Intelligent control method and system for die-cutting machine

Abstract

The invention discloses an intelligent control method and system of a die-cutting machine, and relates to the technical field of intelligent control, a plurality of die-cutting contents are determined based on identification of die-cutting events, and a die-cutting optimization mode of a die-cutting node is determined according to the plurality of die-cutting contents, a real-time image of a die-cutting area and a node importance level of the die-cutting node. And the intelligent control system of the die-cutting machine is determined according to the die-cutting optimization mode of each die-cutting node, the working life of the die-cutting machine and the to-be-die-cut piece, so that the accuracy of the intelligent control system of the die-cutting machine is improved. Meanwhile, multiple optimization characteristics are determined according to detection of the dynamic optimization event, the working mode of the die-cutting machine is determined according to the multiple optimization characteristics, the working efficiency of the die-cutting machine and the corresponding die-cutting progress, and the die-cutting path of the die-cutting machine is dynamically regulated and controlled along with the form change of the part to be subjected to die-cutting; the overall consideration of a plurality of optimization characteristics, the working efficiency of the die-cutting machine and the corresponding die-cutting task is realized, and the accuracy of the working mode of the die-cutting machine is improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology, and in particular to an intelligent control method and system for a die-cutting machine. Background Technology

[0002] With the development of technology, die-cutting machines have been gradually applied in the industrial field. They can accurately cut various sheet materials according to preset graphics. The die-cutting machine performs die-cutting work on the parts to be die-cut and realizes the segmentation of the parts to be die-cut. In the existing technology, multiple working data of the die-cutting machine are collected, the corresponding working status is determined based on the multiple working data of the die-cutting machine, and the die-cutting work of the parts to be die-cut is monitored in real time. The die-cutting machine works along a predetermined route according to preset parameters. However, the die-cutting optimization method of each die-cutting node and the working life of the die-cutting machine are ignored, resulting in low accuracy of the intelligent control system of the die-cutting machine. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the prior art. This invention provides an intelligent control method and system for a die-cutting machine.

[0004] This invention provides an intelligent control method for a die-cutting machine, comprising: During the operation of the die-cutting machine, the corresponding working state is determined based on multiple working data of the die-cutting machine. The die-cutting path of the die-cutting machine is determined according to the working state of the die-cutting machine, the corresponding die-cutting task, and the workpiece to be die-cut. The die-cutting path is a collection of motion instructions and process parameters. Multiple die-cutting nodes are determined based on the identification of the die-cutting path. The die-cutting events corresponding to each die-cutting node are determined according to the node position of each die-cutting node, the corresponding die-cutting requirements, and the die-cutting area corresponding to the part to be die-cut. In each die-cutting node, multiple die-cutting contents are determined based on the identification of die-cutting events. The die-cutting optimization method for the die-cutting node is determined based on the multiple die-cutting contents, the real-time image of the die-cutting area, and the node importance level of the die-cutting node. The intelligent control system of the die-cutting machine is determined based on the die-cutting optimization method of each die-cutting node, the working years of the die-cutting machine, and the workpiece to be die-cut. The intelligent control system includes a set of basic process parameters for the current workpiece to be die-cut, a set of special instructions for specific nodes, a set of general rules extracted from node optimization, and global correction information based on the working years. Based on the identification of this intelligent control system, multiple sub-intelligent control items are identified, and the dynamic optimization events of the die-cutting machine are determined according to the control content of each sub-intelligent control item, the corresponding item priority, and the dynamic working data of the die-cutting machine. Multiple optimization features are determined based on the detection of dynamic optimization events. The working mode of the die-cutting machine is determined based on the multiple optimization features, the working efficiency of the die-cutting machine, and the corresponding die-cutting progress. The die-cutting path of the die-cutting machine is dynamically adjusted as the shape of the part to be die-cut changes.

[0005] This invention provides an intelligent control system for a die-cutting machine, which is applied to the aforementioned intelligent control method for the die-cutting machine.

[0006] Compared with the prior art, the beneficial effects of the present invention are: (1) Based on the identification of die-cutting events, multiple die-cutting contents are determined. The die-cutting optimization method of the die-cutting node is determined according to the multiple die-cutting contents, the real-time image of the die-cutting area and the node importance level of the die-cutting node. The intelligent control system of the die-cutting machine is determined according to the die-cutting optimization method of each die-cutting node, the working years of the die-cutting machine and the die-cutting parts. The die-cutting optimization method of the die-cutting node is controlled, which improves the accuracy of the intelligent control system of the die-cutting machine.

[0007] (2) Multiple optimization features are determined based on the detection of dynamic optimization events. The working mode of the die-cutting machine is determined based on the multiple optimization features, the working efficiency of the die-cutting machine and the corresponding die-cutting progress. The die-cutting path of the die-cutting machine is dynamically adjusted as the shape of the die-cutting part changes. This realizes the overall consideration of multiple optimization features, the working efficiency of the die-cutting machine and the corresponding die-cutting task, and improves the accuracy of the working mode of the die-cutting machine. Attached Figure Description

[0008] Figure 1 This is a flowchart illustrating the intelligent control method for a die-cutting machine in an embodiment of the present invention. Figure 2 This is a flowchart illustrating step S11 in the intelligent control method of the die-cutting machine in this embodiment of the invention. Figure 3 This is a flowchart illustrating step S12 in the intelligent control method of the die-cutting machine in this embodiment of the invention. Figure 4 This is a flowchart illustrating step S13 in the intelligent control method of the die-cutting machine in this embodiment of the invention. Figure 5 This is a flowchart illustrating step S14 of the intelligent control method for a die-cutting machine in an embodiment of the present invention. Figure 6 This is a flowchart illustrating step S15 in the intelligent control method of the die-cutting machine in this embodiment of the invention. Figure 7 This is a schematic diagram of the structure of the intelligent control system of the die-cutting machine in an embodiment of the present invention. Detailed Implementation

[0009] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0010] Please see Figures 1 to 7 A smart control method for a die-cutting machine, applied to intelligent control scenarios; the smart control method for the die-cutting machine includes: Step S11: During the operation of the die-cutting machine, the corresponding working status is determined based on multiple working data of the die-cutting machine, and the die-cutting path of the die-cutting machine is determined according to the working status of the die-cutting machine, the corresponding die-cutting task and the workpiece to be die-cut. Step S12: Based on the identification of the die-cutting path, determine multiple die-cutting nodes, and determine the die-cutting events corresponding to each die-cutting node according to the node position of each die-cutting node, the corresponding die-cutting requirements, and the die-cutting area corresponding to the part to be die-cut. Step S13: In each die-cutting node, multiple die-cutting contents are determined based on the identification of die-cutting events. The die-cutting optimization method of the die-cutting node is determined according to the multiple die-cutting contents, the real-time image of the die-cutting area and the node importance level of the die-cutting node. The intelligent control system of the die-cutting machine is determined according to the die-cutting optimization method of each die-cutting node, the working years of the die-cutting machine and the workpiece to be die-cut. Step S14: Based on the identification of the intelligent control system, multiple sub-intelligent control items are determined, and the dynamic optimization events of the die-cutting machine are determined according to the control content of each sub-intelligent control item, the corresponding item priority, and the dynamic working data of the die-cutting machine. Step S15: Determine multiple optimization features based on the detection of dynamic optimization events, determine the working mode of the die-cutting machine based on the multiple optimization features, the working efficiency of the die-cutting machine and the corresponding die-cutting progress, and dynamically adjust the die-cutting path of the die-cutting machine as the shape of the part to be die-cut changes.

[0011] refer to Figure 2 In step S11, the specific steps are as follows: S111: Collect the working signal of the die-cutting machine, and trigger the working process of the die-cutting machine based on the working signal. Determine multiple working data of the die-cutting machine based on the detection of the die-cutting machine, and determine the working status of the die-cutting machine based on the multiple working data and the dynamic video of the die-cutting module of the die-cutting machine. S112: Determine the die-cutting task list of the die-cutting machine based on the detection of the die-cutting machine, determine multiple die-cutting tasks based on the identification of the die-cutting task list, and determine the die-cutting path of the die-cutting machine based on the multiple die-cutting tasks, the information of the part to be die-cut, and the working status of the die-cutting machine.

[0012] In the embodiments of this application, various types of physical signals are acquired in real time. These signals include, but are not limited to: timing signals, such as pulse signals from the spindle encoder, used to accurately determine the angular position and operating phase of the die-cutting machine; pressure signals, such as pressure sensors installed in the die-cutting pressure cylinder or air circuit, used to monitor the peak value, steady-state value, and fluctuation of the die-cutting pressure in real time; speed signals, such as speed data fed back by the servo motor driver, reflecting the operating speed of the equipment; level signals, such as the material stack height sensor in the feeder section or the stack height sensor in the paper receiving section; and temperature / vibration signals, used to monitor the operating status of key components (such as bearings and motors) as an auxiliary basis for judging the health of the equipment.

[0013] The system does not continuously perform high-intensity calculations, but instead uses a combination of one or more key signals to activate the core control algorithm. For example, when the spindle encoder signal reaches the preset zero phase and the feeder sensor detects the entry of a new die-cut part, the system determines that a valid work cycle has begun, triggering subsequent data acquisition and status evaluation processes. The purpose is to avoid invalid calculations, concentrate computing resources on the actual production process, and improve the real-time performance and efficiency of the system response.

