A method and system for adjusting and controlling stamping dies for mid-frame components

By constructing a deformation distribution field and adjustment response mapping library, the mold adjustment parameters are adjusted in real time, solving the problem of inaccurate mold deformation adjustment in the existing technology, and realizing efficient mold deformation control and production stability.

CN123104147APending Publication Date: 2026-11-17南通诚业德精密组件有限公司
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611612082.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-10-19
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing stamping die adjustment methods rely on static calibration and experience-based adjustments, which cannot track the relationship between die deformation and adjustment amount in real time. This leads to batch quality fluctuations and increased scrap rates during production, and the debugging cycle accounts for a large proportion.

Method used

By acquiring deformation state data of the mold surface, a deformation distribution field is constructed, an adjustment response mapping library is established, and the adjustment position and load amplitude are extracted by reverse tracing of the deformation response topology. A multi-position coordinated adjustment sequence is constructed, and the adjustment parameters are adjusted in real time to achieve the accuracy and continuity of deformation control.

Benefits of technology

It enables real-time monitoring and precise adjustment of mold deformation, reduces debugging cycle, improves adjustment success rate, avoids mold damage, extends mold service life, and improves production batch consistency and control accuracy.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123104147A_ABST
    Figure CN123104147A_ABST
Patent Text Reader

Abstract

This invention provides a method and system for adjusting and controlling a stamping die for a mid-frame component, relating to the field of stamping die technology. The method includes acquiring deformation data of the monitoring area on the die surface during stamping, identifying abnormal regions with deformation gradients exceeding the mean, and extracting spatial distribution locations and deformation amplitudes to construct a deformation distribution field. Through offline calibration, calibration loads are applied to different adjustment positions, and responses are collected to establish an adjustment response mapping library. The similarity between the deformation distribution field and the deformation response topology in the mapping library is calculated and traced back to determine the initial adjustment strategy. Based on the initial adjustment strategy, secondary anomalies are predicted through superposition of secondary deformations in non-abnormal regions, and auxiliary adjustment positions are retrieved to construct a multi-position collaborative adjustment sequence. After executing this sequence, the mapping library is iteratively updated based on the actual deformation deviation, achieving precise collaborative adjustment and adaptive optimization of the stamping die deformation.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of stamping die technology, and in particular to a method and system for adjusting and controlling a stamping die based on a middle frame component. Background Technology

[0002] During continuous production, stamping dies are subjected to high-frequency impact loads, causing complex elasto-plastic deformation of the cavity surface. To ensure the dimensional accuracy of stamped parts, production typically relies on pre-set adjusting bolts or hydraulic cushion mechanisms to apply pressure locally to the die surface. The conventional practice is to adjust each adjustment position one by one while the machine is stopped, based on the measurement results of test stamps or the initial calibration values ​​provided by the die supplier. After each adjustment, a test stamp is performed, and the adjustment effect is judged by manually measuring the part's wall thickness or contour deviation. This method, relying on static calibration and sequential approximation, has long been used for die compensation, but its effectiveness is closely related to the operator's experience level, and the debugging cycle often occupies a large proportion of the die delivery cycle.

[0003] The die surface is not an isolated elastic body; multiple adjustment positions are spatially interconnected. In actual stamping processes, when an adjustment force is applied to a specific local area, material flow and die structural deformation create a ripple effect, causing synchronous changes in stress distribution in adjacent areas and even opposing surfaces. Traditional adjustment methods treat each adjustment point as independent, adjusting them step-by-step according to an empirical sequence of "from large to small, primary to secondary." When the part exhibits multiple positional accuracy deviations simultaneously with inconsistent directions, this sequential adjustment strategy easily falls into a cycle of repeated corrections. The convergence result of the previous adjustment is often disrupted by subsequent adjustments, resulting in an oscillating adjustment process.

[0004] The deformation state of the die surface also slowly drifts with increasing stroke count, rising die temperature, and intensified local wear. Calibration is typically only performed once during new die delivery or maintenance shutdowns, lacking the ability to track and correct the relationship between deformation and adjustment as operating conditions change. If batch quality fluctuations occur during production, process engineers can only stop the machine and re-test the die for correction, unable to adjust the adjustment parameters in a timely manner based on the actual stress state of the die, resulting in increased scrap rates and forced interruptions to continuous production. In summary, this technical field requires an intelligent adjustment scheme that can address the real-time deformation status of the entire die domain, quantitatively describe the impact of adjustment actions, and continuously correct control parameters. Summary of the Invention

[0005] The present invention provides a method and system for adjusting and controlling stamping dies for mid-frame components, which can solve the problems in the prior art.

[0006] A first aspect of the present invention provides a method for adjusting and controlling a stamping die for a mid-frame component, comprising: Acquire deformation state data of each monitoring area on the die surface during the stamping process, identify abnormal areas where the deformation gradient value exceeds the mean, extract the spatial distribution location and deformation amplitude of the abnormal areas, and construct a deformation distribution field. Different amplitude calibration loads are applied to each adjustment position of the mold at an independent offline calibration station. After each calibration, deformation response data of the entire monitoring area of ​​the mold is collected to establish an adjustment response mapping library. Spatial morphological similarity is calculated between the deformation distribution field of the abnormal region and the deformation response topology of each adjustment position in the adjustment response mapping library. The adjustment position and load amplitude corresponding to the highest similarity are extracted by reverse tracing of the deformation response topology to determine the initial adjustment strategy. Using the initial adjustment strategy as the baseline action, the secondary deformation distribution generated by the adjustment position on the non-abnormal region is extracted from the adjustment response mapping library, and superimposed with the current deformation field of the non-abnormal region for pre-simulation. Regions with spatial gradient values ​​exceeding the mean after superposition are extracted as predicted secondary abnormal regions. Auxiliary adjustment positions and offset adjustment amounts are retrieved and screened to construct a multi-position coordinated adjustment sequence. Execute a multi-location coordinated regulation sequence, calculate the deformation deviation distribution between the actual deformation change distribution and the predicted deformation response field, and write the regulation parameters into the regulation response mapping library for iterative updates based on the deviation energy ratio.

[0007] In one optional embodiment, calibration loads of different amplitudes are applied to each adjustment position of the mold at an independent offline calibration station. After each calibration, deformation response data of the entire monitoring area of ​​the mold is collected, and an adjustment response mapping library is established, including: A pre-tightening load is applied to the mold at an independent offline calibration station to bring the mold to the initial calibration state, and the reference deformation field of each monitoring area is collected under the initial calibration state. The calibration traversal sequence is arranged according to the spatial topological relationship of the adjustment position. The calibration load spectrum is applied to each adjustment position in the calibration traversal sequence in a step-increasing manner. The load is kept stable and the transient deformation field of the entire monitoring area is collected each time the calibration load is applied. The deformation response field is obtained by performing spatial field difference operation between the transient deformation field and the reference deformation field. The load is unloaded and the deformation field is monitored to converge and regress to the reference deformation field. Construct a calibration mapping tuple by combining the deformation response field with the corresponding adjustment position identifier and calibration load amplitude, extract the spatial gradient tensor field distribution of the deformation response field, construct the deformation response topological signature and label it to the calibration mapping tuple. The deformation response topology signatures generated at the same adjustment position under all calibration load amplitudes included in the calibration load spectrum are summarized. Calibration mapping tuples in the deformation response topology signatures whose spatial angle between the principal direction feature vectors exceeds a preset tolerance threshold are removed. The remaining calibration mapping tuples are then summarized to construct an adjustment response mapping library.

[0008] In one optional embodiment, constructing a calibration mapping tuple by combining the deformation response field with the corresponding adjustment position identifier and calibration load amplitude, extracting the spatial gradient tensor field distribution of the deformation response field, and constructing and labeling the deformation response topological signature to the calibration mapping tuple includes: The deformation response field, along with the corresponding adjustment position identifier and calibration load amplitude, is encapsulated into a calibration mapping tuple as the basic data structure. Gradient difference operation is performed on the deformation amplitude of each monitoring area in the deformation response field along the spatial coordinate direction to construct a spatial gradient tensor field. Tensor eigenvalue decomposition is performed on the spatial gradient tensor field to extract the eigenvector corresponding to the largest eigenvalue, determine the principal direction eigenvector, and detect the ratio of the second largest eigenvalue to the largest eigenvalue in the spatial gradient tensor field. When the ratio exceeds the preset directional dispersion threshold, the deformation response field is marked as a multimodal response and the secondary direction eigenvector corresponding to the second largest eigenvalue is recorded to the calibration mapping tuple. Identify the spatial coordinates of the peak value of deformation in the deformation response field, determine the centroid coordinates of the peak response, search for the spatial boundary position where the deformation amplitude decays to a preset proportion of the peak amplitude along the principal eigenvector direction, extract the spatial distance from the centroid coordinates of the peak response to the spatial boundary position, and obtain the response influence radius. The deformation response topological signature is constructed by combining the principal directional feature vector, the centroid coordinates of the peak response, and the radius of influence of the response. This signature is then marked in the topological feature field of the calibration mapping tuple, which is associated with the adjustment position identifier and the calibration load amplitude. For calibration mapping tuples marked as multimodal responses, the secondary directional feature vector is also recorded in the topological feature field.

