Welding workshop welding spot behavior simulation and twinborn parameter self-adaptive updating method and system

Abstract

The invention belongs to the technical field of intelligent manufacturing, industrial simulation and digital twinborn modeling, provides a welding workshop welding spot behavior simulation and twinborn parameter self-adaptive updating method and system, and aims to solve the technical problems that a traditional welding simulation system is poor in real-time performance, low in precision, weak in adaptability and the like. A multi-scale online simulation calculation module based on GPU acceleration is constructed, an MBD-FEM combined offline batch analysis platform is established, a parameter adaptive correction mechanism based on multi-source data fusion is designed, cooperative work of online and offline simulation is realized, a system health monitoring and fault self-healing mechanism is established, and the system health monitoring and fault self-healing efficiency is improved. And an effective technical solution is provided for intelligentization of an automobile welding process.

Description

Technical Field

[0001] This invention belongs to the field of intelligent manufacturing, industrial simulation and digital twin modeling technology, and in particular relates to a method and system for simulating weld point behavior and adaptively updating twin parameters in welding workshops. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] In modern automotive body-in-white welding production, the complex structure, dense weld points, and variable fixture constraints easily lead to fluctuations in connection consistency, tooling stiffness deviations, and uncontrolled assembly dimensions during the welding process. Traditional process simulation methods mostly rely on single offline modeling and finite element solutions, which cannot meet the modeling and prediction needs of complex welding environments with variable tasks, high cycle times, and dynamic data.

[0004] To address the aforementioned technical challenges, existing technologies have proposed a welding process aid design approach driven by digital twins. However, the following issues remain: Existing solutions focus on voltage, current, speed, and temperature monitoring and control decision generation for single-pass welds, failing to provide prediction and compensation mechanisms for the overall deformation of the assembly under different welding sequences, the coupling effects of fixture constraints and release springback, etc., thus making it difficult to support global optimization of fixture layout and welding paths; Traditional solutions rely on real-time sensor data and threshold / rule reasoning for quality judgment, without integrating a high-fidelity simulation model of temperature-stress-microstructure multiphysics field, which cannot fully reveal the internal thermo-mechanical-phase transformation coupling mechanism of the weld, resulting in limited prediction accuracy under complex working conditions. Summary of the Invention

[0005] To address at least one of the technical problems mentioned above, this invention provides a method and system for simulating weld joint behavior and adaptively updating twin parameters in a welding workshop. This method considers the collaborative virtual-real linkage between assembly and fixture, and meets the modeling and prediction needs of complex welding environments with variable tasks, high cycle times, and dynamic data.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: The first aspect of the present invention provides a method for simulating weld joint behavior and adaptively updating twin parameters in a welding workshop, comprising the following steps: Based on the constructed GPU-accelerated parallel computing architecture, the multi-scale weld point physics model is simulated and solved to obtain the structural behavior data of each structure during the welding process. The sensitivity index is calculated based on the structural behavior data of the welding process obtained from real-time simulation. Monitoring strategies for each process parameter are generated based on the sensitivity index results. The optimized process scheme is obtained by solving the process parameters obtained under the monitoring strategy and the constructed multi-objective optimization objective. The predicted values ​​of welding process parameters are generated based on the optimized process scheme. Welding process error is estimated by combining the measured values ​​of various process parameters obtained from real-time simulation and the predicted values ​​of welding process parameters. Based on the welding process error estimate, the corresponding adaptive parameter update algorithm is automatically matched according to the system characteristics. The parameters of the digital twin model are updated based on the matched adaptive parameter update algorithm. The updated digital twin model parameters are constrained and verified and their credibility is evaluated to obtain the verified and evaluated digital twin model parameters.

[0007] Furthermore, the simulation and solution of the constructed multi-scale weld point physics model yields structural behavior data for each component during the welding process, including: Establish a multibody system dynamic model of weld point-plate-fixture; The contact pair algorithm is used to process the contact relationship between the weld point and the plate, and between the plate and the fixture, to obtain the weld point constraint, the fixture constraint, and the plate flexible deformation constraint. By combining the multibody system dynamics model of weld point-plate-fixture and boundary constraints, a coupled physical field model spanning the micro, meso, and macro scales is established. The stiffness matrix of the coupled physical field model at the micro, meso, and macro scales is solved to obtain structural behavior data of nodal deformation, stiffness response, and fixture load changes during the welding process.

[0008] Furthermore, the step of generating monitoring strategies for each process parameter based on the sensitivity index results includes: When the total sensitivity index is greater than the set first threshold, it is a high-sensitivity parameter, and high-precision monitoring is adopted to improve the sampling rate. When the total sensitivity index is greater than the first threshold but less than the second threshold, it is considered a medium sensitivity parameter and standard monitoring is used. When the total sensitivity index is less than the second threshold, it is considered a low-sensitivity parameter, and low-frequency monitoring is adopted.

