results from the simulation cannot obtain closed-loop feedback and dynamic adjustment from the actual manufacturing process, and the quality inspection data after forming cannot effectively leverage the role of parameter reverse optimization. On the other hand, as skin designs become larger and more complex, the shortcomings of existing optical or contact-based skin measurement methods become increasingly apparent, rendering them insufficient to provide the necessary collaborative optimization functions for stretch forming equipment. Summary of the Invention

[0003] In view of this, and addressing the technical problems existing in this field, the present invention provides an adaptive iterative optimization control system for skin stretching equipment based on digital twins. The system architecture comprises an execution layer, a detection layer, a digital twin layer, a process knowledge base, and a control layer. The execution layer performs the actual stretching forming action; the detection layer scans and measures any position on the skin surface; the digital twin layer, deployed on a server, establishes a corresponding simulation model of the stretching forming process and obtains skin data and process parameter data from the detection layer and the process knowledge base. It then sequentially performs forward simulation of the stretching process, deviation analysis of the simulation results and detection data, and inverse solution of the process parameters, outputting the corrected and optimized process parameters to the control layer; the control layer implements closed-loop control of the execution layer based on the optimized process parameters; and the process knowledge base obtains the detection data from the detection layer and the optimized process parameters obtained from the digital twin layer, updating and storing relevant process-related experience knowledge through a self-learning mechanism.

[0004] Furthermore, the execution layer specifically includes multi-axis flexible tensioning mechanisms, clamping fixtures, lifting units, and other tensioning action execution components.

[0005] Furthermore, the detection layer consists of a multi-axis truss-type robot and an image acquisition unit. The multi-axis truss-type robot is set at the skin stretching station, and the image acquisition unit is installed on the robot to collect and acquire image data of the skin surface. After data processing, a complete skin point cloud is reconstructed in three dimensions and registered with the design model. Key forming features such as springback, thinning, wrinkling, and mold adhesion of the material at different locations of the skin are extracted from it.

[0006] Furthermore, the simulation model established in the digital twin layer specifically includes: Finite element model is used to simulate the stress, shape and thickness changes of skin. It is solved by explicit dynamic algorithm or implicit static algorithm. The specific modeling method is to select shell element or solid element according to the skin thickness. Material constitutive models are used to simulate the mechanical behavior of skin materials during the stretching process, including elastoplastic models, material anisotropic parameters, and material hardening curves; The contact friction model is used to simulate the frictional behavior between the mold and the skin. It includes the definition of contact type and friction model. The friction coefficient used in the model is dynamically adjusted according to actual working conditions such as pressure and temperature, and can support changes in contact state (stick-slip transition).

[0007] Boundary condition model is used to set various loading methods for the equipment, including the movement trajectory of the clamp (displacement boundary condition) or tensile force curve (force boundary condition), the displacement or force control boundary of the lifting unit, the rigid or elastic connection relationship between the clamp and the skin, and can support multi-stage loading path definition; The digital twin layer also includes functional modules corresponding to forward simulation, deviation analysis, and reverse engineering of process parameters, as well as a virtual verification module. The forward simulation module takes initial process parameters, including clamp tension curves, displacement paths, and mold lifting sequences, as input, and outputs the post-forming geometry, stress-strain distribution, and thickness reduction rate distribution. The deviation analysis module takes measured 3D point cloud data and design CAD models as input, and outputs deviation distribution cloud maps, maximum deviation, average deviation, and root mean square deviation. The reverse engineering of process parameters takes the deviation distribution cloud map output by the deviation analysis module and the current process parameters as input, and outputs corrected optimized process parameters by executing reverse finite element methods, intelligent optimization algorithms, or sensitivity analysis methods. The virtual verification module verifies the corrected optimized process parameters and outputs virtual forming results and deviation prediction values ​​to the control layer.

[0008] Furthermore, the control layer receives and parses the optimized process parameters provided by the digital twin layer, and provides corresponding control quantities for each component of the execution layer, including: the clamp's tension curve, displacement trajectory, spatial attitude and swing angle, the lifting unit's displacement and lifting force, the action sequence of the tensioning mechanism and the lifting unit, etc., thereby realizing synchronous and coordinated control of multiple units in the execution layer, supporting continuous switching and smooth transition of multi-stage process loading paths, monitoring the status of each unit in the execution layer, and performing shutdown protection, alarm, fault recording and backtracking operations when an anomaly occurs.