[0009] In one optional embodiment, the spatial morphological similarity between the deformation distribution field of the abnormal region and the deformation response topology of each adjustment location in the adjustment response mapping library is calculated. The adjustment location and load amplitude corresponding to the highest similarity are extracted through reverse tracing of the deformation response topology to determine the initial adjustment strategy, including: Spatial registration is performed between the deformation distribution field of the abnormal region and each deformation response field in the regulation response mapping library. Spatial field superposition is then performed to identify spatially overlapping and spatially conflicting regions. The coverage of the spatially overlapping regions and the conflict intensity of the spatially conflicting regions are extracted. When the coverage exceeds the preset matching threshold and the conflict intensity is lower than the preset conflict threshold, the corresponding calibration mapping tuple is included in the candidate regulation source set. Multiple sets of deformation response fields corresponding to the adjustment positions of the calibration mapping tuples in the candidate adjustment source set are retrieved under the incremental calibration load spectrum to construct a deformation response evolution sequence. In the deformation response evolution sequence, the target deformation response field that is closest to the deformation distribution field of the abnormal region is searched and the residual deformation difference field is calculated. When the energy ratio of the residual deformation difference field exceeds the preset residual threshold, the deformation response fields adjacent to the target deformation response field are interpolated to generate a compensation deformation response field. The interpolated load amplitude corresponding to the compensation deformation response field is extracted to determine the target value of the adjustment load. Extract the calibration mapping tuple with the highest coverage and lowest conflict intensity from the candidate regulation source set, and encapsulate the regulation location identifier and regulation load target value into an initial regulation strategy.

[0010] In one optional embodiment, performing a spatial field overlay probe to identify spatially overlapping and conflicting regions, and extracting the coverage of the spatially overlapping regions and the conflict intensity of the spatially conflicting regions includes: The deformation distribution field of the abnormal region and each deformation response field are calculated point by point in the overlapping area after spatial registration. When the deformation vector dot product is positive, it is marked as a cooperative action point, and when the deformation vector dot product is negative, it is marked as an antagonistic action point. Perform connectivity analysis on the set of cooperative action points, extract the largest connectivity as the spatially overlapping region, and vector superimpose the deformation distribution field and deformation response field of the abnormal region within the spatially overlapping region. When the residual deformation amplitude after superposition is lower than the deformation distribution field amplitude of the abnormal region before superposition, the spatially overlapping region is confirmed as an effective cancellation region, and the coverage rate of the effective cancellation region to the total area of ​​the abnormal region is calculated. The deformation vector superposition calculation is performed on the set of opposing action points. The spatial region where the deformation amplitude after superposition exceeds the peak value of the deformation distribution field in the abnormal region is extracted as the spatial conflict region. The excess multiple of the superposition deformation amplitude in the spatial conflict region relative to the peak value of the deformation distribution field in the abnormal region is calculated to determine the conflict intensity.

[0011] In one optional embodiment, using the initial adjustment strategy as the baseline action, the secondary deformation distribution generated by the adjustment position on the non-abnormal region is extracted from the adjustment response mapping library, and superimposed with the current deformation field of the non-abnormal region for pre-simulation. Regions where the spatial gradient value exceeds the mean after superposition are extracted as predicted secondary abnormal regions. Auxiliary adjustment positions and offset adjustment amounts are retrieved and screened to construct a multi-position coordinated adjustment sequence, including: Extract the global deformation response field corresponding to the adjustment position in the initial adjustment strategy from the adjustment response mapping library, and separate the secondary deformation distribution outside the abnormal region after scaling according to the target value of the adjustment load; The secondary deformation distribution is superimposed with the current deformation field of the non-abnormal region to obtain the pre-simulated superimposed deformation field. The spatial gradient distribution is calculated on the pre-simulated superimposed deformation field, and the region with the gradient value exceeding the average gradient value of the current deformation field of the non-abnormal region is extracted to obtain the predicted secondary abnormal region. The set of adjustment locations that generate deformation response in the predicted secondary anomaly area is retrieved from the adjustment response mapping library. The deformation response field of each adjustment location is superimposed with the secondary deformation distribution for trial calculation. When the deformation amplitude after superposition is lower than the amplitude of the secondary deformation distribution before superposition, it is marked as an auxiliary adjustment location. The load amplitude optimization solution is performed on the auxiliary adjustment location to minimize the residual deformation energy after superposition. The optimized load amplitude is extracted as the offset adjustment amount. The initial adjustment strategy is taken as the main adjustment action, and the auxiliary adjustment position and the offset adjustment amount are encapsulated as auxiliary adjustment actions into a multi-position coordinated adjustment sequence.

[0012] In one optional embodiment, executing a multi-location coordinated adjustment sequence, calculating the deformation deviation distribution between the actual deformation change distribution and the predicted deformation response field, and iteratively updating the adjustment parameters by writing them into the adjustment response mapping library based on the deviation energy ratio includes: Loads are applied sequentially according to the timing of the main and auxiliary adjustment actions in the multi-position coordinated adjustment sequence. During the load application process, the structural global deformation field data is collected in real time. The difference between the deformation field before and after adjustment is extracted to obtain the actual deformation change distribution. The actual deformation change distribution is spatially compared with the predicted deformation response field before the execution of the multi-position coordinated adjustment sequence, and the deformation deviation distribution is calculated. When the energy proportion of the deformation deviation distribution exceeds the preset deviation threshold, the actual deformation change distribution, the adjustment position identifier and the actual applied load amplitude are encapsulated into a correction calibration mapping tuple. The correction calibration mapping tuple is written into the calibration mapping set of the corresponding adjustment position in the adjustment response mapping library, and the deformation response field of the adjustment position under the current load amplitude is updated. When the energy proportion of the deformation deviation distribution is lower than the deviation threshold, the spatial correlation coefficient between the actual deformation change distribution and the predicted deformation response field is extracted. The spatial correlation coefficient is used as a confidence marker and associated with the calibration mapping tuple of the corresponding adjustment position in the adjustment response mapping library to complete the knowledge iterative update of the adjustment response mapping library.

[0013] A second aspect of the present invention provides a mid-frame stamping die adjustment and control system, comprising: Anomaly identification unit is used to acquire deformation state data of each monitoring area on the surface of the die during the stamping process, identify abnormal areas where the deformation gradient value exceeds the mean, extract the spatial distribution location and deformation amplitude of the abnormal areas, and construct a deformation distribution field. The response mapping unit is used to apply calibration loads of different amplitudes to each adjustment position of the mold at an independent offline calibration station. After each calibration, deformation response data of the entire monitoring area of ​​the mold is collected to establish an adjustment response mapping library. The source matching unit is used to calculate the spatial morphological similarity between the deformation distribution field of the abnormal area and the deformation response topology of each adjustment position in the adjustment response mapping library. By reverse tracing the deformation response topology, the adjustment position and load amplitude corresponding to the highest similarity are extracted to determine the initial adjustment strategy. The collaborative planning unit is used to extract the secondary deformation distribution of the adjustment position to the non-abnormal region from the adjustment response mapping library, take the initial adjustment strategy as the baseline action, superimpose and pre-simulate with the current deformation field of the non-abnormal region, extract the region where the spatial gradient value exceeds the mean after superposition as the predicted secondary abnormal region, search and filter the auxiliary adjustment position and the offset adjustment amount, and construct a multi-position collaborative adjustment sequence. The iterative update unit is used to execute the multi-location coordinated adjustment sequence, calculate the deformation deviation distribution between the actual deformation change distribution and the predicted deformation response field, and write the adjustment parameters into the adjustment response mapping library for iterative update based on the deviation energy ratio.

[0014] A third aspect of the present invention provides an electronic device, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0015] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0016] In this embodiment of the invention, the calibration of the adjustment response mapping library is completed at an independent offline workstation, without occupying actual stamping resources. The debugging cycle is reduced from the traditional repeated mold trial method to two steps: offline library construction and online source tracing. The mold change debugging time is significantly reduced, and the risk of mold damage caused by load testing is avoided, thus extending the service life of the mold. The deformation distribution field of the abnormal area and the deformation response topology of the adjustment position are matched according to the spatial morphological similarity. The optimal adjustment position and load are locked in the multi-dimensional data of the entire field, without being misled by local extrema. For complex deformations with multi-peak coupling or long-distance transmission, the source can be traced. The adjustment decision breaks away from the empirical point-by-point trial and directly hits the real position of deformation, significantly improving the adjustment success rate. The pre-drilling superimposes the secondary disturbance of the benchmark action on the non-abnormal area onto the current deformation field, so that the secondary anomaly is predicted before execution. The auxiliary adjustment position eliminates the disturbance synchronously with the offset amount. The main anomaly suppression and secondary distortion control are integrated into a multi-position collaborative sequence. The coupling effect between adjustment positions is fully decoupled, avoiding the chain reaction of one correction causing deterioration in other places. After execution, the deviation distribution between the actual deformation response and the predicted response is calculated. The correction information is written into the adjustment response mapping library according to the deviation energy ratio. The mapping accuracy is updated on a rolling basis with each production batch. The iteration converges to the deformation characteristics specific to the mold. The residual error is automatically compensated into the next round of strategy. The control accuracy and batch consistency continue to improve with the production cycle. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the adjustment and control method for the stamping die of the middle frame component. Figure 2 Generate a logic flowchart for the deformation adjustment strategy; Figure 3 The deformation response evolution sequence and optimal matching analysis diagram are shown. Figure 4 The graph shows the iterative convergence and adaptive update of the deviation energy ratio. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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.

[0019] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0020] Figure 1 This is a flowchart illustrating the adjustment and control method for the stamping die of the middle frame component according to an embodiment of the present invention. Figure 1 As shown, the method includes: Acquire deformation state data of each monitoring area on the die surface during the stamping process, identify abnormal areas where the deformation gradient value exceeds the mean, extract the spatial distribution location and deformation amplitude of the abnormal areas, and construct a deformation distribution field. Different amplitude calibration loads are applied to each adjustment position of the mold at an independent offline calibration station. After each calibration, deformation response data of the entire monitoring area of ​​the mold is collected to establish an adjustment response mapping library. Spatial morphological similarity is calculated between the deformation distribution field of the abnormal region and the deformation response topology of each adjustment position in the adjustment response mapping library. The adjustment position and load amplitude corresponding to the highest similarity are extracted by reverse tracing of the deformation response topology to determine the initial adjustment strategy. Using the initial adjustment strategy as the baseline action, the secondary deformation distribution generated by the adjustment position on the non-abnormal region is extracted from the adjustment response mapping library, and superimposed with the current deformation field of the non-abnormal region for pre-simulation. Regions with spatial gradient values ​​exceeding the mean after superposition are extracted as predicted secondary abnormal regions. Auxiliary adjustment positions and offset adjustment amounts are retrieved and screened to construct a multi-position coordinated adjustment sequence. Execute a multi-location coordinated regulation sequence, calculate the deformation deviation distribution between the actual deformation change distribution and the predicted deformation response field, and write the regulation parameters into the regulation response mapping library for iterative updates based on the deviation energy ratio.

