An automatic control method and system for changing the stamping head of a stamping die

Abstract

This application relates to the field of stamping die technology, and provides a method and system for automatic replacement control of stamping heads in stamping dies. The method includes: constructing a multi-dimensional sensing network to collect multi-dimensional working data streams of the stamping head; embedding RFID tags on the stamping head and reading its model and usage data; building a three-channel anomaly identification system based on the sensing network and model data, mapping the identification data stream, and outputting anomaly feature sets; predicting the working state based on the usage data and anomaly feature sets, and controlling the stamping head replacement by triggering an automatic replacement mechanism based on the predicted state. This application solves the technical problem of traditional stamping dies relying on manual experience to judge the timing of stamping head replacement, leading to untimely or misjudged replacements, affecting production efficiency and equipment lifespan. It achieves real-time and accurate monitoring and prediction of the stamping head's working state through multi-dimensional sensing and data fusion, improving the timeliness and accuracy of automatic replacement, reducing equipment failure rates, and ensuring continuous production.

Description

Technical Field

[0001] This application relates to the field of stamping die technology, specifically to an automatic control method and system for changing the stamping head of a stamping die. Background Technology

[0002] In precision machining in manufacturing, stamping dies, as core process equipment, are widely used in the mass production of automotive parts, electronic components, and home appliance structural parts. The stamping head, as a key actuator of the stamping die, directly affects product forming accuracy, equipment operational stability, and production efficiency. However, traditional stamping die stamping head replacement control has significant drawbacks: on the one hand, manual judgment relies on the operator's skill level, making it difficult to monitor stamping head wear, deformation, or performance degradation in real time and accurately. This can easily lead to premature replacement, wasting resources, or delayed replacement, causing equipment failure and increased product defect rates. On the other hand, timed online inspections typically use single-sensor monitoring schemes (such as relying solely on displacement or pressure sensors), resulting in a limited data dimension that cannot comprehensively reflect the stamping head's working status, leading to insufficient accuracy in status assessment. Summary of the Invention

[0003] This application provides an automatic replacement control method and system for stamping dies, aiming to solve the technical problem that traditional stamping dies rely on manual experience to judge the timing of stamping head replacement, resulting in untimely or misjudged replacement, which affects production efficiency and equipment life.

[0004] The first aspect disclosed in this application provides an automatic replacement control method for the stamping head of a stamping die. The method includes: constructing a multi-dimensional sensing network, which integrates a laser displacement sensor, a pressure sensor, and a visual recognition system; acquiring multi-dimensional working data streams of the stamping head of a target stamping die through the multi-dimensional sensing network; embedding an RFID tag on the stamping head of the target stamping die; reading stamping head model data and stamping head usage data through the RFID tag; constructing a three-channel stamping head anomaly identification based on the multi-dimensional sensing network and the stamping head model data; mapping and identifying the multi-dimensional working data streams of the stamping head using the three-channel stamping head anomaly identification; and outputting a multi-dimensional anomaly feature set of the stamping head. Based on the stamping head usage data and the multi-dimensional anomaly feature set of the stamping head, the working state is predicted to obtain the predicted working state of the stamping head. The predicted working state triggers the stamping head replacement mechanism for automatic replacement control. According to the multi-dimensional sensing network, an anomaly identification channel architecture for the stamping head is designed, with each channel in the architecture corresponding one-to-one with each sensor type in the multi-dimensional sensing network. Three-dimensional modeling is performed based on the stamping head model data to obtain a standard stamping head model. Anomaly data sets for stamping head displacement, pressure, and appearance are collected. The anomaly types and degrees of the anomaly data sets for the stamping head displacement and pressure are determined respectively. The following steps are taken: First, an abnormal displacement and pressure working sample sets of the stamping head are identified. Then, anomaly recognition training is performed based on these sets to obtain anomaly recognition channels for stamping head displacement and pressure. Next, anomaly recognition training is conducted on the set of abnormal stamping head appearance images using the standard stamping head model to obtain anomaly recognition channel for stamping head appearance. Finally, the three channels—displacement, pressure, and appearance—are connected in parallel to build a three-channel system for stamping head anomaly recognition. Appearance features are extracted from both the standard stamping head model and the set of abnormal stamping head appearance images to obtain a standard stamping head appearance feature set and anomaly recognition channel. The process involves: establishing a standard stamping head appearance feature set; calculating the loss of the standard stamping head appearance feature set and the abnormal stamping head appearance feature set to obtain abnormal stamping head loss feature data; identifying the abnormality type and degree of the abnormal stamping head loss feature data to obtain a stamping head appearance abnormality sample set; training for abnormality recognition based on the stamping head appearance abnormality sample set to obtain the stamping head appearance abnormality recognition channel; determining the key operating points of the stamping head based on the standard stamping head appearance feature set; comparing and analyzing the standard stamping head appearance feature set and the abnormal stamping head appearance feature set to obtain an abnormal stamping head appearance loss feature set; and designing a feature loss function based on the spatial distance set between the abnormal stamping head appearance loss feature set and the key operating points of the stamping head.Based on the aforementioned feature loss function, loss calculation is performed on the appearance loss feature set of the abnormal stamping head to obtain the loss feature data of the abnormal stamping head.

