Defect feedback and digital simulation closed-loop control system and method for stamping process

Abstract

The invention discloses a stamping process-oriented defect feedback and digital simulation closed-loop control system and method, and relates to the technical field of closed-loop control. The system comprises a working condition synchronous sensing and feature coding module, a digital simulation prediction module, a defect feedback and decision control module, a process parameter execution module and an online learning and model evolution module. According to the method, multi-source operation data are collected, a working condition feature packet is generated, material parameters are dynamically matched after a formed digital simulation model is input, initial disturbance is injected, and defect prediction parallel to physical stamping is achieved; calculating a prediction confidence coefficient by combining measured data, and generating a precise process adjustment decision by linking defect risk distribution and applying the decision to a next period; and based on actually measured and predicted errors, continuously updating a material parameter library and forming digital simulation model parameters through online learning. According to the method, the problems that traditional stamping simulation and reality are disjointed, and process adjustment is blind are effectively solved, the product defect rate is greatly reduced, and meanwhile the production efficiency and the quality stability are improved.

Description

Technical Field

[0001] This invention relates to the field of closed-loop control technology, and more specifically, to a defect feedback and digital simulation closed-loop control system and method for stamping processes. Background Technology

[0002] In the field of stamping, in order to reduce the incidence of forming defects such as cracking, wrinkling, and excessive springback, and to improve product quality stability and production efficiency, existing technologies typically collect some operating data of stamping equipment through sensors, combine them with offline digital simulation tools to preset process parameters, and make local adjustments to equipment parameters based on simple feedback logic, thereby adapting to basic changes in production conditions and providing basic quality assurance for the stamping process.

[0003] However, existing technologies still have significant shortcomings in practical applications: data acquisition is often fragmented, lacking synchronous fusion of multi-source data and accurate feature extraction, making it difficult to comprehensively capture multi-dimensional disturbance information such as sheet positioning deviation and mold temperature changes; the material and model parameters relied upon by digital simulation are mostly fixed settings, without dynamic matching with real-time working conditions, resulting in a disconnect between simulation predictions and actual production processes; there is a lack of confidence assessment mechanisms for prediction results, and process parameter adjustment decisions rely heavily on manual experience or single threshold judgments, exhibiting strong arbitrariness; at the same time, simulation models and parameters are not dynamically optimized with the production process, and prediction accuracy gradually declines after long-term use, making it unable to continuously adapt to complex and ever-changing production scenarios, and failing to fundamentally address the core needs of defect prevention and efficiency improvement. Summary of the Invention

[0004] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a defect feedback and digital simulation closed-loop control system and method for the stamping process. The following solutions address the problems mentioned in the background art, such as the disconnect between simulation prediction and actual conditions, high defect rates, and low production efficiency caused by the lack of closed-loop parameter adjustment.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a closed-loop control system and method for defect feedback and digital simulation in the stamping process, comprising: a working condition synchronous sensing and feature encoding module: used to synchronously collect multi-source operating data at the beginning of each stamping cycle, and perform fusion processing and feature encoding to generate working condition codes and corresponding working condition feature packages;

[0006] Digital simulation prediction module: used to receive the working condition feature package, dynamically call the matching material parameter set, inject the initial disturbance conditions into the forming digital simulation model, perform simulation calculations in parallel with the physical stamping process, and output simulation prediction results characterizing the trend of stamping defects.

[0007] Defect feedback and decision control module: used to receive the simulation prediction results and obtain measured data, calculate the simulation prediction confidence through consistency analysis, and generate process parameter adjustment decisions based on defect risk distribution information and simulation prediction confidence.

[0008] Process parameter execution module: used to make adjustment decisions based on the process parameters, update the parameters of the stamping equipment, and apply the adjusted process parameters to the next stamping cycle.

[0009] Online learning and model evolution module: used to acquire actual quality inspection data, compare and analyze it with the corresponding simulation prediction results, and update the online material parameter library and forming digital simulation model parameters on a rolling basis according to the prediction error, forming a closed-loop control system.

