A die cutting machine defect online detection method and system

Abstract

The present application belongs to the technical field of product defect detection, and specifically discloses a die-cutting machine defect online detection method and system. The pressure state of the die-cutting stroke is sensed and abnormally identified in real time during the operation of the die-cutting machine, the abnormal strokes identified are marked with abnormal types, and readable marks are applied on the corresponding material web. Then, the material web is partitioned and a spatial partition index is established before the material enters the visual detection station. When the material enters the visual detection station, different detection modes are triggered according to the identified marks, so that the abnormal types of the pressure abnormal products are associated with the corresponding knife line dense areas or bending transition areas. Imaging resources are actively concentrated and projected to the defect high-incidence areas for local collection, thereby implementing targeted detection. Normal products are subjected to regular detection, so as to improve the overall detection efficiency while ensuring the defect detection accuracy.

Description

Technical Field

[0001] This invention belongs to the field of product defect detection technology, and specifically discloses an online defect detection method and system for die-cutting machines. Background Technology

[0002] Die-cutting machines are processing equipment used in industries such as packaging printing, electronics, and automobiles. Through the cooperation of a die-cutting plate and an imprinting mechanism, they cut sheet or roll materials into specific shapes. During the die-cutting process, due to factors such as fluctuations in die-cutting pressure, appearance defects such as dimensional deviations, edge burrs, and poor indentation are inevitable. Traditional manual visual inspection methods are inefficient, and prolonged operation can easily lead to visual fatigue, causing misjudgments and missed inspections. With the maturity of machine vision technology, more and more die-cutting production lines are beginning to introduce automated inspection equipment based on optical imaging to replace manual inspection.

[0003] For example, patent application CN109387519A discloses a die-cut product appearance inspection machine, which includes a machine body, a feeding shaft, a discharging shaft, a suction platform, a pressure roller mechanism, and an inspection mechanism. The inspection mechanism uses a camera and a vision light source to collect images of the products transported to the inspection area, and then completes the identification of appearance defects through image processing. The above-mentioned machine vision-based inspection equipment has solved the problems of low efficiency and inconsistent inspection standards of manual inspection to a certain extent.

[0004] However, current die-cut product appearance inspection involves acquiring images of the final product after die-cutting is complete. This means the inspection time lags behind the actual formation of defects, ignoring the differences in stress states experienced by the product during die-cutting. These stress state differences are the main cause of die-cutting defects. Because existing inspection equipment passively relies on product images for defect identification only after die-cutting, it is difficult to actively focus imaging resources on truly high-risk areas. Ultimately, high-risk products with abnormal pressure do not receive targeted inspection resources, and subtle defects are easily masked by the material background texture in the overall image, resulting in low overall inspection efficiency. Summary of the Invention

[0005] To solve the above-mentioned technical problems, or at least partially solve them, the present invention provides an online defect detection method and system for die-cutting machines. By sensing the pressure state of the die-cutting stroke and identifying abnormalities, the abnormal products are marked to trigger the local detection mode of the visual inspection station, thereby realizing targeted image acquisition and defect identification of high-risk products.

[0006] The objective of this invention can be achieved through the following technical solutions: Firstly, this invention proposes an online defect detection method for a die-cutting machine, comprising the following steps: Before the die-cutting machine is run, the dense area of ​​the blade lines and the bending transition area are delineated according to the direction of the steel blade cutting edge line and the indentation steel line in the die-cutting plate design drawing, and a spatial partition index of the material sheet is established.

[0007] During the operation of the die-cutting machine, the state of the die-cutting reaction force borne by the die-cutting plate in each die-cutting stroke is sensed and pressure signals are obtained.

[0008] The stroke anomaly is determined by comparing the pressure characteristics in the pressure signal with the normal pressure reference, and when an abnormal stroke is determined, the deviation direction of the pressure characteristics is used to output an anomaly type identifier.

[0009] When an abnormal stroke is determined, a readable mark is applied to a preset non-functional area on the material sheet corresponding to the abnormal stroke.

[0010] When materials enter the visual inspection station, the presence of readable markers is identified, and an inspection mode switching command is generated.

