Digital twin construction method for metal stamping forming and computer system

Abstract

The invention provides a digital twin construction method for metal stamping forming and a computer system, and the method comprises the steps: obtaining a multi-source monitoring data set in a metal stamping forming process, carrying out the collaborative feature extraction processing, and generating a material structure feature set and a process dynamic feature set; a mixed feature space is constructed based on the material structure feature set and the process dynamic feature set, multi-dimensional feature matching processing is executed in the mixed feature space, a feature mapping relation network is generated, a dynamic twin model is constructed according to the feature mapping relation network, and real-time state deduction is conducted on the stamping forming process based on the dynamic twin model. And a forming quality prediction result and a process defect positioning result are generated, and a stamping process parameter adjustment scheme is generated and fed back to a stamping control system to trigger parameter optimization operation. According to the method, the early warning capability of process defects and the precision of process parameter adjustment under complex working conditions are improved, and meanwhile, the real-time performance and engineering feasibility of digital twin construction in the large-scale stamping forming process are guaranteed.

Description

Technical Field

[0001] The present invention relates to the fields of digital control and data processing and analysis. Specifically, it relates to a method for constructing a digital twin for metal stamping forming and a computer system. Background Art

[0002] With the deepening application of intelligent manufacturing technology, digital modeling and process optimization of the metal stamping forming process have become the core links of modern industrial manufacturing. Digital twin technology provides technical support for forming quality prediction and process defect analysis by constructing a virtual mapping model of the physical process. The current mainstream methods mainly rely on monitoring data (such as material thickness changes or equipment pressure curves) to construct static simulation models, and generate process adjustment suggestions through offline parameter matching. However, such methods are difficult to effectively integrate the coupling relationships of material deformation characteristics, equipment dynamic responses, and energy transfer processes, resulting in insufficient sensitivity of the forming quality prediction model to identify hidden defects such as transient stress concentration and abnormal energy dissipation; at the same time, traditional process optimization strategies mostly perform local adjustments on a single parameter (such as stamping speed or die pressure) based on empirical rules, lacking a global collaborative optimization mechanism driven by multi-source data, and prone to parameter conflicts or secondary process defects. In addition, the ability to process spatio-temporal alignment of large-scale heterogeneous data during the construction of the existing digital twin is insufficient, resulting in a lag in model real-time update behind the actual production rhythm, and unable to meet the dynamic optimization requirements under complex working conditions. The above defects seriously restrict the improvement of the quality control accuracy and process iteration efficiency of the stamping forming process, and there is an urgent need for a digital twin construction method that can realize multi-physical field coupling analysis, dynamic closed-loop optimization, and real-time state deduction. Summary of the Invention

[0003] The present invention provides a method for constructing a digital twin for metal stamping forming and a computer system.

[0004] In a first aspect, an embodiment of the present invention provides a method for constructing a digital twin for metal stamping forming, including: obtaining a multi-source monitoring data set during the metal stamping forming process, where the multi-source monitoring data set includes material physical parameter data, equipment operating state data, and dynamic response data of the forming process; performing collaborative feature extraction processing on the multi-source monitoring data set to generate a material structure feature set and a process dynamic feature set, where the material structure feature set characterizes the deformation characteristics and stress distribution characteristics of the stamping material, and the process dynamic feature set characterizes the action timing characteristics and energy transfer characteristics of the stamping equipment; constructing a hybrid feature space based on the material structure feature set and the process dynamic feature set, and performing multi-dimensional feature matching processing in the hybrid feature space to generate a feature mapping relationship network; constructing a dynamic twin model according to the feature mapping relationship network, and performing real-time state deduction on the stamping forming process based on the dynamic twin model to generate a forming quality prediction result and a process defect location result; generating a stamping process parameter adjustment plan according to the forming quality prediction result and the process defect location result, and feeding back the stamping process parameter adjustment plan to the stamping control system to trigger parameter optimization operations.

[0005] In a second aspect, an embodiment of the present invention provides a computer system, including: a memory in which a computer program is stored; a processor for loading the computer program to implement the method for constructing a digital twin for metal stamping forming as described above.

[0006] The digital twin construction method for metal stamping forming provided by the present invention constructs a hybrid feature space covering the material deformation mechanism, equipment action logic, and energy transfer process by integrating multi-source monitoring data of material physical parameters, equipment operating status, and dynamic response during the forming process, and realizes the full-dimensional state deduction of the stamping forming process based on the dynamic mapping relationship network. Through the collaborative feature extraction and spatio-temporal alignment processing of multi-source data, this method integrates the thickness change, stress distribution, and equipment action parameters analyzed in a traditional isolated manner into an associated feature set, significantly improving the coupled analysis ability of forming quality prediction and process defect location; using the real-time parameter correction mechanism of the dynamic twin model, it dynamically matches the simulation of the material deformation propagation path and the equipment operating status monitoring data, effectively identifying transient stress concentration and energy transfer anomalies that are difficult to capture by traditional offline detection; by establishing a direct association between the deformation prediction result and the process parameter adjustment through an optimized policy network, a multi-parameter collaborative optimization scheme is generated based on the dual constraints of material flow uniformity and equipment operating stability, overcoming the parameter conflict and secondary defect risk caused by single-dimensional process adjustment. Without relying on a large number of historical defect sample labels, this method significantly improves the early warning ability of process defects and the accuracy of process parameter adjustment under complex working conditions through the adaptive mapping and deduction mechanism of the feature space, while ensuring the real-time performance and engineering feasibility of the digital twin construction for large-scale stamping forming processes. Description of the Drawings

[0007] Figure 1 is a flowchart of a digital twin construction method for metal stamping forming provided by an embodiment of the present invention.

[0008] Figure 2 is a schematic diagram of the composition of a computer system provided by an embodiment of the present invention. Detailed Embodiment

[0009] Please refer to Figure 1 , Figure 1 , which is a flowchart of a digital twin construction method for metal stamping forming provided by an embodiment of the present invention. This method is executed by a computer system and includes the following steps: Step S100: Obtain a multi-source monitoring data set during the metal stamping forming process. The multi-source monitoring data set includes material physical parameter data, equipment operating status data, and dynamic response data during the forming process.

[0010] The material physical parameter data is the data that describes the physical properties of the stamping material itself, such as the density, elastic modulus, Poisson's ratio, etc. of the material, which reflects the basic properties of the material when it is stressed and deformed. The equipment operating state data is the data about the operating conditions of the stamping equipment, covering information such as the spindle speed of the equipment, the pressure of the hydraulic system, the position and movement state of the die, etc., which can reflect the working state and performance of the equipment during the stamping process. The dynamic response data of the forming process is the dynamic change data generated by the material and the equipment in response to the stamping action during the stamping process, such as the plastic deformation of the material, the vibration and energy transfer of the equipment, etc., which reflects the real-time changes in each link of the stamping process.

[0011] For the material physical parameter data, a thickness sensor can be used to collect the real-time thickness measurement data of the stamping material during the forming process, and a yield strength test device can be used to obtain the yield strength test data of the stamping material at different forming stages. For the equipment operating state data, the spindle speed data of the stamping equipment can be obtained through a spindle speed sensor, and the hydraulic system pressure data can be collected through a hydraulic system pressure sensor. For the dynamic response data of the forming process, a power sensor can be used to calculate the energy consumption rate of material plastic deformation, and an infrared thermal imager can be used to obtain the temperature distribution data of the contact surface between the die and the material.

[0012] Step S200: Perform collaborative feature extraction processing on the multi-source monitoring data set to generate a material structure feature set and a process dynamic feature set. The material structure feature set characterizes the deformation characteristics and stress distribution characteristics of the stamping material, and the process dynamic feature set characterizes the action timing characteristics and energy transfer characteristics of the stamping equipment.

[0013] The collaborative feature extraction processing comprehensively considers the associations and interactions between different types of data in the multi-source monitoring data set, and extracts key features that can reflect the material and process characteristics. The material structure feature set is a set of feature sets used to describe the deformation and stress distribution of the stamping material during the forming process. Among them, the deformation characteristics reflect the shape changes of the material under the action of stamping force, such as the stretching, compression, bending and other deformations of the material; the stress distribution characteristics reflect the magnitude and distribution of the internal stress of the material, such as the distribution pattern of residual stress. The process dynamic feature set is a set of feature sets that describe the action timing and energy transfer of the stamping equipment during the working process. The action timing characteristics involve the movement time sequence and speed changes of each component of the equipment, such as the change curve of the stamping speed, the delay time of pressure transmission, etc.; the energy transfer characteristics reflect the conversion and transfer of energy during the stamping process, such as the energy consumption rate of material plastic deformation, the kinetic energy loss rate of the equipment, etc.

[0014] As an implementation method, step S200 can specifically include the following steps S210~S240: Step S210: Perform structural deformation feature extraction processing on the material physical parameter data to obtain the structural deformation features in the material structure feature set; among them, the structural deformation features include the material thickness change rate, the yield strength change gradient, and the residual stress distribution pattern.

[0015] The structural deformation feature extraction processing aims to extract key features from these data that can reflect the structural deformation of the material. The material thickness change rate is the ratio of the material thickness change over time or position during stamping, reflecting the thinning or thickening of the material under the stamping force. The yield strength change gradient is the change rate of the material's yield strength at different positions or different forming stages, reflecting the change in the mechanical properties of the material during the deformation process. The residual stress distribution pattern is the distribution law of the internal residual stress in the material after stamping, including information such as the magnitude, direction, and distribution area of the residual stress.

[0016] As an implementation method, step S210 may specifically include the following steps S211 to S214: Step S211: Collect the real-time thickness measurement data of the stamping material during the forming process, and perform spatial interpolation processing on the real-time thickness measurement data to generate a thickness change rate distribution cloud map.

[0017] The real-time thickness measurement data are the material thickness data collected in real time by a thickness measurement device during the forming process of the stamping material. These data reflect the thickness information of the material at different times and different positions. The spatial interpolation processing can convert the discrete thickness measurement data into continuous thickness change information. Through this processing, the thickness change rate distribution of the material in the entire forming area can be obtained. The thickness change rate distribution cloud map is a visualization tool that graphically shows the spatial distribution of the material thickness change rate, which can intuitively reflect the thickness change of the material during stamping. Multiple thickness measurement devices can be used to collect the real-time thickness measurement data of the stamping material during the forming process, such as laser thickness sensors, ultrasonic thickness sensors, etc. For example, during the stamping process of a metal pipe, laser thickness sensors can be installed at different positions of the stamping die, and the thickness data of different positions of the pipe can be collected at regular intervals. Multiple interpolation algorithms can be used for the spatial interpolation processing of the real-time thickness measurement data, such as the Kriging interpolation algorithm, the bilinear interpolation algorithm, etc. Specifically, first sort the collected thickness data according to the spatial position, then calculate the distance and correlation between each data point and other data points, and construct a covariance matrix based on this information. Then, use the least squares method to solve the covariance matrix to obtain the weight coefficients of each unknown position. Finally, perform weighted summation on the known thickness data according to the weight coefficients to obtain the thickness prediction value of the unknown position, thereby generating a thickness change rate distribution cloud map.

[0018] Step S212: Obtain the yield strength test data of the stamping material at different forming stages, and perform gradient calculation and processing on the yield strength test data to generate a yield strength change gradient curve.

[0019] The yield strength test data are the yield strength values of the material at different forming stages obtained by performing mechanical property tests on the stamping material. These data reflect the stress magnitude when the material begins to undergo plastic deformation at different degrees of deformation. The yield strength change gradient curve is a visualization tool that graphically shows the change rate of the yield strength with respect to position or forming stage, and can intuitively reflect the change in the mechanical properties of the material during the stamping process.

[0020] Obtaining the yield strength test data of the stamping material at different forming stages can be achieved through mechanical property test methods such as tensile tests and compression tests. For example, during the stamping process of a metal sheet, specimens can be intercepted at different stamping stages for tensile tests to measure their yield strength. The gradient calculation and processing of the yield strength test data can use numerical differentiation methods such as the finite difference method. Specifically, assume that a series of discrete yield strength test data points are known, and each data point corresponds to a forming stage or position. First, calculate the difference in yield strength between two adjacent data points, and then divide this difference by the corresponding position or stage interval to obtain the average gradient value within this interval. By calculating for all adjacent data points, a series of gradient values can be obtained. Finally, sort these gradient values according to the corresponding position or stage and plot the yield strength change gradient curve.

[0021] Step S213: Based on a preset residual stress detection device, collect the surface stress distribution data of the material, and perform pattern recognition processing on the stress distribution data to generate a residual stress distribution pattern feature vector.

[0022] The preset residual stress detection device is a pre-set device for detecting the residual stress distribution on the material surface, such as an X-ray diffractometer, a blind hole method stress detector, etc. Pattern recognition processing is to analyze and process the collected stress distribution data to extract the representative features and patterns. The residual stress distribution pattern feature vector contains the key information of the residual stress distribution, such as the magnitude, direction, and distribution area of the stress, etc.

[0023] Pattern recognition processing of stress distribution data can adopt various algorithms, such as principal component analysis algorithm, clustering analysis algorithm, etc. Taking the principal component analysis algorithm as an example, the collected stress distribution data is standardized to have the same scale and range. Then, the covariance matrix of the data is calculated, and by solving the eigenvalues and eigenvectors of the covariance matrix, the principal components of the data are obtained. The first N principal components with larger eigenvalues are selected as the main features, and the original data is projected onto these principal components to obtain the feature vectors after dimensionality reduction. Finally, these feature vectors are combined to generate the feature vectors of the residual stress distribution pattern.

[0024] Step S214: Perform normalization fusion processing on the thickness change rate distribution cloud map, the yield strength change gradient curve, and the feature vectors of the residual stress distribution pattern to obtain the structural deformation features.

[0025] Normalization fusion processing is to uniformly process feature data of different types and scales to make them have the same dimension and range, and then fuse these features to obtain a comprehensive feature representation. The thickness change rate distribution cloud map, the yield strength change gradient curve, and the feature vectors of the residual stress distribution pattern respectively reflect the structural deformation of the material during the stamping process from different angles, but their numerical ranges and data formats may be different. Through normalization fusion processing, these features can be effectively integrated to obtain a comprehensive feature that can comprehensively and accurately reflect the structural deformation features of the material.

[0026] As an implementation manner, step S214 may specifically include the following steps S2141 to S2147: Step S2141: Perform data format conversion processing on the thickness change rate distribution cloud map, map the pixel point gray values to a continuous change rate numerical sequence, and generate the thickness change rate distribution cloud map with a unified format.