[0021] In one optional embodiment, calibration loads of different amplitudes are applied to each adjustment position of the mold at an independent offline calibration station. After each calibration, deformation response data of the entire monitoring area of ​​the mold is collected, and an adjustment response mapping library is established, including: A pre-tightening load is applied to the mold at an independent offline calibration station to bring the mold to the initial calibration state, and the reference deformation field of each monitoring area is collected under the initial calibration state. The calibration traversal sequence is arranged according to the spatial topological relationship of the adjustment position. The calibration load spectrum is applied to each adjustment position in the calibration traversal sequence in a step-increasing manner. The load is kept stable and the transient deformation field of the entire monitoring area is collected each time the calibration load is applied. The deformation response field is obtained by performing spatial field difference operation between the transient deformation field and the reference deformation field. The load is unloaded and the deformation field is monitored to converge and regress to the reference deformation field. Construct a calibration mapping tuple by combining the deformation response field with the corresponding adjustment position identifier and calibration load amplitude, extract the spatial gradient tensor field distribution of the deformation response field, construct the deformation response topological signature and label it to the calibration mapping tuple. The deformation response topology signatures generated at the same adjustment position under all calibration load amplitudes included in the calibration load spectrum are summarized. Calibration mapping tuples in the deformation response topology signatures whose spatial angle between the principal direction feature vectors exceeds a preset tolerance threshold are removed. The remaining calibration mapping tuples are then summarized to construct an adjustment response mapping library.

[0022] In one specific implementation, deformation state data of each monitoring area on the die surface is first acquired during the stamping process. Using an array of strain sensors or optical measuring devices deployed on the die surface, deformation values ​​of each monitoring area are continuously collected during the stamping stroke. The deformation gradient values ​​between adjacent monitoring areas are calculated, and areas with deformation gradient values ​​exceeding the global mean are marked as abnormal areas. The spatial distribution coordinates and corresponding deformation amplitudes of the abnormal areas are extracted, and the deformation distribution within the abnormal areas is filled using spatial interpolation methods to form a continuous deformation distribution field, providing a target reference for subsequent offline database construction at calibration stations.

[0023] After transferring the mold from the stamping station to the independent offline calibration station, a pre-tightening load is first applied to the mold to ensure it is in a uniformly stressed and stably positioned initial calibration state on the calibration bench. The magnitude of the pre-tightening load is determined based on the mold's own weight and the bench clamping specifications, aiming to eliminate free deformation errors caused by inconsistent mold placement. After applying the pre-tightening load and confirming mold positioning, deformation measurements of the entire monitoring area are collected, and these measurements are used as the reference deformation field. The deformation response under all subsequent calibration loads is based on Differential calculations are performed using a reference baseline.

[0024] The calibration traversal sequence is arranged according to the spatial topology of the adjustable positions. All adjustable positions on the mold are topologically sorted according to their spatial coordinates, forming traversal paths from the edge to the center or from the center to the edge. This avoids mutual contamination of response data due to load superposition effects between adjacent adjustable positions in the same batch of calibration. Following the traversal sequence, the first... A stepped, incremental calibration load spectrum is applied at each adjustment position. The load spectrum consists of several discrete amplitude levels, with increments between adjacent amplitude levels. The load is determined based on the elastic range of the mold material and the upper limit of the adjustment mechanism's stroke, ensuring that each load level is within the mold's elastic deformation range. When applying the first load... Level rated load Then, the load is kept stable until the deformation field readings tend to stabilize, and the transient deformation field of the entire monitoring area at this time is collected. By performing spatial field difference calculations between the transient deformation field and the reference deformation field, the deformation response field at the adjustment position under the load amplitude is obtained. Deformation response field This reflects the net impact distribution of the adjustment position on the deformation of all monitored areas under this load. After data acquisition, the load is unloaded, and the changes in the global deformation field are continuously monitored until the deformation field values ​​converge and return to the reference deformation field. Within the allowable deviation range, the calibration process for the next load level or the next adjustment position can only proceed after confirming that the mold has fully recovered its elasticity. If the deformation field fails to return to its original state after unloading, the load level is deemed to be outside the elastic range, and the corresponding calibration data is marked as invalid and discarded.

[0025] Obtain the deformation response field Then, match it with the corresponding adjustment position label. and calibrated load amplitude Common encapsulation to construct calibration mapping tuples To further characterize the spatial morphological features of the deformation response field, the following was conducted: The spatial gradient tensor field distribution is calculated by determining the partial derivatives of the deformation response field along each spatial direction at each monitoring location, forming the gradient tensor matrix. Principal direction analysis is then performed on the gradient tensor field to extract principal direction eigenvectors and their corresponding gradient intensities. The sequence of principal direction eigenvectors and the gradient intensity distribution are then combined to form the topological signature of the deformation response. and will Label the corresponding calibration mapping tuple, expanding the tuple to... The introduction of deformation response topological signatures enables accurate retrieval based on spatial morphological similarity during subsequent adjustment strategy matching, rather than relying solely on amplitude magnitude.

[0026] In the same adjustment position After completing the calibration of all load amplitude levels, summarize the set of calibration mapping tuples generated at this adjustment position for all effective amplitudes of the load spectrum. ,in To adjust position The effective calibrated load levels. Perform a consistency check on the set and calculate the spatial angle between the principal direction eigenvectors of the deformation response topological signatures of any two tuples in the set. ,like Exceeding the preset tolerance threshold This indicates a directional deviation in the spatial shape of the deformation response between the two calibrations, possibly caused by a slight offset in the mold mounting position or sensor drift. Calibration mapping tuples with excessive included angles should be removed from the set. Tolerance threshold. Based on the sensor accuracy and mold stiffness characteristics, the angle upper limit is usually set to ensure that the response topology at the same adjustment position maintains directional consistency within the linear elastic range. After removing inconsistent tuples, the remaining calibration mapping tuples retain valid calibration data with consistent direction and continuously varying amplitude, which can truly reflect the deformation response law of the adjustment position under different load amplitudes.

[0027] All valid calibration mapping tuples after screening of all regulation positions are compiled and indexed according to regulation position identifier and load amplitude to construct a complete regulation response mapping library. The mapping library is organized by regulation position identifier. and load amplitude Using the deformation response field as the retrieval key and deformation response topological signature To retrieve values, it supports quick querying of the global deformation response distribution at any adjustment position under a specified load amplitude during the subsequent adjustment strategy generation stage. Since the calibration process is completed in an independent offline station, it does not occupy stamping production time, and the calibration results can be reused throughout the entire life cycle of the mold. Local calibration is only re-executed and the corresponding entries are updated after the mold undergoes structural repair or the adjustment mechanism is replaced, ensuring the continuous validity of the mapping library data.

[0028] In one optional embodiment, constructing a calibration mapping tuple by combining the deformation response field with the corresponding adjustment position identifier and calibration load amplitude, extracting the spatial gradient tensor field distribution of the deformation response field, and constructing and labeling the deformation response topological signature to the calibration mapping tuple includes: The deformation response field, along with the corresponding adjustment position identifier and calibration load amplitude, is encapsulated into a calibration mapping tuple as the basic data structure. Gradient difference operation is performed on the deformation amplitude of each monitoring area in the deformation response field along the spatial coordinate direction to construct a spatial gradient tensor field. Tensor eigenvalue decomposition is performed on the spatial gradient tensor field to extract the eigenvector corresponding to the largest eigenvalue, determine the principal direction eigenvector, and detect the ratio of the second largest eigenvalue to the largest eigenvalue in the spatial gradient tensor field. When the ratio exceeds the preset directional dispersion threshold, the deformation response field is marked as a multimodal response and the secondary direction eigenvector corresponding to the second largest eigenvalue is recorded to the calibration mapping tuple. Identify the spatial coordinates of the peak value of deformation in the deformation response field, determine the centroid coordinates of the peak response, search for the spatial boundary position where the deformation amplitude decays to a preset proportion of the peak amplitude along the principal eigenvector direction, extract the spatial distance from the centroid coordinates of the peak response to the spatial boundary position, and obtain the response influence radius. The deformation response topological signature is constructed by combining the principal directional feature vector, the centroid coordinates of the peak response, and the radius of influence of the response. This signature is then marked in the topological feature field of the calibration mapping tuple, which is associated with the adjustment position identifier and the calibration load amplitude. For calibration mapping tuples marked as multimodal responses, the secondary directional feature vector is also recorded in the topological feature field.

[0029] In one specific implementation, after applying calibration loads and acquiring deformation response fields at each adjustment position, the acquired deformation response fields, along with their corresponding adjustment position identifiers and calibration load amplitudes, need to be structured and encapsulated to form a basic data structure of calibration mapping tuples that can be used for subsequent retrieval and matching. During the encapsulation process, the adjustment position identifiers... As a primary key index field, the calibrated load amplitude is used. As a load-level field, the deformation response field The data is stored in matrix form within the data body fields. Each calibration mapping tuple... The basic data structure is established at this stage, and the extracted topological features will be written into the same tuple as additional fields to ensure the integrity and traceability of data association.