[0005] Another aspect of this application discloses an automatic replacement control system for a stamping die's stamping head. The system includes: a working data acquisition module: constructing a multi-dimensional sensing network integrating a laser displacement sensor, a pressure sensor, and a visual recognition system; acquiring multi-dimensional working data streams of the stamping head of the target stamping die through the multi-dimensional sensing network; an RFID tag reading module: embedding RFID tags on the stamping head of the target stamping die; reading stamping head model data and stamping head usage data through the RFID tags; a mapping and identification module: establishing a three-channel stamping head anomaly identification system based on the multi-dimensional sensing network and the stamping head model data; mapping and identifying the multi-dimensional working data stream of the stamping head using the three-channel stamping head anomaly identification system; and outputting a multi-dimensional anomaly feature set of the stamping head; and a replacement control module: predicting the working state based on the stamping head usage data and the multi-dimensional anomaly feature set of the stamping head, obtaining a predicted working state of the stamping head, and triggering a stamping head replacement mechanism for automatic replacement control based on the predicted working state of the stamping head.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0007] The aforementioned method for automatic replacement control of the stamping head of a stamping die constructs a multi-dimensional sensing network integrating a laser displacement sensor, a pressure sensor, and a vision recognition system. This network collects multi-dimensional working data of the stamping head in real time, enabling comprehensive monitoring of its operational status. Simultaneously, RFID tags are installed on the stamping head to read its model and usage information, providing a foundation for data analysis. Subsequently, using this data and model information, a three-channel anomaly identification mechanism is established to analyze the stamping head's working data and identify potential anomalies. Based on these anomalies and the stamping head's usage data, the operating status of the stamping head can be predicted. Finally, when the predicted operating status of the stamping head reaches the replacement standard, the replacement mechanism is automatically triggered, resulting in automatic replacement of the stamping head and thus optimizing production efficiency.

[0008] The above description is merely an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, specific embodiments of this application are given below. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 This is a flowchart illustrating an automatic replacement control method for the stamping head of a stamping die in one embodiment.

[0011] Figure 2 This is a schematic diagram of an automatic changing control system architecture for a stamping die in one embodiment.

[0012] Explanation of reference numerals in the attached diagram: 11 Working data acquisition module, 12 RFID tag reading module, 13 Mapping and identification module, 14 Replacement control module. Detailed Implementation

[0013] This application provides an automatic replacement control method and system for the stamping head of a stamping die, which solves the technical problem that traditional stamping dies rely on manual experience to judge the timing of stamping head replacement, resulting in untimely or misjudged replacement, affecting production efficiency and equipment life.

[0014] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0015] It should be noted that the terms “comprising” and “having”, and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to such process, method, product, or device.

[0016] Example 1, as Figure 1 As shown, this application provides an automatic changing control method for the stamping head of a stamping die, the method comprising:

[0017] A multidimensional sensing network is constructed, which integrates a laser displacement sensor, a pressure sensor, and a visual recognition system. The multidimensional working data stream of the stamping head of the target stamping die is collected through the multidimensional sensing network.

[0018] In this embodiment, a multidimensional sensing network is first constructed, which consists of multiple sensors and a visual recognition system. The sensors include a laser displacement sensor and a pressure sensor. The laser displacement sensor is used to accurately measure the changes in the position and movement of the stamping head during operation, and can detect the displacement data of the stamping head in real time to monitor its movement during the stamping process, such as excessive offset or deformation. The pressure sensor is used to monitor the pressure applied to the material by the stamping head during operation, and helps to determine whether the stamping head is malfunctioning, worn, or damaged. The visual recognition system uses a high-definition camera to detect the appearance of the stamping head in real time, and is used to identify whether there are physical damages such as cracks, wear, and deformation. After deploying the constructed multidimensional sensing network to the key working areas of the target stamping die (such as the worktable, around the stamping head, etc.), the multidimensional sensing network will be used to comprehensively monitor the working status of the stamping head. The displacement data, pressure data, and appearance image data of the stamping head will be collected in real time. These data will be summarized and stored to form a multidimensional working data stream of the stamping head, which will provide a comprehensive view of the stamping head's performance during use and provide basic information for subsequent anomaly identification and fault prediction.

[0019] An RFID tag is embedded in the stamping head of the target stamping die, and the stamping head model data and stamping head usage data are read through the RFID tag.

[0020] In one embodiment, an RFID tag is embedded in the stamping head of the target stamping die. This RFID tag serves as an information carrier, storing the stamping head's model data and usage data. The model data mainly includes basic information such as the stamping head's specifications, material, and design parameters, used to quickly identify the stamping head type and ensure its compatibility with the stamping process. The usage data covers operational information such as the stamping head's cumulative working count, usage duration, historical maintenance records, and replacement records, comprehensively reflecting the stamping head's lifespan and health status. Through an RFID reader installed on the die, the stamping head's model data and usage data can be read in real time during the target stamping die's operation, eliminating the need for manual recording and avoiding omissions and errors in manual management. This provides reliable input data support for subsequent anomaly identification and operational status prediction, achieving efficient and intelligent stamping head management.