[0027] The thickness change rate distribution cloud map is presented in the form of an image, for example, where the gray value of each pixel point represents the thickness change rate information at that position. Data format conversion processing is to convert the gray values of these pixel points into a continuous change rate numerical sequence for subsequent processing and analysis. The continuous change rate numerical sequence is the thickness change rate data represented in numerical form, which can be more conveniently fused and compared with other feature data.

[0028] Data format conversion processing of the cloud map of thickness change rate distribution can be carried out by using the method of mapping function. First, determine the mapping relationship between the gray value and the change rate value. This mapping relationship can be determined through experiments or experience. For example, some positions with known thickness change rates can be selected in the cloud map, measure their corresponding gray values, and then establish a linear or non-linear mapping function between the gray value and the change rate value. Then, traverse each pixel point in the cloud map, and convert its gray value into the corresponding change rate value according to the mapping function. Finally, arrange these converted values in the order of the pixel point positions to generate a continuous change rate value sequence. For example, assume that the linear mapping function between the gray value and the change rate value obtained through experiments is: change rate value = a * gray value + b, where a and b are constants determined through experiments. For each pixel point in the cloud map, read its gray value, substitute it into the mapping function to calculate the corresponding change rate value, and then store these values in an array in sequence to obtain a continuous change rate value sequence.

[0029] Step S2142: Perform sampling point encryption processing on the yield strength change gradient curve, convert the discrete gradient values into a gradient continuous signal consistent with the time scale of the continuous change rate value sequence based on the stamping forming time axis, and generate a yield strength change gradient curve after time alignment.

[0030] The yield strength change gradient curve is obtained, for example, by calculating the gradient of discrete yield strength test data. Therefore, the gradient values on the curve are discrete. Sampling point encryption processing is to insert new points between the discrete gradient values to make the curve smoother and more continuous, so as to better reflect the change trend of the yield strength. The stamping forming time axis is an axis based on the time of the stamping process, which is used to determine the time position corresponding to each data point. Converting the discrete gradient values into a gradient continuous signal consistent with the time scale of the continuous change rate value sequence is to make the yield strength change gradient curve consistent with the cloud map of thickness change rate distribution in terms of time scale, facilitating subsequent feature fusion processing.

[0031] Sampling point encryption processing of the yield strength change gradient curve can be carried out by using the method of linear interpolation. First, determine the sampling interval that needs to be encrypted. For example, insert a new point at every predetermined time interval. Then, for each position where a new point needs to be inserted, calculate the gradient value of the new point according to its adjacent two discrete gradient values using the linear interpolation formula. The linear interpolation formula is: new point gradient value = (1 - t) * previous discrete gradient value + t * next discrete gradient value, where t is the relative position of the new point between the two adjacent discrete gradient values. In this way, new points are inserted between the discrete gradient values to make the curve smoother and more continuous.

[0032] To convert the discrete gradient values into a gradient continuous signal with a time scale consistent with that of the continuous change rate numerical sequence based on the stamping forming time axis, it is necessary to perform time alignment processing on the encrypted curve. First, determine the time scale of the continuous change rate numerical sequence, that is, the time interval corresponding to each data point. Then, according to the stamping forming time axis, adjust each data point in the encrypted yield strength change gradient curve to the same time position as the continuous change rate numerical sequence. If there is no corresponding gradient value at a certain time position, linear interpolation or other interpolation methods can be used for estimation.

[0033] Step S2143: Perform spatial coordinate system transformation processing on the residual stress distribution pattern eigenvector to align the dimension direction of the eigenvector with the spatial grid distribution of the thickness change rate distribution nephogram, and generate a spatially unified residual stress distribution pattern eigenvector.

[0034] The spatial coordinate system transformation processing is to transform the spatial coordinate system of the eigenvector so that its dimension direction is consistent with the spatial grid distribution of the thickness change rate distribution nephogram. The spatial grid distribution of the thickness change rate distribution nephogram is the spatial arrangement of pixel points in the nephogram, which defines a spatial coordinate system. By aligning the dimension direction of the residual stress distribution pattern eigenvector with the spatial grid distribution of the nephogram, these two features can be made consistent in space, facilitating subsequent feature fusion processing.

[0035] The spatial coordinate system transformation processing of the residual stress distribution pattern eigenvector can adopt transformation operations such as rotation and translation. First, determine the original spatial coordinate system of the eigenvector and the spatial grid coordinate system of the nephogram. Then, calculate the rotation angle and translation vector between the two coordinate systems. The rotation angle can be determined by calculating the included angle between the coordinate axes of the two coordinate systems, and the translation vector can be determined by calculating the displacement between the origins of the two coordinate systems. According to the calculated rotation angle and translation vector, perform rotation and translation transformations on the eigenvector. Specifically, for each component in the eigenvector, perform corresponding transformations according to the rotation matrix and translation vector to obtain the transformed eigenvector. Through these transformation operations, align the dimension direction of the residual stress distribution pattern eigenvector with the spatial grid distribution of the thickness change rate distribution nephogram, and generate a spatially unified residual stress distribution pattern eigenvector.

[0036] Step S2144: Perform spatial alignment processing on the thickness change rate distribution nephogram with unified format, the yield strength change gradient curve after time alignment, and the spatially unified residual stress distribution pattern eigenvector, and project each feature data onto the same spatial coordinate grid node based on the stamping material surface grid division rule to generate a multi-feature space superposition dataset.

[0037] Spatial alignment processing is to align different types of feature data spatially so that they have corresponding feature values at the same spatial position. The stamping material surface mesh division rule is a rule for dividing the stamping material surface into several mesh nodes, and these mesh nodes form a spatial coordinate system. Projecting each feature data to the same spatial coordinate mesh node means distributing the feature values in the thickness change rate distribution nephogram after format unification, the yield strength change gradient curve after time alignment, and the residual stress distribution pattern eigenvector after space unification to the corresponding mesh nodes according to the mesh division rule. The multi-feature space superposition data set is to superpose these projected feature data spatially to obtain a data set containing various feature information.

[0038] To perform spatial alignment processing on the thickness change rate distribution nephogram after format unification, the yield strength change gradient curve after time alignment, and the residual stress distribution pattern eigenvector after space unification, the following steps can be taken. First, according to the stamping material surface mesh division rule, determine the spatial coordinates of each mesh node. Then, for the thickness change rate distribution nephogram after format unification, project it to the corresponding mesh node according to the position of each pixel point in the nephogram and the corresponding change rate value. If a pixel point is located between multiple mesh nodes, the interpolation method can be used to distribute its change rate value to the adjacent mesh nodes. For the yield strength change gradient curve after time alignment, project its gradient value to the corresponding mesh node according to the time and spatial position corresponding to each data point in the curve. Similarly, if a data point is located between multiple mesh nodes, the interpolation method can be used for distribution. For the residual stress distribution pattern eigenvector after space unification, project its eigenvalue to the corresponding mesh node according to the spatial coordinates of the eigenvector.

[0039] After the projection of each feature data is completed, superpose the different feature values projected onto the same mesh node to generate a multi-feature space superposition data set.

[0040] Step S2145: Perform dimensionless processing on the thickness change rate value, yield strength gradient value, and residual stress eigenvalue in the multi-feature space superposition data set respectively, and use the dynamic range compression algorithm to adjust the numerical range of each eigenvalue to the same proportional interval to generate a standardized multi-feature data set.

[0041] The dimensionless processing is to remove the dimensions of different eigenvalue quantities to make them comparable. The dynamic range compression algorithm is an algorithm that compresses the numerical range of eigenvalue quantities into a smaller interval. Through this algorithm, the numerical ranges of each eigenvalue quantity can be adjusted to the same proportional interval, avoiding the dominant effect of some eigenvalue quantities with too large numerical ranges on other eigenvalue quantities. The standardized multi-feature data set is a data set obtained after dimensionless processing and dynamic range compression, in which each eigenvalue quantity has the same dimension and numerical range, facilitating subsequent analysis and processing. There are various methods for dimensionless processing, such as normalization methods, standardization methods, etc. The logarithmic compression algorithm can be used to adjust the numerical ranges of each eigenvalue quantity to the same proportional interval by the dynamic range compression algorithm. The logarithmic compression algorithm is an algorithm that performs a logarithmic transformation on the numerical range of eigenvalue quantities, which can compress a relatively large numerical range into a smaller interval. Specifically, for each normalized eigenvalue quantity, take its logarithm to obtain the compressed eigenvalue quantity. For example, assuming the normalized eigenvalue quantity is x, the compressed eigenvalue quantity is y = log(x + 1). Through the logarithmic compression algorithm, the numerical ranges of each eigenvalue quantity can be adjusted to the same proportional interval to generate a standardized multi-feature data set.

[0042] Step S2146: Based on the material deformation sensitivity weight distribution strategy in the stamping forming stage, perform weighted superposition processing on the thickness change rate data, yield strength gradient data, and residual stress data in the standardized multi-feature data set to generate a fused feature distribution map.

[0043] The material deformation sensitivity weight distribution strategy in the stamping forming stage assigns different weights to each feature according to the sensitivity of the material to thickness change, yield strength change, and residual stress in different stamping forming stages. During the stamping process, the mechanical properties and deformation characteristics of the material are different in different stages, so the sensitivities to different features are also different. By reasonably distributing weights, the features with greater influence on material deformation in different stages can be highlighted, improving the accuracy and effectiveness of feature fusion. The weighted superposition processing is to perform weighted summation on the thickness change rate data, yield strength gradient data, and residual stress data in the standardized multi-feature data set according to the assigned weights to obtain a comprehensive eigenvalue quantity. The fused feature distribution map is a visualization tool that graphically shows the spatial distribution of the comprehensive eigenvalue quantity after weighted superposition, which can intuitively reflect the comprehensive deformation situation of the material during the stamping process.

[0044] Specifically, the weight of each feature can be determined based on the mechanical properties and deformation characteristics of the material at different stamping forming stages. For example, in the initial stage of stamping, the thickness change of the material has a greater impact on the overall deformation, so a larger weight can be assigned to the thickness change rate data; in the later stage of stamping, the distribution of residual stress has a greater impact on the quality and performance of the material, so a larger weight can be assigned to the residual stress data. Then, for each grid node in the standardized multi-feature data set, multiply its corresponding thickness change rate data, yield strength gradient data, and residual stress data by the corresponding weights to obtain the weighted feature values. Finally, sum up the weighted feature values to obtain the comprehensive feature value of the grid node. Arrange the comprehensive feature values of each grid node according to their spatial positions to generate a fused feature distribution map.

[0045] Step S2147: Perform local feature enhancement on the fused feature distribution map, detect the fused feature values in the deformation mutation region and perform smooth transition correction to generate structural deformation features.

[0046] Local feature enhancement is to enhance the local regions in the fused feature distribution map to highlight the feature information in these regions. The deformation mutation region is the region where the deformation situation changes suddenly during the stamping process of the material, and these regions may correspond to defects such as cracks and wrinkles in the material. Detecting the fused feature values in the deformation mutation region and performing smooth transition correction is to eliminate the abnormal feature values in these mutation regions, make the fused feature distribution map smoother and more continuous, and at the same time retain the true deformation information of the material. The structural deformation feature is the feature obtained after local feature enhancement and smooth transition correction, which can more accurately reflect the structural deformation situation of the material during the stamping process.

[0047] Edge detection algorithms such as the Sobel operator and Canny operator can be used to perform local feature enhancement on the fused feature distribution map. These algorithms can detect the edges and mutation regions in the fused feature distribution map, enhance the feature values in these regions to make them more obvious. For example, use the Sobel operator to calculate the gradients of the fused feature distribution map in the horizontal and vertical directions, then mark the regions with larger gradient values as edges and mutation regions, and amplify the feature values in these regions.

[0048] Smoothing filter algorithms such as Gaussian filtering and median filtering can be used to detect the fused feature values in the deformation mutation region and perform smooth transition correction. These algorithms can smooth the feature values in the mutation region, eliminate the outliers, and make the feature values more continuous and stable.

[0049] Finally, use the fused feature distribution map after local feature enhancement and smooth transition correction as the structural deformation feature.

[0050] Step S220: Perform action timing decomposition processing on the device operation status data to obtain the device action characteristics in the process dynamic feature set; among them, the device action characteristics include the stamping speed change curve, the pressure transfer delay time, and the friction coefficient of the die contact surface.

[0051] The action timing decomposition processing decomposes the device operation status data in chronological order, extracts the time characteristics and sequence information of each device action. The device operation status data includes various parameters and status information during the operation of the stamping device, such as the spindle speed, the pressure of the hydraulic system, etc. Through the action timing decomposition processing, key information that can reflect the device action characteristics can be extracted from these data. The device action characteristics are part of the process dynamic feature set, which describes the action conditions of the stamping device during the working process, including the change of stamping speed, the pressure transfer delay time, and the friction coefficient of the die contact surface, etc. The stamping speed change curve is a curve that graphically shows the change of stamping speed over time, which reflects the speed change of the device during stamping. The pressure transfer delay time is the time required for the pressure to be transferred from the device to the die and the material after the device applies pressure, which reflects the efficiency and response speed of the device pressure transfer. The friction coefficient of the die contact surface is the ratio of the frictional force to the normal pressure on the contact surface between the die and the material, which affects the forming quality of the material and the service life of the die.

[0052] The time series analysis method, such as Fourier transform, wavelet transform, etc., can be used to perform action timing decomposition processing on the device operation status data.

[0053] As an implementation manner, step S220 may specifically include the following steps S221 to S224: Step S221: Obtain the spindle speed sensor data of the stamping device, and perform time series segmentation processing on the spindle speed sensor data to generate a stamping speed change curve; among them, the time series segmentation processing is synchronized and aligned based on the device action cycle and the material feeding cycle.

[0054] The spindle speed sensor data is the spindle speed data collected by the speed sensor installed on the spindle of the stamping device, and these data reflect the change of the rotational speed of the spindle of the stamping device. The time series segmentation processing is to segment the spindle speed sensor data in chronological order, and extract the time periods related to the device action cycle and the material feeding cycle. The device action cycle is the time required for the stamping device to complete a complete stamping action, and the material feeding cycle is the time interval for each material feeding during stamping. By synchronizing and aligning the time series segmentation processing based on the device action cycle and the material feeding cycle, the spindle speed data corresponding to each stamping action can be accurately extracted, so as to generate a stamping speed change curve.

