New energy vehicle stamping die detection method, device, equipment and medium

Abstract

This application relates to the technical field of stamping die inspection, and particularly to methods, apparatus, equipment, and media for inspecting stamping dies for new energy vehicles. The method includes acquiring die sensing data, which is used to monitor the operating status of the stamping die; extracting spatiotemporal features from the die sensing data to obtain the die's spatiotemporal characteristics; inputting the die's spatiotemporal characteristics into a digital twin corresponding to the stamping die; in the digital twin, performing health checks based on the die's spatiotemporal characteristics according to a preset time series to obtain detection results; the digital twin is used to predict the future failure time point of the stamping die based on the die's spatiotemporal characteristics; and based on the detection results, formulating a monitoring strategy, which provides a monitoring and maintenance strategy for the stamping die before the future failure time point. This method enables real-time monitoring of the stamping die, achieving comprehensive perception of the die's status, early diagnosis of die faults, and reducing finished product loss costs.

Description

Technical Field

[0001] This application relates to the technical field of stamping die testing, and in particular to methods, apparatus, equipment and media for testing stamping dies for new energy vehicles. Background Technology

[0002] Stamping dies are core equipment in modern manufacturing, and their operating status directly affects product quality, production efficiency, and equipment safety. Therefore, online monitoring of stamping dies is of paramount importance.

[0003] Currently, the monitoring of stamping dies generally uses a single signal, such as only monitoring pressure or sound wave signals, which is difficult to fully reflect the complex state of the die. In addition, when the monitored single signal exceeds the threshold alarm, the die has often already suffered irreversible damage or produced a large number of scrap products, resulting in increased costs. Therefore, there is an urgent need for a method, device, equipment and medium for detecting stamping dies for new energy vehicles. Summary of the Invention

[0004] This application provides a method, apparatus, equipment, and medium for detecting stamping dies for new energy vehicles. It can monitor stamping dies in real time to achieve comprehensive perception of the die status, diagnose die faults in advance, and reduce finished product loss costs.

[0005] On one hand, embodiments of this application provide a method for inspecting stamping dies, the method comprising:

[0006] Acquire mold sensing data, which is used to monitor the operating status of stamping dies;

[0007] Spatiotemporal features of the mold are extracted from the mold sensing data to obtain the mold spatiotemporal features;

[0008] The spatiotemporal characteristics of the mold are input into the digital twin corresponding to the stamping mold. In the digital twin, health detection is performed based on the spatiotemporal characteristics of the mold according to a preset time series to obtain the detection results. The digital twin is used to predict the future failure time nodes of the stamping mold based on the spatiotemporal characteristics of the mold.

[0009] Based on the test results, a monitoring strategy is formulated to provide a monitoring and maintenance strategy for stamping dies before the future failure time point.

[0010] Optionally, before acquiring the mold sensing data, the method further includes:

[0011] Obtain the material properties and geometric model of the stamping die;

[0012] Based on the aforementioned material properties and geometric model, multiple discrete units are constructed;

[0013] The unit density of the multiple discrete units is adjusted to obtain a digital twin.

[0014] Optionally, adjusting the unit density of the plurality of discrete units to obtain a digital twin includes:

[0015] The plurality of discrete units are divided into regions to obtain a plurality of expected regions, the expected regions including the die cavity, the rounded corner transition area and the stress concentration area of ​​the stamping die;

[0016] The density of the expected region is finely adjusted to obtain a fine-grained expected region.

[0017] Based on the aforementioned detailed expected region, a digital twin is constructed.

[0018] Optionally, after constructing the digital twin based on the refined expected region, the method further includes:

[0019] Obtain the historical operation and maintenance data and historical health score of the stamping die, wherein the historical operation and maintenance data includes the historical impact loads experienced by the stamping die at a preset historical time;

[0020] The historical impact load and the historical health score are input into the digital twin;

[0021] Within a preset historical time series, the predicted impact loads for each finely anticipated region in the digital twin are obtained;

[0022] Based on the overlap between the predicted impact load and the historical impact load, the digital twin is divided into stress regions to obtain critical regions and non-critical regions. The critical regions include the actual stress concentration regions of each discrete unit of the digital twin.

[0023] The refined expected region is adjusted based on the key region and the non-key region to obtain the region of interest;

[0024] Based on the region of interest, a data acquisition strategy is generated;

[0025] Based on the aforementioned data acquisition strategy, mold perception data is acquired.

[0026] Optionally, the mold sensing data includes vibration signals, acoustic emission signals, and force signals. The vibration signals are used to monitor impact vibrations, the acoustic emission signals are used to monitor stress wave signals of the stamping mold, and the force signals are used to monitor the real-time force-stroke curve during the stamping process.

[0027] The process of extracting spatiotemporal features from the mold sensing data to obtain the mold's spatiotemporal features includes:

[0028] Time-frequency domain feature extraction is performed based on the vibration signal, acoustic emission signal, and force signal to obtain vibration features, acoustic emission features, and force features;

[0029] The vibration characteristics, acoustic emission characteristics, and force characteristics are normalized to obtain standardized sensing characteristics;

[0030] The standardized perceptual features are concatenated to obtain high-dimensional perceptual features;

[0031] The high-dimensional sensing features are reduced in dimensionality and condensed to obtain the spatiotemporal features of the mold.

[0032] Optionally, the digital twin is configured with a machine learning model, and the health detection based on the spatiotemporal features of the mold is performed according to a preset time series to obtain the detection result, including:

[0033] The machine learning model calculates a health index based on the spatiotemporal features of the mold, and obtains a health score.