[0010] A defect feedback and digital simulation closed-loop control method for the stamping process includes:

[0011] S1: At the beginning of each stamping cycle, multi-source operating data is collected synchronously, and fusion processing and feature encoding are performed to generate working condition codes and corresponding working condition feature packages.

[0012] S2: Receive the working condition feature package, dynamically call the matching material parameter set, inject the initial disturbance conditions into the forming digital simulation model, perform simulation calculations in parallel with the physical stamping process, and output simulation prediction results that characterize the trend of stamping defect occurrence.

[0013] S3: Receive the simulation prediction results and obtain the measured data, calculate the simulation prediction confidence through consistency analysis, and generate process parameter adjustment decisions based on defect risk distribution information and simulation prediction confidence.

[0014] S4: Based on the process parameter adjustment decision, update the parameters of the stamping equipment so that the adjusted process parameters are applied to the next stamping cycle;

[0015] S5: Obtain actual quality inspection data, compare and analyze it with the corresponding simulation prediction results, and update the online material parameter library and forming digital simulation model parameters on a rolling basis according to the prediction error to form a closed-loop control system.

[0016] The technical effects and advantages of this invention are as follows:

[0017] 1. This invention dynamically calls a set of matching material parameters, injects initial disturbance conditions into the forming digital simulation model and performs parallel simulation calculations, realizing defect trend prediction synchronized with the physical stamping process, making defect early warning more forward-looking and targeted, and making up for the shortcomings of traditional offline simulation being disconnected from actual production scenarios.

[0018] 2. This invention calculates the confidence level of simulation prediction by comparing simulation prediction results with measured data, and generates differentiated process parameter adjustment decisions by combining defect risk distribution information. This ensures the scientific nature and accuracy of process adjustment, avoids the blindness of traditional process adjustment that relies on human experience, and improves the adaptability of parameter adjustment.

[0019] 3. This invention applies the adjustment decision to the next stamping cycle in a timely manner through the process parameter execution module, and builds a feedback closed-loop control process of "perception-simulation-decision-execution". It can quickly compensate for the disturbance effect in the stamping process, ensure the continuity and stability of the production process, and reduce production interruptions or quality problems caused by parameter mismatch.

[0020] 4. This invention utilizes an online learning mechanism to continuously update the material parameter library and simulation model parameters based on the error between actual quality inspection data and simulation prediction results. This enables continuous optimization and iteration of the model and parameters, allowing the system's prediction accuracy and control effect to gradually improve with the production process. This forms a self-learning closed loop, enabling long-term adaptation to complex and ever-changing stamping production scenarios. Ultimately, this significantly reduces product defect rates and improves production efficiency and quality stability. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the overall system structure of the present invention;

[0022] Figure 2 This is a flowchart illustrating the process of obtaining the working condition feature package according to the present invention.

[0023] Figure 3 This is a flowchart illustrating the process of obtaining the simulation prediction results of the present invention.

[0024] Figure 4 This is a flowchart illustrating the process parameter adjustment decision-making process of the present invention.

[0025] Figure 5 This is a schematic diagram of the overall method steps of the present invention. Detailed Implementation

[0026] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] As attached Figures 1 to 4 The defect feedback and digital simulation closed-loop control system for the stamping process shown includes:

[0028] Synchronous working condition perception and feature encoding module: used to synchronously collect multi-source operating data at the beginning of each stamping cycle, perform fusion processing and feature encoding, and generate working condition codes and corresponding working condition feature packages.

[0029] The multi-source operational data includes the initial positioning deviation of the sheet metal ΔL and the actual pre-compression pressure P of each zone of the pressure ring. actual With the set value P set Deviation ΔP, mold temperature T die The cumulative number of punches N in the mold;

[0030] It should be noted that multi-source operating data is collected synchronously at the beginning of each stamping cycle, and then fused and encoded to generate working condition codes and corresponding working condition feature packages. The specific process is as follows:

[0031] Initial positioning deviation ΔL acquisition of sheet metal: A high-speed industrial camera with a frame rate of no less than 500fps and a resolution of no less than 2MP is selected. Combined with a sub-pixel edge detection algorithm, the positioning deviation ΔL of the sheet metal relative to the mold (representing the displacement deviation vector of the sheet metal in the X and Y directions) is captured in real time. The measurement range is ±0.01 mm to ±5 mm, and the accuracy is ±0.02 mm. The camera is fixed above the stamping machine and faces the sheet metal area. It is equipped with a ring LED light source to ensure uniform illumination. The trigger signal is provided by the slider position sensor. When the slider descends to a distance of 10 mm to 20 mm from the sheet metal, the shooting is started. Image processing uses the OpenCV library to extract contours and compare coordinates, calculate the deviation in the X and Y directions, and output the data in real time through the GigEVision protocol.