[0014] The system synchronously records all sensor data streams within a time window. After preprocessing (such as filtering, noise reduction, and normalization), these raw data form a set of structured working data vectors, such as D_work={Pressure,Speed,Phase,Vibration,…}. At the same time, the system continuously captures images of the key die-cutting area using an industrial high-speed camera. The video stream needs to undergo image preprocessing (including distortion correction, contrast enhancement, and spatial alignment) and region segmentation (defining multiple regions of interest (ROIs) in the image, such as the die-cutting plate area, the feeding area of ​​the part to be die-cut, and the output area of ​​the finished product). The most critical step is multi-source data fusion and working state determination, which associates the non-visual D_work with visual information to determine a comprehensive working state.

[0015] Ensure precise alignment between the timestamps of sensor data and video frames; for example, the pressure peak must correspond precisely to the frame at the moment of die-cutting. Perform feature extraction and association: extract features such as average pressure, pressure fluctuation variance, and speed stability from D_work. From the dynamic video, use computer vision algorithms to extract the positioning features of the part to be die-cut (determine the precise coordinates, rotation angle, and contour deformation of the part to be die-cut in the image through edge detection, template matching, or deep learning target detection algorithms), die-cutting quality features (analyze the edge quality of the die-cut area, whether there are edge bursts, and whether the waste material is completely separated, etc., through image difference and defect detection algorithms), and equipment status visual features (detect whether there are unexpected small displacements in the die-cutting plate or key components through optical flow or template matching).

[0016] The extracted multimodal features are input into a state recognition model, which can be a rule-based expert system or a machine learning model (such as SVM, random forest or shallow neural network). Its output is a quantized working state vector, where each element is a quantized value (such as between 0 and 1), which together describe the current working condition. Specifically, the die-cutting machine is processing a batch of high-precision electronic tags, which contain multiple small integrated circuit (IC) areas that require precise positioning. The die-cutting machine's feeder sensor detects the entry of new label paper, and at the same time, the spindle encoder signal reaches 270 degrees phase (the preset feeding stability point). This combined signal immediately triggers the intelligent control system, which begins to record all data for the next cycle.

[0017] The system synchronously acquired a pressure sensor reading of 50 Bar, a speed of 6000 sheets / hour, a real-time change in spindle phase, and a vibration sensor reading of 0.2g. In the dynamic video analysis stage, a high-speed camera installed directly above the die-cutting station captured video at a frame rate of 2000fps.

[0018] The video processing module analyzes each frame of the image in real time: through the template matching algorithm, it accurately finds the coordinates of the four corner points of the IC area on the label and compares them with the theoretical coordinates. It finds that the current label has a deviation of +0.15mm in the X direction, a deviation of -0.05mm in the Y direction, and a clockwise rotation of 0.1 degrees. At the same time, by analyzing the previous image of the die-cut label, the system finds that there are slight burrs on the die-cut edge of one of the IC corners.

[0019] The system fuses {Pressure:50Bar,Speed:6000s / h,…} with visual features {X_offset:+0.15mm,Y_offset:-0.05mm,Rotation:0.1°,Burr_Detected:True}. The state recognition model ultimately outputs a comprehensive working state vector: S_work={Normality:0.85,Position_Accuracy:0.70,Pressure_Stability:0.95,Quality_Score:0.88}. This state vector clearly indicates that the current equipment is operating relatively smoothly, but the positioning accuracy has deviated, and the quality of the previous finished product has slight defects. This precisely quantified S_work state vector will serve as a key input for planning the die-cutting path in the next step, enabling the system to fine-tune and compensate for the +0.15mm deviation in the next work cycle, and appropriately adjust the pressure parameters to prevent burrs from recurring.

[0020] Furthermore, the die-cutting task list is automatically obtained by the system in one or more ways, such as pulling detailed information of the current work order from the factory's MES system via API or database connection, or reading and parsing standard process files stored in a specific directory, such as CIP3 / CIP4 (PPF) files, Gerber files, or custom XML / JSON format work instruction sheets; for small-batch or non-standard production, operators can also manually input or select task parameters through HMIs such as touch screens.

[0021] A complete die-cutting task list is a structured data set that includes at least the material identifier (a unique ID or batch number of the part to be die-cut), the geometric layout (coordinates, dimensions, rotation angles, and types of all the graphics to be die-cut on the part), the process parameters (ideal die-cutting pressure, speed requirements, and accuracy level for each graphic), the quality standards (tolerance range, allowed defect types and levels for each graphic), and the task priority (defining the order of processing when multiple graphics conflict or compete for resources).

[0022] The system iterates through each geometric layout element in the task list. Each indentation line definition or cutting box definition is instantiated as a specific die-cutting task object. Each instantiated die-cutting task object is assigned a set of detailed attributes, forming a task feature vector T_i. This vector contains geometric attributes (such as {Center_X, Center_Y, Width, Height, Angle, Type}), process attributes (such as {Target_Pressure, Priority, Tolerance_X, Tolerance_Y}), and association attributes (such as {Parent_Material_ID, Graphic_ID}). All identified and instantiated die-cutting task objects together constitute an ordered task set {T_1, T_2, ..., T_n}, which is the direct input for the next step of path planning.

[0023] The path planning algorithm receives three key inputs: a task set {T_1,…,T_n} that defines all objectives and constraints that must be completed, including information about the die-cutting part such as material type, thickness, hardness, expansion rate, and current actual position and orientation; and the quantized working state S_work output by S111. This is typically a multi-objective optimization problem, aiming to find a path that satisfies all constraints and optimizes one or more performance indicators. The final output die-cutting path is not a simple sequence of points, but a set of motion commands and process parameters. The die-cutting path contains a sequence of path points {P_1,P_2,…,P_m} with target coordinates and arrival times, dynamic pressure and speed adjustment commands associated with each path point or task segment, and real-time compensation vectors for material deformation or equipment deviation.

[0024] Specifically, the die-cutting machine continued processing that batch of high-precision electronic tags; S111 output the working status: S_work={Normality:0.85,Position_Accuracy:0.70,Pressure_Stability:0.95,Quality_Score:0.88}, and detected that the current tag has an X-direction deviation of +0.15mm.

[0025] The die-cutting machine's control system retrieved the process document (PPF format) for work order #202511-TagA from the factory's MES system. After parsing the document, the system obtained a task list: the material is 0.1mm thick PET self-adhesive, with 5 IC chips that need to be half-cut, 1 outer frame that needs to be fully cut, and 2 positioning and crease lines.

[0026] The system breaks down the list into eight independent die-cutting tasks: T_IC1 to T_IC5 (type: half-cut), T_Frame (type: full-cut), T_Crease1, and T_Crease2 (type: indentation). Each task is assigned attributes. For example, the attributes of T_IC1 include: center coordinates (50,80) mm, size 5x5 mm, target pressure 30 N, and high priority (because IC is the core functional component).

[0027] The path planning module receives information on 8 task objects, PET material, and the critical S_work status. During the decision-making process, a compensation decision is made: since Position_Accuracy is 0.70 and an X deviation of +0.15mm is detected, the planning algorithm decides to apply a global compensation vector of (-0.15mm, 0) on the execution coordinates of all tasks. Then, a process adjustment decision is made: since Quality_Score is 0.88 (below 1.0) and burrs are detected in S111, the algorithm decides to fine-tune the target pressure of the T_Frame task from the default value of 40N to 38N to reduce stress concentration.

[0028] Calculate the optimal traversal order for all 8 tasks. Since T_IC1 to T_IC5 have high priority, prioritize planning a sub-path that can complete these 5 half-cut tasks with the shortest idle distance, and then perform indentation and the final outer frame cutting. The final generated die-cutting path is described as follows: move to the compensated coordinates (49.85, 80.00) mm of T_IC1, apply 30N pressure to perform half-cut; move to the compensated coordinates of T_IC2 using the optimal path, apply 30N pressure to perform half-cut; ...; move to the compensated coordinates of T_Frame, apply 38N pressure to perform full cut.

[0029] refer to Figure 3 In step S12, the specific steps are as follows: S121: Collect the die-cutting path and mark the path shape of the die-cutting path. Based on the path shape of the die-cutting path, the overall shape of the part to be die-cut, and the previous die-cutting content of the die-cutting machine, determine multiple die-cutting nodes and mark the node positions of multiple die-cutting nodes. S122: Among multiple die-cutting nodes, the corresponding die-cutting requirements are determined based on the node contact of each die-cutting node. The first layer of die-cutting information is determined based on the node position of each die-cutting node and the die-cutting area corresponding to the part to be die-cut. The second layer of die-cutting information is determined based on the die-cutting requirements of each die-cutting node and the die-cutting area corresponding to the part to be die-cut. The die-cutting event corresponding to each die-cutting node is determined based on the first layer of die-cutting information, the second layer of die-cutting information and the corresponding die-cutting node.

[0030] In the embodiments of this application, the system acquires die-cutting path data generated by S112, which consists of a series of high-precision coordinate points. This path usually exists in the form of a parametric form (such as a NURBS curve or a polyline) or a discrete point set. Then, the system enters the path morphology marking stage, which is a process of semanticizing the path. The system analyzes its geometric features through algorithms and describes them with labels.