[0055] Obtaining the spindle speed sensor data of a stamping device can be achieved by installing a speed sensor on the device's spindle, such as an optical encoder, a magnetoelectric encoder, etc. When performing time series segmentation on the spindle speed sensor data, first determine the time lengths of the device operation cycle and the material feeding cycle. These cycle time lengths can be determined by analyzing the working principle and operating parameters of the device, or through actual measurement. Then, according to the time lengths of the device operation cycle and the material feeding cycle, segment the spindle speed sensor data in chronological order, and take the data within each cycle as an independent time series. Finally, process each segmented time series to remove noise and outliers, obtaining the spindle speed data corresponding to each stamping action. Arrange the spindle speed data corresponding to each stamping action in chronological order and plot the stamping speed change curve. For example, with time as the horizontal axis and spindle speed as the vertical axis, plot the spindle speed values at each time point in the coordinate system, and connecting these points will obtain the stamping speed change curve.

[0056] Step S222: Collect the hydraulic system pressure data of the stamping device, and perform delay time calculation and processing on the hydraulic system pressure data to obtain a pressure transfer delay time series, where the delay time calculation and processing is determined based on the difference between the pressure peak arrival time and the die closing time.

[0057] The hydraulic system pressure data is the pressure data collected by a pressure sensor installed in the hydraulic system of the stamping device, and these data reflect the pressure changes in the hydraulic system during the stamping process. The delay time calculation and processing is the process of calculating the pressure transfer delay time based on the hydraulic system pressure data. The pressure transfer delay time is the time required for the pressure to be transferred from the device to the die and the material, reflecting the efficiency and response speed of the device's pressure transfer. The pressure peak arrival time is the time when the hydraulic system pressure reaches the maximum value, and the die closing time is the time when the die is fully closed. By calculating the difference between the pressure peak arrival time and the die closing time, the pressure transfer delay time can be obtained.

[0058] Collecting the hydraulic system pressure data of the stamping device can be achieved by installing pressure sensors at key parts of the hydraulic system, such as strain type pressure sensors, piezoelectric pressure sensors, etc. When performing delay time calculation and processing on the hydraulic system pressure data, first find the pressure peak arrival time from the hydraulic system pressure data. The pressure peak arrival time can be found by analyzing the pressure data and taking the time when the pressure value is the largest as the pressure peak arrival time. Then, determine the die closing time. The die closing time information can be obtained through a position sensor installed on the die or other monitoring devices. Finally, calculate the difference between the pressure peak arrival time and the die closing time to obtain the pressure transfer delay time. Arrange the pressure transfer delay times corresponding to each stamping action in chronological order to obtain the pressure transfer delay time series.

[0059] Step S223: Obtain real-time friction coefficient data through the friction sensor on the mold contact surface, and perform sliding window statistical processing on the real-time friction coefficient data to generate the dynamic change characteristics of the friction coefficient.

[0060] The sliding window statistical processing divides the real-time friction coefficient data according to the set time window, performs statistical analysis on the data within each window, and extracts the dynamic change characteristics of the friction coefficient. The dynamic change characteristics of the friction coefficient are used to describe the characteristics of the friction coefficient changing with time during the stamping process, and can reflect the friction characteristics and wear conditions of the contact surface between the mold and the material. When performing sliding window statistical processing on the real-time friction coefficient data, first determine the size and step length of the sliding window. The size of the sliding window determines the number of data points included in each statistical window, and the step length determines the distance the window moves each time. Then, divide the real-time friction coefficient data according to the sliding window, and perform statistical analysis on the data within each window, such as calculating statistical quantities such as the average value and standard deviation. Finally, arrange the statistical quantities of each window in chronological order to generate the dynamic change characteristics of the friction coefficient.

[0061] Step S224: Perform time series alignment and feature splicing processing on the stamping speed change curve, the pressure transfer delay time series, and the dynamic change characteristics of the friction coefficient to obtain the equipment action characteristics.

[0062] Specifically, step S224 may specifically include the following steps S2241 to S2247: Step S2241: Perform unified processing on the time axis of the stamping speed change curve, determine the main time axis based on the equipment action cycle, and perform synchronous compensation processing on the starting moment of the pressure transfer delay time series and the equipment startup moment of the main time axis to generate the pressure transfer delay time series after time synchronization.

[0063] The unified processing of the time axis is to unify the stamping speed change curve and the pressure transfer delay time series to the same time axis for subsequent time series alignment processing. The equipment action cycle is the time required for the stamping equipment to complete a complete stamping action. Determining the main time axis based on the equipment action cycle can enable different characteristic data to be compared and analyzed on the same time scale. The synchronous compensation processing is to adjust the starting moment of the pressure transfer delay time series and the equipment startup moment of the main time axis to make them consistent in time. The pressure transfer delay time series after time synchronization is the pressure transfer delay time series obtained after synchronous compensation processing, and it is consistent with the stamping speed change curve on the time axis.

[0064] When performing a unified processing of the stamping speed change curve on the time axis, first determine the time length of the equipment operation cycle, which can be determined by analyzing the working principle and operating parameters of the equipment or through actual measurement. Then, starting from the equipment startup moment, divide the main time axis according to the time length of the equipment operation cycle. For the stamping speed change curve, re-mark each data point on the curve according to the time scale of the main time axis to align it with the main time axis. When synchronously compensating the starting moment of the pressure transmission delay time series with the equipment startup moment of the main time axis, first determine the time difference between the starting moment of the pressure transmission delay time series and the equipment startup moment of the main time axis. Then, translate the pressure transmission delay time series according to this time difference so that its starting moment is aligned with the equipment startup moment of the main time axis.

[0065] Step S2242: Perform a sampling rate matching process on the dynamic change characteristics of the friction coefficient, and perform linear interpolation on the discrete friction coefficient data according to the time resolution of the main time axis to generate a continuous friction coefficient change curve.

[0066] The sampling rate matching process is to match the sampling rate of the dynamic change characteristics of the friction coefficient with the time resolution of the main time axis, so that the friction coefficient data is aligned with the main time axis in time. The time resolution of the main time axis is the time interval between two adjacent time points on the main time axis, which determines the time accuracy of the data. The discrete friction coefficient data is the friction coefficient data collected by the friction force sensor, and these data are discrete, that is, there are measurement values only at certain time points. The linear interpolation process is to insert new data points between the discrete friction coefficient data to make the data more continuous in time. The continuous friction coefficient change curve is the friction coefficient change curve obtained after the sampling rate matching process and the linear interpolation process, which is continuous in time and can more accurately reflect the change of the friction coefficient.

[0067] When performing a sampling rate matching process on the dynamic change characteristics of the friction coefficient, first determine the time resolution of the main time axis. According to the equipment operation cycle and the requirements of data acquisition, determine the time interval between two adjacent time points on the main time axis. Then, analyze the sampling rate of the discrete friction coefficient data, that is, the time interval between two adjacent measurement values. If the sampling rate of the discrete friction coefficient data is inconsistent with the time resolution of the main time axis, a sampling rate matching process is required.

[0068] Linear interpolation processing of discrete friction coefficient data according to the time resolution of the main timeline can adopt a linear interpolation formula. Suppose two adjacent discrete friction coefficient data points (t1, f1) and (t2, f2) are known, where t1 and t2 are time points, and f1 and f2 are the corresponding friction coefficient values. For any time point t on the main timeline between t1 and t2, the corresponding friction coefficient value f can be calculated using the linear interpolation formula: f = f1 + (f2 - f1) * (t - t1) / (t2 - t1). In this way, new data points are inserted between the discrete friction coefficient data to make the data more continuous in time. Arrange the friction coefficient values at each time point in chronological order and plot a continuous friction coefficient change curve.

[0069] Step S2243: Input the stamping speed change curve, the time-synchronized pressure transfer delay time series, and the continuous friction coefficient change curve into the multi-channel time window alignment module. Based on the sliding window mechanism, perform timestamp matching processing on the data points of the three to generate the time-aligned stamping speed data series, pressure delay data series, and friction coefficient data series.

[0070] The multi-channel time window alignment module is used to perform time alignment processing on multiple time series data. It can process data from multiple channels simultaneously and perform timestamp matching on data points based on the sliding window mechanism. The sliding window mechanism slides a fixed-size window over the time series data and processes and analyzes the data within the window. Timestamp matching processing is to match the data points in the stamping speed change curve, the time-synchronized pressure transfer delay time series, and the continuous friction coefficient change curve in time so that they have corresponding eigenvalue at the same time point. The time-aligned stamping speed data series, pressure delay data series, and friction coefficient data series are data series obtained after timestamp matching processing. They are aligned in time and are convenient for subsequent analysis and processing.

[0071] After inputting the stamping speed change curve, the time-synchronized pressure transfer delay time series, and the continuous friction coefficient change curve into the multi-channel time window alignment module, the module first determines the size and step of the sliding window. The size of the sliding window determines the number of data points contained in each window, and the step determines the distance the window moves each time. Then, the sliding window is sequentially moved to each time point, and timestamp matching processing is performed on the stamping speed data, pressure delay data, and friction coefficient data within the window. Specifically, for each time point within the window, find the data point closest to this time point from the stamping speed change curve, the time-synchronized pressure transfer delay time series, and the continuous friction coefficient change curve, and use the eigenvalue of these data points as the corresponding eigenvalue at this time point.

[0072] Arrange the matching results at each time point in chronological order to generate a stamping speed data sequence, a pressure delay data sequence, and a friction coefficient data sequence after time alignment. For example, for each time point, record the matched stamping speed value, pressure transfer delay time value, and friction coefficient value in sequence to form three time series.

[0073] Step S2244: Analyze the speed-pressure coupling relationship of the stamping speed data sequence after time alignment, extract the pressure delay characteristic values corresponding to the extreme points of the stamping speed in each time window, and generate a coupling feature vector.

[0074] Specifically, first divide the stamping speed data sequence and the pressure delay data sequence after time alignment according to time windows. The size of each time window can be adjusted according to the actual situation. Then, find the extreme points of the stamping speed within each time window. The maximum and minimum values of the stamping speed within the window can be compared to find the points corresponding to the maximum and minimum values as the extreme points. For each extreme point of the stamping speed, extract the corresponding pressure delay characteristic value from the pressure delay data sequence. Arrange the pressure delay characteristic values corresponding to the extreme points of the stamping speed in each time window in the order of the time windows to generate a coupling feature vector. Through the coupling feature vector, the coupling relationship between the stamping speed and the pressure transfer delay time can be analyzed, providing a basis for the monitoring and optimization of the operating state of the equipment.

[0075] Step S2245: Perform feature dimension expansion processing on the coupling feature vector and the friction coefficient data sequence within the corresponding time window, and splice the friction coefficient mean, fluctuation amplitude, and change trend parameters to the extended dimension of the coupling feature vector to generate a multi-dimensional joint feature matrix.

[0076] Feature dimension expansion processing is to add new feature dimensions on the basis of the coupling feature vector and fuse other relevant feature information into the coupling feature vector.

[0077] Specifically, the mean, fluctuation amplitude, and change trend parameters of the friction coefficient data within the corresponding time window can be calculated first. For the calculation of the mean, all the friction coefficient data within the time window can be added and then divided by the number of data to obtain the average value. For the calculation of the fluctuation amplitude, the maximum and minimum values of the friction coefficient data within the time window can be found, and then their difference can be calculated. For the calculation of the change trend parameter, methods such as linear regression can be used to fit the friction coefficient data within the time window, and the slope of the fitted straight line can be used as the change trend parameter.

[0078] Concatenate the mean friction coefficient, the fluctuation amplitude, and the change trend parameters to the extended dimension of the coupled feature vector. Specifically, add three new dimensions at the end of the coupled feature vector, which are used to store the mean friction coefficient, the fluctuation amplitude, and the change trend parameters respectively. Fill the calculated parameters into the corresponding dimensions in sequence to generate a multi-dimensional joint feature matrix.

[0079] Step S2246: Perform null value filling and outlier smoothing on the multi-dimensional joint feature matrix, and use the adjacent time window feature interpolation method to complete the missing data and perform local weighted average filtering to generate a standardized device action feature set.

[0080] Specifically, first check whether there are null values in the multi-dimensional joint feature matrix. If there are null values, use the adjacent time window feature interpolation method to complete them. Specifically, for the time window with null values, find its two adjacent time windows, and perform linear interpolation calculation based on the feature data of these two time windows to obtain the estimated value of the null value. For example, assume that there is a null value in a certain feature dimension of a certain time window, and the values of this feature dimension in its two adjacent time windows are a and b respectively, the time interval between this time window and the previous time window is t1, and the time interval between this time window and the next time window is t2, then the estimated value of this null value is: Estimated value = a+(b - a)*t1 / (t1 + t2).

[0081] Perform outlier smoothing on the multi-dimensional joint feature matrix using the local weighted average filtering method. For each data point, determine its neighborhood range. For example, select k data points before and after this data point as the neighborhood. Then, according to the distance between the data points in the neighborhood and this data point, assign a weight to each neighborhood data point, and the closer the point, the greater the weight. Finally, perform weighted average calculation on the data points in the neighborhood to obtain the smoothed data value. For example, assume that there are n data points in the neighborhood of a certain data point, and their values are x1, x2,..., xn respectively, and the corresponding weights are w1, w2,..., wn respectively, then the smoothed data value is: Smoothed value = (w1 * x1 + w2 * x2 +... + wn * xn) / (w1 + w2 +... + wn).

[0082] Use the multi-dimensional joint feature matrix after null value filling and outlier smoothing as the standardized device action feature set.

[0083] Step S2247: Perform segment marking processing on the standardized device action feature set based on the device action stage division rule, and perform dynamic weight assignment and connection relationship optimization on the feature subsets corresponding to the acceleration stage, the steady state stage, and the deceleration stage to generate device action features.

[0084] The device action phase division rule is a rule that divides the action process of the device into different phases according to the working principle and operating characteristics of the stamping device. For example, the action process of the stamping device can be divided into an acceleration phase, a steady-state phase, and a deceleration phase. The segmented marker processing is to mark each data point in the standardized device action feature set according to the device action phase division rule to determine the action phase it belongs to. The dynamic weight assignment is to assign different weights to the feature subsets corresponding to each phase according to the influence degree of different action phases on the device performance. The connection relationship optimization is to adjust and optimize the connection relationship between the feature subsets corresponding to different phases to make the association between the feature subsets more reasonable and effective. The device action feature is the feature obtained after segmented marker processing, dynamic weight assignment, and connection relationship optimization, which can more accurately reflect the working state of the stamping device in different action phases.