[0021] Based on the multidimensional sensing network and the stamping head model data, a three-channel stamping head anomaly identification system is constructed. The three-channel stamping head anomaly identification system is used to map and identify the multidimensional working data stream of the stamping head, and output a multidimensional anomaly feature set of the stamping head.

[0022] In one embodiment, after obtaining the stamping head model data, a three-channel architecture for stamping head anomaly identification is designed based on a multi-dimensional sensing network. Then, a 3D model is created using the stamping head model data to form a standard stamping head model. Subsequently, the standard stamping head model is combined with collected historical data to iteratively train the designed three-channel architecture for stamping head anomaly identification, building the final three-channel system for analyzing and detecting displacement, pressure, and appearance anomalies of the stamping head, thereby achieving comprehensive monitoring and fault identification of the stamping head status. Afterward, the real-time collected multi-dimensional working data stream of the stamping head is mapped to the corresponding channels in the three-channel system for stamping head anomaly identification. For example, displacement data is mapped to the stamping head displacement anomaly identification channel. The three-channel system analyzes and identifies the mapped data based on learned knowledge, outputting a multi-dimensional anomaly feature set for the stamping head. This feature set contains various anomaly information that may occur during the operation of the stamping head, providing important basis for subsequent working status prediction and stamping head replacement decisions, ensuring the stable operation of the production line.

[0023] Furthermore, this application provides the aforementioned three-channel system for identifying abnormalities in the stamping head, including:

[0024] Based on the multidimensional sensing network, a stamping head anomaly identification channel architecture is designed, where each channel in the architecture corresponds one-to-one with each sensor type in the multidimensional sensing network. A standard stamping head model is obtained by performing 3D modeling based on the stamping head model data. A stamping head displacement anomaly working dataset, a stamping head pressure anomaly working dataset, and a stamping head appearance anomaly image set are collected. The anomaly types and degrees of the stamping head displacement anomaly working dataset and the stamping head pressure anomaly working dataset are respectively identified to obtain a stamping head displacement anomaly working sample set and a stamping head pressure anomaly working sample set. Anomaly identification training is performed based on the stamping head displacement anomaly working sample set and the stamping head pressure anomaly identification channel to obtain a stamping head displacement anomaly identification channel and a stamping head pressure anomaly identification channel. Anomaly identification training is performed on the stamping head appearance anomaly image set using the standard stamping head model to obtain a stamping head appearance anomaly identification channel. The stamping head displacement anomaly identification channel, the stamping head pressure anomaly identification channel, and the stamping head appearance anomaly identification channel are connected in parallel to build a three-channel stamping head anomaly identification system.

[0025] Preferably, firstly, based on the sensor types in the constructed multidimensional sensing network, a stamping head anomaly recognition channel architecture is designed. For laser displacement sensors, a displacement anomaly recognition channel architecture can be designed based on long short-term memory networks, deep neural networks, etc., to detect the displacement and deformation of the stamping head during operation. For pressure sensors, a pressure anomaly recognition channel architecture can be designed based on multilayer perceptrons, long short-term memory networks, etc., to monitor changes in the pressure applied by the stamping head. For visual recognition systems, an appearance anomaly recognition channel architecture can be designed based on convolutional neural networks, to identify external damage or anomalies based on the appearance image of the stamping head. Subsequently, based on the design parameters, dimensions, material, and other information in the stamping head model data, a standard stamping head model is constructed using 3D modeling software (such as CAD). This standard stamping head model represents the stamping head under ideal conditions and serves as a benchmark for normal operation. It is used in the subsequent anomaly recognition training process to ensure that the normal operating conditions of different stamping head models can be identified. Subsequently, the collected abnormal stamping head displacement data is input into the displacement anomaly identification channel architecture, the collected abnormal stamping head pressure data is input into the pressure anomaly identification channel architecture, and the collected abnormal stamping head appearance images and standard stamping head models are input into the displacement anomaly identification channel architecture. Through iterative training, three channels for identifying stamping head anomalies—displacement, pressure, and appearance—are constructed. Finally, these three channels are connected in parallel to form a complete three-channel system for stamping head anomaly identification. These channels work in parallel to jointly detect the working status of the stamping head, providing comprehensive anomaly identification information. This information serves as a basis for subsequent working status prediction and automatic replacement control, ensuring the smooth operation of the production line.