[0011] In response to the detection mode switching command, if it is in local detection mode, the camera determines the area of ​​interest from the spatial partition index according to the anomaly type identifier and controls the camera to perform local image acquisition; if it is in normal detection mode, the camera is controlled to perform global image acquisition on the material area.

[0012] The acquired images are used to identify defects and generate material defect detection results.

[0013] Secondly, the present invention proposes an online defect detection system for a die-cutting machine, comprising the following modules: a spatial partition index module, used to establish a spatial partition index for the material sheet according to the pattern design drawing of the current batch of die-cutting products before the die-cutting machine is run.

[0014] The pressure sensing module is installed on the load-bearing surface of the die-cutting machine to sense the pressure state of each die-cutting stroke and output a pressure signal.

[0015] The stroke status determination module is used to determine stroke abnormalities by comparing the pressure characteristics in the pressure signal with the normal pressure reference, and outputs an abnormality type identifier by combining the deviation direction of the pressure characteristics when an abnormal stroke is determined.

[0016] The marking application module is used to apply readable markings to a preset non-functional area on the material web when an abnormal stroke is detected.

[0017] The mark recognition module is located at the entrance of the vision inspection station. It is used to identify whether there are readable marks on the material surface and generate inspection mode switching instructions accordingly.

[0018] The imaging control module is used to perform image acquisition operations in response to detection mode switching commands and in conjunction with spatial partition indexes.

[0019] The defect identification module is used to identify defects in the acquired images and generate defect detection results.

[0020] Combining all the above technical solutions, the positive effects of this invention are as follows: 1. This invention can perceive and identify the pressure state of the die-cutting stroke in real time during the operation of the die-cutting machine, mark the abnormal strokes with abnormal types, and apply readable marks to the corresponding material surface. Then, different detection modes are triggered according to the readable marks, thereby performing targeted detection on products with abnormal pressure and performing routine detection on normal products. This improves the overall detection efficiency while ensuring the accuracy of defect detection.

[0021] 2. When performing targeted testing on products with abnormal pressure, this invention partitions the material area and establishes a spatial partition index, associating the abnormality type with the corresponding dense blade line area or bending transition area. This allows the imaging resources to be actively concentrated and projected onto the high-incidence defect area for local acquisition, enabling the subtle defects to be revealed, thereby effectively improving the detection rate. Attached Figure Description

[0022] The present invention will be further described with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present invention. For those skilled in the art, other drawings can be obtained based on the following drawings without creative effort.

[0023] Figure 1 This is a flowchart illustrating the method steps of Embodiment 1 of the present invention;

[0024] Figure 2 This is a flowchart illustrating the implementation process of determining stroke anomalies based on the deviation between the pressure signal and a pre-established normal pressure reference in this invention.

[0025] Figure 3 This is a system module connection diagram of Embodiment 2 of the present invention. Detailed Implementation

[0026] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] Example 1

[0028] This invention proposes an online defect detection method for die-cutting machines, comprising the following steps: S1. Before the die-cutting machine is run, the dense area of ​​the blade lines and the bending transition area are delineated according to the direction of the steel blade cutting edge line and the indentation steel line in the die-cutting plate design drawing, and a spatial partition index of the material sheet is established.

[0029] Considering the differences in structural features across different areas of the die-cutting plate, to facilitate rapid localization of high-defect areas by the camera based on anomaly type during subsequent local inspection, it is necessary to know the specific spatial location of these areas on the material sheet beforehand. Therefore, during the preparation stage before the die-cutting machine runs, a spatial partition index needs to be established on the material sheet according to the plate design drawing.

[0030] In the manner achievable by this invention, establishing a spatial partition index includes the following process: obtaining the pattern design drawing of the current batch of die-cut products, and extracting the positions of each steel blade edge line and the positions and extension directions of each indentation steel line on the die-cutting plate. The steel blade edge lines are used to cut and shape the material; their distribution density directly affects the cutting quality. Areas with denser blade lines experience more concentrated stress during die-cutting, increasing the risk of poor cutting. The indentation steel lines are used to press creases into the material for subsequent folding and shaping; the locations where their direction changes are typically stress concentration areas.