[0032] Pressure deviation ΔP of the blank holder ring is acquired by using high-precision piezoelectric sensors with a range of 0 to 500 kN and an accuracy of ±0.5% FS. These sensors are deployed in various sections of the blank holder ring, typically 4 to 8 independent hydraulic or servo sections, to read the actual preload pressure P in real time. actual (Represents the real-time pressure values ​​measured in each zone), compared with the controller setpoint P. set (Indicates a preset target pressure value) Perform an instantaneous comparison and calculate the deviation. (ΔP represents the pressure deviation vector, calculated independently for each zone, with dimensions in kN), Measurement range Up to +50 kN, with a response time of no more than 1ms, the sensor integrates an automatic temperature compensation function to avoid thermal drift affecting accuracy. Sensor data from each zone is synchronously acquired via CANopen or EtherCAT protocol to ensure consistent timestamps between zones.

[0033] Mold temperature T die Data Acquisition: Infrared array sensor selected, temperature range With an accuracy of ±2℃, a resolution of no less than 640×480, and a response time of no more than 100ms, this device is deployed in key areas of the mold, such as draw beads, fillet areas, and cavity surfaces, to monitor the mold temperature distribution in real time and collect the average temperature T at representative key points. die (Represents the equivalent temperature of the entire mold or a key area, in °C). The sensor is equipped with a protective cover to prevent interference from metal debris. The data sampling frequency is synchronized with vision and outputs via RS485 or Modbus protocol.

[0034] Accumulated number of punches N for the die is acquired directly from the main controller of the stamping machine. The cumulative number of punches N (representing a dimensionless integer count) after the die is installed is read directly from the main controller. The accuracy is 1. The data is acquired in real time via Profibus or OPCUA protocol, and the accuracy of the count is verified by combining the equipment maintenance log.

[0035] Data fusion and feature encoding: The edge computing unit performs synchronous fusion processing on the above multi-source data. First, the Kalman filter algorithm is used to denoise and align the time of each sensor data to ensure that all data correspond to the same stamping cycle start time. Second, feature normalization processing is performed to normalize ΔL, ΔP, and T. die N and N are mapped to the intervals from 0 to 1 respectively, and the normalization formula is: Where x represents the original parameter value to be normalized (such as ΔL, ΔP, T) die or N); x min This represents the historical minimum value of the parameter; x max This represents the historical maximum value of the parameter, and then a high-dimensional feature vector is extracted, including the original value, the partition variance of bias statistics such as ΔP, and cross features such as T. die The product of ×N terms further generates a unique operating condition code (CODE). k .

[0036] The final output is the working condition feature pack Fpack={CODE} k , ΔL, ΔP, T die , N, and the fused high-dimensional feature vector}, this feature package encapsulates all the initial disturbance information of the current stamping cycle.

[0037] Digital simulation prediction module: used to receive the working condition feature package, dynamically call the matching material parameter set, inject the initial disturbance conditions into the forming digital simulation model, perform simulation calculations in parallel with the physical stamping process, and output simulation prediction results characterizing the trend of stamping defects.

[0038] It should be noted that the process of receiving the operating condition feature packet and dynamically calling the matching set of material parameters is as follows:

[0039] Operating condition feature packet reception and parsing: The module receives a unique operating condition feature packet F from the operating condition synchronization sensing and feature encoding module. pack It performs real-time parsing and extracts the operating condition code on a dedicated high-performance computing unit. k Initial positioning deviation of sheet metal ΔL, pressure deviation vector ΔP, mold temperature T die The cumulative number of impulses N and the fused high-dimensional feature vector.