[0009] Furthermore, the process knowledge base specifically includes a database and an adaptive update strategy based on a self-learning mechanism; the database contains: a process parameter record table, a test data record table, an optimization history record table, and a skin material parameter table; the process parameter adaptive update strategy consists of a material parameter adaptive correction learning strategy, a simulation model self-calibration learning strategy, and an optimal parameter correction strategy.

[0010] Accordingly, the present invention also provides an adaptive iterative optimization control method for skin-forming devices based on digital twins, implemented using the above-described system, specifically including the following steps: S1. Import the skin design model and tolerance data, and use the process knowledge base to retrieve similar forming case data to generate default initial parameters; the digital twin layer performs forward simulation based on the initial process parameters to output the initial process parameters, and the control layer controls the execution layer to perform the first round of stretching action based on the initial process parameters. S2. After the detection layer acquires the detection data and reconstructs the complete skin point cloud in 3D, it registers it with the design model, generates a deviation cloud map, extracts key forming features, and outputs them to the digital twin layer. S3. The digital twin layer determines whether the deviation meets the requirements and decides whether to continue stretching based on the initial process parameters or proceed to S4 for reverse solution of process parameters: S4. The digital twin layer maps the deviation data to each simulation model and constructs the corresponding error relationship model. It then executes optimization algorithms to correct and optimize process parameters, and outputs the results to the control layer after completing the virtual simulation verification. S5. The control layer loads and optimizes the process parameters to execute the next round of stretching. The detection layer and the digital twin layer simultaneously enter a new round of iterations of steps S2 to S4. S6. The process knowledge base automatically stores process parameters, test data, and optimization results after each iteration, and completes adaptive updates.

[0011] Accordingly, the present invention also provides a skin stretching device, including the above-mentioned adaptive iterative optimization control system for skin stretching devices based on digital twins.

[0012] The adaptive iterative optimization control system for skin forming equipment provided by this invention establishes a closed-loop architecture encompassing detection, simulation, and control. It utilizes digital twin technology for simulation and inverse optimization calculation of process parameters, significantly improving the efficiency and accuracy of parameter optimization compared to existing forward trial-and-error and open-loop control methods. The system also establishes a process knowledge base based on a self-learning mechanism, continuously acquiring and accumulating experience from each production run, thereby continuously improving the accuracy of the digital twin model and the efficiency of the optimization algorithm. This system has strong applicability and can meet the high production requirements of large and complex skins. Attached Figure Description

[0013] Figure 1 This forms the overall framework of the system provided by the present invention; Figure 2 The method flow for executing the system provided by this invention; Figure 3 A skin stretching device that includes the system provided by the present invention. Detailed Implementation

[0014] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. 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.

[0015] The present invention provides an adaptive iterative optimization control system for skin stretching equipment based on digital twins, such as... Figure 1 As shown, the system architecture consists of an execution layer, a detection layer, a digital twin layer, a process knowledge base, and a control layer. The execution layer performs the actual stretching forming action; the detection layer scans and measures any position on the skin surface; the digital twin layer, deployed on the server, establishes a simulation model of the stretching forming process and obtains skin data and process parameters from the detection layer and the process knowledge base. It then sequentially performs forward simulation of the stretching process, analyzes the deviation between the simulation results and the detection data, and solves for the process parameters in reverse, outputting the corrected and optimized process parameters to the control layer. The control layer implements closed-loop control of the execution layer based on the optimized process parameters. The process knowledge base obtains the detection data from the detection layer and the optimized process parameters from the digital twin layer, updating and storing relevant process-related experience knowledge through a self-learning mechanism.

[0016] In a preferred embodiment of the present invention, the execution layer specifically includes a multi-axis flexible tensioning mechanism, a clamping fixture, a lifting unit, and other tensioning action execution components.

[0017] In a preferred embodiment of the present invention, the detection layer consists of a multi-axis truss-type robot and an image acquisition unit. The multi-axis truss-type robot is set at the skin stretching station, and the image acquisition unit is mounted on the robot to acquire image data of the skin surface. After data processing, a complete skin point cloud is reconstructed in three dimensions and registered with the design model. Key forming features are extracted from the data, such as: maximum springback and its corresponding position, springback area distribution, local thinning rate, wrinkling ripple degree, and mold adhesion degree.

[0018] In a preferred embodiment of the present invention, the digital twin layer can be deployed on a centralized server or cloud server, and the platform can adopt a CPU and GPU collaborative computing architecture and provide an interface for real-time communication with the detection layer, control layer, etc.