[0030] The construction of the topological signature of the deformation response begins with the deformation response field. Calculation of the spatial gradient tensor field. For the deformation amplitude of each monitoring region in the deformation response field, along the spatial coordinates... direction and Gradient difference operations are performed along the direction to obtain the gradient components at each monitoring point, which are then combined to form the second-order spatial gradient tensor at that point. Let a monitoring point be along... The gradient components in the direction are ,along The gradient components in the direction are Then the gradient tensor at that point Can be derived from components and The resulting symmetric tensor representation. By repeating the above operation at all monitoring points across the entire domain, the spatial gradient tensor field covering the entire deformation response field can be obtained.

[0031] After the spatial gradient tensor field is established, eigenvalue decomposition is performed on the tensor field. For the gradient tensor at each monitoring point... This decomposes the data into a set of eigenvalues ​​and a set of eigenvectors. The largest eigenvalue is... Corresponding feature vector This reflects the spatial direction where the deformation gradient changes most dramatically at that point, and is defined as the principal direction feature vector, which is then stored in the calibration mapping tuple. The topological feature field. After determining the principal direction feature vector, the second largest feature value is further detected. With the largest eigenvalue ratio ,Right now .when Exceeding the preset directional dispersion threshold When this occurs, it indicates that the deformation response field has multiple significant gradient directions in space, and its distribution characteristics cannot be described by a single principal direction. In this case, the deformation response field is labeled as a multimodal response, and the second largest eigenvalue is... Corresponding secondary eigenvector The topological feature fields are synchronously recorded to the calibration mapping tuple to retain multi-directional distribution information for use in subsequent morphological similarity calculations.

[0032] After completing the gradient tensor feature analysis, the process proceeds to identify the centroid coordinates of the peak response. This is done within the deformation response field. In the search, the spatial location where the deformation amplitude reaches its maximum value is determined, and the spatial coordinates of this location are defined as the centroid coordinates of the peak response, denoted as . The centroid coordinates of the peak response represent the spatial location where the deformation response generated by the calibration load on the mold surface is most concentrated, and are a key geometric parameter describing the center of the response distribution.

[0033] After determining the centroid coordinates of the peak response, the eigenvectors along the principal direction are... The indicated direction is used to perform a spatial search for deformation amplitude from the center of mass outwards to both sides. During the search, the deformation amplitude at each location is continuously recorded until the deformation amplitude at a certain location decays to a preset proportion of the peak amplitude. (For example, take) When the peak response has decayed to 20% of its peak value, that location is marked as the spatial boundary. The centroid coordinates of the peak response are then extracted. The spatial distance between the location and the spatial boundary is used to obtain the response influence radius. Response radius This quantitatively describes the spatial coverage of the adjustment location's influence on the deformation of the surrounding area under the current load amplitude, and is an important parameter for assessing the sweep range of adjustment operations. For deformation response fields labeled as multimodal responses, it is also necessary to consider the eigenvectors along the secondary direction. Perform the same boundary search in the next direction to obtain the response influence radius in that direction. To fully describe the spatial coverage characteristics of multimodal responses.

[0034] The principal direction feature vector Peak response centroid coordinates With response influence radius The combination of these three elements constitutes the deformation response topological signature of the calibration mapping tuple. The topological signature is stored in the form of a compact geometric descriptor, avoiding data redundancy caused by directly storing the global deformation field, while retaining sufficient spatial topological information to support subsequent morphological similarity matching calculations. The constructed deformation response topological signature is then used. Label to calibration mapping tuple Center and Adjustment Position Marker and calibrated load amplitude The associated topological feature fields complete the full encapsulation of the tuple. For calibration mapping tuples labeled as multimodal responses, the secondary direction feature vectors... and the radius of influence of the secondary direction response Synchronously write to the topological feature field to ensure that the directional distribution information of the multimodal response is completely preserved.

[0035] During actual calibration, the same adjustment position At different load amplitude levels The topological signatures obtained below can be connected through the main direction feature vector. spatial angle Perform a consistency check. When Exceeding the preset tolerance threshold If this occurs, it indicates a significant shift in the principal direction of the response topology at that adjustment location under different load levels, necessitating verification or recalibration of the calibration data to ensure the reliability of the adjustment response mapping library. After the aforementioned full-process handling, each record in the adjustment response mapping library contains complete basic data and topological feature information, providing a reliable data foundation for subsequent reverse source matching of deformation distribution fields and the construction of multi-location collaborative adjustment strategies.

[0036] In one optional embodiment, the spatial morphological similarity between the deformation distribution field of the abnormal region and the deformation response topology of each adjustment location in the adjustment response mapping library is calculated. The adjustment location and load amplitude corresponding to the highest similarity are extracted through reverse tracing of the deformation response topology to determine the initial adjustment strategy, including: Spatial registration is performed between the deformation distribution field of the abnormal region and each deformation response field in the regulation response mapping library. Spatial field superposition is then performed to identify spatially overlapping and spatially conflicting regions. The coverage of the spatially overlapping regions and the conflict intensity of the spatially conflicting regions are extracted. When the coverage exceeds the preset matching threshold and the conflict intensity is lower than the preset conflict threshold, the corresponding calibration mapping tuple is included in the candidate regulation source set. Multiple sets of deformation response fields corresponding to the adjustment positions of the calibration mapping tuples in the candidate adjustment source set are retrieved under the incremental calibration load spectrum to construct a deformation response evolution sequence. In the deformation response evolution sequence, the target deformation response field that is closest to the deformation distribution field of the abnormal region is searched and the residual deformation difference field is calculated. When the energy ratio of the residual deformation difference field exceeds the preset residual threshold, the deformation response fields adjacent to the target deformation response field are interpolated to generate a compensation deformation response field. The interpolated load amplitude corresponding to the compensation deformation response field is extracted to determine the target value of the adjustment load. Extract the calibration mapping tuple with the highest coverage and lowest conflict intensity from the candidate regulation source set, and encapsulate the regulation location identifier and regulation load target value into an initial regulation strategy.

[0037] like Figure 2 The diagram shown illustrates the logic flowchart for generating the deformation adjustment strategy.

[0038] In one specific implementation, during the determination of the initial adjustment strategy, the spatial morphological similarity is calculated between the deformation distribution field of the abnormal region and the deformation response fields of each adjustment position stored in the adjustment response mapping library. By tracing back to the source, the adjustment position and corresponding load amplitude that best match the current abnormal deformation are found. The core of this process lies in spatial registration and candidate screening, rather than simple point-by-point numerical comparison. Therefore, a spatial field superposition and probing mechanism needs to be introduced to complete the initial morphological matching.

[0039] In the spatial registration stage, the deformation distribution field of the anomalous region is regarded as the target deformation field to be matched, and its spatial coordinate system is sequentially aligned with the deformation response field corresponding to each calibration mapping tuple in the adjustment response mapping library. After alignment, a spatial field superposition test is performed, that is, the target deformation field and the candidate deformation response field are compared region by region in the same coordinate system. During the superposition comparison, regions where the two fields have the same deformation direction, the same amplitude sign, and the amplitude difference is within the tolerance range are identified as spatially overlapping regions. Regions where the two fields have opposite deformation directions or opposite amplitude signs and have a significant cancellation effect are identified as spatially conflicting regions. The coverage rate of spatially overlapping regions is defined as the ratio of the area of ​​overlapping regions to the total area of ​​anomalous regions. The conflict intensity of spatially conflicting regions is quantified by the weighted integral of the deformation amplitude difference within the conflict region, reflecting the degree of antagonism between the two fields at the conflict location. When the coverage of a candidate deformation response field exceeds a preset matching threshold and the conflict intensity is lower than a preset conflict threshold, the candidate field is considered to have basic consistency with the current abnormal deformation field in terms of spatial morphology, and its corresponding calibration mapping tuple is included in the candidate regulation source set. Both the preset matching threshold and the preset conflict threshold are preset in the offline calibration stage according to the mold type, material properties, and production process requirements, and are dynamically corrected as the regulation response mapping library is iteratively updated.

[0040] After constructing the candidate regulation source set, the search phase for the deformation response evolution sequence begins. For each calibration mapping tuple in the candidate regulation source set, the complete deformation response fields at its corresponding regulation position under the incremental calibration load spectrum are retrieved and arranged in ascending order of load amplitude to construct the deformation response evolution sequence. This evolution sequence reflects the continuous evolution of the mold's global deformation response field as the load increases when different amplitude loads are applied at the same regulation position. In the evolution sequence, the deformation distribution field of the anomalous region is used as the search target, and the residual deformation difference field between the target deformation distribution field and each deformation response field in the evolution sequence is calculated one by one. The energy ratio of the residual deformation difference field is defined as the ratio of the difference field energy to the total energy of the target deformation distribution field, used to measure the degree of coverage and restoration of the target deformation distribution field by the candidate deformation response field. In the evolution sequence, the deformation response field that minimizes the energy ratio of the residual deformation difference field is selected as the target deformation response field.

[0041] When the proportion of residual deformation difference field energy corresponding to the target deformation response field still exceeds the preset residual threshold, it indicates that the target deformation distribution field cannot be accurately matched under the current discrete calibration load level. Interpolation needs to be performed between the target deformation response field and its adjacent deformation response fields to generate a compensated deformation response field. The interpolation process uses the load amplitude corresponding to the target deformation response field. and the load amplitude corresponding to the adjacent deformation response field Using the endpoints, the interpolation weights are determined according to the energy distribution characteristics of the residual deformation difference field. Interpolated load amplitude Calculated using the following formula: ; in These are the interpolation weights, with a range of values ​​of [value range missing]. The energy ratio is determined by the ratio of the residual deformation difference field to the adjacent deformation response field difference field. The generated compensated deformation response field is used to verify the deformation coverage effect under the interpolated load amplitude. If the residual energy ratio of the compensated deformation response field drops below the preset residual threshold, then... The final determined target value for the adjustment load is used; if it still exceeds the threshold, adjustments are made in finer increments. and Multiple interpolation iterations are performed until the accuracy requirement is met or the maximum number of interpolation iterations is reached.