[0031] Traverse every point or tiny line segment on the path, calculating the first derivative (direction vector) and second derivative (curvature vector) of the path; based on changes in geometric features, divide the continuous path into different morphological segments and assign semantic labels: when the curvature value is close to zero, it is labeled LINEAR; when the curvature value is constant or changes smoothly, it is labeled ARC or CURVE; when the path direction vector undergoes a sudden change (i.e., the curvature shows a peak), it is labeled CORNER, and can be further subdivided into ACUTE_CORNER, RIGHT_CORNER, and OBTUSE_CORNER based on the angle; the start and end positions of the path are labeled START and END; after this step, an original sequence of coordinate points is transformed into a meaningful sequence of morphological labels, for example: {START,LINEAR,ACUTE_CORNER,CURVE,RIGHT_CORNER,LINEAR,END}.

[0032] The system generates a candidate node at the boundary of each shape label, such as at the intersection of LINEAR and ACUTE_CORNER. For longer shape segments that exceed the length threshold, path nodes can be inserted in the middle for segmentation control. Secondly, node generation is based on the overall shape of the part to be die-cut. This strategy ensures that the node settings match the functional structure of the product. The system loads the CAD design model or high-resolution overall image of the part to be die-cut and accurately overlays the die-cutting path with the overall shape of the part through image registration or coordinate system transformation. It identifies key structural features on the overall shape, such as functional area boundaries, pre-fold lines, holes, and material splicing seams. When the die-cutting path intersects with or is close to these key structural features, the system will forcibly generate nodes at or near the intersection.

[0033] By incorporating machine learning and historical experience, the node settings become predictive. The system queries its built-in quality database, which records detailed information on all die-cutting defects in historical production. When the shape and material of the current path match a high-incidence defect scenario in the historical record, the system triggers the risk node generation rule to create a high-priority control node.

[0034] For each node determined by the above fusion strategy, the system will find its nearest coordinate point in the original die-cutting path dataset, and bind the coordinate point (usually the (x,y) coordinate in the die-cutting machine workbench coordinate system, and also the z coordinate if Z-axis floating is involved) with all the attributes of the node (such as node type, generation basis, associated risks, etc.) to form a complete node object; all marked node objects are arranged according to their physical order on the die-cutting path.

[0035] Specifically, the die-cutting path of the electronic tag to be processed by the die-cutting machine includes a square half-cut contour of 5 IC chips and a large outer frame full-cut contour. The system collects path data composed of a series of coordinate points. Through curvature analysis, the system marks the path as: {START,LINEAR,RIGHT_CORNER,LINEAR,RIGHT_CORNER,…} (repeated 4 times to form a square). It is a similar marking sequence of a long LINEAR connecting to the next square, and finally a large rectangle. The 5 small squares are marked as the high-risk SMALL_SQUARE_WITH_SHARP_CORNERS pattern.

[0036] The system generates basic candidate nodes at the four corner points (RIGHT_CORNER) and the start and end points of each side of each small square. Based on the shape of the part to be die-cut, the system loads the design drawing of the electronic tag and finds that each small square path precisely corresponds to the functional area of ​​an IC chip. Therefore, the system forcibly confirms the node at the four corner points of each square path and marks it as CRITICAL_FUNCTION_NODE. Most importantly, based on the decision of previous die-cut content, the system queries the historical database and finds a key record: on PET material, the probability of edge bursting at the right-angle inflection point of a square outline with a side length of less than 5mm is 25%. The current IC outline has a side length of 4mm, so the system additionally marks a HIGH_RISK_CORNER attribute at the 20 corner points of all 5 square outlines.

[0037] The system ultimately determined 29 nodes: 5 squares × 4 corner points + 4 outer frame corner points + other path connection points. Taking the first corner point of the first IC contour as an example, its node object is marked as: Node_ID: N_003; Position: (49.85, 82.00) mm (compensated coordinates); Type: RIGHT_CORNER; Generation_Source:{Path_Morphology,Part_Morphology}; Attribute:{CRITICAL_FUNCTION_NODE,HIGH_RISK_CORNER}; Associated_Risk: For small-sized PET right-angle edge bursting, through S121, the die-cutting machine no longer treats the path as a line, but as a sequence of 29 key control points, and each control point has a clear identity and risk warning.

[0038] Furthermore, in multiple die-cutting nodes, the corresponding die-cutting requirements are determined based on the node contact of each die-cutting node. The first layer of die-cutting information is determined based on the node position of each die-cutting node and the die-cutting area corresponding to the part to be die-cut. The second layer of die-cutting information is determined based on the die-cutting requirements of each die-cutting node and the die-cutting area corresponding to the part to be die-cut. The die-cutting events corresponding to each die-cutting node are determined based on the first layer of die-cutting information, the second layer of die-cutting information, and the corresponding die-cutting nodes. This approach takes into account the overall consideration of the first layer of die-cutting information, the second layer of die-cutting information, and the corresponding die-cutting nodes, ensuring the accuracy of the die-cutting events corresponding to each die-cutting node. At the same time, the die-cutting path of the die-cutting machine is introduced to control multiple die-cutting nodes, taking into account the node position of each die-cutting node, the corresponding die-cutting requirements, and the die-cutting area corresponding to the part to be die-cut, thereby improving the accuracy of the die-cutting events corresponding to each die-cutting node.

[0039] At this point, the system performs kinematic and dynamic simulation analysis on each die-cutting node. This requires combining the local geometric information of the die-cutting path and the kinematic model of the die-cutting machine. By analyzing the changes in the direction vectors of the path before and after the node, the system determines the contact type of the tool at the node, including the path start point, the cutting node where the tool changes from non-contact to contact, the path end point, the cutting node where the tool changes from contact to non-contact, and the continuous nodes in the middle of the path where the tool remains in contact.

[0040] For continuous nodes, the system will further analyze their motion state, such as uniform nodes located in the middle of a straight line segment with constant velocity and direction, turning nodes located at path inflection points with changing velocity vectors, and curvature change nodes located on curve segments with changing radius of curvature with calculated rate of change of acceleration.

[0041] Based on these analysis results, the system generates a set of customized process requirements for each node. These requirements are the direct targets for subsequent optimization. For example, the cut-in requirements include the pressure ramp rate for smooth and rapid pressure establishment and the limitation of impact velocity; the cut-out requirements include the pressure descent rate for smooth pressure release and the control of lift height; the steering requirements include inflection point speed limits, deceleration curves and necessary dwell time; and the curvature requirements include centripetal force limits and speed adaptive rules.

[0042] For the first layer of die-cutting information, the system uses coordinate transformation to map the physical location of the node onto the CAD model or functional zoning diagram of the part to be die-cut. Once a node falls into a predefined die-cutting area, the attributes of that area are associated with the node. The information includes the area ID, area type, material properties, and precision tolerance. The second layer of die-cutting information (process dimension information) is the combination of die-cutting requirements and specific materials. The system matches the determined die-cutting requirements with the material properties obtained from the first layer of information. Through the built-in process knowledge base or model, the abstract requirements are transformed into specific process parameters. The information includes target pressure, speed curve, tool offset, and vibration control level.

[0043] The system creates a die-cutting event object E_i for each die-cutting node, with the structure E_i={Node_i,Info_1_i,Info_2_i}. This object contains all the information required to perform the operation of that node: where (Node_i), what to operate on (Info_1_i), and how to operate (Info_2_i). All die-cutting events are arranged in their physical order on the path, forming a complete, parameterized execution sequence {E_1,E_2,…,E_k}. This sequence is the final output of step S12, which will serve as the basic unit for die-cutting optimization in step S13.

[0044] Specifically, the die-cutting machine has generated 29 die-cutting nodes. We focus on one high-risk node: N_003, which is the right-angle inflection point of the first IC profile and is marked as HIGH_RISK_CORNER. System analysis of N_003 reveals that it is a RIGHT_CORNER (right-angle turning node), where the tool needs to change from moving along the positive X-axis to moving along the negative Y-axis. Therefore, the die-cutting requirements generated by the system are: {Corner_Speed_Limit:0.3m / s,Deceleration_Profile:S-Curve,dwell_time:10ms}, requiring the speed to drop below 0.3m / s at the inflection point, decelerate smoothly in an S-curve, and pause at the center of the inflection point for 10ms to eliminate mechanical vibration.

[0045] The system determines the first layer of information: by mapping the coordinates of N_003 to the functional diagram of the electronic tag, it is determined that it is located in the IC1 chip area; therefore, Info_1 is: {Region_ID:IC1,Region_Type:Functional_Chip,Material:PET,Thickness:0.1mm,Precision_Tolerance:±20μm}. The system generates the second layer of information by combining the die-cutting requirements and material properties: the system queries the process knowledge base and finds that PET material is prone to whitening due to tool extrusion when turning at low speeds. Therefore, the die-cutting requirements are modified, generating Info_2: {Target_Pressure:28N (2N lower than the norm),Speed_Profile:0.28m / s (lower than the requirement),Tool_Offset:+5μm (slight compensation to prevent undercutting),Vibration_Control_Level:High}.

[0046] The die-cutting event E_003 corresponding to N_003 was finally determined as follows: the associated node is N_003at(49.87,82.00)mm; the spatial information is that the 0.1mm thick PET is operated in the IC1 chip area; the process instruction is to decelerate to 0.28m / s with an S-curve, apply a pressure of 28N, stay at the center of the inflection point for 10ms, offset the tool in the positive direction by 5μm, and activate high-level vibration suppression. Through S122, the die-cutting machine transforms a simple turning action at (49.87,82.00) into a complex event that includes precise speed, pressure, compensation, and vibration control.