[0031] Using the material conveying direction as the longitudinal reference and the direction perpendicular to the conveying direction as the transverse reference, each steel blade edge line is traversed one by one. The distance between each pair of adjacent steel blade edge lines is calculated. Areas with a distance less than a set distance threshold are marked as blade line dense areas. Blade line dense areas refer to areas where blade edge lines are densely distributed within a unit area. The denser the blade lines, the greater the cutting load per unit area, and the more significant the impact of abnormal pressure on blade line dense areas.

[0032] Each indentation steel wire is traversed one by one. Since the extension path of each indentation steel wire on the material sheet is not completely straight, the location where the direction changes often corresponds to the stress concentration area. In order to identify the local sections where the direction of the indentation steel wire changes, each indentation steel wire needs to be divided into several local sections of equal length according to its extension direction, and the directional deflection of the extension direction of each section relative to the overall extension direction of the indentation steel wire is calculated. The directional deflection represents the degree of deviation of the local section from the overall direction. The specific calculation method is as follows: taking the unit vector of the overall extension direction of the indentation steel wire as the reference, calculate the angle between the direction vector of each local section and the reference vector. The angle is the directional deflection of the local section.

[0033] The local section where the directional deflection exceeds the set deflection threshold is marked as the bending transition area. The bending transition area is the section where the direction of the indentation steel wire changes. The material in the corresponding area is subjected to combined stresses such as bending and extrusion, and is most sensitive to pressure changes.

[0034] To facilitate the automatic movement of the camera to the target area based on its spatial position during subsequent local image acquisition, it is necessary to map the spatial boundaries of the aforementioned densely defined blade lines and bending transition areas on the material sheet from the design coordinates to the actual transport plane coordinates.

[0035] Specifically, a two-dimensional coordinate reference system is established with the conveying direction and the width direction as the reference. The spatial boundaries of each region on the material width are projected onto the reference system. The extreme coordinates of the boundary curves of each region in the corresponding directions are obtained along the conveying direction and the width direction. The start and end positions of each region in the conveying direction and the offset start and end positions in the width direction are determined by the extreme coordinates.

[0036] Each region is assigned a spatial identifier, which includes the region's sequential position in the transport direction and its offset range in the width direction. The spatial identifiers of all regions together form a spatial partition index of the material sheet, which is used to determine the moving target position and framing range of the camera during subsequent local inspection.

[0037] In the spatial partitioning operation example above, the process of setting the spacing threshold is as follows: Statistically count the spacing values ​​of all adjacent cutter lines on the die-cutting plate to form a one-dimensional spacing dataset. For example, use the K-means clustering algorithm to divide the spacing dataset into two clusters, where the cluster with smaller values ​​corresponds to the cutter line clustering region, and the cluster with larger values ​​corresponds to the cutter line sparse region. The value center of the first cluster is smaller than the value center of the second cluster. The median value between the value centers of the first and second clusters is used as the spacing threshold.

[0038] The process of setting the deflection threshold is as follows: The directional deflection values ​​of all local sections of each indentation steel wire on the die-cutting plate are statistically analyzed to form a one-dimensional deflection dataset. Similarly, a clustering algorithm is used to divide the deflection dataset into two clusters, where the cluster with smaller values ​​corresponds to the straight section and the cluster with larger values ​​corresponds to the turning section. The value center of the first cluster is smaller than the value center of the second cluster. The median value between the value centers of the first and second clusters is used as the deflection threshold.

[0039] S2. During the operation of the die-cutting machine, the state of the die-cutting reaction force borne by the die-cutting plate in each die-cutting stroke is sensed and pressure signals are obtained.