[0034] Based on the health score, plot the change curves of each region in the digital twin;

[0035] Based on the aforementioned change curve and the preset historical change trend, predict the future failure time point;

[0036] And / or, classify the stamping die according to the health score to obtain the operating status;

[0037] The future time point and / or the operating status are determined as the detection result.

[0038] Optionally, the monitoring results include a health score for each discrete unit, and the step of formulating a monitoring strategy based on the monitoring results includes:

[0039] According to a preset color calculation strategy, the mapped color of each discrete unit is calculated based on the health score;

[0040] Each discrete unit is color-rendered according to the mapped color to obtain a rendered digital twin;

[0041] Based on the rendered digital twin and the preset execution actions, a monitoring strategy is generated.

[0042] On the other hand, embodiments of this application provide a stamping die inspection device, the device comprising:

[0043] The acquisition module acquires mold sensing data, which is used to monitor the operating status of the stamping die.

[0044] The extraction module is used to extract spatiotemporal features from the mold sensing data to obtain the spatiotemporal features of the mold.

[0045] The detection module is used to input the spatiotemporal characteristics of the mold into the digital twin corresponding to the stamping mold. In the digital twin, health detection is performed based on the spatiotemporal characteristics of the mold according to a preset time series to obtain the detection results. The digital twin is used to predict the future failure time nodes of the stamping mold based on the spatiotemporal characteristics of the mold.

[0046] A planning module is used to formulate a monitoring strategy based on the detection results. The monitoring strategy is used to provide a monitoring and maintenance strategy for stamping dies before a future failure time point.

[0047] In another aspect, embodiments of this application provide an electronic device, the device including: a processor and a memory storing computer program instructions;

[0048] When the processor executes the computer program instructions, it implements the stamping die detection method as described in the first aspect.

[0049] In another aspect, embodiments of this application provide a computer storage medium storing computer program instructions, which, when executed by a processor, implement the stamping die detection method as described in the first aspect.

[0050] The new energy vehicle stamping die detection method, apparatus, equipment, and medium of this application embodiment can identify the state of the stamping die from multiple aspects and angles through multi-source information fusion. Then, through the spatiotemporal characteristics of the die, it is simulated in a digital twin. Based on the real-time acquired spatiotemporal characteristics of the die, the health of the stamping die is detected, and the detection results are obtained. That is, the stamping die is prevented from failing before failure in the digital twin, making the monitoring and diagnosis process more intuitive and transparent. At the same time, the monitoring strategy is formulated in advance according to the future failure point to ensure the safe use of the stamping die, thereby reducing the generation of scrap and reducing losses. Attached Figure Description

[0051] Figure 1 This is a schematic flowchart of a stamping die inspection method provided in an embodiment of this application;

[0052] Figure 2 This is a schematic diagram of the structure of a stamping die inspection device provided in another embodiment of this application;

[0053] Figure 3 This is a schematic diagram of the structure of an electronic device provided in another embodiment of this application. Detailed Implementation

[0054] The features and exemplary embodiments of various aspects of this application will be described in detail below. To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain this application and not to limit it. For those skilled in the art, this application can be implemented without some of these specific details. The following description of the embodiments is merely to provide a better understanding of this application by illustrating examples.

[0055] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes said element.

[0056] To address the problems of existing technologies, this application provides a method, apparatus, equipment, and medium for detecting stamping dies for new energy vehicles. In this application, the state of the stamping die can be identified from multiple perspectives through multi-source information fusion. Then, based on the spatiotemporal characteristics of the die, simulation is performed in a digital twin. Health checks are conducted on the stamping die based on the real-time acquired spatiotemporal characteristics, yielding the detection results. This allows for preventative measures against die failures before they occur within the digital twin, making the monitoring and diagnosis process more intuitive and transparent. Furthermore, by developing monitoring strategies in advance based on future failure points, the safe use of the stamping die is ensured, thereby reducing scrap and minimizing losses.

[0057] The stamping die testing method provided in the embodiments of this application will be introduced first below.

[0058] Figure 1 A schematic flowchart of a stamping die inspection method according to an embodiment of this application is shown. Figure 1 As shown, the stamping die inspection method may include S101-S104:

[0059] S101, acquire mold sensing data.

[0060] In some embodiments, the mold sensing data is used to monitor the operating status of the stamping die; wherein the mold sensing data includes vibration signals, acoustic emission signals and force signals, the vibration signals are used to monitor impact vibration, the acoustic emission signals are used to monitor the stress wave signals of the stamping die, and the force signals are used to monitor the real-time force-stroke curve during the stamping process.

[0061] Furthermore, the mold sensing data can also include process parameter signals, temperature signals, and optical signals. Specifically, process parameter signals can be obtained through equipment to acquire punch stroke, speed, tonnage, and balance bar pressure. Temperature signals can be obtained by installing infrared temperature sensors inside the stamping die or on the press slide to monitor temperature rise changes. Optical signals can be obtained by periodically taking product images using a high-speed industrial camera to indirectly monitor appearance quality.

[0062] In some embodiments, vibration signals can be obtained by installing accelerometers on the upper and lower dies of the stamping die and the press table; acoustic emission signals can be obtained by installing acoustic emission sensors near the die cavity; and force signals can be obtained by installing force sensors inside the die or on the press block.

[0063] S102, perform spatiotemporal feature extraction on the mold sensing data to obtain the mold spatiotemporal features.