[0047] refer to Figure 4 In step S13, the specific steps are as follows: S131: Real-time monitoring of each die-cutting node and marking of the die-cutting events corresponding to each die-cutting node. Multiple die-cutting contents are determined based on the identification of the die-cutting events. At the same time, the die-cutting machine performs visual inspection on the die-cutting area of ​​the part to be die-cut to collect real-time images of each die-cutting area. S132: Based on the comparison of each die-cutting node, determine the node importance level of each die-cutting node, determine the first optimization coefficient based on the node importance level of each die-cutting node and the real-time image of the corresponding die-cutting area, and determine the second optimization coefficient based on the node position of each die-cutting node and the real-time image of the corresponding die-cutting area. S133: Determine the die-cutting optimization method of the die-cutting node based on the mapping relationship between the first optimization coefficient, the second optimization coefficient, and the die-cutting optimization method; at the same time, collect the working years of the die-cutting machine, and determine the intelligent control system of the die-cutting machine based on the die-cutting optimization method of each die-cutting node, the working years of the die-cutting machine, and the workpiece to be die-cut.

[0048] In the embodiments of this application, each die-cutting node is monitored in real time, and the die-cutting events corresponding to each die-cutting node are marked. Multiple die-cutting contents are determined based on the identification of the die-cutting events. At the same time, the die-cutting machine performs visual inspection on the die-cutting area of ​​the part to be die-cut to collect real-time images of each die-cutting area. This takes into account the overall consideration of the identification of die-cutting events and ensures the accuracy of multiple die-cutting contents.

[0049] At this time, the system obtains real-time position feedback data from the motion controller or servo driver of the die-cutting machine through high-frequency (such as kHz level) bus communication, and runs a digital twin trajectory model synchronized with the physical die-cutting machine internally. This model receives real-time position feedback and continuously tracks the precise position of the current actuator on the preset die-cutting path. The system compares the real-time position coordinates with the ordered node sequence generated in S121 in real time. When the real-time position enters the predefined neighborhood of a node, the system marks the current time as the node time.

[0050] Once a node is marked, the system immediately triggers the event recognition process, queries the sequence of die-cutting events constructed in S122, finds the die-cutting event uniquely bound to the node, and parses the identified die-cutting event to extract its core die-cutting content. Essentially, this simplifies a complex parameterized object into a few key action-target pairs to facilitate rapid decision-making in the future.

[0051] To ensure the precise correspondence between image content and die-cutting action, synchronous triggering at the hardware or software level must be employed. Hardware triggering involves the motion controller outputting a high-precision TTL signal through the digital I / O port to directly trigger the industrial camera exposure when it reaches a node, resulting in the lowest latency and highest accuracy. Software triggering, on the other hand, involves sending a trigger command to the camera via the network after recognizing an event.

[0052] The system dynamically configures the camera's acquisition window, i.e., the dynamic ROI, based on the first die-cutting information of the event, capturing only small patches of images containing the die-cutting area to improve frame rate, reduce data transmission volume, and shorten subsequent processing time. Simultaneously, it dynamically adjusts the camera's exposure time and light source intensity based on the material and surface morphology of the part to be die-cut to obtain optimal image contrast. Before being transmitted to the subsequent analysis module, the acquired raw images undergo a series of real-time preprocessing steps, including geometric correction to eliminate lens distortion, suppression of random noise using median filtering or bilateral filtering algorithms, and enhancement of edge and detail sharpness.

[0053] The system precisely registers the pre-processed real-time image with the standard CAD template of the part to be die-cut. Through feature point matching or cross-correlation algorithms, it calculates the minute displacement, rotation and scaling of the real-time image relative to the template, thereby aligning the image coordinate system with the machine's world coordinate system.

[0054] Specifically, the die-cutting machine is running at high speed and is about to perform the right-angle inflection point die-cut of the first IC chip outline, namely node N_003. The control system of the die-cutting machine receives the X and Y axis position feedback of the cutter head at a frequency of 2kHz via the EtherCAT bus, and the internal digital twin model updates the position of the cutter head on the path in real time. When the feedback coordinates show that the cutter head enters the spatial area of ​​(49.87±0.1, 82.00±0.1)mm, the system immediately marks the current time as node N_003. The system then queries the event library to confirm that N_003 corresponds to the die-cutting event E_003, and parses its core die-cutting content as: {Action:Corner_Turn_Half-Cut,Target:IC1_Functional_Area}.

[0055] At the same moment that node N_003 is marked, the motion controller sends a trigger pulse to the industrial high-speed camera facing the IC1 area through a dedicated digital output port; upon receiving the pulse, the camera immediately performs an exposure with a duration of 20μs; since the system is pre-configured with a dynamic ROI, the camera only captures an image area centered on IC1 with a size of 256x256 pixels.

[0056] The acquired small image is immediately sent to the image processing unit, where the system performs lens distortion correction. The system then performs feature-based registration of the image with the stored IC1 standard CAD template. The registration result shows that the real-time image has a slight offset (Δx=+0.015mm, Δy=-0.003mm) relative to the template. This precise offset information, along with the registered image, is packaged together as the real-time sensing data of node N_003.

[0057] Furthermore, based on the comparison of each die-cutting node, the importance level of each die-cutting node is determined. The first optimization coefficient is determined according to the importance level of each die-cutting node and the real-time image of the corresponding die-cutting area. The second optimization coefficient is determined according to the node position of each die-cutting node and the real-time image of the corresponding die-cutting area. This takes into account the overall consideration of the node position of each die-cutting node and the real-time image of the corresponding die-cutting area, ensuring the accuracy of the second optimization coefficient.

[0058] At this point, the system performs a horizontal comparison between the current node and all other nodes in the task list, extracting multi-dimensional key features. This includes parsing the first die-cutting information corresponding to the node to determine the functional features of its region type, associating the economic features of the product or component value of the node, checking the risk attributes assigned to the node in S121 and the risk features of historical defect data, as well as the geometric features of the node's geometric complexity (such as the size of the radius of curvature). The system uses a multi-attribute decision model to calculate the importance level of the node, assigning a weight to each feature dimension, such as functional weight, risk weight, economic weight, etc. These weights can be customized according to the specific requirements of the product. Then, each feature is normalized and scored (e.g., on a 0-1 scale), for example, the functional chip area gets 1.0 points, and the waste area gets 0.1 points. Finally, through comprehensive calculation, such as weighted summation, a quantitative importance level is obtained, and the priority of each node is finally evaluated within the range of 1 to 10.

[0059] The first optimization coefficient (C_1) is a weighted quality deviation index. The system uses computer vision algorithms to analyze the real-time image of the die-cutting area corresponding to the node, and extracts one or more quality feature vectors, such as edge cleanliness, burr area, and dimensional accuracy. Each feature is quantified into a 0-1 quality score, and a comprehensive quality score is obtained through weighting or worst-case principle. Finally, C_1 is calculated through a mapping function. Its key inputs are the node importance level and the comprehensive quality score. A typical function is C_1=I_i×(1-Q_i_total), which means that the more important the node and the worse its current quality, the larger C_1 is, and the higher the urgency of optimization.

[0060] The second optimization coefficient (C_2) is a geometric deviation index. The system performs pixel-level differential or contour matching analysis between the real-time image registered in S131 and the theoretical CAD model to calculate the deviation vector field of the actual processing contour relative to the theoretical contour. For the current node, the system extracts the deviation vector ΔP_i (including magnitude and direction) at its position. C_2 is calculated through a mapping function, and its core input is the magnitude of the deviation. Usually, C_2 is proportional to |ΔP_i|. This coefficient directly reflects the intensity of the path compensation requirement.

[0061] Specifically, the die-cutting machine has just finished die-cutting node N_003 (the right-angle inflection point of the IC1 area) and has acquired a real-time image of the area. The system analyzes node N_003 and extracts its features: in terms of functional features, it is located in IC1_Functional_Area, with a score of 1.0; in terms of risk features, it is marked as HIGH_RISK_CORNER, with a score of 1.0; in terms of economic features, the electronic tag is a high-value product, with a score of 1.0. The system uses preset weights W_Function=0.4, W_Risk=0.4, and W_Economic=0.2 to perform a weighted calculation: I_003=(1.0×0.4)+(1.0×0.4)+(1.0×0.2)=1.0. After the final mapping, the node importance level of N_003 is determined to be the highest level, 10.

[0062] When calculating the first optimization coefficient, the visual algorithm analyzes the real-time image of N_003 and detects a slight material whitening phenomenon inside the inflection point, with edge sharpness slightly lower than expected. The system calculates a comprehensive quality score Q_003_total = 0.85; therefore, C_1_003 = I_003 × (1 - Q_003_total) = 10 × (1 - 0.85) = 1.5, which is a relatively high optimization coefficient, indicating a quality defect in a very important location; when calculating the second optimization coefficient, image registration... The algorithm shows that the actual inflection point is offset eastward by 15μm compared to the theoretical position, i.e., |ΔP_003|=15μm; the system sets the proportional constant k=0.05 / μm, then C_2_003=0.05×15=0.75, which is a moderately high optimization coefficient, indicating that there is a positional deviation that needs to be corrected; thus, for node N_003, the system has obtained two key quantitative indicators: C_1_003=1.5 (high-quality optimization requirement) and C_2_003=0.75 (moderate positional compensation requirement).