[0019] The simulation model established in the digital twin layer specifically includes: a finite element model, used to simulate the stress, shape and thickness changes of the skin, which is solved using explicit dynamics algorithms or implicit statics algorithms, and the modeling is specifically selected based on the skin thickness, using shell elements or solid elements; Material constitutive models are used to simulate the mechanical behavior of skin materials during the stretching process, including elastoplastic models, material anisotropic parameters, and material hardening curves; The contact friction model is used to simulate the frictional behavior between the mold and the skin. It includes the definition of the contact type (face-to-face contact) and the friction model (Coulomb friction model or shear friction model). The friction coefficient used in the model is dynamically adjusted according to the actual working conditions such as pressure and temperature, and it can support changes in contact state (stick-slip transition).

[0020] Boundary condition model is used to set various loading methods for the equipment, including the movement trajectory of the clamp (displacement boundary condition) or tensile force curve (force boundary condition), the displacement or force control boundary of the lifting unit, the rigid or elastic connection relationship between the clamp and the skin, and can support multi-stage loading path definition; The digital twin layer also includes functional modules corresponding to forward simulation, deviation analysis, and reverse engineering of process parameters, as well as a virtual verification module. The forward simulation module takes initial process parameters, including clamp tension curves, displacement paths, and mold lifting sequences, as input, and outputs the post-forming geometry, stress-strain distribution, and thickness reduction rate distribution. The deviation analysis module takes measured 3D point cloud data and design CAD models as input, and outputs deviation distribution cloud maps, maximum deviation, average deviation, and root mean square deviation. The reverse engineering of process parameters module takes the deviation distribution cloud map output by the deviation analysis module and the current process parameters as input. The optimization algorithm for reverse engineering can be selected from: the inverse finite element method, which uses the measured forming result as the target and determines the loading conditions leading to the shape through reverse engineering to obtain the process correction amount; intelligent optimization algorithms, which search for the parameter combination that minimizes the simulation deviation within the feasible domain of process parameters using genetic algorithms, particle swarm optimization, or reinforcement learning methods; and sensitivity analysis methods, which quickly calculate the parameter correction direction and magnitude by constructing a sensitivity matrix between process parameters and forming errors. The virtual verification module is used to verify the corrected optimized process parameters and output virtual forming results and deviation prediction values ​​to the control layer. It realizes functions such as pre-verification of optimized parameters, avoiding unreasonable parameters from directly affecting actual equipment, and ensuring that the predicted deviation meets the preset tolerance requirements through multiple rounds of simulation iteration.

[0021] In a preferred embodiment of the present invention, the control layer hardware may employ a high-performance programmable logic controller (PLC) and an industrial computer (IPC); a multi-axis motion control card for synchronous control of multiple actuators; a real-time industrial Ethernet communication module; communication interfaces with the actuator servo driver, hydraulic control unit, and various sensors; and a data acquisition module for real-time acquisition of force, displacement, and status parameters. This hardware architecture possesses high real-time performance, high reliability, and multi-axis collaborative control capabilities. The control layer receives and parses the optimized process parameters provided by the digital twin layer and provides corresponding control quantities to each component of the actuator layer, including: the clamp's tension curve, displacement trajectory, spatial posture and swing angle, the lifting unit's displacement and lifting force, and the action sequence of the tensioning mechanism and the lifting unit, thereby achieving synchronous and coordinated control of multiple units in the actuator layer. It can also support continuous switching and smooth transition of multi-stage process loading paths, monitor the status of each unit in the actuator layer, such as the force, displacement, and speed status parameters of each actuator, as well as hydraulic pressure, temperature, and equipment load; and perform shutdown protection, alarm, fault recording, and backtracking operations when anomalies occur, such as over-force, over-travel, or instability.

[0022] In a preferred embodiment of the present invention, the process knowledge base specifically includes a database and an adaptive update strategy based on a self-learning mechanism; wherein, the database contains: a process parameter record table, used to record process parameter information for each batch of production, including: skin model or drawing number, material grade, batch and thickness, mold number, initial process parameters (tension curve, lifting sequence), corrected process parameters after each iteration, parameter source identifier (simulation generation, knowledge base call or manual setting), etc.

[0023] The test data recording table is used to store test results and environmental information, including: test time and corresponding process stage, point cloud data storage path, deviation cloud map storage path, extracted feature parameters (maximum deviation, RMS deviation, etc.), and environmental parameters (temperature, humidity, etc.).

[0024] An optimization history table is used to record the optimization process and its effects, including: number of iterations, input deviation characteristics, optimization algorithm and its parameters used, output corrected process parameters, virtual verification results, and actual forming quality evaluation.

[0025] It also includes a skin material parameter table, which stores parameters related to the material model, including: material grade, hardening curve parameters, anisotropy parameters, elastic modulus and Poisson's ratio, parameter source and confidence level, etc.