[0040] In the die-cutting process, die-cutting pressure is the main process parameter that determines the product forming quality. If the pressure is too low, defects such as insufficient cut depth, burrs or bridging will occur. If the pressure is too high, irreversible damage such as crushing and bursting will occur. In order to carry out targeted detection according to the actual stress state of the product, this invention performs pressure sensing during the operation of the die-cutting machine. Considering that die-cutting is a periodic stamping process, each stroke corresponds to one product, and the stress state of different products is independent of each other. Therefore, this invention performs pressure sensing according to the die-cutting stroke. Each die-cutting stroke refers to the complete action cycle of the die-cutting plate from the beginning of contact with the material to the complete separation from the material during one rotation of the main shaft of the die-cutting machine. The specific sensing process is as follows: S21. Since the main force of the die-cutting machine is transmitted to the die-cutting plate through transmission components such as crankshaft and connecting rod during operation, the bearing surface is the part that directly bears the die-cutting reaction force in the power transmission path. The magnitude of the force is in a deterministic correspondence with the actual pressure applied to the material by the die-cutting plate. Therefore, a pressure sensing element is set on the bearing surface of the main force transmission path of the die-cutting machine, which can truly reflect the actual pressure applied to the material by the die-cutting plate in each stroke. In one embodiment, the selected pressure sensing element is a piezoelectric pressure sensor.

[0041] S22. When the spindle of the die-cutting machine rotates to the pressing start phase of each stroke, the pressure sensing element is triggered to collect pressure signals. The pressing start phase refers to the specific angle position when the spindle rotates to the working surface of the die-cutting plate begins to contact the material surface. At this time, the die-cutting plate begins to apply pressure to the material. The pressure signal recorded from this moment completely covers the entire process of pressure gradually building up from zero.

[0042] S23. When the spindle of the die-cutting machine rotates to the pressing end phase of the stroke, the pressure sensing element is triggered to stop the pressure signal acquisition. The pressing end phase refers to the angle position when the spindle continues to rotate until the die-cutting plate is completely separated from the material surface. At this time, the pressure application process of the current stroke ends. The continuous waveform of pressure change over time acquired from the pressing start phase to the pressing end phase is used as the pressure signal of the current stroke.

[0043] S3. Based on the comparison between the pressure characteristics in the pressure signal and the normal pressure reference, determine the stroke abnormality, and when the stroke is determined to be abnormal, output the abnormality type identifier by combining the deviation direction of the pressure characteristics.

[0044] After acquiring the pressure signal for each stroke through step S2, the pressure signal can be compared with a benchmark representing a normal die-cutting state to determine whether the pressure state of the current stroke is normal. Therefore, before formally performing stroke determination, it is necessary to establish a pressure benchmark that reflects the normal die-cutting state of the current batch as a reference for all subsequent determinations.

[0045] As an optional embodiment of the present invention, the above-mentioned normal pressure benchmark is pre-established through the following process: After the die-cutting machine runs continuously at its rated operating speed for a preset period of time, multiple die-cut products output continuously are subjected to quality inspection according to a preset sampling plan. When the inspection results all meet the factory quality standards of the current batch of products, the current batch of products is confirmed to be qualified. After the confirmation node, multiple consecutive die-cutting strokes are selected as benchmark calibration strokes. This is because at this time, the die-cutting machine has completed initial debugging, the die-cutting pressure is stable, and the running-in of the die-cutting plate and the material has entered a stable period. The collected pressure data can truly reflect the stress characteristics under normal production conditions of the current batch.

[0046] After selecting the benchmark calibration stroke, considering that the complete pressure waveform contains a large number of redundant data points, it is inefficient to directly use the entire waveform as the comparison object. Therefore, it is necessary to extract feature values ​​that can characterize the stress state of the stroke to represent the stroke. The specific extraction process is as follows: Take the continuous waveform from the pressing start phase to the pressing end phase in the pressure signal as the analysis object, locate the maximum value of the waveform amplitude on the continuous waveform, and take the maximum value as the peak pressure value of the benchmark calibration stroke. The peak pressure value reflects the maximum pressure applied to the material by the die-cutting plate in the stroke. It is a direct indicator of the stress intensity of the stroke. Too much pressure will cause the material to crush, and too little pressure will lead to incomplete cutting.

[0047] On a continuous waveform, the change in amplitude per unit time within the rising segment from the initial value to the peak pressure value is taken as the pressure rise edge change rate of the stroke. The pressure rise edge change rate reflects how fast the pressure is built up. If the rate is too fast, it means that the die-cutting plate has a violent impact on the material, which can easily cause local damage to the material. If the rate is too slow, it indicates that there may be gaps or insufficient rigidity in the pressure transmission path, which is also not conducive to the stability of the molding quality.