[0026] Optionally, during channel training, the following datasets are first collected from historical detection records in the sensor and vision recognition system: a working dataset of abnormal stamping head displacement, a working dataset of abnormal stamping head pressure, and a set of images showing abnormal stamping head appearance. The working dataset of abnormal stamping head displacement includes displacement data caused by excessive stamping head offset or deformation, covering displacement performance under different working conditions. The working dataset of abnormal stamping head pressure includes data under normal operating pressure range and abnormal pressure states (such as excessively high or low pressure), reflecting whether the stamping head is malfunctioning due to uneven pressure or overload. The set of images showing abnormal stamping head appearance includes normal and abnormal appearance images of the stamping head (such as cracks, wear, corrosion, etc.), revealing any abnormalities in the stamping head's appearance. Subsequently, the collected working datasets of abnormal stamping head displacement and abnormal stamping head pressure are labeled with anomaly type (such as lateral offset, longitudinal offset, pressure overload, insufficient pressure, etc.) and anomaly degree (such as mild, moderate, severe, extreme), forming a working sample set of abnormal stamping head displacement. Subsequently, the labeled working datasets of stamping head displacement anomalies and stamping head pressure anomalies are input into the displacement anomaly recognition channel architecture and pressure anomaly recognition channel architecture, respectively, for iterative training. Taking a Long Short-Term Memory (LSTM) network as an example, displacement or pressure data is used as input data, and the corresponding anomaly type and degree are used as output data. The LSM network is forward-propagated, and the model accuracy is optimized by adjusting the number of network layers (3 LSTM layers + fully connected layers), learning rate (0.001), and loss function (cross-entropy loss), thus obtaining the final stamping head displacement anomaly recognition channel and stamping head pressure anomaly recognition channel. In addition, features are extracted from the standard stamping head model and the stamping head appearance anomaly image set, and the extracted features are used to iteratively train the appearance anomaly recognition channel architecture, thus obtaining the stamping head appearance anomaly recognition channel. Through the above steps, anomaly recognition channels in three dimensions—displacement, pressure, and appearance—can be constructed to comprehensively monitor the working status of the stamping head and promptly identify displacement anomalies, pressure anomalies, and appearance anomalies, providing support for subsequent automatic stamping head replacement control.

[0027] Furthermore, this application provides the aforementioned channel for identifying abnormalities in the appearance of the stamping head, including:

[0028] Appearance features are extracted from the standard stamping head model and the set of stamping head appearance anomalies, respectively, to obtain a standard stamping head appearance feature set and an abnormal stamping head appearance feature set; loss calculation is performed on the standard stamping head appearance feature set and the abnormal stamping head appearance feature set to obtain abnormal stamping head loss feature data; the abnormal stamping head loss feature data is labeled with anomaly type and degree to obtain a stamping head appearance anomaly sample set; anomaly recognition training is performed based on the stamping head appearance anomaly sample set to obtain the stamping head appearance anomaly recognition channel.

[0029] Optionally, the standard stamping head model is first rendered under various preset lighting conditions and from multiple perspectives (such as frontal view, side view, and tilt angle) to generate multiple standard stamping head images. These images are then input into a pre-trained deep convolutional neural network (such as ResNet or VGG) to extract high-dimensional feature vectors before the last convolutional or fully connected layer. These extracted high-dimensional feature vectors are then integrated to form a standard stamping head appearance feature set. This set defines a standard feature space, representing the range within which the appearance of the stamping head should fall under all normal conditions. Similarly, an abnormal stamping head appearance image set is also input into a deep convolutional neural network to obtain an abnormal stamping head appearance feature set. Besides deep convolutional neural networks, conventional feature extraction algorithms, such as edge detection, texture analysis, and shape matching, can be used to process the standard and abnormal stamping head images to obtain contour features, surface texture, color changes, geometric shapes, and other features, thus constructing the required standard and abnormal stamping head appearance feature sets. Subsequently, the designed feature loss function is used to calculate the difference between the standard stamping head appearance feature set and the abnormal stamping head appearance feature set, assessing the degree of abnormality in the stamping head appearance, thereby generating loss feature data of the stamping head appearance. This abnormal stamping head loss feature data reflects the changes between the abnormal stamping head appearance features and the standard stamping head appearance features. Then, based on the abnormality type (e.g., cracks, wear) and degree of abnormality in the stamping head appearance images, the loss feature data corresponding to the abnormal stamping head appearance images are labeled, constructing a stamping head appearance abnormality sample set. This sample set is then input into the appearance abnormality recognition channel architecture. Iterative training is performed through forward propagation, loss calculation, backpropagation, and parameter optimization. By continuously learning and adjusting parameters, various abnormal features and their severity in the appearance images are identified, resulting in a highly efficient stamping head appearance abnormality recognition channel. This channel can process stamping head appearance images in real time, automatically determine the existence of abnormalities, and accurately identify the type and degree of abnormalities, effectively improving the management and maintenance efficiency of stamping heads.

[0030] Furthermore, this application provides the aforementioned data on obtaining abnormal stamping head loss characteristics, including:

[0031] Based on the standard stamping head appearance feature set, the key operating points of the stamping head are determined; the standard stamping head appearance feature set and the abnormal stamping head appearance feature set are compared and analyzed to obtain the abnormal stamping head appearance loss feature set; based on the abnormal stamping head appearance loss feature set and the spatial distance set of the key operating points of the stamping head, a feature loss function is designed; based on the feature loss function, the loss of the abnormal stamping head appearance loss feature set is calculated to obtain the abnormal stamping head loss feature data.