[0042] Within the candidate regulation source set, there may be multiple calibration mapping tuples that satisfy both coverage and conflict intensity selection criteria. Further selection is needed to choose the optimal one and encapsulate it into an initial regulation strategy. The selection rule prioritizes the highest coverage as the primary ranking criterion. When coverage is the same or minimally different, the lowest conflict intensity is used as a secondary ranking criterion. The top-ranked calibration mapping tuple is extracted, its regulation location identifier is read, and combined with the regulation load target value calculated through interpolation, both are encapsulated into the initial regulation strategy. The initial regulation strategy contains clearly defined spatial coordinates of the regulation location, the direction of action, and precise load amplitude, providing a baseline action for the subsequent construction of multi-location coordinated regulation sequences.

[0043] If there is a certain angular rotational deviation between the deformation distribution field of the abnormal region and the candidate deformation response field during spatial registration, rotational alignment compensation needs to be introduced in the registration stage to eliminate the influence of spatial coordinate system deflection caused by mold clamping angle error on the coverage calculation. Rotational alignment compensation is achieved by angularly aligning the principal direction feature vectors of the two fields. The rotation angle range is limited to the clamping error range allowed by the process to prevent excessive rotation from causing incorrect matching. At the same time, the three key parameters of preset matching threshold, preset conflict threshold, and preset residual threshold need to be jointly calibrated in practical applications in combination with mold material, stamping process parameters, and historical adjustment data. The three together determine the strictness of the selection of candidate adjustment source set and the accuracy of the final initial adjustment strategy. Too high a matching threshold will result in an empty candidate set and no initial adjustment strategy can be output, while too low a conflict threshold may introduce incorrect adjustment sources with contradictory deformation directions. Therefore, the reasonable setting of the three thresholds has a direct impact on the overall adjustment and control effect.

[0044] like Figure 3 As shown, this graph presents the continuous evolution of the global deformation response of a candidate adjustment location in a specific anomaly region under an increasing calibration load spectrum during the reverse tracing stage. The horizontal axis of the graph represents spatial location (in mm), and the vertical axis represents deformation amplitude (in μm). The graph contains five core curves: one solid red line represents the target morphology of the "anomaly distribution field" captured during the current stamping process; the other four lines represent the deformation response evolution sequences extracted from the adjustment response mapping library for this adjustment location under calibration loads of 2kN, 4kN, 6kN, and 8kN. Figure 3At low loads (2kN, 4kN), the response amplitude is low and the spatial distribution trend differs significantly from the target field. As the load increases to 8kN, the blue solid line curve closely matches the red anomalous distribution field in terms of peak position, spatial coverage, and overall contour. This matching process does not rely on single-point values ​​but is based on the topological comparison of the "spatial gradient tensor" and the "principal direction eigenvector". Once the system identifies 8kN as the closest target deformation response field, it further calculates the residual deformation difference field. If the residual energy ratio exceeds the limit, interpolation is performed between adjacent load levels (such as 6kN and 8kN) to generate a compensation response. Figure 3 It intuitively demonstrates the load optimization process from coarse to fine, which is the key decision-making basis for determining the initial adjustment strategy (adjustment position + load target value). It ensures that the adjustment action can accurately hit the real position where deformation occurs from the source, avoiding the inefficiency and misjudgment caused by traditional experience-based trial and error.

[0045] In one optional embodiment, performing a spatial field overlay probe to identify spatially overlapping and conflicting regions, and extracting the coverage of the spatially overlapping regions and the conflict intensity of the spatially conflicting regions includes: The deformation distribution field of the abnormal region and each deformation response field are calculated point by point in the overlapping area after spatial registration. When the deformation vector dot product is positive, it is marked as a cooperative action point, and when the deformation vector dot product is negative, it is marked as an antagonistic action point. Perform connectivity analysis on the set of cooperative action points, extract the largest connectivity as the spatially overlapping region, and vector superimpose the deformation distribution field and deformation response field of the abnormal region within the spatially overlapping region. When the residual deformation amplitude after superposition is lower than the deformation distribution field amplitude of the abnormal region before superposition, the spatially overlapping region is confirmed as an effective cancellation region, and the coverage rate of the effective cancellation region to the total area of ​​the abnormal region is calculated. The deformation vector superposition calculation is performed on the set of opposing action points. The spatial region where the deformation amplitude after superposition exceeds the peak value of the deformation distribution field in the abnormal region is extracted as the spatial conflict region. The excess multiple of the superposition deformation amplitude in the spatial conflict region relative to the peak value of the deformation distribution field in the abnormal region is calculated to determine the conflict intensity.

[0046] In one specific implementation, after determining the initial adjustment strategy, it is necessary to quantitatively evaluate the spatial interaction between the deformation response field of each candidate adjustment location and the deformation distribution field of the anomalous region, thereby providing a reliable basis for constructing a multi-location coordinated adjustment sequence. The core of the spatial field superposition trial is to determine whether the deformation response of a certain adjustment location can effectively offset the deformation of the anomalous region, while identifying whether it will trigger new deformation conflicts in other regions.

[0047] Spatially register the deformation distribution field of the abnormal region with the deformation response field of the candidate adjustment positions, placing them in a unified mold coordinate system. After spatial registration, extract the deformation vector from the deformation distribution field of the abnormal region and the deformation vector from the deformation response field point by point within the overlapping region of the two fields for each monitoring point, and perform a dot product operation on these two vectors. The result of the dot product operation reflects the degree of consistency of the two vectors in direction: when the dot product result is positive, it indicates that the two vectors tend to be consistent in direction at that point, and the adjustment response at that point and the abnormal deformation are superimposed in the same direction, marking this point as a cooperative action point; when the dot product result is negative, it indicates that the two vectors are opposite in direction or there is a significant reverse component, and the adjustment response at that point and the abnormal deformation are antagonistic in direction, marking this point as an antagonistic action point. Through point-by-point marking, two spatial point sets are formed within the overlapping region: a set of cooperative action points and a set of antagonistic action points.

[0048] Connectivity analysis is performed on the set of cooperating points to group spatially adjacent points that are all cooperating points into the same connected region. The connected region analysis employs a grid-neighborhood-based region growing approach. For each cooperating point, it checks if there are other cooperating points within its eight-neighborhood; if so, they are included in the same connected region until all cooperating points have been assigned. The connected region with the largest area among all connected regions is extracted as the spatially overlapping region, representing the largest spatially cooperating coverage of the regulatory response and anomalous deformation.

[0049] After identifying the spatially overlapping region, a vector superposition calculation is performed on the deformation distribution field and deformation response field of the anomalous region within this region. Specifically, the deformation vectors of the two fields are vector-added at each monitoring point to obtain the superimposed residual deformation vector, and then the residual deformation amplitude is calculated. The superimposed residual deformation amplitude is compared with the deformation amplitude of the anomalous region's deformation distribution field at the corresponding point before superposition. If the superimposed residual deformation amplitude is lower than the amplitude of the anomalous region's deformation distribution field before superposition, it indicates that the control response has a counteracting effect on the anomalous deformation in this region, confirming the spatially overlapping region as an effective counteracting region. After confirming the effective counteracting region, the ratio of the area of ​​the effective counteracting region to the total area of ​​the anomalous region is calculated. This ratio is the coverage rate, used to measure the degree of spatial coverage of the anomalous region's deformation by the current candidate control position. The higher the coverage rate, the better the spatial matching of the deformation response of the control position with the deformation distribution of the anomalous region, indicating stronger control targeting.

[0050] For the set of counteracting points, a deformation vector superposition calculation is performed. Specifically, at each counteracting point, the deformation vector of the abnormal region's deformation distribution field is vector-added with the deformation vector of the deformation response field to obtain the superimposed composite deformation vector, and the amplitude of the composite deformation vector is calculated. The set of spatial points whose superimposed deformation amplitude exceeds the peak value of the abnormal region's deformation distribution field is extracted, and the region formed by these points is defined as the spatial conflict region. The identification logic for the spatial conflict region is as follows: if the superimposed deformation amplitude at the counteracting point not only fails to decrease but also exceeds the original deformation peak value of the abnormal region, it means that the adjustment action has introduced a more severe deformation concentration in this region, which is a conflict behavior that needs to be avoided.

[0051] After the spatial conflict region is determined, the excess multiple of the superimposed deformation amplitude at each point within the region relative to the peak value of the deformation distribution field in the anomalous region is calculated. This is done by dividing the superimposed deformation amplitude at each point by the peak value of the deformation distribution field in the anomalous region, yielding the excess multiple distribution for each point. Statistical characteristic values ​​(such as the maximum value or area-weighted average) of the excess multiples at all points within the spatial conflict region are used as a quantitative indicator of conflict intensity. Higher conflict intensity indicates more severe secondary deformation introduced by the candidate adjustment location during adjustment, requiring higher penalty weights in the screening of multi-location coordinated adjustment sequences or compensation through auxiliary adjustment locations.

[0052] By jointly evaluating the coverage rate of spatially overlapping areas and the conflict intensity of spatially conflicting areas, a comprehensive spatial quantitative description of the regulation effect of each candidate regulation location can be achieved. Candidate regulation locations with high coverage and low conflict intensity are preferentially included in the coordinated regulation sequence. For candidate regulation locations with high conflict intensity, auxiliary regulation locations need to be matched in the regulation sequence. The deformation response of the auxiliary regulation locations compensates for and offsets the conflict areas, thereby effectively eliminating the deformation of abnormal areas as a whole, while avoiding the introduction of new deformation anomalies in other areas of the mold. This evaluation mechanism provides a quantitative spatial interaction basis for the subsequent refined construction of multi-location coordinated regulation sequences, ensuring the rationality and stability of the regulation strategy at the global deformation control level.