[0048] The peak pressure value and the rate of change of the rising edge described above describe the force characteristics of the same stroke from the two dimensions of amplitude and process, respectively. Combining the two can more comprehensively depict the pressure state of the stroke.

[0049] After extracting the pressure characteristic values ​​of all benchmark calibration strokes, the values ​​of the same pressure characteristic value in multiple benchmark calibration strokes are arranged in order of numerical magnitude. The minimum and maximum values ​​are used as the lower and upper limits of the pressure characteristic value fluctuation boundary, respectively. The fluctuation boundary defines the allowable fluctuation range of each characteristic value under normal die-stamping pressure conditions. Fluctuations within the boundary are considered normal fluctuations, while fluctuations outside the boundary indicate that the stress state of the stroke has deviated from the normal range, posing a risk of defects.

[0050] The normal pressure reference is formed by the fluctuation boundaries of each pressure characteristic value.

[0051] It should be noted that, considering that the die-cutting plate will gradually wear down during long-term operation of the die-cutting machine, the normal pressure reference should be re-established after each predetermined number of strokes of the die-cutting machine or after each replacement of the die-cutting plate.

[0052] Once the normal pressure reference is established, anomalies can be determined for the pressure state of each current stroke.

[0053] See Figure 2 As a further embodiment of the present invention, the anomaly determination includes the following process: extracting the peak pressure value and the pressure rise rate from the pressure signal of the current stroke.

[0054] The peak pressure value and the rate of change of pressure rise are compared with their respective normal pressure reference values.

[0055] If the characteristic value is within the fluctuation boundary, no deviation has occurred; if the characteristic value is higher than the upper limit of the fluctuation boundary, it indicates that the force intensity or pressure build-up rate of the current stroke exceeds the maximum fluctuation limit in normal production, that is, the material is subjected to a load beyond the normal range, and a positive deviation is determined to have occurred; if the characteristic value is lower than the lower limit of the fluctuation boundary, it indicates that the force intensity or pressure build-up rate of the current stroke has not reached the minimum level in normal production, that is, the material is subjected to insufficient load, and a negative deviation is determined to have occurred.

[0056] After the above comparison is completed, the following judgment is made according to the deviation: a) When the peak pressure value and the rate of change of the pressure rise edge do not deviate, it is judged as a normal stroke.

[0057] b) When only a single feature deviates, it is determined to be an abnormal stroke, and the abnormality type is marked as a single offset abnormality.

[0058] c) When both the peak pressure value and the rate of change of the pressure rise deviate, compare the directions of their deviations: the peak pressure value represents the force amplitude during the stroke, and the rate of change of the pressure rise represents how quickly the pressure is built up. Although there is a certain positive correlation between the two, factors such as changes in die-cutting clearance, material thickness fluctuations, local hard spots, or the introduction of foreign objects can all cause relatively independent changes in both.

[0059] When the two deviate in the same direction, it indicates that the pressure waveform has risen or fallen as a whole, and the shape has not been distorted. This is a continuous deviation anomaly, which is usually caused by factors such as the overall drift of the die-cutting gap.

[0060] When the two deviate in opposite directions, it indicates that the pressure waveform is distorted, with a fast rise rate but a low peak value, or a slow rise rate but a high peak value. The dynamic process and the final amplitude are mismatched. For example, local hard spots in the material may cause the peak value to rise abnormally while the rate remains unchanged, or insufficient support of the die-cutting plate may cause the peak value to be too low. Such situations are considered abnormal pressure abrupt changes.

[0061] The above anomaly determination not only provides a result on whether the stroke is normal, but also further subdivides the abnormal strokes into different types, so that the corresponding areas of interest can be matched according to different types to achieve more targeted imaging detection.

[0062] S4. When an abnormal stroke is determined, a readable mark is applied to the preset non-functional area on the material sheet corresponding to the abnormal stroke.