[0040] Dynamic material parameter set retrieval: based on working condition code (CODE) k Key identifying features, especially mold temperature T die The system dynamically retrieves and loads the best-matching set of material constitutive parameters from the online material parameter library, along with the cumulative number of impacts N. This set includes parameters of the flow stress curve, friction coefficient μ, and anisotropic parameter r. The parameters in the library have been continuously optimized by the online learning and model evolution module based on historical production data.

[0041] Initial disturbance conditions are injected into the forming digital simulation model, and simulation calculations are performed in parallel with the physical stamping process. The simulation prediction results characterizing the trend of stamping defects are output. The specific process is as follows:

[0042] Initial disturbance condition injection: The initial positioning deviation ΔL of the sheet metal is converted into the initial coordinate offset of the sheet metal in the simulation model, and the pressure deviation vector ΔP is converted into the initial blank holder force disturbance boundary conditions for each zone of the blank holder ring. Simultaneously, the mold temperature T is... die The above disturbance conditions are mapped to the initial temperature field distribution of the mold. They are directly injected into the lightweight high-fidelity forming digital simulation model through the finite element preprocessing interface. The model is constructed based on the explicit dynamic finite element algorithm, and the mesh size is controlled between 100,000 and 300,000 elements to ensure computational efficiency.

[0043] Ultra-real-time parallel simulation computing: The simulation computing runs on a dedicated GPU-accelerated computing core or a high-performance edge server, starting strictly in parallel with the physical stamping process. The simulation time step is adaptively adjusted, and the overall simulation time is controlled within 30% to 50% of the physical stamping cycle, usually less than 1 second, to achieve ultra-real-time prediction. During the simulation, the flow, stress-strain evolution, and thickness reduction behavior of the sheet material under disturbance conditions are simulated in real time.

[0044] Defect Occurrence Trend Prediction: After the simulation is completed, the simulation prediction results characterizing the occurrence trend of stamping defects are output, including the predicted thickness reduction rate distribution cloud map, minimum thickness value, maximum thinning rate, principal strain path curve, and predicted forming force-displacement curve F. pred (s), where s represents the slider stroke. Further, based on material failure criteria, the thickness reduction rate and principal strain path are compared to generate a quantitative defect risk matrix, Risk. MapThe risk levels are divided into three levels: low, medium, and high, and potential defect types such as cracking, wrinkling, or excessive springback are marked.

[0045] Final output simulation prediction results report R pre ={Risk Map F pred (s), thickness reduction rate distribution, principal strain path}, this report fully characterizes the defect occurrence trend of this stamping and is output before the physical stamping is completed.

[0046] Defect feedback and decision control module: used to receive the simulation prediction results and obtain measured data, calculate the simulation prediction confidence through consistency analysis, and generate process parameter adjustment decisions based on defect risk distribution information and simulation prediction confidence.

[0047] It should be specifically noted that the process of receiving the simulation prediction results and obtaining the measured data, and then calculating the simulation prediction confidence through consistency analysis, is as follows:

[0048] Simulation Prediction Result Reception and Analysis: The module receives the simulation prediction result report R from the digital simulation prediction module. pre The defect risk matrix (Risk) is extracted in real time by analyzing the data on the real-time control computing unit. Map Predicting the forming force-displacement curve F pred (s), thickness reduction rate distribution, principal strain path.

[0049] Acquisition of measured forming force curve: Real-time communication with the main controller of the stamping press to acquire the actual forming force-displacement curve F during this physical stamping process. actual (s), where s represents the slider stroke, the data sampling frequency is not less than 1000Hz, and the data is acquired synchronously via EtherCAT or Profibus protocol. The curve fully covers the entire process from the slider contacting the sheet metal to the end of stamping, and the acquisition window is strictly aligned with the stroke range predicted by the simulation.