[0064] In some embodiments, the spatiotemporal characteristics of the mold can be key features extracted from the time-series data and spatial distribution of the mold sensing data, i.e., the characteristics of the mold sensing data in the time domain, frequency domain, and time-frequency domain. Specifically, the vibration signal can include RMS value, kurtosis, and envelope spectrum features; the acoustic emission signal can include amplitude, energy, count, and frequency center features; and the force signal can include peak value, area, and shape deviation features.

[0065] S103, input the spatiotemporal characteristics of the mold into the digital twin corresponding to the stamping mold. In the digital twin, health detection is performed based on the spatiotemporal characteristics of the mold according to a preset time series to obtain the detection results.

[0066] In some embodiments, a digital twin can be a virtual mapping model of a physical mold, specifically implemented through discrete unit modeling and real-time data-driven methods. The digital twin is used to predict the future failure time of a stamping die based on its spatiotemporal characteristics.

[0067] By operating the digital twin virtually, health checks on stamping dies can be performed, ensuring that the dies are inspected and repaired in a timely manner before their future failure time, thus preventing damage to the dies from affecting production and reducing losses.

[0068] S104. Based on the test results, a monitoring strategy is developed to provide a monitoring and maintenance strategy for stamping dies before the future failure time point.

[0069] In this embodiment of the application, health detection refers to simulating the long-term operating state of the mold in a virtual environment. Specifically, machine learning models can be used to predict the trend of time series data to achieve early identification of failure nodes.

[0070] In this embodiment, the state of the stamping die can be identified from multiple perspectives and angles through multi-source information fusion. Then, the die's spatiotemporal characteristics are used to simulate its operation in a digital twin. Based on the real-time acquired spatiotemporal characteristics of the die, a health check is performed on the stamping die to obtain the check results. In other words, the stamping die can be prevented from failing before the failure occurs through the digital twin, making the monitoring and diagnosis process more intuitive and transparent. At the same time, monitoring strategies can be formulated in advance based on future failure points to ensure the safe use of the stamping die, thereby reducing the generation of scrap and reducing losses.

[0071] Furthermore, in this embodiment of the application, in order to ensure that the acquired mold sensing data can more realistically reflect the actual situation of the stamping mold, the method may further include the following steps before S101:

[0072] Obtain the material properties and geometric model of the stamping die;

[0073] Based on material properties and geometric models, multiple discrete units are constructed;

[0074] By adjusting the element density of multiple discrete elements, a digital twin is obtained.

[0075] In this embodiment, the material properties can be physical characteristics of the stamping die, such as Young's modulus, Poisson's ratio, and plasticity parameters. These can be measured using material testing equipment or extracted from a material database to define the mechanical behavior of the digital twin. Specifically, the Young's modulus can be approximately 210 GPa, the Poisson's ratio approximately 0.3, and the plasticity parameters can be obtained through material testing to obtain the plastic stress-strain curve.

[0076] In some other embodiments, material properties may also include basic properties such as material density and size.

[0077] The geometric model can be the three-dimensional structural data of the stamping die, such as obtained through 3D scanning or CAD modeling, used to construct the spatial topological relationships of the digital twin.

[0078] In this embodiment, after obtaining the geometric model, minor features that have little impact on the overall stiffness image, such as chamfers, fillets, small threaded holes, oil grooves, and trademarks, can be deleted to reduce the difficulty of element meshing. Additionally, if the mold exhibits a symmetrical structure, only a portion of the model's features can be analyzed, reducing the computational load.

[0079] In this embodiment, the discrete element can be a finite number of mesh or voxel elements decomposed into the geometric model. Specifically, it can be implemented using a finite element meshing algorithm to establish the computational model of the digital twin. Specifically, C3D8R can be used for element meshing, and shell elements can be used for thin-walled structural parts.

[0080] Element density adjustment can be used to adjust the distribution density of discrete elements according to the importance of the region. For example, the number of elements can be increased in stress concentration areas. This can be achieved through an adaptive mesh refinement algorithm to improve the modeling accuracy of key areas.

[0081] Specifically, material properties and geometric models are input into modeling software, where a mesh generation algorithm divides the geometric model into discrete elements. For example, complex curved surfaces are divided into triangular elements, and planar areas are divided into quadrilateral elements. Subsequently, an element density adjustment module locally refines or coarsens the discrete elements according to preset rules; for example, the element side length is adjusted from 2 mm to 0.5 mm on the mold cavity surface, while maintaining the original density in non-critical areas. The adjusted set of discrete elements forms the basic structure of the digital twin, and its element distribution density matches the actual stress characteristics of the stamping die.

[0082] Furthermore, adjusting the element density of multiple discrete elements to obtain a digital twin can include:

[0083] Multiple discrete units are divided into regions to obtain multiple expected regions;

[0084] Fine-tuning the density of the target region yields a refined target region.

[0085] Based on the precise expected region, construct a digital twin.

[0086] In this embodiment, region division can be achieved by dividing the physical structure of the stamping die into different regions according to its function or stress distribution characteristics. Specifically, this can be accomplished using a mesh generation algorithm combined with stress simulation tools, such as determining the boundaries of high-stress regions through finite element analysis. The expected regions include the die cavity, fillet transition area, and stress concentration area of ​​the stamping die. Because the stress gradient in these expected regions is relatively large, fine-grained density adjustments are needed to facilitate more accurate subsequent monitoring of each discrete unit of the stamping die.