[0063] After anomaly detection, if the current stroke is determined to be an abnormal stroke, it means that the product corresponding to the current stroke experienced a stress state deviating from the normal range during the die-cutting process, and there is a high probability that a quality defect corresponding to the anomaly type has occurred. If such products are allowed to flow to the subsequent visual inspection station without markings, the inspection system can only perform uniform, undifferentiated inspection on all products, resulting in low overall inspection efficiency.

[0064] To address the aforementioned issues, when the current stroke is determined to be an abnormal stroke, the present invention applies a readable mark to the material surface corresponding to the abnormal stroke. This allows the abnormal status information to be attached to the product surface in the form of a mark and simultaneously transported to the visual inspection station along with the material. By recognizing the mark before the material enters the inspection field of view, the inspection system can proactively identify abnormal products based on the presence or absence of the mark, thereby triggering a differentiated inspection mode.

[0065] Specifically, applying readable markings includes the following: setting a marker at the die-cutting station. The marker can be a laser marking device with its marking head facing the material surface. When the die-cutting machine finishes the current stroke and the die-cutting plate returns to the die holder, the current product is still within the die-cutting station range. At this time, the marker is in an effective position where marking operations can be performed.

[0066] When the current stroke is determined to be an abnormal stroke, a trigger signal carrying an abnormality type identifier is sent to the marker.

[0067] After receiving the trigger signal, the marker applies a readable mark to the preset non-functional area on the material web corresponding to the abnormal stroke. The preset non-functional area refers to the area on the material web that does not affect the main function of the product, such as the seam blank edge, process edge, and waste area. Applying a mark to such areas will not have any impact on the quality and appearance of the final product.

[0068] In a specific example where the marker is a laser marking device, the readable mark is a QR code mark formed by laser etching. The mark is formed by the laser beam acting on the surface of the material to create a visually recognizable trace, which remains attached to the material surface during subsequent transportation until it enters the visual inspection station.

[0069] S5. When the material enters the visual inspection station, identify the presence of readable markers and generate an inspection mode switching command.

[0070] After the spatial partition index is established and readable tags are applied, the material enters the visual inspection station along with the conveyor platform. This invention uses a tag reader at the entrance of the visual inspection station to trigger a corresponding detection mode switch based on the detection result of readable tags. Specifically, the process includes the following steps: S51. A tag reader is set at the entrance of the visual inspection station. Specifically, the tag reader is a barcode reader with its reading field of view facing a preset non-functional area on the material surface. When the material surface moves to the sensing area of ​​the tag reader along with the conveyor platform, the tag reader scans the preset non-functional area on the material surface, automatically decodes the QR code, reads the information, and determines whether a readable tag exists in the non-functional area.

[0071] S52. If it exists, the presence signal and its corresponding anomaly type identifier are transmitted to the imaging control module; if it does not exist, the absence signal is transmitted.

[0072] S53. The imaging control module completes image acquisition configuration before the material enters the imaging field of view according to the type of received signal: when a marker presence signal is received, it is configured as a local detection mode; otherwise, it is configured as a regular detection mode.

[0073] S6. In response to the detection mode switching command, if it is in local detection mode, the area of ​​interest is determined from the spatial partition index according to the anomaly type identifier, and the camera is controlled to perform local image acquisition; if it is in normal detection mode, the camera is controlled to perform global image acquisition on the material area.

[0074] After the aforementioned material area partitioning and anomaly type identification, a corresponding relationship has been established between the anomaly type and the spatial location of potential defects on the material area. Therefore, the corresponding image acquisition operation can be performed according to the configuration results, specifically including the following process: Before the material enters the imaging field of view, the imaging control module determines the current material image acquisition operation according to the received detection mode switching command: In the normal detection mode, the camera is controlled to perform global image acquisition on the material area. Global image acquisition refers to the camera taking full-coverage images of the entire surface area of ​​the material area according to the preset scanning path. The acquired images contain complete surface information of the material area and can perform comprehensive appearance defect screening. However, due to the large coverage area and large amount of data of the global image, the defect identification efficiency is relatively low.