[0050] Consistency analysis and confidence calculation: predicting the forming force curve F pred (s) and the measured forming force curve F actual (s) Perform rapid comparison at key feature points, including the initial contact point, peak force point, and unloading section start point. Typically, 5 to 10 feature points are selected, and the relative error of each point is calculated. Where i represents the i-th feature point, s i To determine the corresponding travel position, further calculate the overall curve consistency index, for example, using a weighted average of Euclidean distances to obtain the comprehensive force curve deviation δ. F The deviation range is typically controlled within 0% to 15%;

[0051] Simultaneously considering the defect risk matrix RiskMap To ensure the conservatism of the predicted thickness reduction rate distribution, a weighted fusion formula was used to calculate the prediction confidence level C for this simulation. Where w1, w2, and w3 are weighting coefficients (typical values ​​are 0.5, 0.3, and 0.2), δF is the percentage deviation of the force curve, and R... max σ represents the maximum risk value in the risk matrix, β is the penalty coefficient (typically between 0.1 and 0.3), and σ is the maximum risk value in the risk matrix. FLC The value is the standardized value of the closest distance between the main strain path and the boundary of the forming limit diagram. The confidence level C ranges from 0 to 1, and the higher the value, the more reliable the prediction.

[0052] It should be further explained that the process parameter adjustment decision is generated based on defect risk distribution information and confidence level. The specific process is as follows:

[0053] Intelligent decision generation: based on the defect risk matrix Risk Map The maximum thinning rate, the risk of the main strain path, and the calculated confidence level C are used to generate process parameter adjustment decisions through a pre-set lightweight reinforcement learning decision model.

[0054] When the confidence level C ≥ 0.85 and the risk level is medium or high, a clear feedforward compensation instruction is triggered; the decision-making object mainly includes the adjustment amount ΔP of the blank holder force zone in the next cycle. adj (Each zone is independent), slider speed curve fine-tuning coefficient, mold preheating temperature compensation value, the adjustment range is determined by gradient descent optimization, for example, for areas with excessive thickness reduction, the blank holder force is increased by a certain percentage. Where k is the proportional gain coefficient (typically 0.05 to 0.2), R j Let R be the risk value corresponding to the j-th partition. threshold The threshold is set as the safety threshold. If the confidence level C < 0.6, a conservative decision is made, and only the minimum adjustment is performed or the original parameters are maintained. At the same time, it is marked as "requiring manual intervention".

[0055] Final output process parameter adjustment decision CMD={Adjust Target "Next" Cycle , Parameters: [ΔP adj (Adjustment vectors for each partition)}, slider speed curve fine-tuning coefficients}, other compensation parameters}]}.

[0056] Process parameter execution module: used to make adjustment decisions based on the process parameters, update the parameters of the stamping equipment, and apply the adjusted process parameters to the next stamping cycle.

[0057] It should be noted that, based on the aforementioned process parameter adjustment decisions, the parameters of the stamping equipment are updated, and the specific process is as follows:

[0058] Process Parameter Adjustment Decision Reception and Analysis: The module receives process parameter adjustment decisions (CMDs) from the defect feedback and decision control module, analyzes them in real time on the main controller of the stamping press, and extracts the adjustment target (Adjust). Target (Fixed to "Next") Cycle The document includes a list of specific process parameter adjustments, including the adjustment amount ΔP for the blank holder force. adj (Independent adjustment vectors for each zone), slider speed curve fine-tuning coefficient, mold preheating temperature compensation value, and other compensation parameters.

[0059] Process parameter update execution timing control: After the instruction is parsed, the module waits for the key completion events of the current stamping cycle to be triggered, including signals such as workpiece ejection, slide return to top dead center, and pressure ring fully lifted. These events are determined by the stamping machine position sensor and status signals, and usually occur within 50ms to 200ms after the physical stamping ends. Once it is confirmed that the current cycle has ended and the equipment is in a safe standby state, the parameter update process is immediately started to avoid changing parameters during the stamping process, which could lead to safety accidents or quality abnormalities.

[0060] Blanking force zone servo adjustment: For each zone of blanking force adjustment ΔP specified in the instruction set. adj，j The main controller calculates the target blank holder force setpoint for the next cycle. Where j represents the j-th pressing edge partition, This is the current setting value for this cycle; the updated target value is immediately written to the pressure / force closed-loop register of the servo motor controller. The servo system uses a PID algorithm to achieve fast tracking, adjusts the response time to within 20ms, and maintains the force control accuracy at ±1%FS.