[0026] The adaptive update strategy for process parameters consists of an adaptive correction learning strategy for material parameters, a self-calibration learning strategy for the simulation model, and an optimal parameter correction strategy.

[0027] The material parameter adaptive correction learning strategy is carried out in the following steps: collecting forming data of the same material grade in different batches; comparing the deviation between simulation prediction results and actual test results; using reverse identification methods to reverse calculate material parameters (such as hardening curves and friction coefficients); updating the material parameter table and updating the parameter confidence level simultaneously.

[0028] The simulation model self-calibration learning strategy proceeds sequentially: a mapping relationship between "process parameters and forming quality" is constructed based on historical data from the knowledge base; a surrogate model is established using machine learning methods; the surrogate model is fused with the finite element model to correct simulation deviations; and the simulation prediction accuracy and computational efficiency are improved.

[0029] Therefore, the optimal parameter correction strategy is implemented by: statistically analyzing the optimal correction strategy corresponding to different deviation characteristics; establishing a mapping rule between "deviation mode and optimization strategy"; and continuously optimizing the strategy selection as data accumulates to improve convergence speed and stability.

[0030] Accordingly, the present invention also provides an adaptive iterative optimization control method for skin-forming devices based on digital twins, implemented using the above-described system, such as... Figure 2 As shown, the specific steps include: S1. Import the skin design model and tolerance data, and use the process knowledge base to retrieve similar forming case data to generate default initial parameters; the digital twin layer performs forward simulation based on the initial process parameters to output the initial process parameters, and the control layer controls the execution layer to perform the first round of stretching action based on the initial process parameters. S2. After the detection layer acquires the detection data and reconstructs the complete skin point cloud in 3D, it registers it with the design model, generates a deviation cloud map, and extracts key forming features to output to the digital twin layer. S3. The digital twin layer determines whether the deviation meets the requirements and decides whether to continue stretching based on the initial process parameters or proceed to S4 for reverse solution of process parameters: S4. The digital twin layer maps the deviation data to each simulation model and constructs the corresponding error relationship model. It then executes optimization algorithms to correct and optimize process parameters, and outputs the results to the control layer after completing the virtual simulation verification. S5. The control layer loads and optimizes the process parameters to execute the next round of stretching action, and the detection layer and digital twin layer simultaneously enter a new round of iterations of steps S2 to S4. S6. The process knowledge base automatically stores process parameters, test data, and optimization results after each iteration, and completes adaptive updates.

[0031] Accordingly, the present invention also provides a skin stretching device, such as... Figure 3As shown, the adaptive iterative optimization control system for the skin stretching equipment based on digital twins includes an execution layer comprising a stretching frame 1 and two symmetrical stretching units 2. Each stretching unit contains six independent motion control axes: horizontal movement (P) along the X-axis, oscillation (R) along the Z-axis, oscillation (R) along the Y-axis, movement (P) along the X-axis of the jaws, rotation (R) along the Z-axis of the jaws, and bending (R) along the X-axis of the jaws. A liftable lifting unit 3 can work in conjunction with the clamps to achieve various process modes such as "stretching before covering" or "stretching while lifting". The detection layer 4 consists of a five-axis truss-type robot and similar components. A mold 6 is set on the lifting unit 3, and the skin to be processed 5 is placed on the mold 6. The digital twin layer is deployed on a server 7.

[0032] It should be understood that the sequence number of each step in the embodiments of the present invention does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0033] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An adaptive iterative optimization control system for skin stretching equipment based on digital twins, characterized in that: The system architecture consists of an execution layer, a detection layer, a digital twin layer, a process knowledge base, and a control layer. The execution layer performs the actual stretching forming action; the detection layer scans and measures any location on the skin surface; the digital twin layer, deployed on the server, establishes a simulation model of the stretching forming process and obtains skin data and process parameters from the detection layer and the process knowledge base. It then sequentially performs forward simulation of the stretching process, analyzes the deviation between the simulation results and the detection data, and solves for the process parameters in reverse, outputting the corrected and optimized process parameters to the control layer. The control layer implements closed-loop control of the execution layer based on the optimized process parameters. The process knowledge base obtains the detection data from the detection layer and the optimized process parameters from the digital twin layer, updating and storing relevant process-related experience knowledge through a self-learning mechanism.

2. The system as described in claim 1, characterized in that: The execution layer consists of a tensioning action execution component including a multi-axis flexible tensioning mechanism, a clamping fixture, and a lifting unit.