[0009] Furthermore, after obtaining the optimized process plan, feature extraction is performed based on the optimized process plan to identify failure modes. Based on the feature vectors of the failure modes, the risk level of the current process status is evaluated in real time. When the risk probability is greater than the set threshold, immediate parameter adjustment is triggered. When the risk probability is less than the set threshold, it is a medium-risk state, and preventive measures are initiated. At the same time, equipment maintenance strategies are formulated based on the occurrence probability of failure modes.

[0010] Furthermore, the formula for estimating welding process error is as follows: , , , in, For parameter error estimation, S The error-parameter sensitivity matrix, Represents the j-th parameter For the i-th state variable The impact, Represents the state deviation vector. These are the measured values ​​of each process parameter. Here are the simulation prediction values ​​for each process parameter, and W is the weight matrix.

[0011] Furthermore, the adaptive parameter update algorithm that automatically matches the system characteristics includes: If the system linearity is greater than the set linearity and the noise is less than the set noise value, choose the recursive least squares method. If the system has moderate nonlinearity and low noise level, choose the extended Kalman filter algorithm. If the system is a complex nonlinear system, choose the Bayesian estimation algorithm.

[0012] Furthermore, the method also includes constructing a multi-layered system monitoring and self-healing mechanism. The multi-layered system monitoring includes hardware monitoring, software monitoring, and business monitoring; the self-healing mechanism includes a service-level self-healing mechanism, a system-level self-healing mechanism, and an architecture-level self-healing mechanism.

[0013] A second aspect of the present invention provides a system for simulating weld joint behavior and adaptively updating twin parameters in a welding workshop, comprising: The multi-scale simulation module is used to simulate and solve the constructed multi-scale weld point physics model based on the constructed GPU-accelerated parallel computing architecture to obtain the structural behavior data of each structure during the welding process. The offline analysis module is used to calculate the sensitivity index based on the structural behavior data of the welding process obtained from real-time simulation, generate monitoring strategies for each process parameter based on the sensitivity index results, solve the optimized process scheme based on the process parameters obtained under the monitoring strategy and the constructed multi-objective optimization objective, and generate predicted values ​​of welding process parameters based on the optimized process scheme. Welding process error is estimated by combining the measured values ​​of various process parameters obtained from real-time simulation and the predicted values ​​of welding process parameters. Based on the welding process error estimate, the corresponding adaptive parameter update algorithm is automatically matched according to the system characteristics. The parameters of the digital twin model are updated based on the matched adaptive parameter update algorithm. The updated digital twin model parameters are constrained and verified and their credibility is evaluated to obtain the verified and evaluated digital twin model parameters.

[0014] A third aspect of the present invention provides a computer-readable storage medium.

[0015] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the above-described method for simulating weld joint behavior in a welding workshop and adaptively updating twin parameters.

[0016] A fourth aspect of the present invention provides a computer device.

[0017] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps in the above-described method for simulating weld joint behavior in a welding workshop and adaptively updating twin parameters.

[0018] Compared with the prior art, the beneficial effects of the present invention are: This invention considers the collaborative virtual-real linkage of assembly and fixture, combines multi-physics high-fidelity simulation model for multi-scale modeling, introduces parameter adaptive update mechanism, and through intelligent parameter correction mechanism, can automatically adapt to factors such as equipment aging and environmental changes. The model maintains stable accuracy over long-term operation, meeting the modeling and prediction needs of complex welding environments with variable tasks, high cycle time and dynamic data.

[0019] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0020] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0021] Figure 1 This is a flowchart of the welding workshop weld joint behavior simulation and twin parameter adaptive update method provided in the embodiments of the present invention; Figure 2 This is a schematic diagram of a coupled physical field model spanning three scales—microscopic, mesoscopic, and macroscopic—provided in an embodiment of the present invention. Detailed Implementation

[0022] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0023] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0024] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0025] Example 1 like Figure 1 As shown, this embodiment provides a method for simulating weld joint behavior and adaptively updating twin parameters in a welding workshop, including the following steps: Step 1: Based on the constructed GPU-accelerated parallel computing architecture, the constructed multi-scale weld point physics model is simulated and solved to obtain the structural behavior data of each structure during the welding process; Specifically, the steps include the following: Step 101: Establish a multibody system dynamic model of the weld point-plate-fixture; In this embodiment, the process of establishing the multibody system dynamic model of weld point-plate-fixture includes: automatically generating multi-scale finite element meshes based on the CAD model, using a refined mesh (element size 0.2-0.5mm) in the weld point region and a coarser mesh (element size 2-5mm) in non-critical regions, and establishing the multibody system dynamic equations based on generalized coordinates: Specifically, the dynamic equations of a multibody system based on generalized coordinates are established: , , in, For generalized coordinate vectors, This is a quality matrix, including solder joint quality, sheet metal quality, and fixture quality. The matrix of Coriolis force and centrifugal force. Here is the stiffness matrix. Let Lagrange multiplier vectors be used. These are the constraint equations; Step 102: Use the contact pair algorithm to process the contact relationship between the weld point and the plate, and between the plate and the fixture to obtain the weld point constraint, the fixture constraint, and the plate flexible deformation constraint. In this embodiment, the weld point constraint is that the weld point rigidly connects the two plate nodes, including positional constraints and rotational constraints, as shown below: , , in, Indicates positional constraints. Indicates corner constraint, Indicates the first i The spatial coordinates of each plate node (or weld node) Indicates and The spatial location of another connected node (e.g., an adjacent plate node or a clamp node). This represents the rotational degree of freedom of the i-th node. Indicates and The rotation state of the connected nodes.