[0032] Optionally, firstly, based on the preset appearance feature range of key working points (such as the edge of the contact surface between the stamping head and the die, threaded connections, and thin-walled transition areas), the appearance feature set of the standard stamping head is matched to determine the key working points of the stamping head. These key working points represent the geometric areas that have the greatest impact on stamping performance. Subsequently, for each feature point in both the standard and abnormal stamping head appearance feature sets, feature points are matched using Fast Nearest Neighbor (FLANN) search, and mismatched points are eliminated using a random sampling consensus algorithm, retaining valid matching pairs. For successfully matched feature point pairs, the distance deviation between the feature point pairs is calculated using Euclidean distance to identify differences in shape, texture, cracks, wear, etc., thus constructing the appearance loss feature set of the abnormal stamping head. Next, Euclidean distance is used to calculate the distance between corresponding feature points and key operating points of the stamping head in the abnormal stamping head appearance loss feature set, constructing a spatial distance set. Based on this spatial distance set, a weight for each feature point is defined using a Gaussian decay model. A feature loss function is then designed based on these weights and the abnormal stamping head appearance loss feature set. This feature loss function is a weighted summation formula of the abnormal stamping head appearance loss features based on the feature point weights. Then, the designed feature loss function is used to calculate the loss on the abnormal stamping head appearance loss feature set. That is, each feature point is weighted and fused to obtain abnormal stamping head loss feature data containing all damage features, used to represent the appearance loss of the stamping head, providing a basis for subsequent fault prediction, working condition assessment, and stamping head replacement mechanism.

[0033] Based on the usage data of the stamping head and the multidimensional abnormal feature set of the stamping head, the working state is predicted to obtain the predicted working state of the stamping head, and the stamping head replacement mechanism is triggered by the predicted working state of the stamping head for automatic replacement control.

[0034] In one embodiment, after obtaining the multidimensional anomaly feature set of the stamping head, the feature set is weighted and fused to determine the anomaly severity level. If this anomaly severity level does not reach the anomaly severity threshold, the stamping head's operating status is predicted using the stamping head usage data and this anomaly severity level. This predictive operating status provides a basis for determining whether the stamping head is within its normal operating range and assesses the potential for failure. If the predicted operating status indicates that the replacement conditions are met, the stamping head replacement mechanism is automatically activated based on the stamping head replacement mechanism, while simultaneously notifying the operator and reducing manual intervention. Through this method, the system can replace the stamping head in a timely and efficient manner, avoiding production stoppages caused by stamping head damage and ensuring the stable operation and safety of the production line.

[0035] Furthermore, this application provides the method for obtaining the predicted working state of the stamping head, including:

[0036] The multidimensional abnormal feature set of the stamping head is weighted and fused and the degree of abnormality is evaluated to determine the level of abnormality in the operation of the stamping head. If the level of abnormality in the operation of the stamping head does not reach the preset abnormality threshold, the working state is predicted based on the usage data of the stamping head and the level of abnormality in the operation of the stamping head to obtain the predicted working state of the stamping head.

[0037] Preferably, for the obtained multidimensional abnormal feature set of the stamping head, a weighting factor is used to weight each abnormal feature (displacement, pressure, appearance, etc.), and the weighting result is quantified using a stamping head abnormality level system to determine the level of abnormality in the stamping head's operation. If this level of abnormality in the stamping head's operation does not reach the preset abnormality threshold, it indicates that the stamping head is still in a normal or slightly abnormal state. At this time, the usage data of the stamping head and the level of abnormality in the stamping head's operation are combined, and an LSTM network is used to predict the future working state of the stamping head, obtaining the predicted working state of the stamping head. This predicted working state of the stamping head provides a quantitative result about the health status of the stamping head, which can reflect whether the stamping head can continue to operate safely, whether maintenance or replacement is required, and helps to identify potential problems of the stamping head in advance and take corresponding measures to ensure the stability and safety of production.

[0038] Furthermore, this application provides the method for determining the degree of abnormality in the operation of the stamping head, including:

[0039] An influence degree analysis is performed on each dimension of the multidimensional abnormal feature set of the stamping head to determine the influence weight factor of the abnormal dimension; the influence weight factor of the abnormal dimension is used to perform weighted fusion of the multidimensional abnormal feature set of the stamping head to generate the stamping head abnormal fusion feature set; a stamping head abnormality level system is defined to quantitatively evaluate the abnormality degree of the stamping head abnormal fusion feature set and determine the abnormality degree level of the stamping head operation.

[0040] Optionally, based on the reserved key name for each feature, dimensional features such as displacement, pressure, and appearance features are identified from the multidimensional anomaly feature set of the stamping head. Subsequently, based on domain experts and actual business needs, the influence degree of each feature is analyzed to determine the anomaly dimension influence weight factor for each feature. In addition, the correlation between each feature and the anomaly degree of the stamping head can be calculated using the Pearson correlation coefficient based on the historical data corresponding to each feature, and the anomaly dimension influence weight factor for each feature can be set accordingly. Then, the anomaly features in the multidimensional anomaly feature set of the stamping head are weighted and fused using the determined anomaly dimension influence weight factor to form a comprehensive stamping head anomaly fusion feature set, representing the overall working anomaly degree of the stamping head. It should be noted that before all weighting, maximum-minimum value normalization and z-score methods are used to ensure that these features are all under the same dimension. Then, the abnormal fusion feature set of the stamping head is compared with the predefined abnormality level system of the stamping head to determine which level range of the abnormality fusion feature of the stamping head falls into, thereby matching a level of abnormality of the stamping head operation. This level is used to quantify the abnormality of the current stamping head, reflect the current health status of the stamping head, and provide support for subsequent production decisions and automatic replacement control.