[0063] Therefore, the die-cutting optimization method for this die-cutting node is determined based on the mapping relationship between the first optimization coefficient, the second optimization coefficient, and the die-cutting optimization method. Simultaneously, the working years of the die-cutting machine are collected. Based on the die-cutting optimization methods of each die-cutting node, the working years of the die-cutting machine, and the workpiece to be die-cut, the intelligent control system of the die-cutting machine is determined. This system considers the overall factors of the die-cutting optimization methods of each die-cutting node, the working years of the die-cutting machine, and the workpiece to be die-cut, ensuring the accuracy of the intelligent control system. Furthermore, by controlling the die-cutting optimization method of this die-cutting node, the optimization of each die-cutting node is achieved, realizing the overall consideration of the die-cutting optimization methods of each die-cutting node, the working years of the die-cutting machine, and the workpiece to be die-cut, thus improving the accuracy of the intelligent control system of the die-cutting machine.

[0064] At this point, the system maintains a high-dimensional decision matrix. The input to this matrix is ​​the first optimization coefficient C_1 and the second optimization coefficient C_2, and the output is the specific modulus optimization method. For ease of decision-making, C_1 and C_2 are usually discretized into several levels, such as Low, Medium, and High. The decision logic follows a clear mapping relationship: When {C_1:High,C_2:Low}, it indicates poor quality but accurate positioning, with the problem lying in the process parameters. The system will choose the process parameter optimization method. When {C_1:Low,C_2:High}, it indicates good quality but off-center positioning, indicating that the current process parameters are appropriate but there is a systematic deviation in the execution path. The system will choose the real-time path compensation method. When {C_1:High,C_2:High}, this is the most complex case, indicating poor quality and off-center positioning. The system will choose a hybrid advanced control mode, simultaneously performing parameter optimization and path compensation. When {C_1:Low,C_2:Low}, it indicates that everything is normal, and the system will choose to maintain the status quo or standard mode.

[0065] Each optimization method corresponds to a specific set of control algorithms and parameter adjustment instructions. For example, process parameter optimization includes adjusting pressure and speed, real-time path compensation generates a compensation vector to correct the coordinates of subsequent path points, while hybrid advanced control combines the above two and triggers more advanced strategies, such as activating vibration suppression or reducing global speed.

[0066] The system obtains the working years or more precise total operating hours of the die-cutting machine from the equipment management database or human-machine interface, and transforms this raw data into a quantified equipment performance degradation factor D. This factor can be a linear function or a nonlinear model based on maintenance records, used to characterize the degree of performance degradation caused by aging, wear and other factors.

[0067] The system integrates and constructs the control system. It collects the optimization methods corresponding to all nodes in the task and analyzes these node-level optimization methods to find common patterns. For example, if multiple inflection point nodes in the IC region are determined to require process parameter tuning, the system will generate a global rule: all inflection point nodes located in the IC region will have their pressure reduced by 5% by default. At the same time, the attenuation factor D is applied to a global compensation model, such as multiplying all pressure commands by a coefficient to offset the performance drift caused by equipment aging. The information of the part to be die-cut (such as material batch number, thickness, hardness) is also integrated into the system as the starting point for all optimizations. The final intelligent control system is a dynamic and parameterized control model. The intelligent control system includes a set of basic process parameters for the current part to be die-cut, a set of special commands for specific nodes, a set of general rules extracted from node optimization, and a global correction information based on years of service.

[0068] Specifically, the die-cutting machine has calculated C_1_003=1.5 and C_2_003=0.75 for node N_003; the system maps C_1_003=1.5 to High and C_2_003=0.75 to Medium; querying the decision matrix, for the input {C_1:High,C_2:Medium}, the mapping relationship is determined to be hybrid high-level control; the system then generates specific optimization instructions for N_003: based on C_2, a compensation vector ΔP_compensation=(-15μm,0) is generated, ready to be applied to subsequent paths; based on C_1 and the whitening phenomenon detected in the image, the instruction {Pressure_Adjustment:-2N} is generated; since it is hybrid control, the system adds an additional instruction: {Activate_Local_Vibration_Damping}.

[0069] The system read from the maintenance log of the die-cutting machine that it had been running for 5 years and calculated the performance degradation factor D=0.05. During the control system integration phase, the system recorded the optimization method of N_003. When continuing to execute subsequent nodes, it was found that similar situations occurred at the other 4 IC inflection points. Therefore, a global optimization rule was extracted: for all inflection point nodes in the IC region, the default pressure was reduced by 2N and local vibration suppression was activated. The system applied the degradation factor D to the global compensation model and calculated that due to equipment aging, all pressure commands should be increased by an additional 1% on the original plan to compensate for the decrease in hydraulic system efficiency.

[0070] The intelligent control system of the die-cutting machine has been completed. It includes: a basic pressure / speed curve for the current PET; a global rule to reduce the IC inflection point pressure by 2N; a device aging compensation function of +1% for all pressures; and a -15μm path compensation vector for N_003. This system is distributed to the controller of the die-cutting machine. It no longer executes a fixed program, but dynamically and adaptively completes the entire electronic tag die-cutting task under an intelligent framework based on real-time feedback and historical experience.

[0071] refer to Figure 5 In step S14, the specific steps are as follows: S141: Real-time monitoring of the intelligent control system, dynamic identification of the intelligent control system, and output of multiple sub-intelligent control items. At this time, each intelligent control item has a corresponding relationship with multiple die-cutting nodes, and the control content of each intelligent control item is marked. S142: The working data of the die-cutting machine changes dynamically as the die-cutting machine works, so as to output the dynamic working data of the die-cutting machine. The first dynamic optimization content is determined based on the dynamic working data of the die-cutting machine and the control content of each sub-intelligent control project. S143: Determine the second level of dynamic optimization content based on the dynamic working data of the die-cutting machine and the project priority of each sub-intelligent control project, and determine the dynamic optimization event of the die-cutting machine based on the first level of dynamic optimization content and the second level of dynamic optimization content.

[0072] In the embodiments of this application, the system does not passively wait for instructions, but continuously listens to and polls the intelligent control system generated by step S13 through a background daemon process or real-time operating system task. The monitored objects are all control elements within the system, including the basic process parameter set, the global optimization rule set, the node-level optimization instruction set, and the equipment attenuation compensation function. The system will check the activation status, scope of action, and dependencies of these elements.

[0073] The system employs rule-based or pattern-matching algorithms to identify monitored control elements. Its identification strategies include: based on the scope of action, identifying which elements affect the entire system, which affect specific node classes, and which affect a single node; based on the control type, identifying which are process parameter controls, which are geometric path controls, and which are equipment status controls; and based on triggering conditions, identifying which are immediately effective and which are activated only when specific conditions are met. Through this multi-dimensional identification, the system understands a static system as a series of logical entities with clear attributes and behaviors.

[0074] Each identified independent control strategy is instantiated by the system as a Sub-Intelligent-Control-Project (SICP) object. Each SICP object is created in memory and assigned a unique ID, a process similar to creating instances of classes in object-oriented programming. When establishing the mapping between the modular nodes, the system creates one or more references or pointers to the modular nodes for each SICP object. This includes one-to-one mapping, where a path compensation project for a specific node establishes a strong binding relationship with that node; one-to-many mapping, where a pressure adjustment project for a type of node establishes a relationship with all nodes of that type; and global mapping, where a global speed adjustment project, whose scope is marked as Global, has a potential association with all nodes.

[0075] The system parses the original control policy corresponding to SICP into a structured, machine-readable control content data structure. This structure is usually a key-value pair or a JSON object, containing at least key fields such as control type, control target, adjustment value, trigger condition, and priority. The system stores the above structured data content as a core attribute of the SICP object. At this point, an SICP object becomes complete: it knows who it is (ID), where it is to act (NodeReferences), and what it is to do (ControlContent).

[0076] Specifically, the intelligent control system of the die-cutting machine has been completed. This system includes a global rule and a node-level instruction. The background monitoring process of the die-cutting machine's control system scans the control system generated by S13 and identifies two independent control strategies: Strategy 1 reduces the default pressure by 2N and activates local vibration suppression for all inflection point nodes in the IC region. The system identifies it as a multi-node composite control strategy based on its scope of action (multiple IC inflection points) and control type (parameters + status). Strategy 2 applies path compensation of (-15μm,0) to node N_003. The system identifies it as a single-node path compensation strategy based on its scope of action (single node) and control type (geometric path).

[0077] The system creates SICP objects for these two strategies: SICP_A (ID: SICP_001) represents IC inflection point composite control, and SICP_B (ID: SICP_002) represents N_003 path compensation. During the relationship marking stage, the system establishes a set mapping relationship between SICP_A and all 5 IC inflection point nodes {N_003, N_007, N_011, N_015, N_019}, and at the same time establishes a one-to-one mapping relationship between SICP_B and a single node N_003.

[0078] The control content of SICP_A is marked as: {Control_Type:Hybrid_Control,Target:[Pressure,Vibration_Damping],Adjustment_Value:{Pressure:Vibration_Damping:High},Trigger_Condition:On_Entering_Any_Node_in_Set,Priority:High}; The control content of SICP_B is marked as follows: {Control_Type:Path_Compensation,Target:Tool_Path_Vector,Adjustment_Value:{X:-15,Y:0,Unit:μm},Trigger_Condition:Immediately,Priority:Medium}, Through S141, the die-cutting machine's control system now has a clear, structured task list, which can handle the two specific control items SICP_A and SICP_B independently and in parallel.