[0087] Specifically, for the fine-grained expected region, fine-grained density adjustment is required. Fine-grained density adjustment can be used to differentiate the density of discrete element meshes in a specific region. This can be achieved by using an adaptive mesh refinement algorithm, such as increasing the element density to twice the base density in stress concentration areas.

[0088] As an example, a mold cavity refers to the forming cavity in a stamping die that is in direct contact with the workpiece; a rounded transition area refers to the arc-shaped transition part where the curved surface and the flat surface connect in the mold; and a stress concentration area refers to the area that bears the maximum stress during the stamping process.

[0089] Specifically, in constructing the digital twin, the spatial coordinate range of the mold cavity, fillet transition area, and stress concentration area is first determined through finite element simulation. A region growing algorithm is then used to cluster discrete elements into corresponding expected regions. Subsequently, independent density adjustments are performed on each expected region; for example, 0.1 mm precision elements are used on the mold cavity surface, and 0.05 mm precision elements are used in the stress concentration area. These adjusted, finely detailed expected regions are then topologically connected to form a complete digital twin, with the element density in critical regions significantly higher than in non-critical regions.

[0090] In this embodiment, by making differentiated adjustments in different regions, the unit density of key parts such as stress concentration areas is increased to two to three times that of non-critical areas while ensuring computational efficiency, thereby more accurately reflecting the actual stress state of the mold.

[0091] In other embodiments, after constructing the digital twin based on a finely anticipated region, the method further includes:

[0092] Obtain historical operation and maintenance data and historical health scores for stamping dies;

[0093] Input historical impact loads and historical health scores into the digital twin;

[0094] Within a preset historical time series, the predicted impact loads for each detailed expected region in the digital twin are obtained;

[0095] The digital twin is divided into stress regions based on the overlap between predicted and historical impact loads, resulting in critical and non-critical regions.

[0096] The refined target area is adjusted based on key and non-key areas to obtain the area of ​​interest;

[0097] Generate a data acquisition strategy based on the region of interest;

[0098] Based on the data acquisition strategy, mold perception data is acquired.

[0099] In some embodiments, historical operation and maintenance data can be historical impact loads, temperature, and vibration parameters recorded during the historical operation of the stamping die. Specifically, this data can be collected by sensors installed on the stamping die and stored in a database to provide training data for the digital twin. Historical health scores can be quantitative indicators of the die's health status assessed based on historical operation and maintenance data. Specifically, a weighted scoring algorithm can be used to calculate the degree of historical damage to reflect the actual health status of the stamping die at historical points in time. Predicted impact loads can be the expected load distribution simulated by the digital twin based on historical data. Specifically, this can be achieved through stress analysis of discrete elements using a finite element simulation model. The data is used to compare and verify with actual historical data; the overlap degree can be used to determine the degree of matching between predicted data and historical data in terms of load distribution patterns. Specifically, it can be quantified using Pearson correlation coefficient or Euclidean distance to evaluate the simulation accuracy of the digital twin; the key area can include the actual stress concentration areas of each discrete unit of the digital twin. Specifically, it can be marked by clustering algorithms for areas with overlap exceeding a threshold to identify parts that need to be monitored; the area of ​​interest can be the data acquisition range redefined after the key area screening. Specifically, it can be formed by merging key areas with adjacent non-key areas to form an extended monitoring range to optimize sensor deployment strategies.

[0100] In this embodiment, after the digital twin is constructed, the impact load sequence from historical operation and maintenance data is input into the digital twin for reverse verification. For example, for the mold cavity region, the digital twin generates a predicted load distribution curve for this region at historical time points based on material properties, and compares it point by point with the actual recorded load curve. When the overlap between the two in terms of peak position, fluctuation frequency, and other characteristics reaches a preset threshold, the region is marked as a critical region. Furthermore, adjacent units in non-critical regions that have a mechanical relationship with critical regions can be included in the region of interest, such as the connection between the rounded corner transition area and the stress concentration area. The data acquisition strategy dynamically adjusts the sensor sampling frequency according to the spatial distribution of the region of interest, for example, deploying high-frequency vibration sensors in critical regions and using an intermittent sampling mode in non-critical regions.

[0101] Furthermore, in order to ensure that the acquired mold sensing data can more accurately reflect the operation of the stamping die, the deployment of sensors is also crucial. The rationality of the sensor deployment can be verified by the historical operation of the stamping die, thereby further improving the data acquisition strategy. It is worth noting that the data acquisition strategy refers to the deployment strategy of sensors on the stamping die to obtain mold sensing data that is closer to reality.

[0102] In this embodiment, the actual stress concentration area can be dynamically identified through the membership data of the digital twin. Different sensor deployment strategies and sampling modes can be adopted for critical and non-critical areas to expand the monitoring range while reducing the amount of data collected.

[0103] In addition, it is worth noting that through the dynamic operation of the digital twin, the stress distribution changes caused by material performance degradation can be dynamically captured, and key and non-key areas can be dynamically adjusted to achieve adaptive matching between sensor deployment and the actual damage mode of the stamping die, thus enabling more intelligent detection of the stamping die's operating status.

[0104] In some embodiments, S102 may specifically include:

[0105] Time-frequency domain feature extraction is performed based on vibration signals, acoustic emission signals, and force signals to obtain vibration features, acoustic emission features, and force features;

[0106] The vibration characteristics, acoustic emission characteristics, and force characteristics are normalized to obtain standardized sensing characteristics;

[0107] Standardized perceptual features are concatenated to obtain high-dimensional perceptual features;

[0108] The high-dimensional perceptual features are reduced and condensed to obtain the spatiotemporal features of the mold.