[0053] In one optional embodiment, using the initial adjustment strategy as the baseline action, the secondary deformation distribution generated by the adjustment position on the non-abnormal region is extracted from the adjustment response mapping library, and superimposed with the current deformation field of the non-abnormal region for pre-simulation. Regions where the spatial gradient value exceeds the mean after superposition are extracted as predicted secondary abnormal regions. Auxiliary adjustment positions and offset adjustment amounts are retrieved and screened to construct a multi-position coordinated adjustment sequence, including: Extract the global deformation response field corresponding to the adjustment position in the initial adjustment strategy from the adjustment response mapping library, and separate the secondary deformation distribution outside the abnormal region after scaling according to the target value of the adjustment load; The secondary deformation distribution is superimposed with the current deformation field of the non-abnormal region to obtain the pre-simulated superimposed deformation field. The spatial gradient distribution is calculated on the pre-simulated superimposed deformation field, and the region with the gradient value exceeding the average gradient value of the current deformation field of the non-abnormal region is extracted to obtain the predicted secondary abnormal region. The set of adjustment locations that generate deformation response in the predicted secondary anomaly area is retrieved from the adjustment response mapping library. The deformation response field of each adjustment location is superimposed with the secondary deformation distribution for trial calculation. When the deformation amplitude after superposition is lower than the amplitude of the secondary deformation distribution before superposition, it is marked as an auxiliary adjustment location. The load amplitude optimization solution is performed on the auxiliary adjustment location to minimize the residual deformation energy after superposition. The optimized load amplitude is extracted as the offset adjustment amount. The initial adjustment strategy is taken as the main adjustment action, and the auxiliary adjustment position and the offset adjustment amount are encapsulated as auxiliary adjustment actions into a multi-position coordinated adjustment sequence.

[0054] In one specific implementation, the potential collateral effects of the strategy on non-abnormal areas of the mold during execution are further evaluated. Due to the overall continuity of the mold structure, the load applied at any adjustment position not only acts on the target abnormal area but also generates a diffuse deformation response across the entire domain. Therefore, the global deformation response field corresponding to the adjustment position of the initial adjustment strategy is extracted from the adjustment response mapping library. This response field is then linearly scaled according to the determined target value of the adjustment load, ensuring its amplitude corresponds to the magnitude of the actual load to be applied. After scaling, the portion of the global deformation response field within the coverage area of ​​the abnormal area is removed, retaining the response components outside the boundary of the abnormal area, which constitutes the secondary deformation distribution. The secondary deformation distribution reflects the "side effect" deformation produced by the initial adjustment action in the non-target area and is the core input for subsequent pre-analysis.

[0055] The secondary deformation distribution is superimposed point-by-point with the current deformation field in the non-abnormal region to obtain the pre-simulated superimposed deformation field. This superposition operation simulates the deformation state that the non-abnormal region will exhibit after the initial adjustment strategy is implemented. It is a purely computational prediction and deduction, without involving any actual physical adjustment actions. The gradient distribution of the pre-simulated superimposed deformation field is calculated spatially, using the same method as the extraction of deformation gradient values. That is, the deformation change rate in the neighborhood of each monitoring point in the superimposed deformation field is calculated to form a complete gradient distribution map. This gradient distribution is compared point-by-point with the gradient mean of the current deformation field in the non-abnormal region. The set of monitoring points whose gradient values ​​exceed the mean is extracted, and the spatial area covered by these points is marked as the predicted secondary anomaly region. The predicted secondary anomaly region represents the deformation concentration zone that may emerge in the non-abnormal region if only the initial adjustment strategy is implemented without any compensation measures. It is an important basis for the design of the collaborative adjustment strategy.

[0056] For the identified predicted secondary anomaly regions, a set of control positions capable of generating effective deformation responses in these regions is retrieved from the control response mapping library. The retrieval criterion is that the deformation response field of the control position has significant amplitude coverage within the predicted secondary anomaly region, meaning that the response influence range of the control position spatially overlaps with the predicted secondary anomaly region. Each candidate control position is tested: the deformation response field of each candidate control position is superimposed with the secondary deformation distribution, and the change in deformation amplitude within the predicted secondary anomaly region after superposition is observed. If the deformation amplitude in the superimposed region is lower than the amplitude of the secondary deformation distribution in the same region before superposition, it indicates that the response direction of the candidate control position and the secondary deformation distribution have a certain canceling relationship in this region, and it is marked as an auxiliary control position. If the amplitude increases after superposition, it indicates that the candidate position will exacerbate the secondary anomaly and is excluded from the auxiliary control set.

[0057] For all candidate locations marked as auxiliary adjustment positions, load amplitude optimization is performed to minimize the residual deformation energy within the predicted secondary anomaly region after the combined effect of all auxiliary adjustment positions. Residual deformation energy is defined as the sum of squares of the deformation amplitudes at each monitoring point within the predicted secondary anomaly region after superimposing the response fields and secondary deformation distributions of all auxiliary adjustment positions. Let there be a total of [number missing] auxiliary adjustment positions. The auxiliary adjustment position, the first The load scaling factor to be optimized for each auxiliary adjustment position is: Its deformation response field vector within the predicted secondary anomaly region is The vector of secondary deformation distribution in this region is The residual deformation energy It can be represented as .right Regarding each To find the minimum value, a system of linear equations can be established, and the optimal load scaling factor for each auxiliary adjustment position can be solved using the least squares method. Multiplying the optimal scaling factor by the calibrated load amplitude for the corresponding adjustment position in the adjustment response mapping library yields the offset adjustment amount for each auxiliary adjustment position. To ensure that the offset adjustment amount is physically feasible, the solution results must be constrained within the load amplitude range covered by the adjustment response mapping library; if the solution results exceed the range, they are truncated to the boundary value.

[0058] After identifying the auxiliary adjustment positions and determining the offset adjustment amounts, the initial adjustment strategy is encapsulated as the main adjustment action, and each auxiliary adjustment position and its corresponding offset adjustment amount are encapsulated as an auxiliary adjustment action. Following the principle of prioritizing the main adjustment action and sequentially following the auxiliary adjustment actions, a multi-position coordinated adjustment sequence is constructed. The execution order of each adjustment action in the sequence is determined based on the spatial coupling relationship between the adjustment positions: if the response fields of the auxiliary adjustment positions and the main adjustment positions have significant spatial overlap, the auxiliary adjustment actions are scheduled to be executed after the main adjustment action is completed and stabilized to avoid nonlinear interference from response superposition; if the response fields of the auxiliary adjustment positions and the main adjustment positions are spatially separated, they can be arranged to be executed in parallel or near-synchronously to improve overall adjustment efficiency.

[0059] The construction of a multi-location coordinated adjustment sequence allows the entire adjustment process to move beyond isolated interventions at a single location, integrating the primary adjustment objective and the suppression of secondary anomalies into a unified adjustment scheme. By identifying secondary anomaly risks in advance and pre-setting offset adjustment amounts during the pre-implementation phase, the concentrated deformation in non-target areas caused by structural coupling can be suppressed simultaneously while executing the initial adjustment strategy. This enables overall optimization of the die deformation state within a single adjustment cycle, effectively reducing the frequency of repeated adjustments and improving the adjustment and control efficiency and accuracy of stamping dies.

[0060] In one optional embodiment, executing a multi-location coordinated adjustment sequence, calculating the deformation deviation distribution between the actual deformation change distribution and the predicted deformation response field, and iteratively updating the adjustment parameters by writing them into the adjustment response mapping library based on the deviation energy ratio includes: Loads are applied sequentially according to the timing of the main and auxiliary adjustment actions in the multi-position coordinated adjustment sequence. During the load application process, the structural global deformation field data is collected in real time. The difference between the deformation field before and after adjustment is extracted to obtain the actual deformation change distribution. The actual deformation change distribution is spatially compared with the predicted deformation response field before the execution of the multi-position coordinated adjustment sequence, and the deformation deviation distribution is calculated. When the energy proportion of the deformation deviation distribution exceeds the preset deviation threshold, the actual deformation change distribution, the adjustment position identifier and the actual applied load amplitude are encapsulated into a correction calibration mapping tuple. The correction calibration mapping tuple is written into the calibration mapping set of the corresponding adjustment position in the adjustment response mapping library, and the deformation response field of the adjustment position under the current load amplitude is updated. When the energy proportion of the deformation deviation distribution is lower than the deviation threshold, the spatial correlation coefficient between the actual deformation change distribution and the predicted deformation response field is extracted. The spatial correlation coefficient is used as a confidence marker and associated with the calibration mapping tuple of the corresponding adjustment position in the adjustment response mapping library to complete the knowledge iterative update of the adjustment response mapping library.

[0061] In one specific implementation, when executing a multi-position coordinated adjustment sequence, the main adjustment action and the auxiliary adjustment action apply loads sequentially according to a pre-arranged time sequence. Throughout the application process, the deformation field of the entire structure is continuously acquired in real time, and the sampling frequency must be sufficient to capture the complete transition process of the deformation field stabilizing and converging after each load application. Using the deformation field snapshot before each adjustment action as a reference, the difference is calculated between this snapshot and the stable state of the deformation field after the action, yielding the actual deformation change distribution corresponding to that action. The actual deformation change distributions of all adjustment actions are spatially superimposed to form the total actual deformation change distribution after the entire multi-position coordinated adjustment sequence has been completed.