[0026] The fixture constraint is a unilateral constraint applied by the fixture to the plate, which is achieved using the Kuhn-Tucker condition. The flexible deformation constraint of the sheet metal is represented by a modal-based reduced-order model, as follows: , in, To constrain the flexible deformation of the sheet metal, The modal matrix, For modal coordinates, Step 103: Combine the multibody system dynamics model of weld point-plate-fixture and boundary constraints to establish a coupled physical field model spanning the micro-meso-macro scales; In this embodiment, the microscale physical field model (0.1-10μm) is a grain-level finite element model that considers crystal orientation, grain boundary effects, and phase transition dynamics. Specifically, the Johnson-Cook constitutive relation is used to describe the mechanical behavior of materials at high temperature and high strain rate. The grain growth dynamics described by the Arrhenius equation and the martensitic phase transformation model are integrated to calculate the phase transformation strain and residual stress distribution, providing boundary conditions for the microscale. Microscale modeling (10μm-1mm): Combining the boundary constraints obtained in step 102, and based on the Rosenthal analytical solution, a temperature field distribution model of the heat-affected zone is established. Based on the established temperature field distribution model of the heat-affected zone, the microstructure type and distribution are determined according to the continuous cooling transition (CCT) diagram. The hardness distribution and mechanical property gradient of the heat-affected zone are calculated, and a homogenization model from microstructure to microstructure is established. Macroscale modeling (1mm-1m): Material property distribution and microscopic residual stress field distribution are obtained from mesoscale modeling. Using the microscopic residual stress field distribution as the initial stress, a macroscopic finite element model considering microstructure inhomogeneity is established by combining the multibody system dynamics model of weld point-plate-fixture. This achieves automatic transfer and coupled solution of multi-scale information, ultimately resulting in a coupled physical field model spanning the micro-meso-macroscale. Figure 2 As shown.

[0027] Step 104: Solve the stiffness matrix of the coupled physical field model at the micro, meso, and macro scales to obtain the structural behaviors such as nodal deformation, stiffness response, and fixture load changes during the welding process; In this embodiment, the CUDA architecture is used to parallelize the stiffness matrix assembly and solution, and supports the sparse matrix storage format (CSR format); multi-GPU collaborative computing is realized, with a single GPU processing up to 1 million nodes, and multi-GPU expansion supports a scale of 10 million nodes; asynchronous computing flow technology is adopted, and GPU computing and data transmission are executed in parallel, improving computing efficiency by 60%.

[0028] Step 2: Calculate the sensitivity index based on the historical welding task data obtained from real-time simulation, generate monitoring strategies for each process parameter based on the sensitivity index results, solve the optimized process scheme based on the process parameters obtained under the monitoring strategies and the constructed multi-objective optimization objective, and generate predicted values ​​of welding process parameters based on the optimized process scheme. Specifically, the steps include the following: Step 201: Obtain historical welding task data and preprocess the obtained historical welding task data; In this embodiment, the historical welding task data includes robot motion trajectory (6-DOF pose sequence), welding process parameters (current I, voltage U, speed v, pressure F), fixture configuration parameters, and quality inspection data. The acquired historical welding task data is preprocessed, including data quality assessment and repair: outliers are identified using the 3σ criterion, missing data is filled using cubic spline interpolation, and data integrity is required to be >95%; Finally, a process parameter-quality index mapping database was established, with a cumulative sample size of >10,000 solder joints.

[0029] Step 202: Establish a multi-body system model of robot-fixture-workpiece, perform simulation analysis based on finite element structure, and obtain structural response data by coupled solution; In this embodiment, when constructing the robot-fixture-workpiece multibody system model, nonlinear factors such as joint stiffness, friction, and clearance are considered, and the model is constructed in conjunction with the motion equations. The motion equations adopt the Lagrange equations: Where L is the Lagrange function, For generalized coordinates, For generalized force; In this embodiment, when performing analysis based on the finite element structure, an explicit dynamic solver is used, and the time integration adopts the central difference method: ,in M For the quality matrix, F As an external force, Fint For internal force, This represents the node displacement vector at the current moment; The specific strategy for coupled solution is to use a strongly coupled iterative algorithm: the MBD solver calculates the fixture motion, and the FEM solver calculates the structural response. Data exchange is achieved through force-displacement boundary conditions.