[0041] Furthermore, this application provides the method for obtaining the predicted working state of the stamping head, including:

[0042] An LSTM network is used to standardize and predict the working status of the stamping head usage data to obtain a basic working prediction status. Based on the level of abnormality in the working of the stamping head, an anomalous influence enhancement is applied to the basic working prediction status to obtain the working prediction status of the stamping head.

[0043] Optionally, when predicting the working status, the stamping head usage data is first standardized. Standardization aims to map all data to a uniform scale, typically by converting the data to zero mean and unit variance, or by using maximum-minimum normalization. After standardization, the standardized stamping head usage data is input into a pre-trained Long Short-Term Memory (LSTM) network. This LSTM network has been iteratively trained through forward propagation, loss calculation, backpropagation, and parameter optimization. It can predict a basic working status based on the received stamping head usage data, using patterns learned from historical data. This basic working status reflects the health score of the stamping head without considering abnormal influences. Subsequently, a score deduction item is matched according to the level of stamping head abnormality. This score deduction item is then used to enhance the abnormal influence on the basic working status; that is, the basic working status is subtracted from the score deduction item, making the prediction result more accurately reflect the actual working status of the stamping head under the current abnormal condition. This yields the final stamping head working status prediction, helping to provide early warning of potential faults or performance degradation, thereby improving the reliability and efficiency of the production line.

[0044] Furthermore, this application provides an automatic replacement control mechanism triggered by the predicted working state of the stamping head, including:

[0045] When the predicted working state of the stamping head reaches the preset replacement threshold, the stamping head replacement mechanism is triggered to start the stamping head replacement mechanism; based on the stamping head replacement mechanism, the stamping head of the target stamping die is automatically replaced.

[0046] Preferably, when the health status score in the predicted working state of the stamping head is less than or equal to the replacement threshold, it indicates that the current working state of the stamping head is approaching the replacement condition. At this time, a warning signal is triggered. This warning signal is used to activate the stamping head replacement mechanism and provides relevant information about the stamping head, such as its model, location, and current working status. After receiving the warning signal, the stamping head replacement mechanism will activate the corresponding stamping head replacement mechanism based on the received relevant information and begin to perform the replacement operation. During this process, the existing stamping head will be located and disassembled by a robotic arm or automated equipment. After disassembly, the new stamping head will be automatically removed and precisely installed on the mold, restoring the production line to normal operation and improving production efficiency and safety.

[0047] In summary, the embodiments of this application have at least the following technical effects:

[0048] This embodiment first constructs a multi-dimensional sensing network, which integrates a laser displacement sensor, a pressure sensor, and a visual recognition system. The multi-dimensional working data stream of the stamping head of the target stamping die is collected through this network. Then, an RFID tag is embedded in the stamping head of the target stamping die, and the stamping head model data and usage data are read through the RFID tag. Next, based on the multi-dimensional sensing network and the stamping head model data, a three-channel system for stamping head anomaly identification is established. This system is used to map and identify the multi-dimensional working data stream of the stamping head, outputting a multi-dimensional anomaly feature set. Finally, based on the stamping head usage data and the multi-dimensional anomaly feature set, the working state is predicted to obtain the predicted working state of the stamping head. The predicted working state triggers an automatic replacement mechanism for the stamping head. These technologies collectively solve the technical problem that traditional stamping dies rely on manual experience to judge the timing of stamping head replacement, leading to untimely or misjudged replacements that affect production efficiency and equipment lifespan. They achieve the technical effect of real-time and accurate monitoring and prediction of the working status of stamping heads through multi-dimensional perception and data fusion, improving the timeliness and accuracy of automatic replacement, reducing equipment failure rate, and ensuring production continuity.

[0049] Example 2, based on the same inventive concept as the automatic changing control method for the stamping head of a stamping die in the foregoing examples, such as... Figure 2 As shown, this application provides an automatic replacement control system for the stamping head of a stamping die. The system includes: a working data acquisition module 11: constructing a multi-dimensional sensing network, which integrates a laser displacement sensor, a pressure sensor, and a visual recognition system, and acquiring multi-dimensional working data streams of the stamping head of the target stamping die through the multi-dimensional sensing network; an RFID tag reading module 12: embedding RFID tags on the stamping head of the target stamping die, and reading stamping head model data and stamping head usage data through the RFID tags; a mapping and identification module 13: building a three-channel stamping head anomaly identification based on the multi-dimensional sensing network and the stamping head model data, using the three-channel stamping head anomaly identification to map and identify the multi-dimensional working data stream of the stamping head, and outputting a multi-dimensional anomaly feature set of the stamping head; and a replacement control module 14: predicting the working state based on the stamping head usage data and the multi-dimensional anomaly feature set of the stamping head, obtaining the predicted working state of the stamping head, and triggering a stamping head replacement mechanism for automatic replacement control based on the predicted working state of the stamping head.