[0085] Specifically, first determine the basis for dividing the device action phases. The action phases can be divided according to the changes in parameters such as the spindle speed and pressure of the stamping device. For example, when the spindle speed gradually increases from 0, this phase can be marked as the acceleration phase; when the spindle speed stabilizes near a fixed value, this phase can be marked as the steady-state phase; when the spindle speed gradually decreases from the stable value, this phase can be marked as the deceleration phase. Then, for each data point in the standardized device action feature set, determine the action phase it belongs to according to the changes in the corresponding spindle speed, pressure, etc. parameters and mark it.

[0086] Perform dynamic weight assignment and connection relationship optimization on the feature subsets corresponding to the acceleration phase, steady-state phase, and deceleration phase. For dynamic weight assignment, different weights can be assigned to the feature subsets corresponding to each phase according to the influence degree of different action phases on the device performance. For example, in the acceleration phase, the acceleration and speed changes of the device have a greater impact on the device performance, so a larger weight can be assigned to the feature subset of the acceleration phase; in the steady-state phase, the stability and accuracy of the device have a greater impact on the product quality, so an appropriate weight can be assigned to the feature subset of the steady-state phase; in the deceleration phase, the braking performance and the accuracy of the stop position of the device have a greater impact on production safety and efficiency, so a corresponding weight can be assigned to the feature subset of the deceleration phase.

[0087] For the optimization of connection relationships, methods such as graph neural networks can be used to model and optimize the connection relationships between the feature subsets corresponding to different stages. Specifically, the feature subsets corresponding to the acceleration stage, steady-state stage, and deceleration stage are used as nodes in the graph, and edges between the nodes are constructed according to the chronological order and physical associations between different stages. Then, the graph-structured data is input into the graph neural network for training, and by optimizing the parameters of the network, the connection relationships between the nodes become more reasonable and effective. During the training process of the graph neural network, supervised learning or unsupervised learning methods can be used. If there is labeled training data, the supervised learning method can be used, taking the labeled information as the training target, and optimizing the network parameters by minimizing the error between the prediction result and the labeled information. If there is no labeled training data, the unsupervised learning method can be used, and the network parameters are optimized by learning the internal structure and feature representation of the data. For example, using unsupervised learning models such as autoencoders to encode and decode the graph-structured data, enabling the network to learn the latent feature representation of the data, thereby optimizing the connection relationships between the nodes.

[0088] After dynamic weight allocation and connection relationship optimization, the feature subsets corresponding to each stage are combined according to the optimized connection relationships to generate device action features. Specifically, for each feature dimension, the feature values corresponding to different stages are weighted and summed according to the allocated weights to obtain the final feature value of this feature dimension. The final feature values of all feature dimensions are combined together to form the device action features.

[0089] Step S230: Perform energy transfer analysis and processing on the dynamic response data of the forming process to obtain the energy transfer features in the process dynamic feature set; among them, the energy transfer features include the energy consumption rate of material plastic deformation, the kinetic energy loss rate of the device, and the heat dissipation distribution.

[0090] The dynamic response data of the forming process are various data that reflect the dynamic changes of materials and equipment during the metal stamping forming process, such as data on the deformation of materials, the vibration of equipment, and energy transfer.

[0091] As an implementation manner, step S230 may specifically include the following steps S231 to S234: Step S231: Calculate the energy consumption rate of material plastic deformation based on the power sensor data of the stamping equipment, where the energy consumption rate of plastic deformation is determined by the ratio of the energy absorbed by the material per unit time to the total input energy.

[0092] Specifically, first obtain the total input power \(P_{total}\) from the power sensor data. Then, calculate the energy absorbed by the material \(P_{absorbed}\) per unit time. To calculate the energy absorbed by the material, parameters such as the deformation amount and mechanical properties of the material can be measured and calculated in combination with the mechanical property model of the material. For example, for the stamping process of a metal sheet, devices such as strain gauges can be used to measure the strain change of the sheet per unit time, and according to the stress-strain curve and deformation volume of the sheet, the energy absorbed by the material per unit time can be calculated. Specifically, assuming that the stress-strain curve of the sheet can be represented by the function \(\sigma(\epsilon)\), where \(\sigma\) is the stress and \(\epsilon\) is the strain, and the deformation volume of the sheet is \(V\), then the energy absorbed by the material can be calculated by integration: \(P_{absorbed}=\int\sigma(\epsilon)d\epsilon*V\), where the integration interval is the strain change range of the sheet per unit time.

[0093] Finally, divide the energy absorbed by the material \(P_{absorbed}\) per unit time by the total input energy \(P_{total}\) to obtain the energy consumption rate of material plastic deformation: Energy consumption rate = \(P_{absorbed} / P_{total}\).

[0094] Step S232: Collect the kinetic energy loss data of the transmission system of the stamping equipment, and perform frequency domain decomposition processing on the kinetic energy loss data to obtain the spectral distribution characteristics of the equipment kinetic energy loss rate.

[0095] The kinetic energy loss data of the transmission system of the stamping equipment are the data related to the kinetic energy loss caused by various factors such as friction and vibration during the operation of the stamping equipment, such as the vibration acceleration and torque change data of the transmission system. The frequency domain decomposition processing is to convert the kinetic energy loss data in the time domain into frequency domain data and analyze the energy distribution of different frequency components. The spectral distribution characteristics of the equipment kinetic energy loss rate are obtained through frequency domain decomposition processing, which reflect the characteristics of the kinetic energy loss rate distribution of the equipment at different frequencies.

[0096] Frequency domain analysis methods such as Fourier transform can be used to perform frequency domain decomposition processing on the kinetic energy loss data. According to the results of the Fourier transform, calculate the spectral distribution characteristics of the equipment kinetic energy loss rate. Specifically, for each frequency component, calculate the proportion of its corresponding energy in the total energy to obtain the kinetic energy loss rate at that frequency. Arrange the kinetic energy loss rates at different frequencies in the order of frequency to obtain the spectral distribution characteristics of the equipment kinetic energy loss rate. For example, plot the spectral distribution characteristics as a spectrogram, with the abscissa representing the frequency and the ordinate representing the kinetic energy loss rate. Through the spectrogram, the kinetic energy loss of the equipment at different frequencies can be intuitively observed. By analyzing the spectral distribution characteristics of the equipment kinetic energy loss rate, the frequency components with relatively large energy loss in the transmission system can be found, providing a basis for the maintenance and optimization of the equipment.

[0097] Step S233: Obtain the temperature distribution data of the contact surface between the mold and the material through an infrared thermal imager, and perform heat conduction simulation processing on the temperature distribution data to generate a heat dissipation distribution map.

[0098] The thermal conduction simulation of the temperature distribution data can be carried out by numerical simulation methods such as finite element analysis. Specifically, it is a numerical calculation method that discretizes a continuous physical system into a finite number of elements, and can simulate the heat conduction process by solving the heat conduction equation. Specifically, first, a heat conduction model of the mold and the material is established, including defining the thermal conduction performance parameters of the material (such as thermal conductivity, specific heat capacity, etc.) and boundary conditions (such as the temperature of the contact surface between the mold and the material, the ambient temperature, etc.). Then, the temperature distribution data is input into the heat conduction model as the initial condition, and the finite element software is used to solve the heat conduction equation to obtain the temperature distribution at different times and different positions. According to the results of the heat conduction simulation, a heat energy dissipation distribution map is generated. The method of color mapping can be used to map different temperature values to different colors, and then the temperature distribution image, that is, the heat energy dissipation distribution map, is drawn in a two-dimensional or three-dimensional space.

[0099] Step S234: Perform space-time joint coding processing on the spectrum distribution characteristics of the plastic deformation energy consumption rate, the equipment kinetic energy loss rate, and the heat energy dissipation distribution map to obtain the energy transfer characteristics.

[0100] As an implementation manner, step S234 may specifically include the following steps S2341 to S2347: Step S2341: Perform time discretization processing on the plastic deformation energy consumption rate, and based on the time nodes of the stamping forming stage, divide the continuous energy consumption rate data into an equally spaced energy consumption rate time series to generate a discretized plastic deformation energy consumption rate set.

[0101] The energy consumption rate of plastic deformation is the ratio of the energy consumed by the material during plastic deformation to the total input energy in the metal stamping process, and it is a continuously varying quantity. Time discretization processing is to divide the continuous energy consumption rate data of plastic deformation according to the set time interval and convert it into a discrete time series. The time nodes in the stamping process are determined according to the characteristics and requirements of the stamping process. The entire stamping process is divided into several stages, and the start and end times of each stage are the time nodes. The discrete set of energy consumption rates of plastic deformation is obtained after time discretization processing and consists of a series of discrete energy consumption rate values of plastic deformation. During time discretization processing, first determine the time nodes in the stamping process, which can be divided according to the specific situation of the stamping process. For example, the time nodes can be determined based on factors such as the action cycle of the stamping equipment and the deformation process of the material. Then, obtain the continuous energy consumption rate data of plastic deformation. The energy input during the stamping process and the deformation of the material can be measured in real time through devices such as power sensors, and the continuous energy consumption rate of plastic deformation can be calculated. For each time node ti, extract the energy consumption rate value Pi corresponding to this time node from the continuous energy consumption rate data. Arrange these energy consumption rate values in chronological order to form a discrete set of energy consumption rates of plastic deformation {P0, P1, P2, ..., Pn-1} with a length of n.

[0102] Through time discretization processing, the continuous energy consumption rate data of plastic deformation is converted into a discrete time series, which is convenient for subsequent processing and analysis.

[0103] Step S2342: Perform spatial mapping processing on the spectral distribution characteristics of the equipment kinetic energy loss rate, map the main frequency band energy value of the spectral characteristics to the three-dimensional space coordinate points of the stamping equipment transmission system, and generate a kinetic energy loss spatial distribution map.

[0104] The spectral distribution characteristics of the equipment kinetic energy loss rate are obtained through frequency domain decomposition processing of the kinetic energy loss data of the stamping equipment transmission system, and they are the characteristics reflecting the distribution of the kinetic energy loss rate of the equipment at different frequencies. Spatial mapping processing is to correspond the main frequency band energy value in the spectral distribution characteristics to the three-dimensional space coordinate points of the stamping equipment transmission system and allocate the energy value to the corresponding spatial positions.

[0105] Specifically, first determine the three-dimensional space coordinate system of the stamping equipment transmission system. A three-dimensional rectangular coordinate system can be established with a certain fixed point of the transmission system as the origin to determine the position coordinates of each component in this coordinate system. Then, analyze the spectral distribution characteristics of the equipment kinetic energy loss rate and find the main frequency band energy value. The main frequency band is the frequency range where the energy in the spectrum is mainly concentrated, and the main frequency band energy value is the sum of the energy within this frequency range. For each main frequency band energy value, determine the transmission system component corresponding to this energy value according to its corresponding frequency component and the working principle of the transmission system. For example, if the frequency component corresponding to a certain main frequency band energy value is related to the rotation frequency of a certain gear, then the component corresponding to this energy value is this gear.

[0106] Map the main frequency band energy value to the corresponding three-dimensional space coordinate point. For each component, distribute its corresponding main frequency band energy value to the coordinate point of this component in three-dimensional space. The energy value can be visualized using methods such as color mapping or numerical annotation.

[0107] Integrate the energy value information of all coordinate points to generate a kinetic energy loss spatial distribution map. Three-dimensional drawing software or visualization tools can be used to plot the coordinate points and the corresponding energy value information to form a kinetic energy loss spatial distribution map. Through the kinetic energy loss spatial distribution map, the kinetic energy loss situation at different positions in the transmission system can be intuitively observed, providing a basis for the maintenance and optimization of the equipment. For example, it can be found in the figure that a certain component has a large kinetic energy loss, which may indicate that there is wear or a fault in this component and it needs to be repaired and replaced in a timely manner.

[0108] Step S2343: Perform time-axis resampling processing on the heat dissipation distribution map, adjust the time resolution of the temperature data according to the time nodes of the discretized plastic deformation energy consumption rate set, and generate a heat dissipation distribution map with unified time.

[0109] The heat dissipation distribution map is obtained through equipment such as an infrared thermal imager, and it is a distribution map reflecting the heat dissipation situation of the contact surface between the mold and the material and the surrounding space, containing temperature distribution information at different time points. The time-axis resampling processing is to resample the temperature data in the heat dissipation distribution map on the time axis, adjust the time resolution of the data to be consistent with the time nodes of the discretized plastic deformation energy consumption rate set. The time nodes of the discretized plastic deformation energy consumption rate set are the time points determined during the time discretization process of the plastic deformation energy consumption rate, and these time points divide the stamping forming process into several stages. The heat dissipation distribution map with unified time is obtained after time-axis resampling processing, and it is a heat dissipation distribution map where the time resolution of the temperature data is consistent with the time nodes of the discretized plastic deformation energy consumption rate set.

[0110] When performing time-axis resampling on the heat energy dissipation distribution map, first determine the time nodes of the discretized plastic deformation energy consumption rate set. Then, analyze the time resolution of the temperature data in the heat energy dissipation distribution map, that is, the time interval between two adjacent time points. If the time resolution of the temperature data is inconsistent with the time nodes of the discretized plastic deformation energy consumption rate set, resampling processing is required. For each time node ti of the discretized plastic deformation energy consumption rate set, find the two closest time points tj and tj+1 in the heat energy dissipation distribution map, where tj ≤ ti ≤ tj+1. According to the temperature data of these two time points, use the interpolation method to calculate the temperature distribution corresponding to the time node ti. Taking linear interpolation as an example, assuming that the temperature distributions corresponding to the time points tj and tj+1 are Tj and Tj+1 respectively, the temperature distribution Ti corresponding to the time node ti can be calculated by the following formula: Ti = Tj + (Ti - Tj) * (ti - tj) / (tj+1 - tj). Arrange the temperature distributions corresponding to the time nodes of each discretized plastic deformation energy consumption rate set in chronological order to generate the heat energy dissipation distribution map with unified time.

[0111] Step S2344: Input the discretized plastic deformation energy consumption rate set, the kinetic energy loss spatial distribution map, and the heat energy dissipation distribution map with unified time into the spatio-temporal encoding structure. Extract the spatial dimension features through the convolutional layer and capture the time dimension correlation through the recurrent layer to generate the spatio-temporal joint encoding feature tensor.