[0061] Slider speed curve fine-tuning: If the instruction set includes a slider speed curve fine-tuning coefficient (usually a proportional coefficient k) v If the value is typically between 0.95 and 1.05, then the preset speed curve for the next cycle will be scaled overall or adjusted locally. For example, speed correction v can be applied in the deep drawing sensitive stage (such as the rounded corner forming section). next (t)=k v ·v current (t), or differential adjustment for a specific travel range; the adjusted speed curve is regenerated by the motion planner inside the controller and loaded onto the servo drive, and the speed following error is controlled within ±2%, avoiding vibration or unstable forming caused by sudden speed changes.

[0062] Other compensation parameter execution: For additional instructions such as mold preheating temperature compensation value, the module updates the mold heating system set temperature through the temperature controller interface; if auxiliary parameters such as lubrication amount and air cushion pressure are involved, they are written into the corresponding actuator register. All update operations are completed before the start of the next cycle (usually all are completed within 100ms to 300ms after the workpiece is ejected), and the update is confirmed to be successful through status feedback signal.

[0063] Some experimental data are shown in the table below:

[0064] Online learning and model evolution module: used to acquire actual quality inspection data, compare and analyze it with the corresponding simulation prediction results, and update the online material parameter library and digital simulation model parameters on a rolling basis according to the prediction error, forming a closed-loop control system.

[0065] It should be noted that the specific process for obtaining actual quality inspection data and comparing it with the corresponding simulation prediction results is as follows:

[0066] Actual quality inspection data acquisition: After the stamping cycle is completed, the workpiece is ejected and transferred to the online quality inspection station. This module triggers the inspection process through the conveyor belt interface; it mainly collects the actual geometric quality data Q of the workpiece. actual The measurements include a 3D thickness distribution cloud map, minimum thickness value, maximum thinning rate, principal strain path in key areas, and springback amount. The testing equipment uses a laser 3D scanner or structured light scanning system with an accuracy of ±0.01mm and a scanning resolution of no less than 0.1mm. The testing cycle is controlled to be completed within 5 to 15 seconds after the end of the current stamping cycle. All measurement data are linked to a unique working condition code (CODE). k Binding, forming a complete data pair {CODE k Q actual}

[0067] Simulation prediction result traceability and binding: The module is based on the operating condition code (CODE). k Quickly retrieve the simulation prediction result report corresponding to the current cycle from historical cache or database. pre Key prediction indicators are extracted, including the predicted thickness reduction rate distribution cloud map, the minimum thickness prediction value, the maximum thinning rate prediction value, and the main strain path prediction curve, to ensure that the predicted data strictly correspond to the actual measured data under the same working conditions; if the data is not found in the cache, the output record of the digital simulation prediction module is traced back through the system log.

[0068] Prediction error calculation: based on actual quality data Q actual The prediction results are compared point by point with the simulation results to calculate the multi-dimensional prediction error, which mainly includes:

[0069] Thickness reduction rate error vector: Where k represents the grid or partition number, T actual，k and T pred，k These are the actual and predicted thickness reduction rates (percentages), respectively.

[0070] Absolute error of maximum thinning rate: ;

[0071] The principal strain path deviation is quantified using either the Fraser distance or the point-by-point Euclidean distance.

[0072] The forming force curve error (if a full measured curve is available) is expressed as the root mean square error (RMSE). Where M is the number of sampling points;

[0073] The overall error index E is obtained by weighted summation: Weight w t w m w f Based on the defect severity preset (e.g., thickness reduction has the highest weight) and satisfying w t +w m +w f =1;

[0074] After the error calculation is completed, an error package {CODE} is generated. k E, Ed etail}, where E detail Detailed error components for each dimension.