[0109] In this embodiment, the vibration signal can be obtained by using time-domain analysis combined with fast Fourier transform to extract spectral features, which is used to capture the transient impact and periodic vibration modes of the die during the stamping process; the acoustic emission signal can be the stress wave signal captured by the piezoelectric sensor, which can be obtained by using wavelet transform to extract high-frequency stress release features, which is used to detect micro-stress changes caused by crack propagation or material fatigue inside the die; the force signal can be the real-time load data of the die and workpiece contact during the stamping process recorded by the force sensor, which can be obtained by using the piecewise integration method to extract the slope features of the force-stroke curve, which is used to reflect the dynamic correlation between the stamping process parameters and the force on the die.

[0110] Specifically, after obtaining the mold perception data, feature alignment and data cleaning are required, which may include:

[0111] Time synchronization: Ensure that all sensor data (vibration, acoustic emission, force) are aligned in timestamps. Typically, a common sampling clock or the main cycle of the device (e.g., one stamping cycle) is used as a reference, and a set of features is extracted for each stamping cycle.

[0112] Outlier Handling: Check the extracted raw feature values ​​for outliers caused by momentary sensor malfunction or significant external interference. Simple statistical methods or proximity-based methods can be used for identification and correction. For example, the statistical method could be the 3σ principle, and the correction method could be filling with the mean of preceding and following periods.

[0113] After data processing, vibration features, acoustic emission features, and force features can be normalized. Normalization can map vibration features, acoustic emission features, and force features with different dimensions to the same numerical range. Specifically, it can be achieved by using maximum-minimum scaling or Z-score normalization methods to eliminate scale differences between multi-source data and enhance feature comparability in order to obtain consistent standardized perceptual features.

[0114] For example, the peak value of the force-stroke curve may be on the order of 1000 kN, the kurtosis of the vibration signal may be a dimensionless number, usually around 3, and the count of the acoustic emission signal may reach tens of thousands. Z-score normalization and Min-Max normalization can be used.

[0115] In this embodiment, after obtaining the standardized sensing features, the standardized sensing features of each dimension are spliced ​​together. That is, feature splicing can be a series combination of normalized multidimensional features along the feature dimensions. Specifically, it can be implemented by tensor splicing or stacking to construct a high-dimensional feature space containing comprehensive information of vibration, acoustic emission and force signals.

[0116] While the stitched high-dimensional perceptual features are rich in information, they may suffer from information redundancy and the curse of dimensionality. Dimensionality reduction and condensation can reduce computational burden, eliminate collinearity among features, and improve the generalization ability and robustness of subsequent machine learning models. Specifically, dimensionality reduction and condensation involve compressing high-dimensional features using principal component analysis or autoencoders. This can be achieved using linear projection or nonlinear mapping methods to remove redundant information and extract core features representing the spatiotemporal state of the model.

[0117] As an example, dimensionality reduction and condensation can be achieved through principal component analysis or novel discriminant analysis.

[0118] Specifically, after vibration, acoustic emission, and force signals are acquired synchronously, vibration, acoustic emission, and force features are obtained respectively through time-frequency domain feature extraction methods. For example, the time-frequency domain features of the vibration signal may include the root mean square value, peak factor, and spectral centroid; the acoustic emission features may include energy count, rise time, and amplitude distribution; and the force features may include maximum load, impact force gradient, and stroke correlation. After normalization, these features are concatenated into high-dimensional sensing features containing multi-dimensional information. Subsequently, noise is removed and key spatiotemporal features are extracted using a dimensionality reduction algorithm. The resulting spatiotemporal features of the mold can simultaneously reflect the mold's vibration mode, stress wave propagation characteristics, and dynamic changes in impact force, providing a comprehensive data foundation for subsequent health monitoring.

[0119] In some embodiments, the digital twin is configured with a machine learning model, which may be configured with a gradient boosting tree (GBDT), a long short-term memory network (LSTM), or an autoencoder (AE) to achieve health detection.

[0120] Specifically, S103 can include:

[0121] A health index is calculated based on the spatiotemporal characteristics of the mold to obtain a health score;

[0122] Based on health scores, plot the change curves of various regions in the digital twin;

[0123] Predict future failure time points based on the change curve and preset historical change trends;

[0124] And / or, classify the stamping die into its operating status based on a health score;

[0125] The future time point and / or operating status will be determined as the detection result.

[0126] In this embodiment, the health score can be a quantitative indicator reflecting the current health status of the stamping die. Specifically, it can be obtained by weighting the vibration characteristics, acoustic emission characteristics, and force characteristics. For example, after normalizing each characteristic, it can be input into a regression model to generate a score between 0 and 1. The change curve can be a trend graph of the health score changing over time. Specifically, a line graph or heat map can be used to show the score changes in different regions to quantify the degree of degradation of the stamping die.

[0127] The future failure time point can be used to predict when the mold may fail. Specifically, this can be predicted by analyzing the correlation between the change curve and historical failure data. For example, an early warning can be triggered when the rate of decline of the health score exceeds a preset threshold. Status classification refers to dividing the mold's operating status into different categories based on the health score. This can be done using clustering algorithms or threshold segmentation methods. For example, molds with a score below 0.5 can be classified as abnormal.