[0062] The total actual deformation distribution is spatially compared point-by-point with the predicted deformation response field obtained from the regulation response mapping library before the execution of the multi-location coordinated regulation sequence, and the deformation deviation distribution is calculated. The deformation deviation distribution reflects the spatial difference between the actual execution result and the theoretical prediction, where positive deviation regions indicate that the actual deformation response is greater than the predicted value, and negative deviation regions indicate that the actual deformation response is less than the predicted value. To quantify the overall deviation level, a deviation energy percentage index is introduced. The spatial integral energy of the deformation deviation distribution is divided by the spatial integral energy of the predicted deformation response field to obtain the normalized deviation energy percentage. . The calculation method is the ratio of the sum of squares of the deformation deviation values ​​at each monitoring point in the deviation field to the sum of squares of the deformation response values ​​at each monitoring point in the predicted deformation response field, i.e. ,in For the first Deformation deviation values ​​at each monitoring point For the first Predicted deformation response values ​​at each monitoring point.

[0063] when Exceeding the preset deviation threshold If this occurs, it indicates a significant deviation between the deformation response field at the corresponding adjustment location in the current adjustment response mapping library and the actual response, necessitating a forced correction and update of the mapping library. In this case, the actual deformation distribution collected during the current execution, the unique identifier of the corresponding adjustment location, and the actual applied load amplitude are encapsulated into a correction calibration mapping tuple. ,in To adjust the location identifier index, This is the index of the load amplitude level corresponding to the applied load, indicated by the superscript. This indicates that the tuple originated from online calibration rather than offline calibration. Map the calibration calibration tuple. Write the corresponding adjustment position into the adjustment response mapping library The calibration mapping set, replacing or supplementing the adjustment position at the current load amplitude level. The deformation response field record is used to perform a mandatory overwrite update of the entry. If the load amplitude level already exists in the original calibration mapping set, the correction tuple is used for priority overwrite; if the load amplitude level is a newly appearing amplitude, it is added to the set as a new entry, and the adjustment position is updated synchronously. Effective calibration load level .

[0064] when Below the deviation threshold This indicates that the actual response of this adjustment is in overall agreement with the predicted deformation response field, eliminating the need for forced overwriting of the deformation response field data in the mapping library. However, the results of this adjustment can be used to quantify the reliability of the original calibration mapping tuples. The spatial correlation coefficient between the actual deformation distribution and the predicted deformation response field is extracted. The calculation method is to divide the covariance of the two data points across the entire monitoring area by the product of their respective standard deviations, i.e. ,in This is the actual deformation distribution vector. To predict the deformation response field vector. The range of values ​​is The closer the value is to 1, the higher the spatial morphology of the actual response matches the predicted response.

[0065] Will As a confidence level marker, it is associated with the corresponding regulation position in the regulation response mapping library. calibration mapping tuple The confidence level is updated using a cumulative weighted average method. That is, after multiple verification runs of the calibration mapping tuple at the same adjustment location and with the same load amplitude, the spatial correlation coefficients calculated in each run are weighted and averaged according to the number of runs to reflect the overall reliability of the entry across multiple actual runs. Calibration mapping tuples with high confidence levels have higher priority in subsequent deformation response topological similarity calculations and adjustment strategy generation, helping to make more reliable decisions among multiple candidate adjustment locations with similarity.

[0066] In actual execution, if the deformation field fails to converge to a stable state within a preset stabilization waiting time after a certain adjustment action is applied, it is determined that the adjustment action has structural response lag. The sampling waiting window needs to be extended until the rate of change of the deformation field is lower than the preset stability criterion threshold before the actual deformation change distribution is extracted. For cases where the acquisition time is offset due to lag, the time difference between the actual acquisition time and the standard time is recorded, and this time difference information is appended to the correction calibration mapping tuple or confidence level flag as auxiliary reference information for subsequent analysis of the dynamic response characteristics of the adjustment position.

[0067] The iterative update mechanism drives the continuous evolution of the regulation response mapping library through two paths—forced overwrite update and confidence label update. As the number of executions accumulates, the deformation response field data for each regulation location in the mapping library gradually converges to the true response under actual operating conditions, and the confidence distribution of the calibration mapping tuples gradually stabilizes. When all effective load amplitude level entries for a certain regulation location have accumulated a sufficient number of high-confidence verification records, the response characteristics of that regulation location can be considered fully characterized. Subsequent regulation strategy generation can directly use high-confidence entries as the primary basis, reducing reliance on the original offline calibration data, thereby achieving an adaptive improvement in regulation control performance with increasing usage.

[0068] like Figure 4 As shown, the graph illustrates the decreasing trend of the "deviation energy ratio" between the actual deformation response and the theoretical predicted response as the number of iterations increases in a mid-frame stamping die adjustment and control system across multiple production cycles. The horizontal axis represents the iteration cycle (cycle 1 to cycle 12), and the vertical axis represents the deviation energy ratio (percentage, %). The graph includes a downward trend curve and a red dashed line representing a preset deviation threshold (10%). Figure 4 Data shows that the initial round (round 1) had a deviation energy as high as approximately 45%, indicating a significant deviation between the original offline calibration data and the actual working conditions. After each round of adjustment, the system selectively performs "forced correction writing" (when the deviation energy percentage is higher than 10%) or "confidence mark accumulation" (when the deviation energy percentage is lower than 10%) based on the deviation between the actual deformation distribution and the predicted field. The curve shows a rapid downward trend, converging quickly to around 18% in the first six rounds, stabilizing below the 10% threshold after the 10th round, and approaching 4% in the 12th round. This intuitively demonstrates that through continuous knowledge iteration, the deformation response data of the mapping library gradually converges to the true physical characteristics of the mold, and the control accuracy improves with each production batch. When the curve crosses the red threshold line, it prioritizes high-confidence entries for strategy generation, marking a smooth transition from "data-driven correction" to the advanced stage of "knowledge-driven autonomous decision-making" in the adjustment control, perfectly solving the batch quality fluctuation problem caused by temperature drift and increased wear in traditional stamping dies.

[0069] A second aspect of the present invention provides a mid-frame stamping die adjustment and control system, comprising: Anomaly identification unit is used to acquire deformation state data of each monitoring area on the surface of the die during the stamping process, identify abnormal areas where the deformation gradient value exceeds the mean, extract the spatial distribution location and deformation amplitude of the abnormal areas, and construct a deformation distribution field. The response mapping unit is used to apply calibration loads of different amplitudes to each adjustment position of the mold at an independent offline calibration station. After each calibration, deformation response data of the entire monitoring area of ​​the mold is collected to establish an adjustment response mapping library. The source matching unit is used to calculate the spatial morphological similarity between the deformation distribution field of the abnormal area and the deformation response topology of each adjustment position in the adjustment response mapping library. By reverse tracing the deformation response topology, the adjustment position and load amplitude corresponding to the highest similarity are extracted to determine the initial adjustment strategy. The collaborative planning unit is used to extract the secondary deformation distribution of the adjustment position to the non-abnormal region from the adjustment response mapping library, take the initial adjustment strategy as the baseline action, superimpose and pre-simulate with the current deformation field of the non-abnormal region, extract the region where the spatial gradient value exceeds the mean after superposition as the predicted secondary abnormal region, search and filter the auxiliary adjustment position and the offset adjustment amount, and construct a multi-position collaborative adjustment sequence. The iterative update unit is used to execute the multi-location coordinated adjustment sequence, calculate the deformation deviation distribution between the actual deformation change distribution and the predicted deformation response field, and write the adjustment parameters into the adjustment response mapping library for iterative update based on the deviation energy ratio.

[0070] A third aspect of the present invention provides an electronic device, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0071] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0072] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0073] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for adjusting and controlling a middle frame member stamping die, characterized by, include: Acquire deformation state data of each monitoring area on the die surface during the stamping process, identify abnormal areas where the deformation gradient value exceeds the mean, extract the spatial distribution location and deformation amplitude of the abnormal areas, and construct a deformation distribution field. Different amplitude calibration loads are applied to each adjustment position of the mold at an independent offline calibration station. After each calibration, deformation response data of the entire monitoring area of ​​the mold is collected to establish an adjustment response mapping library. Spatial morphological similarity is calculated between the deformation distribution field of the abnormal region and the deformation response topology of each adjustment position in the adjustment response mapping library. The adjustment position and load amplitude corresponding to the highest similarity are extracted by reverse tracing of the deformation response topology to determine the initial adjustment strategy. Using the initial adjustment strategy as the baseline action, the secondary deformation distribution generated by the adjustment position on the non-abnormal region is extracted from the adjustment response mapping library, and superimposed with the current deformation field of the non-abnormal region for pre-simulation. Regions with spatial gradient values ​​exceeding the mean after superposition are extracted as predicted secondary abnormal regions. Auxiliary adjustment positions and offset adjustment amounts are retrieved and screened to construct a multi-position coordinated adjustment sequence. Execute a multi-location coordinated regulation sequence, calculate the deformation deviation distribution between the actual deformation change distribution and the predicted deformation response field, and write the regulation parameters into the regulation response mapping library for iterative updates based on the deviation energy ratio.

2. The method of claim 1, wherein, Different amplitude calibration loads were applied to each adjustment position of the mold at an independent offline calibration station. After each calibration, deformation response data of the entire monitoring area of ​​the mold were collected, and an adjustment response mapping library was established, including: A pre-tightening load is applied to the mold at an independent offline calibration station to bring the mold to the initial calibration state, and the reference deformation field of each monitoring area is collected under the initial calibration state. The calibration traversal sequence is arranged according to the spatial topological relationship of the adjustment position. The calibration load spectrum is applied to each adjustment position in the calibration traversal sequence in a step-increasing manner. The load is kept stable and the transient deformation field of the entire monitoring area is collected each time the calibration load is applied. The deformation response field is obtained by performing spatial field difference operation between the transient deformation field and the reference deformation field. The load is unloaded and the deformation field is monitored to converge and regress to the reference deformation field. Construct a calibration mapping tuple by combining the deformation response field with the corresponding adjustment position identifier and calibration load amplitude, extract the spatial gradient tensor field distribution of the deformation response field, construct the deformation response topological signature and label it to the calibration mapping tuple. The deformation response topology signatures generated at the same adjustment position under all calibration load amplitudes included in the calibration load spectrum are summarized. Calibration mapping tuples in the deformation response topology signatures whose spatial angle between the principal direction feature vectors exceeds a preset tolerance threshold are removed. The remaining calibration mapping tuples are then summarized to construct an adjustment response mapping library.