[0079] Furthermore, the working data of the die-cutting machine changes dynamically as the die-cutting machine operates, so as to output the dynamic working data of the die-cutting machine. Based on the dynamic working data of the die-cutting machine and the control content of each sub-intelligent control project, the first dynamic optimization content is determined. This takes into account the overall consideration of the dynamic working data of the die-cutting machine and the control content of each sub-intelligent control project, ensuring the accuracy of the first dynamic optimization content.

[0080] At this point, the system synchronously acquires data from different subsystems at high frequency through industrial buses and sensor networks, including motion control system data (such as real-time position, speed, acceleration, torque, and current of servo drives), process data (such as real-time pressure values ​​from pressure sensors and critical area temperatures from temperature sensors), and mechanical system status data (such as vibration spectrum and amplitude data acquired through accelerometers or vibration sensors). All acquired data streams are sent to a data fusion module, which adds a high-precision timestamp to each data packet to ensure that all data are strictly synchronized in time. The final output dynamic working data is a multi-dimensional vector of time series.

[0081] Based on the deviation between real-time data and the predetermined target, an immediate correction instruction is generated; the system compares the key variables in the dynamic working data with the target values ​​set in the SICP control content in real time, and calculates the deviation Error(t) = Target_Value - Actual_Value(t); the system calculates the required adjustment amount based on the deviation. For example, for single-variable control such as pressure and speed, the system will run an independent PID controller and calculate an output based on the deviation and its historical rate of change.

[0082] For known and predictable disturbances, the system pre-calculates a compensation amount based on the model and superimposes it with feedback control to improve response speed. For more complex systems such as vibration suppression, the system establishes a mathematical model and uses a state observer to estimate the true system state from noisy vibration data. It then uses a model predictive control algorithm to predict the vibration trend over a future period and calculates the optimal control input sequence. The output of the control algorithm is encapsulated as the first layer of dynamic optimization, which is usually a series of fine-tuning instructions for the underlying actuators.

[0083] Specifically, the die-cutting machine is executing SICP_A (IC inflection point pressure reduction 2N) and SICP_B (N_003 path compensation); the die-cutting machine's control system outputs a dynamic working data stream at a frequency of 2kHz; at a certain time t, a snapshot of the data stream is as follows: {Pos:(49.87,82.01),Vel:0.28m / s,Torque:1.2Nm,Pressure:26.5N,Vibration_Amplitude:0.08g}.

[0084] For the optimization of SICP_A, the system makes a comparison: the control content of SICP_A requires a pressure of 28N (basic 30N-2N), while the actual pressure in the dynamic data is 26.5N; the deviation is calculated as Error_Pressure=28N-26.5N=1.5N; the system's embedded pressure PID controller detects this negative deviation of 1.5N and calculates a correction command based on its P, I, and D parameters; therefore, the first level of dynamic optimization includes a correction item for SICP_A: {Pressure_Feedback_Correction:+1.5N_to_Actuator}.

[0085] Regarding the optimization of SICP_B, although its control content is to apply a geometric compensation and it does not have a dynamic target value, the system will monitor the status during the compensation process. Dynamic data shows that when the compensation path is executed, the torque of the X-axis servo motor is 1.2Nm, which is close to the safe threshold of 1.5Nm at this speed.

[0086] After assessing the risks, it was decided to adopt a feedforward control strategy based on torque limits. Without affecting the overall progress, the acceleration and deceleration curves of subsequent paths are fine-tuned to reduce torque peaks. Therefore, another item was added to the first dynamic optimization: {Motion_Profile_Adjustment:Reduce_X_Axis_Jerk_by_10%}.

[0087] Therefore, the second level of dynamic optimization content is determined based on the dynamic working data of the die-cutting machine and the project priority of each sub-intelligent control project. The dynamic optimization events of the die-cutting machine are determined based on the first and second levels of dynamic optimization content, which takes into account the overall consideration of the first and second levels of dynamic optimization content and ensures the accuracy of the dynamic optimization events of the die-cutting machine.

[0088] At this point, conflict detection is performed. The system continuously monitors dynamic working data to look for global events that lead to resource contention or system state deterioration, such as total power consumption approaching the upper limit, multiple axis servo motors simultaneously approaching torque saturation, overall vibration level continuously rising or system temperature being too high, and performance conflicts such as a high-priority project requiring low speed to ensure accuracy while a low-priority project requires high speed to ensure efficiency.

[0089] When a conflict or performance bottleneck is detected, the system initiates a priority-based scheduling algorithm. Its decision-making principle is to ensure the performance of high-priority items, even if it means sacrificing the performance of low-priority items and downgrading medium- and low-priority items, such as reducing their target speed, relaxing their accuracy requirements, or even temporarily suspending their execution. The scheduling decision is also manifested in the adjustment of global parameters, such as reducing the overall speed limit of the spindle to release system resources. The result of the scheduling decision is encapsulated as a second layer of dynamic optimization, which is usually a set of macroscopic, global policy instructions.

[0090] The system receives the first level of dynamic optimization content (a series of fine-tuning instructions for specific control parameters) from S142 and the second level of dynamic optimization content (a series of global strategies) from this step. The fusion process is a process of instruction synthesis and conflict resolution. If there is a conflict between the two, the second level (high priority) instructions take precedence. The final fused instruction set is encapsulated into a dynamic optimization event, which is a structured data packet containing all the actions that the die-cutting machine needs to perform in the next control cycle. It typically includes the global strategy layer from the second level of optimization, as well as the project execution layer for each active SICP, containing the synthesis result of its basic control content and the first level of optimization content. The event also has a precise timestamp and validity period. The final dynamic optimization event is sent to the real-time motion controller and PLC of the die-cutting machine. The controller parses the event and precisely drives the servo motors, pressure valves, and other actuators to complete a complete dynamic optimization cycle.

[0091] Specifically, the die-cutting machine is running, and SICP_A (IC inflection point composite control, priority High) and SICP_B (N_003 path compensation, priority Medium) are both active. The system monitors dynamic work data and finds that due to uneven tension of the incoming material roll, the Y-axis servo motor is running at a high load of 95% torque when performing a long-distance movement. At this time, a new low-priority project, SICP_C (quick cutting in the scrap area, priority Low), is activated, which requires the Y-axis to run at a higher acceleration. The system detects a resource conflict because the demand of SICP_C will further exacerbate the overload risk of the Y-axis. After querying the priorities, the system decides that in order to ensure the stable execution of the high-priority SICP_A and SICP_B, the resource consumption of SICP_C must be limited. Therefore, the system generates a second layer of dynamic optimization, namely the global policy: {Global_Y_Axis_Acceleration_Limit:70%of_Nominal,Resource_Policy:Throttle_Low_Priority_Tasks}.

[0092] The system merges the first-level optimization content generated by S142 (such as the +1.5N pressure correction for SICP_A) with the newly generated second-level optimization content and constructs the final dynamic optimization event package. This event package includes a global policy: the upper limit of Y-axis acceleration is set to 70% of the rated value. For the project instruction of SICP_A (high priority), the final instruction is {Apply_Pressure:28N,Apply_Feedback_Correction:+1.5N,Activate_Vibration_Damping:High}.

[0093] For SICP_B (medium priority) project instructions, the final instruction is {Apply_Path_Compensation:(15μm,0),Motion_Profile_Adjustment:Reduce_X_Axis_Jerk_by_10%}; and for SICP_C (low priority) project instructions, the final instruction is: The original requirement of `Apply_Speed_Profile:Standard_Profile,But_With_Y_Axis_Acceleration_Capped_At_70%}` was trimmed by the second level of optimization.

[0094] refer to Figure 6 In step S15, the specific steps are as follows: S151: Detect dynamic optimization events and determine multiple dynamic optimization contents based on the detection of dynamic optimization events. Construct optimization content combinations based on multiple dynamic optimization contents and corresponding die-cutting nodes. Determine corresponding optimization features based on the identification of each optimization content combination to output multiple optimization features. S152: Determine the working efficiency of the die-cutting machine based on the efficiency detection of the die-cutting machine, collect the die-cutting progress of the die-cutting machine, determine the first working coefficient based on multiple optimization features and the working efficiency of the die-cutting machine, determine the second working coefficient based on multiple optimization features and the die-cutting progress of the die-cutting machine, and determine the working mode of the die-cutting machine based on the mapping relationship between the first working coefficient, the second working coefficient and the working mode. S153: Mark the die-cutting task time of the die-cutting machine, determine the replacement event of the part to be die-cut based on the traceability of the die-cutting task time of the die-cutting machine, determine the morphological change content of the part to be die-cut based on the identification of the replacement event of the part to be die-cut, and trigger the adjustment of the die-cutting path of the die-cutting machine based on the morphological change content of the part to be die-cut and the working mode of the die-cutting machine.

[0095] In the embodiments of this application, dynamic optimization events are detected, and multiple dynamic optimization contents are determined based on the detection of dynamic optimization events. Optimization content combinations are constructed based on the multiple dynamic optimization contents and corresponding die-cutting nodes. Corresponding optimization features are determined based on the identification of each optimization content combination, so as to output multiple optimization features. This approach takes into account the overall consideration of the identification of each optimization content combination and ensures the accuracy of the corresponding optimization features.