[0128] Specifically, during the operation of the digital twin, the machine learning model continuously receives spatiotemporal features of the mold from the sensor array. These features, after preprocessing, are input into a trained prediction model to calculate a health score. For example, when high-frequency components appear in the vibration characteristics of the mold cavity region, the model lowers the health score for that region. The system automatically plots the score data as a curve in chronological order and compares it with curves under similar conditions in a historical database. When the current curve's downward slope is detected to be more than twice the historical average, the prediction module generates a warning message indicating potential failure within the next two weeks. Simultaneously, the classification module compares the health score with a preset threshold; if the score for the principal stress region is below 0.3 three times consecutively, the mold status is marked as "requiring immediate maintenance." These detection results are integrated and displayed simultaneously on a visual interface, showing the predicted time point and the current status classification.

[0129] In this embodiment, by fusing multi-dimensional spatiotemporal features through a machine learning model, early abnormal signs in complex fault modes can be identified. That is, by analyzing historical trends, a dynamic early warning mechanism is established, significantly improving prediction accuracy. Furthermore, unlike traditional state classification methods that rely on fixed grading standards set by human experience, this embodiment uses a data-driven approach to achieve automatic state classification, adapting to the characteristic differences of different mold types.

[0130] In some other embodiments, in order to reduce losses before the stamping die is damaged, the monitoring results may include a health score for each discrete unit, and S104 may include:

[0131] According to the preset color calculation strategy, the color mapped to each discrete unit is calculated based on the health score;

[0132] Each discrete unit is color-rendered according to the mapped color to obtain a rendered digital twin;

[0133] A monitoring strategy is generated based on the rendered digital twin and preset execution actions.

[0134] In this embodiment, the preset color calculation strategy can be a rule that associates health scores with color gradients. Specifically, it can be implemented using linear interpolation algorithms or piecewise function mapping. For example, the 0-100 score range can be mapped to a red-yellow-green gradient color system. The mapped color can be a visual color identifier that converts the health score into a visual representation of the health status of the discrete unit. Rendering the digital twin involves color rendering the corresponding discrete unit according to the mapped color. Specifically, it can be implemented using a 3D model shading engine to make the status of key areas visible.

[0135] As an example, the preset execution action can be a predefined set of maintenance operation instructions, such as triggering a mold shutdown and maintenance process when the proportion of the red area exceeds a threshold.

[0136] Specifically, after obtaining the health scores of discrete units, the scores are converted into corresponding RGB color values ​​using a color calculation strategy; for example, a unit with a health score of 80 is mapped to yellow. The color data is then input into a 3D modeling tool to color the surface of each discrete unit in the digital twin, generating a visual model with color layers. Maintenance personnel can quickly identify high-loss areas by observing the color distribution, such as continuous red blocks appearing at the edge of a mold cavity. Based on the rendered digital twin, the system automatically matches preset maintenance actions; for example, when more than 50% of the units in a stress concentration area turn red, a maintenance plan for mold surface polishing and material reinforcement is generated.

[0137] Based on the stamping die inspection method provided in the above embodiments, this application also provides specific implementation methods of the stamping die inspection device. Please refer to the following embodiments.

[0138] First see Figure 2 The stamping die inspection device 200 provided in this application embodiment may include:

[0139] The acquisition module 201 acquires mold sensing data, which is used to monitor the operating status of the stamping die.

[0140] Extraction module 202 is used to extract spatiotemporal features from mold sensing data to obtain the spatiotemporal features of the mold;

[0141] The detection module 203 is used to input the spatiotemporal characteristics of the mold into the digital twin corresponding to the stamping mold. In the digital twin, health detection is performed based on the spatiotemporal characteristics of the mold according to a preset time series to obtain the detection results. The digital twin is used to predict the future failure time nodes of the stamping mold based on the spatiotemporal characteristics of the mold.

[0142] Module 204 is used to develop a monitoring strategy based on the detection results. The monitoring strategy is used to provide a monitoring and maintenance strategy for stamping dies before the future failure time point.

[0143] As an optional implementation, the acquisition module 201 can also be used for:

[0144] Obtain the material properties and geometric model of the stamping die;

[0145] Based on material properties and geometric models, multiple discrete units are constructed;

[0146] By adjusting the element density of multiple discrete elements, a digital twin is obtained.

[0147] As an optional implementation, the acquisition module 201 can also be used for:

[0148] Multiple discrete units are divided into regions to obtain multiple expected regions, including the die cavity, fillet transition area and stress concentration area of ​​the stamping die;

[0149] Fine-tuning the density of the target region yields a refined target region.

[0150] Based on the precise expected region, construct a digital twin.

[0151] As an optional implementation, the acquisition module 201 can also be used for:

[0152] Obtain historical operation and maintenance data and historical health scores of stamping dies. The historical operation and maintenance data includes the historical impact loads experienced by the stamping die at preset historical moments.

[0153] Input historical impact loads and historical health scores into the digital twin;

[0154] Within a preset historical time series, the predicted impact loads for each detailed expected region in the digital twin are obtained;

[0155] Based on the overlap between predicted impact loads and historical impact loads, the digital twin is divided into stress regions to obtain critical and non-critical regions. The critical regions include the actual stress concentration regions of each discrete unit of the digital twin.

[0156] The refined target area is adjusted based on key and non-key areas to obtain the area of ​​interest;

[0157] Generate a data acquisition strategy based on the region of interest;

[0158] Based on the data acquisition strategy, mold perception data is acquired.