[0030] Step 203: Calculate the sensitivity index of each process parameter to the quality index, generate a monitoring strategy for each process parameter based on the sensitivity index results, and obtain the optimized process scheme based on the process parameters obtained under the monitoring strategy and the constructed multi-objective optimization objective. Specifically, the steps include the following: Step 2031: Calculate the sensitivity index of each process parameter to the quality index, and generate a monitoring strategy for each process parameter based on the sensitivity index results; In this embodiment, when calculating the sensitivity index of each process parameter to the quality index, the Sobol variance decomposition method is used to calculate the sensitivity index of each process parameter to the quality index. A priority list is generated based on the sensitivity index structure, and different simulation parameter monitoring strategies are adopted according to different priorities. Specifically, when the total sensitivity index is >0.1, it is considered a high-sensitivity parameter, and high-precision monitoring is adopted, with the sampling rate increased to 1000Hz; When the total sensitivity index ranges from 0.05 to 0.1, it is considered a moderately sensitive parameter, and standard monitoring should be used. When the total sensitivity index is <0.05, it is considered a low-sensitivity parameter, and low-frequency monitoring should be used. High-sensitivity parameters are updated in batches with a convergence threshold of 0.001; medium-sensitivity parameters are updated daily with a convergence threshold of 0.005; and low-sensitivity parameters are updated weekly with a convergence threshold of 0.01. Computational resources are dynamically allocated based on the importance of parameters. High-sensitivity parameters have a greater weight than medium-sensitivity parameters, which in turn have a greater weight than low-sensitivity parameters. For example, high-sensitivity parameters are allocated 80% of the weight to ensure the accurate calculation of key parameters.

[0031] Step 2032: Based on the process parameters obtained under the monitoring strategy and the constructed multi-objective optimization objective, the optimized process scheme is obtained; In this embodiment, the constructed multi-objective optimization objectives include minimizing deformation, homogenizing stress, and minimizing cycle time; Based on the constructed multi-objective optimization objectives, and based on process parameters, objective function values, dominance level, congestion distance, etc., the NSGA-II algorithm is used to extract the Pareto optimal solution and the corresponding optimized process scheme, providing the optimal welding process parameter recommendation scheme for real-time production; Step 2033: Based on the optimized process scheme, feature extraction is performed to identify failure modes. According to the probability of occurrence of failure modes, corresponding maintenance plans are formulated and failure mode knowledge is integrated into the parameter correction constraints of the coupled physical field model at three scales. Feature extraction is performed based on the optimized process scheme. The extracted features are combined with machine learning algorithms such as Support Vector Machine (SVM) to obtain classification results. The classification results are compared with the names and corresponding feature vectors of the set failure modes to identify potential failure modes. Based on the feature vectors of the failure modes, the risk level of the current process status is evaluated in real time. High-risk status (probability > 0.3) triggers immediate parameter adjustment. When the risk probability is less than the set threshold, it is a medium-risk status and preventive measures are initiated. Equipment maintenance strategies are formulated according to the occurrence probability of failure modes. High-probability failure modes (probability > 0.1) are formulated with high-frequency maintenance plans to reduce the risk of failure. Failure mode knowledge is integrated into parameter correction constraints to limit the range of parameters that are prone to failure and improve system stability.

[0032] Step 3: Combine the measured values ​​of each process parameter obtained from real-time simulation with the predicted values ​​of welding process parameters to obtain the welding process error estimate. Based on the welding process error estimate, automatically match the corresponding adaptive parameter update algorithm according to the system characteristics. Update the digital twin model parameters based on the matched adaptive parameter update algorithm. Perform constraint verification and credibility evaluation on the updated digital twin model parameters to obtain the verified and evaluated digital twin model parameters. Specifically, the steps include the following: Step 301: Calculate the state deviation vector by combining the measured values ​​and predicted values ​​of each process parameter obtained from real-time simulation, and obtain the parameter error estimate by combining the state deviation vector and the established error-parameter sensitivity matrix. Define the state deviation vector ,in, These are the measured values ​​of each process parameter. These are predicted values ​​for welding process parameters; Establish the error-parameter sensitivity matrix S, where , Let represent the element in the i-th row and j-th column of matrix S, and let represent the j-th parameter. For the i-th state variable The impact; Error estimation for each process parameter: , where W is the weight matrix, taking into account differences in measurement accuracy.