[0096] At this point, the system captures and parses the dynamic optimization events generated by S143 in real time, decomposing them into a series of basic optimization actions that can be analyzed. The system subscribes to and receives all dynamic optimization events published by the S143 decision module in real time through a high-speed message queue or event bus. The monitoring is continuous and non-blocking, ensuring that no optimization event is missed. Each event has a precise timestamp and event ID. Whenever a dynamic optimization event is captured, the system immediately parses it. This event is a structured data packet containing instructions at multiple levels. The parsing process recursively traverses the event structure to extract all independent, atomic-level dynamic optimization content, which are the specific adjustment actions made by the system. The extracted content can usually be divided into several categories, such as process parameter correction, motion profile adjustment, global strategy change, and equipment status control.

[0097] Discrete, time-sequential optimizations are reorganized into logical sets associated with physical spaces (modular nodes) for aggregation analysis. Each dynamic optimization carries its source information upon generation, indicating which sub-intelligent control project triggered it. SICP is associated with one or more modular nodes, and the system uses this association chain to attach each dynamic optimization to the modular node it affects. The system maintains a dynamic, parallel data structure to the node sequence to store optimization combinations. When a new dynamic optimization is parsed and associated with a node, it is added to the corresponding combination set for that node, forming a node-level combination. Simultaneously, all optimizations marked as global are aggregated into a single global-level combination. To prevent data from growing indefinitely, the system typically uses a sliding time window to manage these combinations; older data exceeding the window is archived or discarded, ensuring the analysis focuses on recent system behavior.

[0098] The system applies a series of feature extraction algorithms to each combination of optimization content. Key feature dimensions include: frequency features (the number of times a certain type of optimization content appears within a time window; high frequency indicates persistent instability in that stage); amplitude features (calculating the average, maximum, variance, and root mean square for numerical optimization content; large or high variance adjustments indicate significant deviations or fluctuations); diversity / complexity features (calculating the number of different types of optimization content in a combination; multiple optimizations coexisting increase complexity); and persistence features (the proportion of the cumulative time an optimization instruction is activated to the total time). All calculated features are integrated into a multi-dimensional optimization feature vector.

[0099] Furthermore, the working efficiency of the die-cutting machine is determined based on the efficiency detection of the die-cutting machine, and the die-cutting progress of the die-cutting machine is collected. A first working coefficient is determined based on multiple optimization features and the working efficiency of the die-cutting machine, and a second working coefficient is determined based on multiple optimization features and the die-cutting progress of the die-cutting machine. The working mode of the die-cutting machine is determined based on the mapping relationship between the first working coefficient, the second working coefficient, and the working mode. This comprehensive consideration of the mapping relationship between the first working coefficient, the second working coefficient, and the working mode ensures the accuracy of the working mode of the die-cutting machine. At the same time, it realizes the comprehensive consideration of multiple optimization features, the working efficiency of the die-cutting machine, and the corresponding die-cutting tasks, thereby improving the accuracy of the working mode of the die-cutting machine.

[0100] At this point, work efficiency is typically determined using the core concept of Overall Equipment Effectiveness (OEE). The system obtains real-time production data from the production execution system or directly from the die-cutting machine's controller. By calculating the product of performance efficiency (the ratio of actual output speed to theoretical design speed) and quality efficiency (the ratio of qualified products to total output), a comprehensive work efficiency index is obtained. Die-cutting progress is collected based on the die-cutting event sequence identified in S131, tracking the number of completed nodes or the length of executed paths in real time. By calculating the percentage of completed parts to the total task, the current task's position in the entire production batch is intuitively reflected.

[0101] For the first working coefficient, the system calculates C_w1 using a multivariate mapping model. It analyzes the relationship between the internal optimization pressure reflected by the optimization feature vector (such as high frequency and large magnitude corrections indicating that the system is under great pressure) and the working efficiency. For example, if the internal pressure is high but the efficiency is still high, it means that the system is extremely robust and C_w1 is high; if the internal pressure is high and the efficiency is low, it means that the system is overburdened and C_w1 is low.

[0102] For the second working coefficient, the system needs time series analysis capabilities. It will compare the optimized feature vector of the current window with the vector of the historical window. If the amplitude and frequency of the optimized features continue to increase as the die-cutting progresses, it indicates that the system performance is declining and the sustainability is poor, and C_w2 is low. Conversely, if the optimized features are stable or decrease, it indicates that the state is stable or improving and the sustainability is strong, and C_w2 is high.

[0103] The system internally maintains a decision matrix, whose input is the quantization levels of C_w1 and C_w2, and whose output is the specific working mode. The mode definition and decision logic follow a clear mapping relationship: when {C_w1: High, C_w2: High}, the system is currently healthy and sustainable, and the decision is to explore potential, entering Aggressive_Mode to try to moderately increase global speed or pressure to pursue higher extreme efficiency; when {C_w1: Low, C_w2: High}, the system is currently in a poor state but the trend is stable, and the decision is to conservatively treat the situation, entering Stabilization_Mode. In the first scenario, the system prioritizes production stability by reducing the overall speed. When {C_w1:High,C_w2:Low}, the system's current efficiency is acceptable, but a performance degradation trend has emerged. The decision is to provide early warning and preparation, entering Predictive_Maintenance_Mode to issue a maintenance warning to the operator and slightly reduce performance parameters to slow down the degradation. When {C_w1:Low,C_w2:Low}, the system's current state is poor and the trend is worsening. The decision is to take emergency precautions, entering Safe_Mode to significantly reduce the operating speed to prevent equipment damage or large-scale scrap.

[0104] Therefore, by marking the die-cutting task time of the die-cutting machine, the replacement event of the part to be die-cut is determined based on the tracing of the die-cutting task time of the die-cutting machine, the morphological change content of the part to be die-cut is determined based on the identification of the replacement event of the part to be die-cut, and the die-cutting path of the die-cutting machine is adjusted based on the morphological change content of the part to be die-cut and the working mode of the die-cutting machine. This approach is compatible with the overall consideration of tracing the die-cutting task time of the die-cutting machine and ensures the accuracy of the replacement event of the part to be die-cut.

[0105] At this point, the system creates a usage cycle record for each independent die-cutting unit, marking the start time of its die-cutting task with a high-precision timestamp, and recording the end time of the die-cutting task after recognizing that all die-cutting events of the unit have been completed, thereby continuously maintaining a time-series data stream composed of die-cutting task cycle values; the determination of replacement events based on traceability adopts cycle time analysis. The system performs real-time statistical analysis on the time series of the die-cutting task cycle to calculate the statistical characteristics of the normal production cycle time, such as the average value and standard deviation; when the system monitors that before the start of a new task, the interval between the end time of the previous task and the start time of the new task is much longer than the normal cycle time, the system determines that an abnormal interruption has occurred. The determination rule is usually that the interval time is greater than the average value plus several times the standard deviation. When this condition is met, the system triggers a replacement event for the part to be die-cut.

[0106] The system performs multimodal information acquisition, including triggering an industrial camera installed at the feed end to perform a high-resolution scan of the first segment of the new material to obtain visual features, attempting to communicate with the RFID tag or QR code on the new material roll to read material metadata, and automatically performing rapid physical parameter measurements if equipped with online sensors. The system compares the acquired new material information with the previous material batch or standard material model item by item, with comparison dimensions covering geometric dimensions (such as thickness and width), physical property dimensions (such as material type and surface hardness), and visual feature dimensions (such as color and gloss). The comparison results are quantified and structured into a morphological change content dataset.

[0107] The system's decision-making is based on two key inputs: the morphological changes of the part to be die-cut and the current operating mode of the die-cutting machine. The fusion logic is that the control strategy is jointly determined by these two factors. The operating mode provides the tone for control (whether it's aggressive or conservative), while the morphological changes provide the specific parameters for control. The system queries its internal process knowledge base based on the morphological changes. This knowledge base stores a large number of material attribute-process parameter mapping models and applies weights according to the current operating mode. For example, in the aggressive_mode, 100% of the recommended adjustment value is applied, while in the conservative_mode, only 70-80% of the recommended value is applied with an increased safety margin. The final generated control strategy is encapsulated as a die-cutting path control event. This is a recalibration or replanning, including instructions such as process parameter reset, path compensation recalibration, and control system update. The system will use this new system to guide subsequent production.

[0108] Please see Figure 7 , Figure 7 This is a schematic diagram of the structural composition of the intelligent control system of the die-cutting machine in an embodiment of the present invention; the intelligent control system of the die-cutting machine includes: The die-cutting path module 21 is used to determine the corresponding working status based on multiple working data of the die-cutting machine during the working process of the die-cutting machine, and to determine the die-cutting path of the die-cutting machine according to the working status of the die-cutting machine, the corresponding die-cutting task and the workpiece to be die-cut. The die-cutting event module 22 is used to determine multiple die-cutting nodes based on the identification of the die-cutting path, and to determine the die-cutting event corresponding to each die-cutting node according to the node position of each die-cutting node, the corresponding die-cutting requirements, and the die-cutting area corresponding to the part to be die-cut. The intelligent control system module 23 is used to determine multiple die-cutting contents based on the identification of die-cutting events in each die-cutting node, determine the die-cutting optimization method of the die-cutting node based on the multiple die-cutting contents, the real-time image of the die-cutting area and the node importance level of the die-cutting node, and determine the intelligent control system of the die-cutting machine based on the die-cutting optimization method of each die-cutting node, the working years of the die-cutting machine and the workpiece to be die-cut. The dynamic optimization event module 24 is used to identify multiple sub-intelligent control items based on the identification of the intelligent control system, and to determine the dynamic optimization event of the die-cutting machine according to the control content of each sub-intelligent control item, the corresponding item priority and the dynamic working data of the die-cutting machine. The working mode module 25 is used to determine multiple optimization features based on the detection of dynamic optimization events, determine the working mode of the die-cutting machine based on the multiple optimization features, the working efficiency of the die-cutting machine and the corresponding die-cutting progress, and dynamically adjust the die-cutting path of the die-cutting machine as the shape of the part to be die-cut changes.