[0159] The mold sensing data includes vibration signals, acoustic emission signals, and force signals. Vibration signals are used to monitor impact vibration, acoustic emission signals are used to monitor stress wave signals of the stamping die, and force signals are used to monitor the real-time force-stroke curve during the stamping process. The extraction module 202 can also be used for:

[0160] Spatiotemporal feature extraction is performed on the mold sensing data to obtain the mold's spatiotemporal features, including:

[0161] Time-frequency domain feature extraction is performed based on vibration signals, acoustic emission signals, and force signals to obtain vibration features, acoustic emission features, and force features;

[0162] The vibration characteristics, acoustic emission characteristics, and force characteristics are normalized to obtain standardized sensing characteristics;

[0163] Standardized perceptual features are concatenated to obtain high-dimensional perceptual features;

[0164] The high-dimensional perceptual features are reduced and condensed to obtain the spatiotemporal features of the mold.

[0165] As an alternative implementation, the detection module 203 can also be used for:

[0166] A machine learning model calculates a health index based on the spatiotemporal features of the model, resulting in a health score.

[0167] Based on health scores, plot the change curves of various regions in the digital twin;

[0168] Predict future failure time points based on the change curve and preset historical change trends;

[0169] And / or, classify the stamping die into its operating status based on a health score;

[0170] The future time point and / or operating status will be determined as the detection result.

[0171] As an alternative implementation, the design module 204 can also be used for:

[0172] According to the preset color calculation strategy, the color mapped to each discrete unit is calculated based on the health score;

[0173] Each discrete unit is color-rendered according to the mapped color to obtain a rendered digital twin;

[0174] A monitoring strategy is generated based on the rendered digital twin and preset execution actions.

[0175] Figure 3 A schematic diagram of the hardware structure of the electronic device provided in an embodiment of this application is shown.

[0176] An electronic device may include a processor 301 and a memory 302 storing computer program instructions.

[0177] Specifically, the processor 301 may include a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0178] Memory 302 may include mass storage for data or instructions. For example, and not limitingly, memory 302 may include a hard disk drive (HDD), floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or Universal Serial Bus (USB) drive, or a combination of two or more of these. In one instance, memory 302 may include removable or non-removable (or fixed) media, or memory 302 may be non-volatile solid-state memory. Memory 302 may be internal or external to the integrated gateway disaster recovery device.

[0179] In one instance, memory 302 may be read-only memory (ROM). In one instance, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), an electrically rewritable ROM (EAROM), or flash memory, or a combination of two or more of these.

[0180] Memory 302 may include read-only memory (ROM), random access memory (RAM), disk storage media device, optical storage media device, flash memory device, electrical, optical, or other physical / tangible memory storage device. Therefore, typically, memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the stamping die inspection method according to the first aspect of this disclosure.

[0181] The processor 301 reads and executes computer program instructions stored in the memory 302 to achieve... Figure 1 A stamping die inspection method is shown in the embodiment.

[0182] In one example, the electronic device may also include a communication interface 303 and a bus 304. For example, Figure 3 As shown, the processor 301, memory 302, and communication interface 303 are connected through bus 304 and complete communication with each other.

[0183] The communication interface 303 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this application.

[0184] Bus 304 includes hardware, software, or both, that couples components of an electronic device together. For example, and not as a limitation, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Microchannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses, or a combination of two or more of these. Where appropriate, bus 304 may include one or more buses. Although specific buses are described and illustrated in embodiments of this application, this application contemplates any suitable bus or interconnect.

[0185] This electronic device can perform the stamping die detection method in the embodiments of this application, thereby achieving the combination Figures 1-2 The described method and apparatus for inspecting stamping dies.

[0186] Furthermore, in conjunction with the stamping die inspection method in the above embodiments, this application embodiment can provide a computer storage medium for implementation. The computer storage medium stores computer program instructions; when these computer program instructions are executed by a processor, they implement any of the stamping die inspection methods in the above embodiments.

[0187] In an optional embodiment, in conjunction with the stamping die inspection method in the above embodiments, this application embodiment can provide a computer program product to implement it. The instructions in the computer program product are executed by the processor of an electronic device, enabling the electronic device to implement any of the stamping die inspection methods in the above embodiments.

[0188] It should be clarified that this application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.

[0189] The functional blocks shown in the above block diagram can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this application are programs or code segments used to perform the required tasks. Programs or code segments can be stored on a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, radio frequency (RF) links, etc. Code segments can be downloaded via computer networks such as the Internet, intranets, etc.

[0190] It should also be noted that the exemplary embodiments mentioned in this application describe methods or systems based on a series of steps or apparatus. However, this application is not limited to the order of the above steps; that is, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.

[0191] It should be clarified that this application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.

[0192] The functional blocks shown in the above-described structural diagram can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this application are programs or code segments used to perform the required tasks. Programs or code segments can be stored on a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, radio frequency (RF) links, etc. Code segments can be downloaded via computer networks such as the Internet, intranets, etc.

[0193] It should also be noted that the exemplary embodiments mentioned in this application describe methods or systems based on a series of steps or apparatus. However, this application is not limited to the order of the above steps; that is, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.

[0194] The aspects of this disclosure have been described above with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It should be understood that each block in the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that these instructions, executable via the processor of the computer or other programmable data processing apparatus, enable the implementation of the functions / actions specified in one or more blocks of the flowchart illustrations and / or block diagrams. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field-programmable logic circuit. It is also understood that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can also be implemented by special-purpose hardware performing the specified functions or actions, or can be implemented by a combination of special-purpose hardware and computer instructions.

[0195] The above description is merely a specific implementation of this application. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, modules, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. It should be understood that the protection scope of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the protection scope of this application.