[0050] Furthermore, the mapping and recognition module 13 is also used to perform the following method:

[0051] Based on the multidimensional sensing network, a stamping head anomaly identification channel architecture is designed, where each channel in the stamping head anomaly identification channel architecture corresponds one-to-one with each sensor type in the multidimensional sensing network; a standard stamping head model is obtained by performing three-dimensional modeling based on the stamping head model data; the stamping head anomaly identification channel architecture is trained for anomaly identification using the standard stamping head model to obtain a stamping head displacement anomaly identification channel, a stamping head pressure anomaly identification channel, and a stamping head appearance anomaly identification channel; the stamping head displacement anomaly identification channel, the stamping head pressure anomaly identification channel, and the stamping head appearance anomaly identification channel are connected in parallel to build a three-channel stamping head anomaly identification system.

[0052] Furthermore, the mapping and recognition module 13 is also used to perform the following method:

[0053] A dataset of abnormal stamping head displacement, a dataset of abnormal stamping head pressure, and a dataset of abnormal stamping head appearance images are collected. The abnormality types and degrees of the datasets are identified to obtain a sample set of abnormal stamping head displacement and a sample set of abnormal stamping head pressure. Anomaly recognition training is performed based on these sample sets to obtain anomaly recognition channels for stamping head displacement and pressure. Finally, anomaly recognition training is performed on the dataset of abnormal stamping head appearance images using the standard stamping head model to obtain anomaly recognition channel for stamping head appearance.

[0054] Furthermore, the mapping and recognition module 13 is also used to perform the following method:

[0055] Appearance features are extracted from the standard stamping head model and the set of stamping head appearance anomalies, respectively, to obtain a standard stamping head appearance feature set and an abnormal stamping head appearance feature set; loss calculation is performed on the standard stamping head appearance feature set and the abnormal stamping head appearance feature set to obtain abnormal stamping head loss feature data; the abnormal stamping head loss feature data is labeled with anomaly type and degree to obtain a stamping head appearance anomaly sample set; anomaly recognition training is performed based on the stamping head appearance anomaly sample set to obtain the stamping head appearance anomaly recognition channel.

[0056] Furthermore, the mapping and recognition module 13 is also used to perform the following method:

[0057] Based on the standard stamping head appearance feature set, the key operating points of the stamping head are determined; the standard stamping head appearance feature set and the abnormal stamping head appearance feature set are compared and analyzed to obtain the abnormal stamping head appearance loss feature set; based on the abnormal stamping head appearance loss feature set and the spatial distance set of the key operating points of the stamping head, a feature loss function is designed; based on the feature loss function, the loss of the abnormal stamping head appearance loss feature set is calculated to obtain the abnormal stamping head loss feature data.

[0058] Furthermore, the replacement control module 14 is also used to perform the following method:

[0059] The multidimensional abnormal feature set of the stamping head is weighted and fused and the degree of abnormality is evaluated to determine the level of abnormality in the operation of the stamping head. If the level of abnormality in the operation of the stamping head does not reach the preset abnormality threshold, the working state is predicted based on the usage data of the stamping head and the level of abnormality in the operation of the stamping head to obtain the predicted working state of the stamping head.

[0060] Furthermore, the replacement control module 14 is also used to perform the following method:

[0061] An influence degree analysis is performed on each dimension of the multidimensional abnormal feature set of the stamping head to determine the influence weight factor of the abnormal dimension; the influence weight factor of the abnormal dimension is used to perform weighted fusion of the multidimensional abnormal feature set of the stamping head to generate the stamping head abnormal fusion feature set; a stamping head abnormality level system is defined to quantitatively evaluate the abnormality degree of the stamping head abnormal fusion feature set and determine the abnormality degree level of the stamping head operation.

[0062] Furthermore, the replacement control module 14 is also used to perform the following method:

[0063] An LSTM network is used to standardize and predict the working status of the stamping head usage data to obtain a basic working prediction status. Based on the level of abnormality in the working of the stamping head, an anomalous influence enhancement is applied to the basic working prediction status to obtain the working prediction status of the stamping head.

[0064] Furthermore, the replacement control module 14 is also used to perform the following method:

[0065] When the predicted working state of the stamping head reaches the preset replacement threshold, the stamping head replacement mechanism is triggered to start the stamping head replacement mechanism; based on the stamping head replacement mechanism, the stamping head of the target stamping die is automatically replaced.