[0075] It should be further explained that the online material parameter library and digital simulation model parameters are updated on a rolling basis according to the prediction error to form a closed-loop control system. The specific process is as follows:

[0076] Parameter rolling update: Based on the calculated prediction error E, an online gradient descent algorithm is used to fine-tune the corresponding parameter set in the online material parameter library; the updated objects include flow stress curve parameters, friction coefficient μ, anisotropic parameter r value, etc.; the update rules follow... Where θ is the material parameter vector, η is the learning rate (typically 1e-4 to 1e-3, adaptively adjusted), and L(E) = E 2 (Mean squared error loss function) The gradient of the error with respect to the parameters is approximated using the finite difference method: for the parameter vector The gradient of the i-th parameter is ( (The step size for parameter perturbation); the update range is constrained by the error magnitude and confidence level C. When the confidence level C is low, the step size is reduced or the update is paused. The updated parameters are immediately written into the current working condition sub-library of the online material parameter library and marked with the update timestamp and error improvement index.

[0077] Model Evolution and Knowledge Accumulation: After each parameter update, the module evaluates the overall improvement in prediction accuracy. If the error decreases for several consecutive cycles, the model is confirmed to have converged to the actual production system. Simultaneously, a rolling window of data pairs (typically the most recent 50 to 200 operating cycle periods) is maintained for periodic batch fine-tuning or meta-learning initialization. For similar operating conditions (via CODE...) k The feature distance judgment uses Euclidean distance, and the most recently updated parameters are loaded first to form a self-learning closed loop that becomes more accurate with use. All update processes are recorded in a complete log, including parameter change trajectory and error evolution curve, which facilitates subsequent offline analysis and system diagnosis.

[0078] As attached Figure 5 The defect feedback and digital simulation closed-loop control method for the stamping process shown includes:

[0079] S1: At the beginning of each stamping cycle, multi-source operating data is collected synchronously, and fusion processing and feature encoding are performed to generate working condition codes and corresponding working condition feature packages.

[0080] S2: Receive the working condition feature package, dynamically call the matching material parameter set, inject the initial disturbance conditions into the forming digital simulation model, perform simulation calculations in parallel with the physical stamping process, and output simulation prediction results that characterize the trend of stamping defect occurrence.

[0081] S3: Receive the simulation prediction results and obtain the measured data, calculate the simulation prediction confidence through consistency analysis, and generate process parameter adjustment decisions based on defect risk distribution information and simulation prediction confidence.

[0082] S4: Based on the process parameter adjustment decision, update the parameters of the stamping equipment so that the adjusted process parameters are applied to the next stamping cycle;

[0083] S5: Obtain actual quality inspection data, compare and analyze it with the corresponding simulation prediction results, and update the online material parameter library and forming digital simulation model parameters on a rolling basis according to the prediction error to form a closed-loop control system.

[0084] Secondly: The accompanying drawings of the embodiments disclosed in this invention only involve the structures involved in the embodiments disclosed in this invention. Other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of this invention can be combined with each other.

[0085] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A defect feedback and digital simulation closed-loop control system for the stamping process, characterized in that, include: Synchronous working condition perception and feature encoding module: used to synchronously collect multi-source operating data at the beginning of each stamping cycle, perform fusion processing and feature encoding, and generate working condition codes and corresponding working condition feature packages. Digital simulation prediction module: used to receive the working condition feature package, dynamically call the matching material parameter set, inject the initial disturbance conditions into the forming digital simulation model, perform simulation calculations in parallel with the physical stamping process, and output simulation prediction results characterizing the trend of stamping defects. Defect feedback and decision control module: used to receive the simulation prediction results and obtain measured data, calculate the simulation prediction confidence through consistency analysis, and generate process parameter adjustment decisions based on defect risk distribution information and simulation prediction confidence. Process parameter execution module: used to make adjustment decisions based on the process parameters, update the parameters of the stamping equipment, and apply the adjusted process parameters to the next stamping cycle. Online learning and model evolution module: used to acquire actual quality inspection data, compare and analyze it with the corresponding simulation prediction results, and update the online material parameter library and forming digital simulation model parameters on a rolling basis according to the prediction error, forming a closed-loop control system.

2. The defect feedback and digital simulation closed-loop control system for the stamping process according to claim 1, characterized in that: The collection of multi-source operational data includes: The initial positioning deviation of the sheet metal is collected by a high-speed industrial camera, the deviation between the actual pre-pressure of each zone of the pressure ring and the set value is collected by a high-precision piezoelectric sensor, the mold temperature is collected by an infrared array sensor, and the cumulative number of punches of the mold is read by the main controller of the stamping machine.