[0112] In the spatio-temporal encoding structure, the convolutional layer can be used to extract the spatial dimension features in the kinetic energy loss spatial distribution map and the heat energy dissipation distribution map with unified time. For example, use a two-dimensional convolutional layer to perform convolution operations on the kinetic energy loss spatial distribution map and the heat energy dissipation distribution map, and extract local features of different scales and directions through different convolutional kernels. It can be understood that the convolutional kernel size, stride, and padding method can be set according to actual needs. Each convolutional kernel slides on the input data to calculate the convolution result and obtain a series of feature maps. These feature maps reflect the feature information of the input data at different positions and scales.

[0113] In the spatio-temporal encoding structure, the recurrent layer can be used to process the discretized plastic deformation energy consumption rate set. Through the iterative operation of the recurrent unit, learn the time series pattern of the data. The recurrent layer is, for example, a long short-term memory network (LSTM), a gated recurrent unit (GRU), etc.

[0114] Fuse the spatial - dimensional features extracted by the convolutional layer and the temporal - dimensional correlation captured by the recurrent layer to generate a spatio - temporal joint - encoded feature tensor. Specifically, the feature map output by the convolutional layer and the hidden state output by the recurrent layer can be concatenated or weighted - summed to obtain a comprehensive feature representation. This feature representation contains the spatial - dimensional features and temporal - dimensional correlation of the input data, and can more comprehensively reflect the energy transfer situation during the stamping process.

[0115] Step S2345: Perform feature - channel fusion processing on the spatio - temporal joint - encoded feature tensor, and perform cross - channel weighted superposition on the feature maps of the plastic - deformation energy - consumption channel, kinetic - energy loss channel, and heat - energy dissipation channel to generate a preliminary fusion energy feature map.

[0116] Specifically, first determine the weight of each channel. The weight assignment can be adjusted according to the importance of different channels for energy transfer. For example, through experiments or experience, the weights of the plastic - deformation energy - consumption channel, kinetic - energy loss channel, and heat - energy dissipation channel are determined to be w1, w2, and w3 respectively, and w1 + w2+w3 = 1. These weights represent the importance of each channel in the fusion process. Then, perform weighted processing on the feature map of each channel. For the feature map F1 of the plastic - deformation energy - consumption channel, the feature map F2 of the kinetic - energy loss channel, and the feature map F3 of the heat - energy dissipation channel, multiply them by the corresponding weights w1, w2, and w3 respectively to obtain the weighted feature maps w1*F1, w2*F2, and w3*F3.

[0117] Perform cross - channel superposition on the weighted feature maps. Specifically, add w1*F1, w2*F2, and w3*F3 in the channel dimension to obtain the preliminary fusion energy feature map F. That is, F = w1*F1+w2*F2+w3*F3. Through this cross - channel weighted superposition method, the feature maps of different channels are fused to obtain a preliminary fusion energy feature map containing various energy - transfer information.

[0118] The preliminary fusion energy feature map can more comprehensively reflect the energy transfer situation during the stamping process, providing richer feature information for subsequent energy - distribution analysis and optimization. For example, in the preliminary fusion energy feature map, the energy - distribution situation in different regions and at different time points, as well as the mutual relationship between different energy - transfer channels, can be observed more clearly. By further analyzing the preliminary fusion energy feature map, the regions and stages with large energy losses can be found, providing a basis for optimizing the stamping process and improving energy - utilization efficiency.

[0119] Step S2346: Based on the energy - absorption characteristics of the stamping material, perform regional feature enhancement processing on the preliminary fusion energy feature map, detect the fusion features of high - energy - density regions and perform edge - smoothing correction to generate an optimized energy - distribution feature.

[0120] The energy absorption characteristics of stamping materials refer to the ability of materials to absorb and dissipate energy during the stamping process. Different stamping materials have different energy absorption characteristics. Regional feature enhancement processing is to perform targeted feature enhancement on different regions in the preliminary fused energy feature map according to the energy absorption characteristics of the stamping materials, highlighting the important features related to the energy absorption characteristics of the materials. High-energy density regions are regions with relatively high energy values in the preliminary fused energy feature map. These regions may correspond to the deformation concentration areas of the materials, the high-load areas of the equipment, etc. Edge smoothing correction processing is to smooth the edge parts of the high-energy density regions to avoid sudden changes in eigenvalue and make the energy distribution characteristics more continuous and natural. The optimized energy distribution characteristics are the energy distribution characteristics obtained after regional feature enhancement processing and edge smoothing correction, which can more accurately reflect the energy distribution related to the energy absorption characteristics of the materials during the stamping process.

[0121] The process of regional feature enhancement processing can be specifically as follows: First, analyze the energy absorption characteristics of the stamping materials. Through methods such as experiments or theoretical calculations, understand the energy absorption law of the materials under different deformation conditions, such as the stress-strain curve and energy absorption coefficient of the materials. According to the energy absorption characteristics of the materials, determine the feature regions that need to be enhanced. For example, if the material has strong energy absorption ability in certain deformation regions, then enhance the features corresponding to these regions. Detect the high-energy density regions in the preliminary fused energy feature map. Threshold segmentation and other methods can be used to mark the regions with energy values higher than a certain threshold as high-energy density regions. For the detected high-energy density regions, perform enhancement processing on their fused features. For example, by increasing the eigenvalue of these regions, enhancing the contrast of the features, etc., to highlight the feature information of these regions. Perform edge smoothing correction processing on the edge parts of the high-energy density regions. Smoothing filtering algorithms such as Gaussian filtering and median filtering can be used to smooth the eigenvalue of the edge region. These filtering algorithms can adjust the current eigenvalue according to the eigenvalues in the neighborhood, making the eigenvalue change in the edge region more gentle.

[0122] After regional feature enhancement processing and edge smoothing correction, the optimized energy distribution characteristics are generated. This optimized energy distribution characteristics can more accurately reflect the energy distribution related to the energy absorption characteristics of the materials during the stamping process, providing a more targeted basis for the optimization of stamping processes and the selection of materials.

[0123] Step S2347: Match and verify the optimized energy distribution characteristics with the energy conservation constraint conditions of equipment operation, eliminate the feature regions that violate the law of energy conservation, and complete the missing eigenvalues to generate energy transfer characteristics.

[0124] Specifically, first calculate the total input energy \(E_{total}\) based on the input power and operating time of the device. Then, extract energy information such as the energy consumption of material plastic deformation \(E_{plastic}\), the kinetic energy loss of the device \(E_{kinetic}\), and the heat dissipation \(E_{heat}\) from the optimized energy distribution characteristics. Add these energy values together to obtain the calculated total energy \(E_{calculated}=E_{plastic}+E_{kinetic}+E_{heat}\).

[0125] Compare the magnitudes of \(E_{total}\) and \(E_{calculated}\). If the difference between \(E_{calculated}\) and \(E_{total}\) exceeds a preset error range, it is considered that there are characteristic regions violating the law of conservation of energy. For these violating regions, further analysis and processing are required. It is possible to check whether there are errors in the data acquisition and processing processes, or whether there are energy loss factors that have not been considered. If it is determined that the problem lies in the characteristic regions themselves, these regions are removed from the optimized energy distribution characteristics.

[0126] Complete the missing characteristic values in the optimized energy distribution characteristics. Interpolation methods such as linear interpolation and spline interpolation can be used to estimate the missing characteristic values based on the characteristic values of the surrounding regions. For example, for a region with a missing characteristic value, find several adjacent regions with known characteristic values, and use the linear interpolation formula to calculate the estimated value of the missing characteristic value according to the characteristic values and spatial position relationships of these regions.

[0127] After matching verification, removing the violating regions, and completing the missing values, an energy transfer characteristic is generated. This energy transfer characteristic accurately reflects the energy transfer situation during the stamping process and provides a reliable basis for the optimization of the stamping process and the improvement of the device performance.

[0128] Step S240: Perform feature dimension unification processing on the structural deformation characteristics, device motion characteristics, and energy transfer characteristics, so that each feature maintains dimensional consistency in terms of time scale and space scale.

[0129] Feature dimension unification processing is to process these different types of features so that they have the same dimension and data format in terms of time scale and space scale for subsequent comprehensive analysis and processing. Specifically, first unify the time scale, analyze the time resolution and time range of the structural deformation characteristics, device motion characteristics, and energy transfer characteristics, and determine a unified time scale. For example, the device motion cycle can be selected as the unified time scale, and the respective feature data are resampled and aligned according to the device motion cycle. For the structural deformation characteristics, interpolate or sample its time series according to the device motion cycle to make its time resolution consistent with the device motion cycle. For the device motion characteristics and energy transfer characteristics, similarly adjust the time axis so that they are aligned with the structural deformation characteristics in terms of time.

[0130] Secondly, unify the spatial scale, analyze the spatial resolution and spatial range of the structural deformation characteristics, equipment motion characteristics, and energy transfer characteristics, and determine a unified spatial scale. A suitable spatial grid division rule can be determined according to the size and shape of the stamping material, as well as the structure and layout of the equipment. Project and interpolate each feature data according to the unified spatial grid division rule to make them have the same resolution and coordinate system in space. For example, for the cloud map of the thickness change rate distribution and the eigenvector of the residual stress distribution pattern in the structural deformation characteristics, project them onto the unified spatial grid nodes; for the stamping speed change curve and the pressure transfer delay time series in the equipment motion characteristics, divide and interpolate them according to the spatial position to align them with the unified spatial grid; for the spatial distribution map of kinetic energy loss and the heat dissipation distribution map in the energy transfer characteristics, also adjust and interpolate the spatial coordinates to make them consistent with the unified spatial scale.

[0131] After completing the unification of the time scale and the spatial scale, process the dimensions of each feature. Check the dimensions of the structural deformation characteristics, equipment motion characteristics, and energy transfer characteristics, and perform dimensionless processing on the characteristics with different dimensions to make them comparable. Methods such as normalization and standardization can be used for dimensionless processing. For example, for characteristics such as the thickness change rate and the yield strength change gradient in the structural deformation characteristics, use the normalization method to adjust their numerical range to the interval [0, 1]; for characteristics such as the stamping speed and the pressure transfer delay time in the equipment motion characteristics, use the standardization method to convert them into data with a mean of 0 and a standard deviation of 1.

[0132] Step S300: Construct a hybrid feature space based on the material structure feature set and the process dynamic feature set, and perform multi-dimensional feature matching processing in the hybrid feature space to generate a feature mapping relationship network.

[0133] The hybrid feature space is formed by integrating the features in the material structure feature set and the process dynamic feature set, and is a high-dimensional space containing various feature information. The multi-dimensional feature matching processing is to analyze and match the correlation and mapping relationship between different features in the hybrid feature space to find the internal connection between the features. The feature mapping relationship network is obtained through multi-dimensional feature matching processing and reflects the mapping relationship between different features. It can be used to describe the interaction and influence between the material structure features and the process dynamic features during the stamping process.

[0134] Specifically, first, the features in the material structure feature set and the process dynamic feature set are merged to form a feature vector containing all features. For example, the structure deformation feature, the equipment action feature, and the energy transfer feature are concatenated to obtain a high-dimensional feature vector. Then, according to the nature and characteristics of the features, the feature vector is dimensionally divided and organized to form the structure of the hybrid feature space. Multiple methods can be used to perform multi-dimensional feature matching processing in the hybrid feature space, such as correlation analysis, clustering analysis, neural networks, etc. Taking correlation analysis as an example, the correlation coefficient between different features in the hybrid feature space can be calculated to find out the feature pairs with strong correlation. The correlation coefficient can be calculated using methods such as Pearson correlation coefficient and Spearman correlation coefficient. For feature pairs with strong correlation, it can be considered that there is a mapping relationship between them.

[0135] According to the results of the multi-dimensional feature matching processing, a feature mapping relationship network is generated. Each feature in the hybrid feature space can be used as a node in the network, and the mapping relationship between the features is used as the edge in the network. The weight of the edge can be set according to the correlation strength between the features. The stronger the correlation, the greater the weight of the edge. Through the feature mapping relationship network, the mutual relationship between different features during the stamping process can be intuitively displayed, providing a basis for the subsequent construction of the dynamic twin model and the optimization of the stamping process.

[0136] As an implementation manner, step S300 may specifically include the following steps S310 to S340: Step S310: Perform feature cross-correlation processing on the structure deformation feature in the material structure feature set and the equipment action feature in the process dynamic feature set to generate a first-level feature correlation matrix.

[0137] Feature cross-correlation processing is to combine and analyze the structure deformation feature and the equipment action feature to find out the correlation and association relationship between them. The first-level feature correlation matrix is obtained through feature cross-correlation processing and reflects the association relationship between the structure deformation feature and the equipment action feature.

[0138] Specifically, first determine the feature vectors of the structural deformation characteristics and the equipment action characteristics. The structural deformation feature vector can be composed of features such as the thickness change rate, the yield strength change gradient, and the residual stress distribution pattern. The equipment action feature vector can be composed of features such as the stamping speed change curve, the pressure transfer delay time, and the die contact surface friction coefficient. Then, calculate the correlation between the structural deformation feature vector and the equipment action feature vector. Correlation analysis methods such as the Pearson correlation coefficient and the Spearman correlation coefficient can be used to calculate the correlation coefficients between each structural deformation feature and each equipment action feature. The calculated correlation coefficients are formed into a matrix, that is, the first-level feature correlation matrix. The rows of the matrix represent the structural deformation features, the columns represent the equipment action features, and each element in the matrix represents the correlation coefficient between the corresponding structural deformation feature and the equipment action feature.

[0139] Step S320: Perform spatio-temporal alignment processing on the residual stress distribution pattern feature vector in the material structure feature set and the energy transfer feature in the process dynamic feature set to generate a second-level feature correlation matrix.

[0140] The residual stress distribution pattern feature vector in the material structure feature set reflects the distribution law of the internal residual stress of the stamping material after forming. The energy transfer feature in the process dynamic feature set describes the energy conversion and transfer during the stamping process. The spatio-temporal alignment processing is to align the residual stress distribution pattern feature vector and the energy transfer feature in terms of time and space, so that they have corresponding feature values at the same time points and spatial positions. The second-level feature correlation matrix is obtained through spatio-temporal alignment processing and correlation analysis, and is a matrix reflecting the correlation relationship between the residual stress distribution pattern feature vector and the energy transfer feature.