3. The method of claim 2, wherein, The deformation response field is used to construct a calibration mapping tuple with the corresponding adjustment position identifier and calibration load amplitude. The spatial gradient tensor field distribution of the deformation response field is extracted, and the topological signature of the deformation response is constructed and labeled to the calibration mapping tuple, including: The deformation response field, along with the corresponding adjustment position identifier and calibration load amplitude, is encapsulated into a basic data structure called a calibration mapping tuple. Gradient difference operation is performed on the deformation amplitude of each monitoring area in the deformation response field along the spatial coordinate direction to construct a spatial gradient tensor field. Tensor eigenvalue decomposition is performed on the spatial gradient tensor field to extract the eigenvector corresponding to the largest eigenvalue, determine the principal direction eigenvector, and detect the ratio of the second largest eigenvalue to the largest eigenvalue in the spatial gradient tensor field. When the ratio exceeds the preset directional dispersion threshold, the deformation response field is marked as a multimodal response and the secondary direction eigenvector corresponding to the second largest eigenvalue is recorded to the calibration mapping tuple. Identify the spatial coordinates of the peak value of deformation in the deformation response field, determine the centroid coordinates of the peak response, search for the spatial boundary position where the deformation amplitude decays to a preset proportion of the peak amplitude along the principal eigenvector direction, extract the spatial distance from the centroid coordinates of the peak response to the spatial boundary position, and obtain the response influence radius. The deformation response topological signature is constructed by combining the principal directional feature vector, the centroid coordinates of the peak response, and the radius of influence of the response. This signature is then marked in the topological feature field of the calibration mapping tuple, which is associated with the adjustment position identifier and the calibration load amplitude. For calibration mapping tuples marked as multimodal responses, the secondary directional feature vector is also recorded in the topological feature field.

4. The method of claim 1, wherein, The spatial morphological similarity between the deformation distribution field of the abnormal region and the deformation response topology of each adjustment location in the adjustment response mapping library is calculated. By tracing the deformation response topology backwards, the adjustment location and load amplitude corresponding to the highest similarity are extracted to determine the initial adjustment strategy, including: Spatial registration is performed between the deformation distribution field of the abnormal region and each deformation response field in the regulation response mapping library. Spatial field superposition is then performed to identify spatially overlapping and spatially conflicting regions. The coverage of the spatially overlapping regions and the conflict intensity of the spatially conflicting regions are extracted. When the coverage exceeds the preset matching threshold and the conflict intensity is lower than the preset conflict threshold, the corresponding calibration mapping tuple is included in the candidate regulation source set. Multiple sets of deformation response fields corresponding to the adjustment positions of the calibration mapping tuples in the candidate adjustment source set are retrieved under the incremental calibration load spectrum to construct a deformation response evolution sequence. In the deformation response evolution sequence, the target deformation response field that is closest to the deformation distribution field of the abnormal region is searched and the residual deformation difference field is calculated. When the energy ratio of the residual deformation difference field exceeds the preset residual threshold, the deformation response fields adjacent to the target deformation response field are interpolated to generate a compensation deformation response field. The interpolated load amplitude corresponding to the compensation deformation response field is extracted to determine the target value of the adjustment load. Extract the calibration mapping tuple with the highest coverage and lowest conflict intensity from the candidate regulation source set, and encapsulate the regulation location identifier and regulation load target value into an initial regulation strategy.

5. The method of claim 4, wherein, Perform spatial field overlay probing to identify overlapping and conflicting spatial regions, and extract the coverage of overlapping regions and the conflict intensity of conflicting regions, including: The deformation distribution field of the abnormal region and each deformation response field are calculated point by point in the overlapping area after spatial registration. When the deformation vector dot product is positive, it is marked as a cooperative action point, and when the deformation vector dot product is negative, it is marked as an antagonistic action point. Perform connectivity analysis on the set of cooperative action points, extract the largest connectivity as the spatially overlapping region, and vector superimpose the deformation distribution field and deformation response field of the abnormal region within the spatially overlapping region. When the residual deformation amplitude after superposition is lower than the deformation distribution field amplitude of the abnormal region before superposition, the spatially overlapping region is confirmed as an effective cancellation region, and the coverage rate of the effective cancellation region to the total area of ​​the abnormal region is calculated. The deformation vector superposition calculation is performed on the set of opposing action points. The spatial region where the deformation amplitude after superposition exceeds the peak value of the deformation distribution field in the abnormal region is extracted as the spatial conflict region. The excess multiple of the superposition deformation amplitude in the spatial conflict region relative to the peak value of the deformation distribution field in the abnormal region is calculated to determine the conflict intensity.

6. The method of claim 1, wherein, Using the initial adjustment strategy as the baseline action, the secondary deformation distribution of the adjustment position on the non-abnormal region is extracted from the adjustment response mapping library, and superimposed with the current deformation field of the non-abnormal region for pre-simulation. Regions with spatial gradient values ​​exceeding the mean after superposition are extracted as predicted secondary abnormal regions. Auxiliary adjustment positions and offset adjustment amounts are retrieved and screened to construct a multi-position coordinated adjustment sequence, including: Extract the global deformation response field corresponding to the adjustment position in the initial adjustment strategy from the adjustment response mapping library, and separate the secondary deformation distribution outside the abnormal region after scaling according to the target value of the adjustment load; The secondary deformation distribution is superimposed with the current deformation field of the non-abnormal region to obtain the pre-simulated superimposed deformation field. The spatial gradient distribution is calculated on the pre-simulated superimposed deformation field, and the region with the gradient value exceeding the average gradient value of the current deformation field of the non-abnormal region is extracted to obtain the predicted secondary abnormal region. The set of adjustment locations that generate deformation response in the predicted secondary anomaly area is retrieved from the adjustment response mapping library. The deformation response field of each adjustment location is superimposed with the secondary deformation distribution for trial calculation. When the deformation amplitude after superposition is lower than the amplitude of the secondary deformation distribution before superposition, it is marked as an auxiliary adjustment location. The load amplitude optimization solution is performed on the auxiliary adjustment location to minimize the residual deformation energy after superposition. The optimized load amplitude is extracted as the offset adjustment amount. The initial adjustment strategy is taken as the main adjustment action, and the auxiliary adjustment position and the offset adjustment amount are encapsulated as auxiliary adjustment actions into a multi-position coordinated adjustment sequence.

7. The method according to claim 1, characterized in that, Execute a multi-location coordinated regulation sequence, calculate the deformation deviation distribution between the actual deformation change distribution and the predicted deformation response field, and iteratively update the regulation parameters into the regulation response mapping library based on the deviation energy ratio, including: Loads are applied sequentially according to the timing of the main and auxiliary adjustment actions in the multi-position coordinated adjustment sequence. During the load application process, the structural global deformation field data is collected in real time. The difference between the deformation field before and after adjustment is extracted to obtain the actual deformation change distribution. The actual deformation change distribution is spatially compared with the predicted deformation response field before the execution of the multi-position coordinated adjustment sequence, and the deformation deviation distribution is calculated. When the energy proportion of the deformation deviation distribution exceeds the preset deviation threshold, the actual deformation change distribution, the adjustment position identifier and the actual applied load amplitude are encapsulated into a correction calibration mapping tuple. The correction calibration mapping tuple is written into the calibration mapping set of the corresponding adjustment position in the adjustment response mapping library, and the deformation response field of the adjustment position under the current load amplitude is updated. When the energy proportion of the deformation deviation distribution is lower than the deviation threshold, the spatial correlation coefficient between the actual deformation change distribution and the predicted deformation response field is extracted. The spatial correlation coefficient is used as a confidence marker and associated with the calibration mapping tuple of the corresponding adjustment position in the adjustment response mapping library to complete the knowledge iterative update of the adjustment response mapping library.

8. A mid-frame stamping die adjustment and control system for implementing the method as described in any one of claims 1-7, characterized in that, include: Anomaly identification unit is used to acquire deformation state data of each monitoring area on the surface of the die during the stamping process, identify abnormal areas where the deformation gradient value exceeds the mean, extract the spatial distribution location and deformation amplitude of the abnormal areas, and construct a deformation distribution field. The response mapping unit is used to apply calibration loads of different amplitudes to each adjustment position of the mold at an independent offline calibration station. After each calibration, deformation response data of the entire monitoring area of ​​the mold is collected to establish an adjustment response mapping library. The source matching unit is used to calculate the spatial morphological similarity between the deformation distribution field of the abnormal area and the deformation response topology of each adjustment position in the adjustment response mapping library. By reverse tracing the deformation response topology, the adjustment position and load amplitude corresponding to the highest similarity are extracted to determine the initial adjustment strategy. The collaborative planning unit is used to extract the secondary deformation distribution of the adjustment position to the non-abnormal region from the adjustment response mapping library, take the initial adjustment strategy as the baseline action, superimpose and pre-simulate with the current deformation field of the non-abnormal region, extract the region where the spatial gradient value exceeds the mean after superposition as the predicted secondary abnormal region, search and filter the auxiliary adjustment position and the offset adjustment amount, and construct a multi-position collaborative adjustment sequence. The iterative update unit is used to execute the multi-location coordinated adjustment sequence, calculate the deformation deviation distribution between the actual deformation change distribution and the predicted deformation response field, and write the adjustment parameters into the adjustment response mapping library for iterative update based on the deviation energy ratio.

9. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 7.