Claims

1. A method for inspecting stamping dies, characterized in that, include: Acquire mold sensing data, which is used to monitor the operating status of stamping dies; Spatiotemporal features of the mold are extracted from the mold sensing data to obtain the mold spatiotemporal features; The spatiotemporal characteristics of the mold are input into the digital twin corresponding to the stamping mold. In the digital twin, health detection is performed based on the spatiotemporal characteristics of the mold according to a preset time series to obtain the detection results. The digital twin is used to predict the future failure time node of the stamping mold based on the spatiotemporal characteristics of the mold. Based on the detection results, a monitoring strategy is formulated to provide a monitoring and maintenance strategy for stamping dies before the future failure time point; Prior to acquiring the mold sensing data, the method further includes: Obtain the material properties and geometric model of the stamping die; Based on the aforementioned material properties and geometric model, multiple discrete units are constructed; The unit density of the plurality of discrete units is adjusted to obtain a digital twin; The step of adjusting the unit density of the plurality of discrete units to obtain a digital twin includes: The plurality of discrete units are divided into regions to obtain a plurality of expected regions, the expected regions including the die cavity, the rounded corner transition area and the stress concentration area of ​​the stamping die; The density of the expected region is finely adjusted to obtain a fine-grained expected region. Based on the aforementioned detailed expected region, a digital twin is constructed.

2. The method according to claim 1, characterized in that, After constructing the digital twin based on the refined expected region, the method further includes: Obtain the historical operation and maintenance data and historical health score of the stamping die, wherein the historical operation and maintenance data includes the historical impact loads experienced by the stamping die at a preset historical time; The historical impact load and the historical health score are input into the digital twin; Within a preset historical time series, the predicted impact loads for each finely anticipated region in the digital twin are obtained; Based on the overlap between the predicted impact load and the historical impact load, the digital twin is divided into stress regions to obtain critical regions and non-critical regions. The critical regions include the actual stress concentration regions of each discrete unit of the digital twin. The refined expected region is adjusted based on the key region and the non-key region to obtain the region of interest; Based on the region of interest, a data acquisition strategy is generated; Based on the aforementioned data acquisition strategy, mold perception data is acquired.

3. The method according to claim 1, characterized in that, The mold sensing data includes vibration signals, acoustic emission signals, and force signals. The vibration signals are used to monitor impact vibrations, the acoustic emission signals are used to monitor stress wave signals of the stamping mold, and the force signals are used to monitor the real-time force-stroke curve during the stamping process. The process of extracting spatiotemporal features from the mold sensing data to obtain the mold's spatiotemporal features includes: Time-frequency domain feature extraction is performed based on the vibration signal, acoustic emission signal, and force signal to obtain vibration features, acoustic emission features, and force features; The vibration characteristics, acoustic emission characteristics, and force characteristics are normalized to obtain standardized sensing characteristics; The standardized perceptual features are concatenated to obtain high-dimensional perceptual features; The high-dimensional sensing features are reduced in dimensionality and condensed to obtain the spatiotemporal features of the mold.

4. The method according to claim 1, characterized in that, The digital twin is equipped with a machine learning model, and the health detection based on the spatiotemporal features of the mold is performed according to a preset time series to obtain the detection results, including: The machine learning model calculates a health index based on the spatiotemporal features of the mold, and obtains a health score. Based on the health score, plot the change curves of each region in the digital twin; Based on the aforementioned change curve and the preset historical change trend, predict the future failure time point; And / or, classify the stamping die according to the health score to obtain the operating status; The future failure time point and / or the operating status are determined as the detection result.

5. The method according to any one of claims 1-4, characterized in that, The detection results include a health score for each discrete unit, and the monitoring strategy formulated based on the detection results includes: According to a preset color calculation strategy, the mapped color of each discrete unit is calculated based on the health score; Each discrete unit is color-rendered according to the mapped color to obtain a rendered digital twin; Based on the rendered digital twin and the preset execution actions, a monitoring strategy is generated.

6. A stamping die inspection device, characterized in that, The device includes: The acquisition module acquires mold sensing data, which is used to monitor the operating status of the stamping die. The extraction module is used to extract spatiotemporal features from the mold sensing data to obtain the spatiotemporal features of the mold. The detection module is used to input the spatiotemporal characteristics of the mold into a digital twin corresponding to the stamping mold. In the digital twin, health detection is performed based on the spatiotemporal characteristics of the mold according to a preset time series to obtain the detection results. The digital twin is used to predict the future failure time node of the stamping mold based on the spatiotemporal characteristics of the mold. A planning module is used to formulate a monitoring strategy based on the detection results. The monitoring strategy is used to provide a monitoring and maintenance strategy for stamping dies before the future failure time point. The acquisition module is also used for: Obtain the material properties and geometric model of the stamping die; Based on the aforementioned material properties and geometric model, multiple discrete units are constructed; The unit density of the plurality of discrete units is adjusted to obtain a digital twin; The acquisition module is also used for: The plurality of discrete units are divided into regions to obtain a plurality of expected regions, the expected regions including the die cavity, the rounded corner transition area and the stress concentration area of ​​the stamping die; The density of the expected region is finely adjusted to obtain a fine-grained expected region. Based on the aforementioned detailed expected region, a digital twin is constructed.

7. An electronic device, characterized in that, The device includes: a processor and a memory storing computer program instructions; When the processor executes the computer program instructions, it implements the stamping die detection method as described in any one of claims 1-5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer program instructions, which, when executed by a processor, implement the stamping die inspection method as described in any one of claims 1-5.

Images