[0033] Step 302: Based on the error estimation of each process parameter, and combined with the characteristics of the system, recommend a scheme to automatically match the corresponding adaptive parameter update algorithm. Update the digital twin model parameters based on the matched adaptive parameter update algorithm, and perform constraint verification and credibility evaluation on the updated digital twin model parameters to obtain the verified and evaluated digital twin model parameters. In this embodiment, the system's specific characteristics include: If the system linearity is greater than the first set linearity (>0.9) and the noise is less than the set noise value (<0.1), select the recursive least squares (RLS) algorithm to update the parameters; If the system has moderate nonlinearity (less than the first set linearity) and low noise level (<0.1): Select the Extended Kalman Filter (EKF) algorithm to update the parameters; If the system is a complex nonlinear system: select the Bayesian estimation algorithm to update the parameters; It should be noted that the specific linearity value and noise threshold can be set by those skilled in the art according to the actual scenario; Specifically, when performing constraint verification on the updated parameters, the Sequential Quadratic Programming (SQP) algorithm is used to solve the constrained optimization problem: , in, To fit the error target, The results obtained from the simulation response, This refers to the physical constraints, or the reasonable range of values ​​for material parameters. To design constraints; The rationality of the parameters is verified by detecting the magnitude of change, the convergence test, and the physical consistency test. Specifically, the magnitude of change test is: |Δθ| / |θ| < 20%; the convergence test is: the variance decreases after 10 consecutive updates; and the physical consistency test is: the parameters satisfy the basic relations of material mechanics. The reliability assessment of the updated parameters includes: statistical confidence level, uncertainty quantification, and model validation metrics; specifically, statistical confidence level: calculating a 95% confidence interval based on the prediction error; uncertainty quantification: evaluating the propagation of parameter uncertainty using the Monte Carlo method; model validation metrics: coefficient of determination. >0.90; Root mean square error: RMSE <0.05mm; Mean absolute percentage error: MAPE <5%; By establishing a parameter identification and adaptive correction system based on multi-source data fusion, dynamic matching between the twin model and the physical object can be achieved.

[0034] Step 4: Establish a two-way data flow and collaborative working mechanism for online simulation and offline analysis; A central data management bus is constructed to uniformly manage data exchange between all steps, including four core data storage components: real-time database, historical database, model parameter library, and configuration parameter library; a standardized data flow protocol is established to ensure effective collaboration between each step. Data flow between Step 1 and Step 2: The structural behavior data (deformation field, stress field, process parameters, quality indicators) from real-time simulation are transferred to the offline analysis module for historical data accumulation and in-depth analysis; Data flow between step 1 and step 3: The deviation data between the predicted value and the measured value is transmitted to the parameter correction module as the basis for updating the model parameters; Step 2 → Step 1 data flow: Optimized process parameters, parameter monitoring priorities, and failure warning thresholds are transmitted to real-time simulation to guide simulation configuration and monitoring strategies; The data flow between Step 2 and Step 3 is as follows: the weight matrix of sensitivity analysis, the parameter constraints of multi-objective optimization, and the correction strategy of failure mode are passed to the parameter correction module. Step 3 → Step 1 data flow: The corrected material properties, contact stiffness, boundary conditions and other model parameters are passed to the real-time simulation to update the simulation model.

[0035] Data synchronization and coordination include timestamp synchronization, data format standardization, and load balancing. Timestamp synchronization specifically adopts the IEEE 1588 Precision Time Protocol (PTP), with synchronization accuracy reaching the microsecond level; Unified data format, establishing a standardized data exchange format (JSON / XML) to support heterogeneous system integration; Load balancing dynamically allocates simulation tasks to different computing nodes based on computing resource usage.

[0036] Model consistency maintenance includes: version control: implementing version management of model parameters, supporting parameter rollback and historical traceability; Incremental update: Only the changed parameters are transmitted, reducing network traffic and update time; Conflict detection: When there are conflicts between online and offline model parameters, a weighted fusion strategy is used to resolve them.

[0037] Step 5: Construct a multi-layered system monitoring and self-healing mechanism to ensure the stable operation of the simulation system; System status monitoring: Hardware monitoring: CPU utilization, memory usage, GPU utilization, network bandwidth, and storage space; Software monitoring: process status, service response time, computational accuracy, model convergence; Business monitoring: simulation success rate, prediction accuracy, parameter update frequency, and user satisfaction.

[0038] Since we can obtain the following from step 2: High-risk failure mode (probability of occurrence > 0.1): Set high-frequency monitoring, increase the monitoring frequency by 2 times, and set strict early warning thresholds; Medium-risk failure modes (probability 0.05-0.1): Standard monitoring frequency, establish preventive maintenance plans; Low-risk failure modes (probability < 0.05): Low-frequency monitoring, with a focus on long-term trend changes; Constructing a multi-layered self-healing mechanism: Level 1 - Service-level self-healing mechanism, including: Service restart: Abnormal services will be restarted automatically, with a maximum of 3 attempts; Parameter rollback: Automatically rolls back to the most recent stable parameter state; Algorithm switching: The correction algorithm in step 3 is automatically switched (RLS→EKF→Bayes). Level 2 - System-level self-healing mechanism, including: Load Shifting: Tasks from a failed node are automatically transferred to a healthy node; Resource reallocation: Dynamically adjust the resource allocation ratio for each step; Degraded operation: When resources are insufficient, the simulation accuracy is reduced to ensure system availability; Level 3 - Architecture-level self-healing mechanism, including: Backup and recovery: Regularly back up critical data and models for rapid recovery in case of failure; Redundancy failover: Enable standby compute nodes and data storage; Emergency Mode: Enables simplified simulation mode to ensure basic functions are available.