3. The defect feedback and digital simulation closed-loop control system for the stamping process according to claim 1, characterized in that: The method for obtaining the operating condition feature package is as follows: Kalman filtering is performed on the collected multi-source operating data for noise reduction and time alignment, then feature normalization is performed to extract feature vectors, and finally operating condition codes and corresponding operating condition feature packages are generated.

4. The defect feedback and digital simulation closed-loop control system for the stamping process according to claim 1, characterized in that: The initial disturbance conditions include: the initial coordinate offset of the sheet in the forming digital simulation model formed by converting the initial positioning deviation of the sheet, the initial pressure force disturbance boundary conditions of each partition of the pressure ring formed by converting the pressure deviation, and the initial temperature field distribution of the mold formed by mapping the mold temperature.

5. The defect feedback and digital simulation closed-loop control system for the stamping process according to claim 1, characterized in that: The simulation prediction results are obtained as follows: the received working condition feature package is parsed, and a matching set of material parameters is dynamically retrieved and loaded from the material parameter library. This set of material parameters includes flow stress curve parameters, friction coefficient, and anisotropy parameters. The initial disturbance conditions are injected into the forming digital simulation model constructed based on the explicit dynamic finite element algorithm through the finite element preprocessing interface. The simulation calculation is started in parallel with the physical stamping process. During the simulation, the sheet flow, stress-strain evolution, and thickness reduction behavior are simulated. After the simulation is completed, the simulation prediction results containing the defect risk matrix, thickness reduction rate distribution, principal strain path curve, and predicted forming force-displacement curve are output.

6. The defect feedback and digital simulation closed-loop control system for the stamping process according to claim 1, characterized in that: The process parameter adjustment decision is obtained by receiving the simulation prediction result and obtaining the measured forming force curve, comparing the feature points of the two and calculating the simulation prediction confidence level in combination with the defect risk distribution information, and generating the process parameter adjustment decision through a lightweight reinforcement learning decision model based on the confidence level and the defect risk distribution information.

7. The defect feedback and digital simulation closed-loop control system for the stamping process according to claim 1, characterized in that: The parameter update process is as follows: analyze the process parameter adjustment decision, and after the workpiece is ejected, the slide returns to the top dead center and the blank holder is fully lifted in the current stamping cycle, while the equipment is in a safe standby state, adjust the blank holder force zone, slide speed curve and other compensation parameters so that the adjusted process parameters take effect in the next stamping cycle.

8. The defect feedback and digital simulation closed-loop control system for the stamping process according to claim 1, characterized in that: The rolling update process is as follows: acquire actual quality inspection data and compare it with the corresponding simulation prediction results, calculate multi-dimensional prediction errors, use online gradient descent algorithm to fine-tune the parameters of online material parameter library and forming digital simulation model, the update range is constrained by the magnitude of prediction error and confidence level, and at the same time maintain a rolling data window to continuously optimize model parameters based on historical data.

9. A defect feedback and digital simulation closed-loop control method for the stamping process, used to implement the defect feedback and digital simulation closed-loop control system for the stamping process as described in any one of claims 1-8, characterized in that, include: S1: At the beginning of each stamping cycle, multi-source operating data is collected synchronously, and fusion processing and feature encoding are performed to generate working condition codes and corresponding working condition feature packages. S2: Receive the working condition feature package, dynamically call the matching material parameter set, inject the initial disturbance conditions into the forming digital simulation model, perform simulation calculations in parallel with the physical stamping process, and output simulation prediction results that characterize the trend of stamping defect occurrence. S3: Receive the simulation prediction results and obtain the measured data, calculate the simulation prediction confidence through consistency analysis, and generate process parameter adjustment decisions based on defect risk distribution information and simulation prediction confidence. S4: Based on the process parameter adjustment decision, update the parameters of the stamping equipment so that the adjusted process parameters are applied to the next stamping cycle; S5: Obtain actual quality inspection data, compare and analyze it with the corresponding simulation prediction results, and update the online material parameter library and forming digital simulation model parameters on a rolling basis according to the prediction error to form a closed-loop control system.

Images