[0141] Specifically, first unify the time scales of the residual stress distribution pattern feature vector and the energy transfer feature. According to the time nodes of the stamping process, the time series of the residual stress distribution pattern feature vector and the energy transfer feature can be resampled and aligned to make them have the same resolution and range in time. Unify the spatial scales of the residual stress distribution pattern feature vector and the energy transfer feature. According to the size and shape of the stamping material, as well as the structure and layout of the equipment, determine a unified spatial grid division rule. Project and interpolate the residual stress distribution pattern feature vector and the energy transfer feature according to the unified spatial grid division rule to make them have the same resolution and coordinate system in space. For example, project the residual stress distribution pattern in the residual stress distribution pattern feature vector onto the unified spatial grid nodes, and also adjust and interpolate the spatial coordinates of the kinetic energy loss spatial distribution map and the heat dissipation spatial distribution map in the energy transfer feature to align them with the unified spatial grid.

[0142] After completing the spatio-temporal alignment process, calculate the correlation between the eigenvectors of the residual stress distribution pattern and the energy transfer characteristics. For example, using the correlation analysis method, form a matrix with the calculated correlation coefficients, which is the second-level feature correlation matrix. The rows of the matrix represent the characteristics of the residual stress distribution pattern, the columns represent the energy transfer characteristics, and each element in the matrix represents the correlation coefficient between the corresponding residual stress distribution pattern feature and the energy transfer feature.

[0143] Through the second-level feature correlation matrix, the correlation between the eigenvectors of the residual stress distribution pattern and the energy transfer characteristics can be analyzed.

[0144] Step S330: Perform matrix fusion processing on the first-level feature correlation matrix and the second-level feature correlation matrix to obtain a hybrid feature correlation topological structure.

[0145] As an implementation, step S330 may specifically include the following steps S331 to S334: Step S331: Perform singular value decomposition on the first-level feature correlation matrix to obtain the first eigenvector set and the first singular value distribution.

[0146] The singular value decomposition of the first-level feature correlation matrix can adopt numerical calculation methods such as the Jacobi method and the QR decomposition method. Taking the Jacobi method as an example, the matrix is gradually transformed into a diagonal matrix through iteration to obtain the singular values and eigenvectors. First, initialize the matrices U and V as the identity matrix, and Σ as the first-level feature correlation matrix A. Then, in each iteration, select the non-diagonal element with the largest absolute value in the matrix Σ, and transform Σ through a rotation matrix to make this non-diagonal element become 0. At the same time, update the U and V matrices. Repeat this process until the matrix Σ is approximately a diagonal matrix. The column vectors of the finally obtained U matrix are the first eigenvector set, and the diagonal elements of the Σ matrix are the first singular value distribution. The first eigenvector set reflects the main feature directions of the first-level feature correlation matrix, and the first singular value distribution reflects the importance of these feature directions. The larger the singular value, the greater the role played by the corresponding eigenvector in the matrix.

[0147] Step S332: Perform principal component analysis on the second-level feature correlation matrix to obtain the second eigenvector set and the principal component contribution rate.

[0148] Specifically, first standardize the second-level feature correlation matrix so that its mean is 0 and its standard deviation is 1 to eliminate the influence of the dimension of different features and improve the accuracy of analysis. Then, calculate the covariance matrix of the standardized second-level feature correlation matrix. The covariance matrix reflects the correlation between different features. Next, solve the eigenvalues and eigenvectors of the covariance matrix. The eigenvalues represent the variance size explained by each principal component, and the eigenvectors represent the directions of the principal components. Methods such as eigenvalue decomposition or singular value decomposition can be used to solve the eigenvalues and eigenvectors. Select the eigenvectors with larger eigenvalues as the principal components to obtain the second eigenvector set. Through principal component analysis, the information of the second-level feature correlation matrix can be compressed and extracted to obtain a few important principal components.

[0149] Step S333: Perform orthogonal projection processing on the first eigenvector set and the second eigenvector set to generate a joint eigenvector space.

[0150] Specifically, first determine a common space dimension. A suitable space dimension can be selected according to the dimensions of the first eigenvector set and the second eigenvector set, so that both eigenvector sets can be projected into this space. Then, calculate the projection matrices of the first eigenvector set and the second eigenvector set on the common space dimension. The projection matrix can be obtained by calculating the inner products between the first eigenvector set and the second eigenvector set and the basis vectors of the common space dimension.

[0151] Merge the projected first eigenvector set and the second eigenvector set to generate a joint eigenvector space. In the joint eigenvector space, the vectors are orthogonal to each other, which makes the different feature directions independent of each other and facilitates subsequent weighted fusion processing.

[0152] Step S334: Perform weighted fusion processing on the joint eigenvector space based on the first singular value distribution and the principal component contribution rate to obtain a mixed feature correlation topological structure.

[0153] Specifically, first normalize the first singular value distribution and the principal component contribution rate so that the sum of their weights is 1. For the first singular value distribution , calculate the normalized singular value weight ; for the principal component contribution rate , calculate the normalized principal component weight .

[0154] Then, according to the normalized weights, assign weights to the projection vectors in the joint eigenvector space. The projection vectors in the joint eigenvector space include the projection vectors of the first eigenvector set and the projection vectors of the second eigenvector set For the projection vectors of the first set of eigenvectors , its weight is ; for the projection vectors of the second set of eigenvectors , its weight is .

[0155] Combine the weighted projection vectors. A hybrid feature correlation topology structure can be obtained by concatenating the weighted projection vectors column by column into a matrix.

[0156] Through this weighted fusion process, the importance degrees of the correlation relationships between the structural deformation features and the device motion features, and between the residual stress distribution pattern eigenvectors and the energy transfer features are comprehensively considered, enabling the hybrid feature correlation topology structure to more accurately reflect the complex correlation relationships between different features. This topology structure provides a key foundation for constructing the feature mapping relationship network subsequently, facilitating in-depth analysis of the interaction and influence between the material structure features and the process dynamic features during the stamping process.

[0157] Step S340: Construct a dynamic weight assignment network based on the hybrid feature correlation topology structure, and optimize the connection relationships of the feature nodes in the hybrid feature space through the dynamic weight assignment network to generate a feature mapping relationship network.

[0158] The hybrid feature correlation topology structure describes the complex correlation relationships between the material structure features and the process dynamic features, and it provides a basis for constructing the dynamic weight assignment network. The dynamic weight assignment network is a network structure that can dynamically adjust the connection weights between feature nodes according to the correlation degrees and actual situations between different features. Feature nodes are the various features in the hybrid feature space, and the optimization of connection relationships is achieved by adjusting the connection weights between feature nodes to make the mapping relationships between features more reasonable and effective. The feature mapping relationship network is obtained after the optimization of connection relationships and reflects the accurate mapping relationships between different features.

[0159] The construction of the dynamic weight assignment network based on the hybrid feature correlation topology structure can adopt the architecture of a neural network. First, take the hybrid feature correlation topology structure as the input layer of the network, and the nodes of the input layer correspond to the various feature nodes in the hybrid feature space. Then, add hidden layers, and the number of nodes in the hidden layers can be set according to the actual situation. In the hidden layers, each node is connected to the nodes of the input layer through connection weights, and these connection weights can be initialized to random values.

[0160] To achieve dynamic weight allocation, a weight adjustment mechanism is introduced into the network. Adaptive learning algorithms such as gradient descent algorithm, Adam algorithm, etc. can be used to continuously adjust the connection weights according to the output results of the network and the objective function. The objective function can be designed according to specific application requirements. For example, it can aim to minimize the mapping error between feature nodes, enabling the network to learn a more accurate feature mapping relationship.

[0161] Optimize the connection relationship of feature nodes in the hybrid feature space through the dynamic weight allocation network. During the training process of the network, continuously input the data of the hybrid feature association topology structure, calculate the output error of the network according to the objective function, and then use the adaptive learning algorithm to adjust the connection weights. As the training progresses, the connection weights will gradually converge to an optimal value, making the connection relationship between feature nodes more reasonable.

[0162] After the connection relationship optimization is completed, a feature mapping relationship network is generated, which can be represented as a graph structure. The nodes represent the feature nodes in the hybrid feature space, the edges represent the connection relationships between feature nodes, and the weights of the edges represent the connection strengths. Through the feature mapping relationship network, the mapping relationships between different features can be intuitively displayed, providing an important basis for the subsequent construction of the dynamic twin model and the real-time state deduction of the stamping process.

[0163] Step S400: Construct a dynamic twin model based on the feature mapping relationship network, and perform real-time state deduction on the stamping forming process based on the dynamic twin model to generate forming quality prediction results and process defect location results.

[0164] The feature mapping relationship network reflects the accurate mapping relationship between the material structure features and process dynamic features in the stamping process. Based on this network, a dynamic twin model can be constructed. The dynamic twin model is a virtual model that can real-time simulate the stamping forming process and keep in sync with the actual stamping process by continuously updating the model parameters. Real-time state deduction is to use the dynamic twin model to predict the state of the stamping process at future moments according to the current stamping state and feature mapping relationship. The forming quality prediction result is the prediction of the forming quality of the stamping product obtained through real-time state deduction, such as the thickness deviation of the material, the internal stress distribution, etc. The process defect location result is to find the location and cause of possible process defects in the stamping process through real-time state deduction.

[0165] As an implementation, step S400 can specifically include the following steps S410~S460: Step S410: Convert the node connection relationship in the feature mapping relationship network into a dynamic differential equation system, determine the coupling coefficient matrix of the equation system based on the feature transfer path between nodes, and generate an initial state transition model.

[0166] The nodes in the feature mapping relationship network represent various features in the stamping process, and the edges represent the connection relationships between the features. These connection relationships reflect the interactions and influences between the features. The dynamic differential equation system is a mathematical model used to describe the dynamic changes of a system, representing the relationship between the various variables in the system over time through a set of differential equations. The coupling coefficient matrix is the coefficient matrix between the various equations in the dynamic differential equation system, reflecting the coupling degree between different variables. The initial state transition model is obtained by converting the feature mapping relationship network into a dynamic differential equation system and is used to describe the initial model of the stamping process state transition.

[0167] Specifically, first define a variable for each node in the feature mapping relationship network. These variables represent various features in the stamping process, such as material thickness, stamping speed, pressure, etc. Then, based on the connection relationships between the nodes, establish differential equations between the variables. For each node, the rate of change of its variable can be expressed as a function of the variables of other nodes connected to it.

[0168] Determine the coupling coefficient matrix of the equation system based on the feature transfer paths between the nodes. The feature transfer path is the order and manner of feature transfer between the nodes, which determines the coupling degree between different variables. For each equation in the dynamic differential equation system, determine its coefficients with other equations according to the feature transfer path.

[0169] Combine the dynamic differential equation system and the coupling coefficient matrix to obtain the initial state transition model. The initial state transition model can be expressed as , where is the variable vector and A is the coupling coefficient matrix. The model describes the relationship between the various features in the stamping process over time and provides a basis for subsequent real-time state deduction.

[0170] Step S420: Collect the real-time pressure data and material deformation data of the current operating stage of the stamping equipment, and input the real-time pressure data and material deformation data into the observation variable interface in the initial state transition model to generate a set of model input parameters.

[0171] The observation variable interface is the interface in the initial state transition model for receiving external observation data. By inputting the real-time pressure data and material deformation data into this interface, the actual stamping state information can be introduced into the model. The set of model input parameters is obtained by inputting the real-time pressure data and material deformation data into the initial state transition model and is a set of parameters used to drive the model for state deduction.

[0172] Input the real-time pressure data and the material deformation data into the observation variable interface in the initial state transition model. The observation variable interface in the initial state transition model is, for example, a vector. Combine the real-time pressure data and the material deformation data in sequence into a vector and input it into the observation variable interface. For example, if the observation variable interface of the initial state transition model requires a three-dimensional vector, where the first element represents pressure, and the second and third elements represent the thickness change and strain of the material, then combine the collected real-time pressure data, material thickness change data, and strain data in this sequence into a three-dimensional vector and input it into the observation variable interface.

[0173] Step S430: Invoke the dynamic parameter correction algorithm to perform online adjustment on the coupling coefficient matrix of the initial state transition model, calculate the parameter correction amount based on the residual between the real-time data stream and the model prediction value, and generate an updated state transition model.

[0174] When invoking the dynamic parameter correction algorithm to perform online adjustment on the coupling coefficient matrix of the initial state transition model, first obtain the real-time data stream and the model prediction value. During the stamping process, continuously collect real-time data, and at the same time use the initial state transition model to make predictions based on the current input parameters to obtain the model prediction value. Calculate the residual between the real-time data stream and the model prediction value. Different methods can be used to calculate the residual, such as mean square error, absolute error, etc.

[0175] As an implementation manner, step S430 may specifically include the following steps S431 to S436: Step S431: Obtain the predicted value of the material thickness output by the initial state transition model and the real-time value of the thickness measured by the actual sensor, and calculate the residual sequence of the two within a continuous time window.

[0176] The predicted value of the material thickness output by the initial state transition model is the value predicted by the model for the thickness of the material at a future moment based on the current input parameters and the coupling coefficient matrix. The real-time value of the thickness measured by the actual sensor is the actual thickness value of the material collected in real time through a thickness sensor installed on the stamping die, such as a laser thickness sensor, an ultrasonic thickness sensor, etc. The residual sequence is the difference sequence between the predicted value of the material thickness and the actual measured value within a continuous time window, which reflects the degree of difference between the model prediction value and the actual value. When calculating the residual sequence, within a continuous time window, sequentially obtain the predicted value of the material thickness and the actual measured value according to the time sequence, and calculate the difference between them. By calculating the residual sequence, the difference between the model prediction value and the actual value can be intuitively observed. If the value of the residual sequence is large and fluctuates frequently, it indicates that the prediction accuracy of the model is low and the parameters of the model need to be adjusted; if the value of the residual sequence is small and stable, it indicates that the prediction accuracy of the model is high.

[0177] Step S432: Conduct autocorrelation analysis on the residual sequence to identify the periodic characteristics of the residual fluctuations and the positions of abnormal mutation points, and generate a residual characteristic analysis report.

[0178] Specifically, first calculate the autocorrelation function of the residual sequence. The autocorrelation function represents the correlation of the residual sequence at different time lags, and it can be calculated by the following formula: , where r i is the i-th element of the residual sequence, is the mean value of the residual sequence, and k is the time lag.