[0039] The following case illustrates the specific implementation process: S1. Deploy the system of this invention on the side panel automatic welding production line of a certain automobile manufacturing enterprise: 1. Hardware configuration and deployment: Edge computing device: It adopts the NVIDIA Jetson AGX Xavier industrial computing platform, equipped with a 512-core Volta GPU and an 8-core ARM v8.2 processor, and features 32GB of LPDDR4x memory and 512GB of NVMe SSD storage; Sensor network: Deploy 20 high-precision laser displacement sensors (accuracy ±2μm), 15 six-dimensional force / torque sensors (accuracy 0.1%FS), and 30 temperature sensors (accuracy ±0.1°C); Network architecture: It adopts Gigabit industrial Ethernet, supports IEEE 802.1AS time synchronization protocol, and has a network latency of <5ms.

[0040] 2. Simulation model establishment: Geometric model: Import the side CAD model, which contains 183 weld points, automatically generate a hybrid mesh with a total of approximately 1.5 million nodes and approximately 800,000 elements; Material model: The Johnson-Cook constitutive model of high-strength steel DP590 is adopted, with parameters: A=620MPa, B=550MPa, C=0.025, n=0.41, m=1.0; Boundary conditions: Set 12 clamping constraint points, consider the flexible deformation of the clamps, and the elastic modulus is 200GPa.

[0041] 3. Real-time simulation implementation: Computational optimization: A GPU-accelerated PCG solver is used, and the preconditioners employ incomplete Cholesky decomposition. Parallel strategy: Divide the grid into 8 subdomains and use the domain decomposition method for parallel computation; Time step: Adaptive time step, initial value Δt=1ms, dynamically adjusted according to convergence.

[0042] 4. Performance verification results: Calculation speed: The single simulation iteration time is 35ms, which meets the requirements of real-time calculation; Accuracy verification: Compared with high-precision offline simulation, displacement error <0.08mm, stress error <5%; Stability: It ran continuously for 72 hours without failure, and the calculation accuracy remained stable.

[0043] S2. Offline batch optimization analysis of multiple process schemes For the development of welding processes for a new vehicle model, offline batch analysis functionality was employed. 1. Data preparation and modeling: Historical data: Collect historical process data for 5000 solder joints, including current 8-12kA, voltage 2-4V, soldering time 200-800ms, and electrode pressure 2-6kN; Multiple-option design: orthogonal experimental design was adopted. Consideration should be given to factors such as welding sequence, clamping force, and welding parameters; MBD Model: A multibody system model containing 6 robots and 24 sets of grippers is established, with a total of 18,000 degrees of freedom.

[0044] 2. Batch simulation execution: Parallel computing: 20 CPU cores are used for parallel computing, and the computing time for a single solution is 4 hours; Simulation management: Automatically submits and monitors simulation tasks, supports breakpoint resume and result collection; Post-processing of results: Automatically generates deformation field distribution, stress concentration areas, and interference inspection reports.

[0045] 3. Application of intelligent analysis results: Sensitivity analysis: The sensitivity index of welding sequence to final deformation is 0.73, and the sensitivity of clamping force is 0.41. Based on this, the real-time monitoring priority is set. Multi-objective optimization: Using the NSGA-II algorithm, after 200 generations of evolution, the Pareto optimal solution set is obtained, generating a process library containing 15 high-quality solutions, 12 high-efficiency solutions, and 18 balanced solutions; Failure Mode Identification: Identifies three main failure modes: "electrode wear", "thermal deformation" and "fixture failure", and sets corresponding warning thresholds.

[0046] 4. Subsequent applications of the analysis results: The parameter monitoring strategy has been updated, improving the welding current monitoring accuracy to ±0.1A and the clamping force monitoring frequency to 500Hz. Set calibration weights; calibrate welding sequence-related parameters for each batch, and calibrate other parameters daily. Update health monitoring rules, double the frequency of electrode wear monitoring, and set up preventive maintenance plans.

[0047] III. Engineering Verification of Parameter Adaptive Correction Verify the effectiveness of the parameter adaptive correction mechanism during the production process: 1. Initial model calibration: Parameter identification: Based on test data from 200 standard weld points, key parameters were identified: fixture stiffness. N / m, contact stiffness K_c=1.8×10^8 N / m, damping coefficient ζ=0.04; Model validation: Initial model prediction accuracy RMSE = 0.12 mm, coefficient of determination =0.89.

[0048] 2. Adaptive correction process: Data acquisition: 30 days of continuous production data collection, including measurement results of 12,000 solder joints; Algorithm selection: Based on the analysis results, the system linearity is 0.85, so the EKF algorithm is selected for correction; Parameter update: Extended Kalman filter is used, and the model parameters are updated once a day; Constraint check: All parameter updates meet the physical constraints, and the change range is <15%.

[0049] 3. Verification of calibration effect: Accuracy Improvement: After 30 days of adaptive correction, the prediction accuracy RMSE = 0.045mm. =0.96; Stability verification: The parameters converge well, with parameter changes of less than 2% in the last 10 days; Robustness test: The system can still maintain high accuracy under interference such as fixture wear and changes in ambient temperature.