[0109] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. An intelligent control method for a die-cutting machine, characterized in that, include: During the operation of the die-cutting machine, the corresponding working state is determined based on multiple working data of the die-cutting machine. The die-cutting path of the die-cutting machine is determined according to the working state of the die-cutting machine, the corresponding die-cutting task, and the workpiece to be die-cut. The die-cutting path is a collection of motion instructions and process parameters. Multiple die-cutting nodes are determined based on the identification of the die-cutting path. The die-cutting events corresponding to each die-cutting node are determined according to the node position of each die-cutting node, the corresponding die-cutting requirements, and the die-cutting area corresponding to the part to be die-cut. In each die-cutting node, multiple die-cutting contents are determined based on the identification of die-cutting events. The die-cutting optimization method for the die-cutting node is determined based on the multiple die-cutting contents, the real-time image of the die-cutting area, and the node importance level of the die-cutting node. The intelligent control system of the die-cutting machine is determined based on the die-cutting optimization method of each die-cutting node, the working years of the die-cutting machine, and the workpiece to be die-cut. The intelligent control system includes a set of basic process parameters for the current workpiece to be die-cut, a set of special instructions for specific nodes, a set of general rules extracted from node optimization, and global correction information based on the working years. Based on the identification of this intelligent control system, multiple sub-intelligent control items are identified, and the dynamic optimization events of the die-cutting machine are determined according to the control content of each sub-intelligent control item, the corresponding item priority, and the dynamic working data of the die-cutting machine. Multiple optimization features are determined based on the detection of dynamic optimization events. The working mode of the die-cutting machine is determined based on the multiple optimization features, the working efficiency of the die-cutting machine, and the corresponding die-cutting progress. The die-cutting path of the die-cutting machine is dynamically adjusted as the shape of the part to be die-cut changes.

2. The intelligent control method for a die-cutting machine according to claim 1, characterized in that, During the operation of the die-cutting machine, the corresponding working state is determined based on multiple working data of the die-cutting machine. The die-cutting path of the die-cutting machine is determined according to the working state, the corresponding die-cutting task, and the workpiece to be die-cut, including: The system collects the working signals of the die-cutting machine and triggers the working process of the die-cutting machine based on the working signals. It determines multiple working data of the die-cutting machine based on the detection of the die-cutting machine, and determines the working status of the die-cutting machine based on the multiple working data and the dynamic video of the die-cutting module. The die-cutting task list of the die-cutting machine is determined based on the detection of the die-cutting machine. Multiple die-cutting tasks are determined based on the identification of the die-cutting task list. The die-cutting path of the die-cutting machine is determined based on the multiple die-cutting tasks, the information of the parts to be die-cut, and the working status of the die-cutting machine.

3. The intelligent control method for a die-cutting machine according to claim 1, characterized in that, The process involves identifying multiple die-cutting nodes based on the die-cutting path, and determining the corresponding die-cutting events for each die-cutting node based on its node position, corresponding die-cutting requirements, and the die-cutting area corresponding to the part to be die-cut. This includes: Collect the die-cutting path and mark the path shape of the die-cutting path. Based on the path shape of the die-cutting path, the overall shape of the part to be die-cut, and the previous die-cutting content of the die-cutting machine, determine multiple die-cutting nodes and mark the node positions of multiple die-cutting nodes. In multiple die-cutting nodes, the corresponding die-cutting requirements are determined based on the node contact of each die-cutting node. The first layer of die-cutting information is determined based on the node position of each die-cutting node and the die-cutting area corresponding to the part to be die-cut. The second layer of die-cutting information is determined based on the die-cutting requirements of each die-cutting node and the die-cutting area corresponding to the part to be die-cut. The die-cutting event corresponding to each die-cutting node is determined based on the first layer of die-cutting information, the second layer of die-cutting information and the corresponding die-cutting node.

4. The intelligent control method for a die-cutting machine according to claim 1, characterized in that, In each die-cutting node, multiple die-cutting contents are determined based on the identification of die-cutting events. The die-cutting optimization method for that node is determined based on the multiple die-cutting contents, the real-time image of the die-cutting area, and the node importance level. The intelligent control system of the die-cutting machine is determined based on the die-cutting optimization method of each node, the working years of the die-cutting machine, and the workpiece to be die-cut, including: The system monitors each die-cutting node in real time and marks the die-cutting events corresponding to each die-cutting node. Based on the identification of the die-cutting events, multiple die-cutting contents are determined. At the same time, the die-cutting machine performs visual inspection on the die-cutting area of ​​the part to be die-cut to collect real-time images of each die-cutting area.

5. The intelligent control method for a die-cutting machine according to claim 4, characterized in that, In each die-cutting node, multiple die-cutting contents are determined based on the identification of die-cutting events. The die-cutting optimization method for that node is determined based on the multiple die-cutting contents, the real-time image of the die-cutting area, and the node importance level. The intelligent control system of the die-cutting machine is determined based on the die-cutting optimization method of each node, the working years of the die-cutting machine, and the workpiece to be die-cut. This also includes: Based on the comparison of each die-cutting node, determine the node importance level of each die-cutting node, determine the first optimization coefficient based on the node importance level of each die-cutting node and the real-time image of the corresponding die-cutting area, and determine the second optimization coefficient based on the node position of each die-cutting node and the real-time image of the corresponding die-cutting area. The die-cutting optimization method for the die-cutting node is determined based on the mapping relationship between the first optimization coefficient, the second optimization coefficient, and the die-cutting optimization method. At the same time, the working years of the die-cutting machine are collected, and the intelligent control system of the die-cutting machine is determined based on the die-cutting optimization method of each die-cutting node, the working years of the die-cutting machine, and the workpiece to be die-cut.

6. The intelligent control method for a die-cutting machine according to claim 1, characterized in that, The system identifies multiple sub-intelligent control items based on the intelligent control system. Dynamic optimization events for the die-cutting machine are determined according to the control content of each sub-intelligent control item, its corresponding priority, and the dynamic working data of the die-cutting machine. These events include: The system monitors the intelligent control system in real time, dynamically identifies the system, and outputs multiple sub-intelligent control items. At this time, each intelligent control item has a corresponding relationship with multiple die-cutting nodes, and the control content of each intelligent control item is marked.

7. The intelligent control method for a die-cutting machine according to claim 6, characterized in that, The process of identifying multiple sub-intelligent control items based on the intelligent control system, and determining the dynamic optimization events of the die-cutting machine according to the control content of each sub-intelligent control item, the corresponding item priority, and the dynamic working data of the die-cutting machine, also includes: The working data of the die-cutting machine changes dynamically as the die-cutting machine operates, so as to output the dynamic working data of the die-cutting machine. Based on the dynamic working data of the die-cutting machine and the control content of each sub-intelligent control project, the first level of dynamic optimization content is determined. The second level of dynamic optimization content is determined based on the dynamic working data of the die-cutting machine and the project priority of each sub-intelligent control project. The dynamic optimization events of the die-cutting machine are then determined based on the first and second levels of dynamic optimization content.

8. The intelligent control method for a die-cutting machine according to claim 1, characterized in that, The process of determining multiple optimization features based on the detection of dynamic optimization events, determining the working mode of the die-cutting machine based on these features, the working efficiency of the die-cutting machine, and the corresponding die-cutting progress, and dynamically adjusting the die-cutting path of the die-cutting machine according to changes in the shape of the part to be die-cut, includes: Dynamic optimization events are detected, and multiple dynamic optimization contents are determined based on the detection of dynamic optimization events. Optimization content combinations are constructed based on multiple dynamic optimization contents and corresponding die-cutting nodes. Optimization features are determined based on the identification of each optimization content combination, so as to output multiple optimization features. The working efficiency of the die-cutting machine is determined based on the efficiency detection of the die-cutting machine, and the die-cutting progress of the die-cutting machine is collected. A first working coefficient is determined based on multiple optimization features and the working efficiency of the die-cutting machine, and a second working coefficient is determined based on multiple optimization features and the die-cutting progress of the die-cutting machine. The working mode of the die-cutting machine is determined based on the mapping relationship between the first working coefficient, the second working coefficient and the working mode.

9. The intelligent control method for a die-cutting machine according to claim 8, characterized in that, The process of determining multiple optimization features based on the detection of dynamic optimization events, determining the working mode of the die-cutting machine based on the multiple optimization features, the working efficiency of the die-cutting machine, and the corresponding die-cutting progress, and dynamically adjusting the die-cutting path of the die-cutting machine according to the shape changes of the part to be die-cut, further includes: The die-cutting task time of the die-cutting machine is marked. The replacement event of the part to be die-cut is determined based on the traceability of the die-cutting task time. The morphological change of the part to be die-cut is determined based on the identification of the replacement event. The die-cutting path of the die-cutting machine is adjusted based on the morphological change of the part to be die-cut and the working mode of the die-cutting machine.

10. An intelligent control system for a die-cutting machine, characterized in that, The intelligent control system of the die-cutting machine is applied to the intelligent control method of the die-cutting machine as described in any one of claims 1-9.

Images