[0179] Then, based on the results of the autocorrelation function, identify the periodic characteristics of the residual fluctuations. Observe the autocorrelation function curve and find the time lags k at which the curve shows peaks. These time lags corresponding to the peaks are the periods of the residual fluctuations. At the same time, by observing and analyzing the residual sequence, identify the positions of abnormal mutation points. A threshold can be set, and when the absolute value of an element in the residual sequence exceeds this threshold, this point is considered an abnormal mutation point.

[0180] The residual characteristic analysis report contains information such as the periodic characteristics of the residual fluctuations and the positions of abnormal mutation points, and analyzes and interprets this information. For example, for the periodic characteristics, the reasons for their generation can be analyzed, whether they are related to factors such as the working cycle of the stamping equipment and the characteristics of the material; for the positions of abnormal mutation points, the relevant data of the stamping process at this time point can be further checked to find possible fault reasons.

[0181] Step S433: Adjust the update frequency of the coupling coefficient matrix based on the periodic characteristics in the residual characteristic analysis report, and trigger an emergency correction mechanism at the positions of abnormal mutation points.

[0182] The periodic characteristics in the residual characteristic analysis report reflect the periodic change law of the model prediction error. By adjusting the update frequency of the coupling coefficient matrix, the model can better adapt to this periodic change and improve the prediction accuracy of the model. The positions of abnormal mutation points correspond to abnormal situations in the stamping process. Triggering an emergency correction mechanism at these positions can timely adjust the model parameters to avoid excessive prediction errors caused by abnormal situations.

[0183] Specifically, first determine the update period of the coupling coefficient matrix according to the periodic characteristics. If the residual sequence has obvious periodic fluctuations and the period is T, then the update period of the coupling coefficient matrix can be set to an integer multiple of T, which can ensure that the model can update the parameters within each period to adapt to the periodic change of the prediction error.

[0184] Then, adjust the execution frequency of the dynamic parameter correction algorithm according to the update period. The dynamic parameter correction algorithm is used to calculate the correction amount of the coupling coefficient matrix, and its execution frequency is adjusted to be consistent with the update period, that is, the dynamic parameter correction algorithm is executed once in each update period.

[0185] Trigger the emergency correction mechanism at the position of the abnormal mutation point. When an abnormal mutation point in the residual sequence is detected, immediately pause the current stamping process and trigger the emergency correction mechanism. The emergency correction mechanism may include the following steps: First, conduct a detailed analysis of the real-time data before and after the abnormal mutation point to find the cause of the abnormal situation, such as equipment failure, material defect, etc. Then, according to the analysis results, manually adjust the parameters of the coupling coefficient matrix, or use a more aggressive dynamic parameter correction algorithm, such as increasing the correction step size, changing the correction strategy, etc., to quickly adjust the model parameters and reduce the prediction error.

[0186] Step S434: Use the recursive least squares method to iteratively optimize the dynamic weight parameters in the coupling coefficient matrix, and calculate the parameter adjustment step size according to the current residual value and the historical correction amount in each iteration.

[0187] The recursive least squares method is an algorithm for online estimating model parameters. By continuously updating the estimated parameters, the sum of the squares of the residuals is minimized. The dynamic weight parameters in the coupling coefficient matrix are the values of each element in the matrix, and these parameters determine the coupling degree between different variables in the model. Iterative optimization is to gradually adjust the values of the dynamic weight parameters through multiple iterations, so that the prediction error of the model is continuously reduced. The parameter adjustment step size is the adjustment amplitude of the dynamic weight parameters in each iteration, and it determines the speed and stability of parameter update.

[0188] Specifically, first initialize the dynamic weight parameters and related matrices. Initialize the dynamic weight parameters in the coupling coefficient matrix to an initial value, and at the same time initialize the matrices required for the recursive least squares method, such as the gain matrix k, the covariance matrix P, etc.

[0189] In each iteration, calculate the parameter adjustment step size according to the current residual value and the historical correction amount. The current residual value is the difference between the predicted value and the actual measured value of the material thickness at the current time point. The historical correction amount is the adjustment amount of the dynamic weight parameters in the previous iteration. The following formula can be used to calculate the parameter adjustment step size: , where is the parameter adjustment step size, K is the gain matrix, and r is the current residual value. The gain matrix K can be updated according to the covariance matrix P and the input data, and the specific update formula is: , where is the input data vector, is the forgetting factor, which is used to control the weight of historical data.

[0190] Update the dynamic weight parameter by adjusting the step size according to the parameter. Add the current dynamic weight parameter to the parameter adjustment step size to obtain the updated dynamic weight parameter. The covariance matrix P reflects the uncertainty of parameter estimation and needs to be updated according to the input data and the parameter adjustment step size in each iteration. The update formula is: , where I is the identity matrix.

[0191] Repeat the above steps until the iteration termination condition is met. The iteration termination condition can be set according to the actual situation, such as the sum of the squares of the residuals being less than a certain threshold, the number of iterations reaching the maximum limit, etc.

[0192] Step S435: Substitute the optimized dynamic weight parameter into the dynamic differential equations to re-solve the material deformation propagation equation and generate a corrected state transition trajectory.

[0193] Specifically, first update the optimized dynamic weight parameter to the coupling coefficient matrix. The elements in the coupling coefficient matrix are the dynamic weight parameters. Replace the original parameter values with the optimized parameter values to obtain the updated coupling coefficient matrix A'.

[0194] Substitute the updated coupling coefficient matrix A' into the dynamic differential equations , where x is a vector representing various state variables in the stamping process, such as material thickness, strain, pressure, etc. The material deformation propagation equation is part of the dynamic differential equations and can be extracted from the equations.

[0195] Select a suitable numerical solution method to solve the material deformation propagation equation, such as the Euler method, the Runge-Kutta method, etc. Taking the Euler method as an example, for the material deformation propagation equation , where x is the material deformation variable and t is the time, the iteration formula of the Euler method is: , where x n is the material deformation value at the nth time step, h is the time step size, and t n is the time at the nth time step.

[0196] Starting from the initial state, calculate according to the iteration formula of the numerical solution method to obtain the material deformation values at each time step. Arrange these values in chronological order to obtain the corrected state transition trajectory. The corrected state transition trajectory can more accurately reflect the actual state change of the material in the stamping process.

[0197] Step S436: Verify the matching degree between the corrected state transition trajectory and the actual measurement data, and terminate the correction process when the residual decrease rate of continuous S iterations is less than the preset convergence threshold to generate an updated state transition model, S≥3.

[0198] Verifying the matching degree between the corrected state transition trajectory and the actual measurement data is to evaluate whether the model can more accurately reflect the actual state of the stamping process after parameter correction. The residual is the difference between the predicted value in the corrected state transition trajectory and the actual measurement data, and the residual reduction rate is the reduction ratio of the residuals in two adjacent iterations. The preset convergence threshold is a preset value used to judge whether the model converges. When the residual reduction rate of three or more consecutive iterations is less than the preset convergence threshold, it indicates that the parameters of the model have been basically stable and the correction process can be terminated. The updated state transition model is obtained after parameter correction and verification, and can more accurately reflect the state transition of the stamping process.

[0199] Specifically, first obtain the actual measurement data. During the stamping process, the state data of the material, such as thickness, strain, pressure, etc., are collected in real time through various sensors, and these data serve as the actual measurement data. Compare the predicted value in the corrected state transition trajectory with the actual measurement data to calculate the residual. Different methods can be used to calculate the residual, such as mean square error, absolute error, etc. After each iteration, calculate the residual reduction rate and judge whether the residual reduction rate of three consecutive iterations is less than the preset convergence threshold. If the residual reduction rate of three consecutive iterations is less than the preset convergence threshold, it is considered that the model has converged and the correction process is terminated.

[0200] Step S440: Simulate the deformation propagation path of the stamping material in the next forming stage through the updated state transition model, extract the predicted surface of the material thickness change and the internal stress diffusion trajectory, and generate a deformation process simulation data set.

[0201] Simulating the deformation propagation path of the stamping material in the next forming stage is to use the updated state transition model to predict the deformation of the material in the future forming stage based on the current material state and process parameters. Specifically, first determine the current material state and process parameters. The current material state includes information such as the thickness, strain, and stress of the material, and the process parameters include information such as the stamping speed, pressure, and die shape. Use this information as the input of the updated state transition model. According to the structure and parameters of the updated state transition model, solve the model. Numerical solution methods, such as the Runge-Kutta method and the finite difference method, can be used to solve the state change of the model in the future forming stage. During the solution process, record the thickness change and internal stress distribution of the material at different positions and times.

[0202] Based on the recorded data, a prediction surface for material thickness variation and a trajectory of internal stress diffusion are generated. For the prediction surface of material thickness variation, the thickness values of the material at different positions and times can be used as the height of the surface to draw a three-dimensional surface plot. For the trajectory of internal stress diffusion, the magnitude and direction of the internal stress can be used as vectors to draw a trajectory plot of stress diffusion in three-dimensional space. Information such as the prediction surface of material thickness variation and the trajectory of internal stress diffusion is integrated together to generate a simulation dataset for the deformation process.

[0203] Step S450: Compare the maximum thickness deviation value in the simulation dataset for the deformation process with a preset forming tolerance threshold. If the deviation value exceeds the threshold, mark it as a deformation non-compliance area and generate a forming quality prediction result.

[0204] Specifically, first extract the material thickness data from the simulation dataset for the deformation process. The material thickness data can be stored in the form of a three-dimensional array or matrix, where each element represents the thickness value of the material at a certain position and time. Calculate the deviation value between the material thickness at each position and the target thickness. The target thickness is the thickness value that the material should reach according to the stamping process design requirements. Traverse the deviation values at all positions to find the maximum value. Compare the maximum thickness deviation value with the preset forming tolerance threshold. If the maximum thickness deviation value exceeds the threshold, it is considered that there is a deformation non-compliance area. In the three-dimensional space of the material, find the positions where the thickness deviation value exceeds the threshold and mark these positions as deformation non-compliance areas. Different colors or symbols can be used to mark the non-compliance areas for intuitive display. The forming quality prediction result can be presented in the form of a report or an image. The report should include information such as whether there is a deformation non-compliance area, the position and scope of the non-compliance area, etc. If there is a deformation non-compliance area, it indicates that there may be problems with the forming quality of the stamping product.

[0205] Step S460: Identify the stress concentration area based on the curvature change characteristics of the internal stress diffusion trajectory, and perform a similarity matching process between the stress concentration area and the crack initiation modes in the historical defect database to generate a process defect location result.

[0206] Specifically, the curvature of the internal stress diffusion trajectory can be calculated first, and the stress concentration area can be identified according to the curvature value. A larger curvature value indicates a higher degree of stress concentration. A curvature threshold can be set. When the curvature value of a certain point on the trajectory exceeds this threshold, the area where this point is located is marked as the stress concentration area. The stress concentration area is subjected to a similarity matching process with the crack initiation modes in the historical defect database. Feature extraction and matching algorithms can be used, such as stress concentration area descriptors (such as the magnitude of the maximum principal stress gradient, the dispersion of stress directions, the radius of curvature, etc.), machine learning algorithms, etc. The features of the stress concentration area are compared with the features of the crack initiation modes in the historical defect database, and the similarity score between the two is calculated (such as using the dynamic time warping (DTW) algorithm to calculate the trajectory similarity), and the mode with the highest similarity score is found. According to the similarity matching result, the process defect location result is generated. The process defect location result should include information such as the possible types of process defects, the location and scope of the defects, etc. For example, if the crack initiation mode with the highest similarity is an edge crack, it can be judged that an edge crack may occur in the stress concentration area, and the specific location and scope where the crack may occur are determined. Through the process defect location result, the process defects that may occur during the stamping process can be detected in time, and corresponding measures can be taken for prevention and repair to improve the quality of the stamping products.

[0207] Step S500: Generate a stamping process parameter adjustment plan based on the forming quality prediction result and the process defect location result, and feedback the stamping process parameter adjustment plan to the stamping control system to trigger parameter optimization operations.

[0208] The forming quality prediction result contains evaluation information on the forming quality of the stamping product, such as whether there are areas with excessive deformation and the location and scope of the excessive deformation areas, etc. The process defect location result clarifies the location and type of process defects that may occur during the stamping process.

[0209] As an implementation method, step S500 may specifically include the following steps S510 to S540: Step S510: Determine the stamping speed adjustment coefficient and the pressure compensation amount according to the information on the areas with excessive deformation in the forming quality prediction result.

[0210] The information on the areas with excessive deformation in the forming quality prediction result includes information such as the location, scope, and thickness deviation value of the excessive deformation areas. The stamping speed adjustment coefficient is a coefficient used to adjust the stamping speed of the stamping equipment, and the magnitude of the stamping speed can be adjusted by changing this coefficient. The pressure compensation amount is the additional pressure that needs to be applied to the stamping equipment to compensate for the possible pressure change due to excessive deformation.

[0211] Specifically, first analyze the thickness deviation values and distribution in the area where the deformation exceeds the standard. If the thickness deviation values are large and the distribution is relatively concentrated, it indicates that the deformation of the material during the stamping process is uneven, which may be caused by too fast stamping speed or unreasonable pressure distribution.

[0212] According to the analysis results, determine the stamping speed adjustment coefficient. If the deformation exceeding the standard is caused by too fast stamping speed, the stamping speed can be appropriately reduced. The stamping speed adjustment coefficient can be determined according to the magnitude of the thickness deviation value and the range of the area where the standard is exceeded.

[0213] The pressure compensation amount can be calculated according to the material properties and thickness deviation values in the area where the deformation exceeds the standard. The larger the thickness deviation value, the greater the pressure that needs to be compensated. A relationship model between the thickness deviation value and the pressure compensation amount can be established through experiments or simulations, and the pressure compensation amount can be calculated according to the model. Record the determined stamping speed adjustment coefficient and pressure compensation amount as part of the stamping process parameter adjustment plan. By adjusting the stamping speed and pressure compensation amount, the deformation uniformity of the material can be improved, the occurrence of areas where the deformation exceeds the standard can be reduced, and the quality of the stamping products can be improved.

[0214] Step S520: Based on the coordinates of the stress concentration area in the process defect location result, generate an optimized die contact surface plan and a corrected vector for the material feeding path.

[0215] Specifically, first determine the location and characteristics of the stress concentration area, and determine the optimization method for the die contact surface. If the stress concentration area is located at a certain edge or corner of the die, the die surface in this area can be polished, buffed, or filleted to improve the stress distribution. If the stress concentration area is caused by unreasonable die shape or structure, the die can be redesigned or modified.