[0050] 4. Systematic application of calibration results: Feedback to S1: Update material properties to adjust the elastic modulus from 200 GPa to 195 GPa, and the contact stiffness from... Adjusted to N / m; Feedback to S2: Correction success rate 95.2%, parameters stable, it is recommended to maintain the current analysis strategy; S4. Industrial Validation of Micro-deformation Modeling and Assembly Deviation Prediction For a welding line of a battery pack tray for a new energy vehicle, the multi-scale modeling and assembly deviation prediction techniques were verified: 1. Implementation of multi-scale modeling: Microscale: A microstructure model containing 500,000 grains with a grain size of 5-50 μm was established, taking into account phase transformation kinetics and residual stress evolution; Detailed scale: Modeling the heat-affected zone, considering the microstructure and property gradient within a range of 0-10mm from the weld, and establishing a hardness distribution map; Macro scale: Overall pallet assembly model, including 156 welding points, 2 million nodes, and integrating multi-scale information.

[0051] 2. Verification of assembly deviation prediction: A multi-source error coupling model was established: manufacturing error ±0.05mm, thermal deformation ±0.08mm, assembly deformation ±0.06mm, and aging deformation ±0.03mm. Monte Carlo simulation: 100,000 sampling analyses predict the assembly deviations at the four corners of the pallet to be +0.08mm, -0.06mm, +0.11mm, and -0.09mm, respectively. Actual measurement verification: The measured values ​​were +0.09mm, -0.07mm, +0.12mm, and -0.08mm, respectively, and the prediction error was <0.02mm, which meets the accuracy requirement of ±0.1mm.

[0052] 3. Fixture compensation effect: Fixture deformation prediction: The maximum deformation point is 0.18mm, and a flexible compensation strategy is established; Compensation Implementation: By optimizing the clamping force distribution at the 12 clamping points, the assembly accuracy was improved from ±0.25mm to ±0.08mm after compensation. Fixture compensation effectiveness: 67% accuracy improvement, fixture springback prediction error <±0.05mm.

[0053] 4. Comparison with traditional methods: Traditional rigid body method: prediction error ±0.32mm, only considering the accumulation of geometric tolerances; The method of this invention has a prediction error of ±0.06 mm and takes into account multi-scale physical mechanisms. Accuracy improvement: An 81% improvement demonstrates the superiority of multi-scale modeling.

[0054] 5. Production effect verification: Quality improvements: Scrap rate decreased from 2.3% to 0.4%, and rework rate decreased from 5.1% to 1.2%; Economic benefits: Quality costs are reduced by approximately 2.8 million yuan annually, and equipment utilization rate is increased by 12%; Technical specifications: Real-time simulation response time 32ms, parameter correction accuracy improved by 73%, system operated stably for 6 months without failure.

[0055] Example 2 This embodiment provides a system for simulating weld joint behavior and adaptively updating twin parameters in a welding workshop, including: The multi-scale simulation module is used to simulate and solve the constructed multi-scale weld point physics model based on the constructed GPU-accelerated parallel computing architecture to obtain the structural behavior data of each structure during the welding process. The offline analysis module is used to calculate the sensitivity index based on the structural behavior data of the welding process obtained from real-time simulation, generate monitoring strategies for each process parameter based on the sensitivity index results, solve the optimized process scheme based on the process parameters obtained under the monitoring strategy and the constructed multi-objective optimization objective, and generate predicted values ​​of welding process parameters based on the optimized process scheme. Welding process error is estimated by combining the measured values ​​of various process parameters obtained from real-time simulation and the predicted values ​​of welding process parameters. Based on the welding process error estimate, the corresponding adaptive parameter update algorithm is automatically matched according to the system characteristics. The parameters of the digital twin model are updated based on the matched adaptive parameter update algorithm. The updated digital twin model parameters are constrained and verified and their credibility is evaluated to obtain the verified and evaluated digital twin model parameters.

[0056] It should be noted that the specific implementation of the welding workshop weld joint behavior simulation and twin parameter adaptive update system in this embodiment of the invention is similar to the specific implementation of the welding workshop weld joint behavior simulation and twin parameter adaptive update method in this embodiment of the invention. Please refer to the description in the method section for details. In order to reduce redundancy, it will not be repeated here.

[0057] Example 3 This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in the above-described method for simulating weld joint behavior in a welding workshop and adaptively updating twin parameters.

[0058] Example 4 This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the above-described method for simulating weld joint behavior in a welding workshop and adaptively updating twin parameters.

[0059] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of hardware embodiments, software embodiments, or embodiments combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0060] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0061] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0062] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0063] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0064] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for simulating weld joint behavior and adaptively updating twin parameters in a welding workshop, characterized in that, Includes the following steps: Based on the constructed GPU-accelerated parallel computing architecture, the multi-scale weld point physics model is simulated and solved to obtain the structural behavior data of each structure during the welding process. The sensitivity index is calculated based on the structural behavior data of the welding process obtained from real-time simulation. Monitoring strategies for each process parameter are generated based on the sensitivity index results. The optimized process scheme is obtained by solving the process parameters obtained under the monitoring strategy and the constructed multi-objective optimization objective. The predicted values ​​of welding process parameters are generated based on the optimized process scheme. Welding process error is estimated by combining the measured values ​​of various process parameters obtained from real-time simulation and the predicted values ​​of welding process parameters. Based on the welding process error estimate, the corresponding adaptive parameter update algorithm is automatically matched according to the system characteristics. The parameters of the digital twin model are updated based on the matched adaptive parameter update algorithm. The updated digital twin model parameters are constrained and verified and their credibility is evaluated to obtain the verified and evaluated digital twin model parameters.