3. The system as described in claim 1, characterized in that: The inspection layer consists of a multi-axis truss-type robot and an image acquisition unit. The multi-axis truss-type robot is set at the skin stretching station, and the image acquisition unit is mounted on the robot to collect and acquire image data of the skin surface. After data processing, a complete skin point cloud is reconstructed in three dimensions and registered with the design model. From this, key forming features at different locations of the skin, including material springback, thinning, wrinkling, and mold fit, are extracted.

4. The system as described in claim 1, characterized in that: The simulation model established in the digital twin layer specifically includes: Finite element model is used to simulate the stress, shape and thickness changes of skin. It is solved by explicit dynamic algorithm or implicit static algorithm. The specific modeling method is to select shell element or solid element according to the skin thickness. Material constitutive models are used to simulate the mechanical behavior of skin materials during the stretching process, including elastoplastic models, material anisotropic parameters, and material hardening curves; The contact friction model is used to simulate the frictional behavior between the mold and the skin, including the definition of contact type and friction model. The friction coefficient used in the model is dynamically adjusted according to the actual pressure and temperature conditions, and can support changes in contact state. Boundary condition model, used to set various loading methods of the equipment, including the motion trajectory or tensile force curve of the clamp, the displacement or force control boundary of the lifting unit, the rigid or elastic connection relationship between the clamp and the skin, and can support multi-stage loading path definition; The digital twin layer also includes functional modules corresponding to forward simulation, deviation analysis, and reverse engineering of process parameters, as well as a virtual verification module. The forward simulation module takes initial process parameters, including clamp tension curves, displacement paths, and mold lifting sequences, as input, and outputs the post-forming geometry, stress-strain distribution, and thickness reduction rate distribution. The deviation analysis module takes measured 3D point cloud data and the design model as input, and outputs deviation distribution cloud maps, maximum deviation, average deviation, and root mean square deviation. The reverse engineering of process parameters takes the deviation distribution cloud map output by the deviation analysis module and the current process parameters as input, and outputs corrected optimized process parameters by executing the reverse finite element method, intelligent optimization algorithm, or sensitivity analysis method. The virtual verification module verifies the corrected optimized process parameters and outputs virtual forming results and deviation prediction values ​​to the control layer.

5. The system as described in claim 1, characterized in that: The control layer receives and parses the optimized process parameters provided by the digital twin layer, and provides corresponding control quantities for each component of the execution layer, including: the clamp's tension curve, displacement trajectory, spatial attitude and swing angle, the lifting unit's displacement and lifting force, and the action sequence of the tensioning mechanism and the lifting unit. This enables synchronous and coordinated control of multiple units in the execution layer, supports continuous switching and smooth transition of multi-stage process loading paths, monitors the status of each unit in the execution layer, and performs shutdown protection, alarm, fault recording and backtracking operations when an anomaly occurs.

6. The system as described in claim 1, characterized in that: The process knowledge base specifically includes a database and an adaptive update strategy based on a self-learning mechanism. The database contains: a process parameter record table, a test data record table, an optimization history record table, and a skin material parameter table. The adaptive update strategy for process parameters consists of an adaptive correction learning strategy for material parameters, a self-calibration learning strategy for simulation models, and an optimal parameter correction strategy.

7. An adaptive iterative optimization control method for a skin-forming device based on digital twins, implemented using the system described in any one of claims 1-6, specifically comprising the following steps: S1. Import the skin design model and tolerance data, and use the process knowledge base to retrieve similar forming case data to generate default initial parameters; The digital twin layer performs forward simulation based on the initial process parameters and outputs the initial process parameters. The control layer controls the execution layer to perform the first round of stretching action based on the initial process parameters. S2. After the detection layer acquires the detection data and reconstructs the complete skin point cloud in 3D, it registers it with the design model, generates a deviation cloud map, and extracts key forming features to output to the digital twin layer. S3. The digital twin layer determines whether the deviation meets the requirements and decides whether to continue stretching based on the initial process parameters or proceed to S4 for reverse solution of process parameters: S4. The digital twin layer maps the deviation data to each simulation model and constructs the corresponding error relationship model. It then executes optimization algorithms to correct and optimize process parameters, and outputs the results to the control layer after completing the virtual simulation verification. S5. The control layer loads and optimizes the process parameters to execute the next round of stretching action, and the detection layer and digital twin layer simultaneously enter a new round of iterations of steps S2 to S4. S6. The process knowledge base automatically stores process parameters, test data, and optimization results after each iteration, and completes adaptive updates.

8. A skin stretching device comprising the system described in any one of claims 1-6.

Images