[0216] The optimized die contact surface plan includes information such as specific optimization measures, optimized parts, and expected effects after optimization. For example, if it is decided to polish a certain edge of the die, the plan should describe the polishing range, the required roughness of the polishing, and the expected improvement in stress distribution after polishing. According to the location of the stress concentration area, determine the corrected vector for the material feeding path. If the stress concentration area is located at a certain position of the material, the feeding direction and distance of the material can be adjusted to avoid this area. The corrected vector for the material feeding path can be represented as a two-dimensional or three-dimensional vector, where the direction of the vector represents the direction in which the material needs to be adjusted, and the length of the vector represents the distance by which the material needs to be adjusted.

[0217] Step S530: Perform multi-objective optimization on the stamping speed adjustment coefficient, pressure compensation amount, optimized die contact surface plan, and corrected vector for the material feeding path to generate a candidate parameter adjustment set.

[0218] Specifically, first determine the optimization objectives, which may include improving the quality of stamping products, reducing production costs, and increasing production efficiency. For example, improving the quality of stamping products can be achieved by reducing the occurrence of areas with excessive deformation and process defects; reducing production costs can be achieved by reducing die wear and material waste; increasing production efficiency can be achieved by shortening the stamping cycle.

[0219] Then, assign weights to each optimization objective. The weight represents the importance of each objective in the optimization process and can be adjusted according to the actual situation. For example, if more emphasis is placed on product quality, a larger weight can be assigned to the objective of improving product quality.

[0220] Next, establish an optimization model. The optimization model can adopt methods such as mathematical programming models, genetic algorithms, and simulated annealing algorithms. Taking the mathematical programming model as an example, the stamping speed adjustment coefficient, pressure compensation amount, die contact surface optimization scheme, and material feed path correction vector can be used as decision variables, and the optimization objective can be used as the objective function. At the same time, various constraint conditions are considered, such as the performance limitations of the equipment and the feasibility of the process.

[0221] Use the optimization algorithm to solve the optimization model. According to the selected optimization algorithm, solve the optimization model to obtain a set of optimal parameter combinations. During the solution process, the algorithm will continuously search and compare different parameter combinations to find a solution that achieves the optimal balance among multiple objectives. The obtained optimal parameter combination and some parameter combinations similar to it are used as the candidate parameter adjustment set. Each solution in the candidate parameter adjustment set achieves a balance among different objectives and can be selected and applied according to the actual situation.

[0222] Step S540: Perform a feasibility screening process on the candidate parameter adjustment set through a preset process stability evaluation model to obtain a stamping process parameter adjustment plan.

[0223] The process stability evaluation model can be a model established based on machine learning algorithms, such as neural network models and support vector machine models, or a model established based on physical models, such as finite element models. According to the output result of the process stability evaluation model, judge the feasibility of each solution. The output result of the model can be an evaluation index, such as a process stability score or a defect occurrence rate. A feasibility threshold can be set. When the evaluation index of a solution exceeds this threshold, the solution is considered feasible. Screen out the solutions in the candidate parameter adjustment set whose evaluation indexes exceed the feasibility threshold to form a set of feasible solutions. Select the optimal solution from the set of feasible solutions as the stamping process parameter adjustment plan. According to the actual situation, the solution with the optimal evaluation index can be selected, or the most suitable solution can be selected by comprehensively considering other factors, such as cost and efficiency.

[0224] It can be understood that in the above introductions of the embodiments of the present invention, various common well-known algorithms involved, such as Euclidean distance, cosine distance, Kriging interpolation algorithm, genetic algorithm, simulated annealing algorithm, etc., can obtain their calculation formulas or methods from relevant content in the prior art. To save space, they will not be elaborated too much in the embodiments of the present invention. In addition, those skilled in the art can make detailed supplements according to the common general knowledge in the art when implementing the solution of the present invention. For example, according to the general knowledge in the art, normalization can be used to eliminate the dimensional conflict before feature fusion, interpolation can be used to eliminate the dimensional difference, the threshold can be reasonably set in combination with historical data or actual requirements, and the model can be trained based on the general model training method, etc. The present invention will no longer give redundant introductions to the overly detailed implementation process.

[0225] Please refer to Figure 2 , Figure 2 FIG. is a schematic structural diagram of a computer system provided by an embodiment of the present invention. The computer system at least includes a processor 101, a communication interface 102, and a memory 103. Among them, the processor 101, the communication interface 102, and the memory 103 can be connected through a bus or other means. Among them, the processor 101 (or the Central Processing Unit (CPU)) is the computing core and control core of the computer system, which can parse various instructions in the computer system and process various data in the computer system. The communication interface 102 can optionally include a standard wired interface, a wireless interface (such as WI-FI, mobile communication interface, etc.), and can be controlled by the processor 101 to be used for sending and receiving data; the communication interface 102 can also be used for the transmission and interaction of internal data in the computer system. The memory 103 (Memory) is a memory device in the computer system, used to store programs and data. It can be understood that the memory 103 here can include both the built-in memory of the computer system and, of course, the extended memory supported by the computer system. The memory 103 provides a storage space, and the operating system of the computer system is stored in this storage space, which is not limited in the present invention.

[0226] In one embodiment, the processor 101 executes the method for constructing a digital twin for metal stamping forming provided above in the embodiments of the present invention by running the computer program in the memory 103.

Claims

1. A method for constructing a digital twin for metal stamping forming, characterized in that, Including: Obtain a multi-source monitoring data set during the metal stamping forming process, where the multi-source monitoring data set includes material physical parameter data, equipment operation status data, and forming process dynamic response data; Perform collaborative feature extraction processing on the multi-source monitoring data set to generate a material structure feature set and a process dynamic feature set. The material structure feature set characterizes the deformation characteristics and stress distribution characteristics of the stamping material, and the process dynamic feature set characterizes the action timing characteristics and energy transfer characteristics of the stamping equipment; Construct a hybrid feature space based on the material structure feature set and the process dynamic feature set, and perform multi-dimensional feature matching processing in the hybrid feature space to generate a feature mapping relationship network; Construct a dynamic twin model according to the feature mapping relationship network, and perform real-time state deduction on the stamping forming process based on the dynamic twin model to generate a forming quality prediction result and a process defect location result; Generate a stamping process parameter adjustment plan according to the forming quality prediction result and the process defect location result, and feedback the stamping process parameter adjustment plan to the stamping control system to trigger parameter optimization operations.

2. The method according to claim 1, characterized in that, The performing collaborative feature extraction processing on the multi-source monitoring data set to generate a material structure feature set and a process dynamic feature set includes: Perform structural deformation feature extraction processing on the material physical parameter data to obtain the structural deformation features in the material structure feature set; among them, the structural deformation features include the material thickness change rate, yield strength change gradient, and residual stress distribution pattern; Perform action timing decomposition processing on the equipment operation status data to obtain the equipment action features in the process dynamic feature set; among them, the equipment action features include the stamping speed change curve, pressure transfer delay time, and die contact surface friction coefficient; Perform energy transfer analysis processing on the forming process dynamic response data to obtain the energy transfer features in the process dynamic feature set; among them, the energy transfer features include the material plastic deformation energy consumption rate, equipment kinetic energy loss rate, and heat dissipation distribution; Perform feature dimension unification processing on the structural deformation features, the equipment action features, and the energy transfer features, so that each feature maintains dimensional consistency in the time scale and space scale.

3. The method according to claim 2, wherein The performing structural deformation feature extraction processing on the material physical parameter data to obtain the structural deformation features in the material structure feature set includes: Collect real-time thickness measurement data of the stamping material during the forming process, and perform spatial interpolation processing on the real-time thickness measurement data to generate a thickness change rate distribution cloud map; Obtain the yield strength test data of the stamping material at different forming stages, and perform gradient calculation processing on the yield strength test data to generate a yield strength change gradient curve; Collect material surface stress distribution data based on a preset residual stress detection device, and perform pattern recognition processing on the stress distribution data to generate a residual stress distribution pattern feature vector; Normalize and fuse the thickness change rate distribution contour map, the yield strength change gradient curve, and the residual stress distribution pattern eigenvector to obtain the structural deformation characteristics.

4. The method according to claim 2, wherein Perform action timing decomposition processing on the device operation state data to obtain the device action characteristics in the process dynamic feature set, including: Obtain the spindle speed sensor data of the stamping device, and perform time series segmentation processing on the spindle speed sensor data to generate a stamping speed change curve; wherein, the time series segmentation processing is synchronously aligned based on the device action cycle and the material feeding cycle; Collect the hydraulic system pressure data of the stamping device, and perform delay time calculation processing on the hydraulic system pressure data to obtain a pressure transfer delay time series; wherein, the delay time calculation processing is determined based on the difference between the pressure peak arrival time and the mold closing time; Obtain real-time friction coefficient data through the friction force sensor on the mold contact surface, and perform sliding window statistical processing on the real-time friction coefficient data to generate a dynamic change feature of the friction coefficient; Perform time series alignment and feature splicing processing on the stamping speed change curve, the pressure transfer delay time series, and the dynamic change feature of the friction coefficient to obtain the device action characteristics.

5. The method according to claim 2, wherein Perform energy transfer analysis processing on the dynamic response data of the forming process to obtain the energy transfer characteristics in the process dynamic feature set, including: Calculate the energy consumption rate of material plastic deformation based on the power sensor data of the stamping device, wherein the energy consumption rate of plastic deformation is determined by the ratio of the energy absorbed by the material per unit time to the total input energy; Collect the kinetic energy loss data of the transmission system of the stamping device, and perform frequency domain decomposition processing on the kinetic energy loss data to obtain the spectral distribution characteristics of the device kinetic energy loss rate; Obtain the temperature distribution data of the contact surface between the mold and the material through an infrared thermal imager, and perform heat conduction simulation processing on the temperature distribution data to generate a heat dissipation distribution map; Perform space-time joint coding processing on the energy consumption rate of plastic deformation, the spectral distribution characteristics of the device kinetic energy loss rate, and the heat dissipation distribution map to obtain the energy transfer characteristics.

6. The method according to claim 1, wherein Construct a hybrid feature space based on the material structure feature set and the process dynamic feature set, and perform multi-dimensional feature matching processing in the hybrid feature space to generate a feature mapping relationship network, including: Perform feature cross-correlation processing on the structural deformation characteristics in the material structure feature set and the device action characteristics in the process dynamic feature set to generate a first-level feature correlation matrix; Perform space-time alignment processing on the residual stress distribution pattern eigenvector in the material structure feature set and the energy transfer characteristics in the process dynamic feature set to generate a second-level feature correlation matrix; Perform matrix fusion processing on the first-level feature correlation matrix and the second-level feature correlation matrix to obtain a hybrid feature correlation topological structure; Construct a dynamic weight assignment network based on the hybrid feature association topology structure, and optimize the connection relationship of the feature nodes in the hybrid feature space through the dynamic weight assignment network to generate the feature mapping relationship network.

7. The method according to claim 6, wherein The matrix fusion process of the first-level feature association matrix and the second-level feature association matrix to obtain the hybrid feature association topology structure includes: Perform singular value decomposition on the first-level feature association matrix to obtain the first eigenvector set and the first singular value distribution; Perform principal component analysis on the second-level feature association matrix to obtain the second eigenvector set and the principal component contribution rate; Perform orthogonal projection on the first eigenvector set and the second eigenvector set to generate a joint eigenvector space; Perform weighted fusion on the joint eigenvector space based on the first singular value distribution and the principal component contribution rate to obtain the hybrid feature association topology structure.

8. The method according to claim 6, characterized in that Construct a dynamic twin model according to the feature mapping relationship network, and perform real-time state deduction on the stamping process based on the dynamic twin model to generate the forming quality prediction result and the process defect location result, including: Convert the node connection relationship in the feature mapping relationship network into a dynamic differential equation system, determine the coupling coefficient matrix of the equation system based on the feature transfer path between nodes, and generate an initial state transition model; Collect the real-time pressure data and material deformation data at the current operation stage of the stamping equipment, and input the real-time pressure data and the material deformation data into the observation variable interface in the initial state transition model to generate a set of model input parameters; Call the dynamic parameter correction algorithm to perform online adjustment on the coupling coefficient matrix of the initial state transition model, calculate the parameter correction amount based on the residual between the real-time data stream and the model prediction value, and generate an updated state transition model; Simulate the deformation propagation path of the stamping material in the next forming stage through the updated state transition model, extract the predicted surface of the material thickness change and the internal stress diffusion trajectory, and generate a deformation process simulation data set; Compare the maximum thickness deviation value in the deformation process simulation data set with the preset forming tolerance threshold. If the deviation value exceeds the threshold, mark it as a deformation exceeding standard area to generate the forming quality prediction result; Identify the stress concentration area based on the curvature change characteristics of the internal stress diffusion trajectory, and perform similarity matching on the stress concentration area and the crack initiation mode in the historical defect database to generate the process defect location result.

9. The method according to claim 8, wherein The calling of the dynamic parameter correction algorithm to perform online adjustment on the coupling coefficient matrix of the initial state transition model, calculate the parameter correction amount based on the residual between the real-time data stream and the model prediction value, and generate an updated state transition model includes: Obtain the predicted value of the material thickness output by the initial state transition model and the real-time thickness value measured by the actual sensor, and calculate the residual sequence of the two within a continuous time window; Perform autocorrelation analysis on the residual sequence to identify the periodic characteristics of the residual fluctuation and the position of abnormal mutation points, and generate a residual characteristic analysis report. Adjust the update frequency of the coupling coefficient matrix based on the periodic characteristics in the residual characteristic analysis report, and trigger an emergency correction mechanism at the position of the abnormal mutation point; Use the recursive least squares method to iteratively optimize the dynamic weight parameters in the coupling coefficient matrix, and calculate the parameter adjustment step size according to the current residual value and the historical correction amount during each iteration; Substitute the optimized dynamic weight parameters into the dynamic differential equations to re-solve the material deformation propagation equation and generate a corrected state transition trajectory; Verify the matching degree between the corrected state transition trajectory and the actual measurement data, and terminate the correction process when the residual decrease rate of continuous S iterations is less than the preset convergence threshold to generate the updated state transition model, where S≥3.

10. A computer system, characterized in that, Including: A memory in which a computer program is stored; A processor for loading the computer program to implement the method for constructing a digital twin for metal stamping forming according to any one of claims 1-9.

Images