2. The method for simulating weld joint behavior and adaptively updating twin parameters in a welding workshop as described in claim 1, characterized in that, The simulation and solution of the constructed multi-scale weld point physics model yields various structural behavior data during the welding process, including: Establish a multibody system dynamic model of weld point-plate-fixture; The contact pair algorithm is used to process the contact relationship between the weld point and the plate, and between the plate and the fixture, to obtain the weld point constraint, the fixture constraint, and the plate flexible deformation constraint. By combining the multibody system dynamics model of weld point-plate-fixture and boundary constraints, a coupled physical field model spanning the micro, meso, and macro scales is established. The stiffness matrix of the coupled physical field model at the micro, meso, and macro scales is solved to obtain structural behavior data of nodal deformation, stiffness response, and fixture load changes during the welding process.

3. The method for simulating weld joint behavior and adaptively updating twin parameters in a welding workshop as described in claim 1, characterized in that, The strategy for generating monitoring strategies for each process parameter based on the sensitivity index results includes: When the total sensitivity index is greater than the set first threshold, it is a high-sensitivity parameter, and high-precision monitoring is adopted to improve the sampling rate. When the total sensitivity index is greater than the first threshold but less than the second threshold, it is considered a medium sensitivity parameter and standard monitoring is used. When the total sensitivity index is less than the second threshold, it is considered a low-sensitivity parameter, and low-frequency monitoring is adopted.

4. The method for simulating weld joint behavior and adaptively updating twin parameters in a welding workshop as described in claim 1, characterized in that, After obtaining the optimized process plan, feature extraction is performed based on the optimized process plan to identify failure modes. Based on the feature vectors of the failure modes, the risk level of the current process status is evaluated in real time. When the risk probability is greater than the set threshold, immediate parameter adjustment is triggered. When the risk probability is less than the set threshold, it is a medium-risk state, and preventive measures are initiated. At the same time, equipment maintenance strategies are formulated based on the occurrence probability of failure modes.

5. The method for simulating weld joint behavior and adaptively updating twin parameters in a welding workshop as described in claim 1, characterized in that, The formula for estimating welding process error is: ， ， ， in, For parameter error estimation, S The error-parameter sensitivity matrix, Represents the j-th parameter For the i-th state variable The impact, Represents the state deviation vector. These are the measured values ​​of each process parameter. Here are the simulation prediction values ​​for each process parameter, and W is the weight matrix.

6. The method for simulating weld joint behavior and adaptively updating twin parameters in a welding workshop as described in claim 1, characterized in that, The adaptive parameter update algorithm that automatically matches the system characteristics includes: If the system linearity is greater than the set linearity and the noise is less than the set noise value, choose the recursive least squares method. If the system has moderate nonlinearity and low noise level: choose the extended Kalman filter algorithm; If the system is a complex nonlinear system, choose the Bayesian estimation algorithm.

7. The method for simulating weld joint behavior and adaptively updating twin parameters in a welding workshop as described in claim 1, characterized in that, The method also includes constructing a multi-layered system monitoring and self-healing mechanism. The multi-layered system monitoring includes hardware monitoring, software monitoring, and business monitoring; the self-healing mechanism includes service-level self-healing mechanism, system-level self-healing mechanism, and architecture-level self-healing mechanism.

8. A simulation system for weld point behavior in a welding workshop and an adaptive update system for twin parameters, characterized in that, include: The multi-scale simulation module is used to simulate and solve the constructed multi-scale weld point physics model based on the constructed GPU-accelerated parallel computing architecture to obtain the structural behavior data of each structure during the welding process. The offline analysis module is used to calculate the sensitivity index based on the structural behavior data of the welding process obtained from real-time simulation, generate monitoring strategies for each process parameter based on the sensitivity index results, solve the optimized process scheme based on the process parameters obtained under the monitoring strategy and the constructed multi-objective optimization objective, and generate predicted values ​​of welding process parameters based on the optimized process scheme. Welding process error is estimated by combining the measured values ​​of various process parameters obtained from real-time simulation and the predicted values ​​of welding process parameters. Based on the welding process error estimate, the corresponding adaptive parameter update algorithm is automatically matched according to the system characteristics. The parameters of the digital twin model are updated based on the matched adaptive parameter update algorithm. The updated digital twin model parameters are constrained and verified and their credibility is evaluated to obtain the verified and evaluated digital twin model parameters.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps in the welding workshop weld joint behavior simulation and twin parameter adaptive update method as described in any one of claims 1-7.

10. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the welding workshop weld joint behavior simulation and twin parameter adaptive update method as described in any one of claims 1-7.

Images