[0066] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0067] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0068] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A punch head automatic replacement control method of a press die, characterized by, The method comprises: constructing a multi-dimensional perception network integrating a laser displacement sensor, a pressure sensor and a visual recognition system, collecting a multi-dimensional working data stream of a punch head of a target stamping die through the multi-dimensional perception network; embedding an RFID tag on the punch head of the target stamping die, reading punch head model data and punch head use data through the RFID tag; building a punch head abnormality identification three-channel according to the multi-dimensional perception network combined with the punch head model data, mapping and identifying the punch head multi-dimensional working data stream through the punch head abnormality identification three-channel, and outputting a punch head multi-dimensional abnormality feature set; based on the punch head use data and the punch head multi-dimensional abnormality feature set, predicting the working state to obtain a punch head working prediction state, and triggering a punch head replacement mechanism through the punch head working prediction state to control automatic replacement; the punch head abnormality identification three-channel comprises: designing a punch head abnormality identification channel architecture according to the multi-dimensional perception network, each channel in the punch head abnormality identification channel architecture corresponding to each sensor type in the multi-dimensional perception network; based on the punch head model data, a three-dimensional model is established to obtain a standard punch head model; collecting a punch head displacement abnormal working data set, a punch head pressure abnormal working data set and a punch head appearance abnormal image set; respectively identifying the types and degrees of abnormality of the punch head displacement abnormal working data set and the punch head pressure abnormal working data set to obtain a punch head displacement abnormal working sample set and a punch head pressure abnormal working sample set; based on the punch head displacement abnormal working sample set and the punch head pressure abnormal working sample set, abnormality identification training is performed to obtain a punch head displacement abnormality identification channel and a punch head pressure abnormality identification channel; combined with the standard punch head model, the punch head appearance abnormal image set is subjected to abnormality identification training to obtain a punch head appearance abnormality identification channel; the punch head displacement abnormality identification channel, the punch head pressure abnormality identification channel and the punch head appearance abnormality identification channel are connected in parallel to build a punch head abnormality identification three-channel; the punch head appearance abnormality identification channel comprises: respectively extracting the appearance features of the standard punch head model and the punch head appearance abnormal image set to obtain a standard punch head appearance feature set and an abnormal punch head appearance feature set; loss calculation is performed on the standard punch head appearance feature set and the abnormal punch head appearance feature set to obtain abnormal punch head loss feature data; the abnormal punch head loss feature data is subjected to abnormal type and degree identification to obtain a punch head appearance abnormal sample set; based on the punch head appearance abnormal sample set, abnormality identification training is performed to obtain the punch head appearance abnormality identification channel; the abnormal punch head loss feature data comprises: determining a punch head key working point according to the standard punch head appearance feature set; comparing and analyzing the standard punch head appearance feature set and the abnormal punch head appearance feature set to obtain an abnormal punch head appearance loss feature set; According to the abnormal stamping head appearance loss feature set and the spatial distance set of the stamping head key working point, a feature loss function is designed; Based on the feature loss function, loss calculation is performed on the abnormal stamping head appearance loss feature set to obtain abnormal stamping head loss feature data.

2. The method of claim 1, wherein the punch head is automatically replaced by the punch head replacement control means when the punch head is worn out. The obtained stamping head working prediction state comprises: The stamping head multi-dimensional abnormal feature set is weighted and fused and the abnormal degree is evaluated to determine the stamping head working abnormal degree level; If the stamping head working abnormal degree level does not reach a preset abnormal degree threshold, a working state is predicted based on the stamping head use data and the stamping head working abnormal degree level to obtain a stamping head working prediction state.

3. The method of claim 2, wherein the punch head is automatically replaced by the punch head replacement control means when the punch head is worn out. The determination of the stamping head working abnormal degree level comprises: The influence degree of each dimension feature in the stamping head multi-dimensional abnormal feature set is analyzed to determine an abnormal dimension influence weight factor; The stamping head multi-dimensional abnormal feature set is weighted and fused using the abnormal dimension influence weight factor to generate a stamping head abnormal fusion feature set; An abnormal level system of the stamping head is defined to quantitatively evaluate the abnormal degree of the stamping head abnormal fusion feature set to determine the stamping head working abnormal degree level.

4. The method of claim 3, wherein the punch head is automatically replaced by the punch head replacement control means when the punch head is worn out. The obtained stamping head working prediction state comprises: An LSTM network is used to standardize the stamping head use data and predict the working state to obtain a basic working prediction state; Based on the stamping head working abnormal degree level, the abnormal influence of the basic working prediction state is enhanced to obtain the stamping head working prediction state.

5. The method of claim 1, wherein the punch head is automatically replaced by the punch head replacement control means when the punch head is worn out. The automatic replacement control of the stamping head replacement mechanism triggered by the stamping head working prediction state comprises: When the stamping head working prediction state reaches a preset replacement threshold, the stamping head replacement mechanism is triggered to start the stamping head replacement mechanism; Based on the stamping head replacement mechanism, the stamping head of the target stamping die is automatically replaced and controlled.

6. A punch head automatic replacement control system of a press die, characterized by, The stamping head automatic replacement control system is used to execute the stamping head automatic replacement control method of the stamping die according to any one of claims 1-5, and the stamping head automatic replacement control system comprises: A working data acquisition module: a multi-dimensional perception network is constructed, the multi-dimensional perception network integrates a laser displacement sensor, a pressure sensor and a visual recognition system, and the multi-dimensional working data stream of the stamping head of the target stamping die is acquired through the multi-dimensional perception network; An RFID tag reading module: an RFID tag is embedded and installed on the stamping head of the target stamping die, and the stamping head model data and the stamping head use data are read through the RFID tag; A mapping recognition module: according to the multi-dimensional perception network combined with the stamping head model data, a stamping head abnormal identification three-channel is built, the stamping head multi-dimensional working data stream is mapped and recognized through the stamping head abnormal identification three-channel, and a stamping head multi-dimensional abnormal feature set is output; A replacement control module: based on the stamping head use data and the stamping head multi-dimensional abnormal feature set, a working state is predicted to obtain a stamping head working prediction state, and the stamping head working prediction state is used to trigger the stamping head replacement mechanism for automatic replacement control.

Images