[0075] In localized detection mode, the area of ​​interest for sampling is first determined based on the anomaly type identifier: for pressure mutation anomalies, the area of ​​interest is the region with dense blade lines. This correspondence is not based on the cause of the mutation originating from the region with dense blade lines, but rather on the consideration that pressure mutation anomalies reflect instantaneous distortion of the pressure waveform. Distortion can occur at any position on the die-cutting plate, but the region with dense blade lines has a concentrated distribution of cutting edges, high stress levels, and complex geometric features, making it most sensitive to distortion. Minor cutting quality deviations are more likely to manifest as detectable surface defects in the region with dense blade lines. Therefore, prioritizing detection resources in the region with dense blade lines can achieve the highest defect detection efficiency within a limited localized detection range.

[0076] The area of ​​concern for continuous pressure deflection anomalies is the bending transition region. This correspondence is also because the material in the bending transition region is subjected to multiple stress components such as bending and extrusion during the die-cutting process, and is most sensitive to the overall drift of the die-cutting pressure. Therefore, detection resources are preferentially concentrated in this region.

[0077] A single offset anomaly involves only a deviation of a single pressure feature, and it is impossible to determine which spatial influence attribute it is biased towards. Therefore, it is necessary to collect data from both the dense tool line area and the bending transition area at the same time.

[0078] Then, the spatial identifier corresponding to the region of interest is read from the spatial partition index. The spatial identifier contains the sequential position of the region of interest in the conveying direction and the offset interval in the width direction. The sequential position in the conveying direction indicates the arrangement number of the region of interest in the material conveying path, and the offset interval in the width direction indicates the starting and ending boundaries of the region of interest in the transverse direction of the material sheet. Based on the above information, the spatial coordinates of each region of interest can be uniquely determined in the two-dimensional coordinate reference system of the material sheet.

[0079] Finally, the camera is controlled to move sequentially above each area of ​​interest according to the transport direction sequence. The starting and ending positions of the camera's lateral framing are determined according to the offset range in the width direction. The lateral framing range of the camera is adjusted to fully cover the area of ​​interest. Then, local image acquisition is performed. A local image is an image that only contains surface information of the area of ​​interest. The spatial range of the image is smaller than that of the global image, and the defect detection efficiency is relatively high.

[0080] As a preferred implementation, the camera is mounted on a two-dimensional motion platform that can reciprocate along the conveying direction and along the width direction, wherein the two-dimensional motion platform is driven by a servo motor; the imaging control module generates motion control commands containing displacement in the conveying direction and displacement in the width direction based on the spatial coordinates of the area of ​​interest read from the spatial partition index, and sends the motion control commands to the servo driver of the two-dimensional motion platform to drive the camera to move above the area of ​​interest.

[0081] S7. Perform defect identification on the acquired images to generate material defect detection results.

[0082] The specific defect identification process is as follows: In the normal detection mode, defect feature extraction is performed on the global image, including contour size, edge burrs, indentation depth, etc. After extraction, the extracted defect features are compared with the design form features defined according to the pattern design drawing. If they do not match the design form features, a defect is determined to exist; otherwise, it is determined to be normal.

[0083] In local detection mode, defect features are extracted from local images of each region of interest. After extraction, the features are compared with the design features of the material sheet defined by the pattern design drawing in the same way to identify whether there are defects in the region of interest.

[0084] It is important to understand that the correspondence between the above-mentioned anomaly types and areas of interest is not a causal limitation on the mechanism of anomaly generation, but rather a regional priority division based on the detection strategy. That is, given the limited detection resources, imaging resources are prioritized for observation in the areas with the highest probability of defect occurrence. If no anomaly is found in the local detection, it is supplemented and confirmed by global image acquisition.

[0085] Example 2

[0086] See Figure 3 As shown, the present invention proposes an online defect detection system for a die-cutting machine, comprising the following modules: a spatial partition index module, used to establish a spatial partition index for the material sheet based on the pattern design drawing of the current batch of die-cutting products before the die-cutting machine runs.

[0087] The pressure sensing module is installed on the load-bearing surface of the die-cutting machine to sense the pressure state of each die-cutting stroke and output a pressure signal.

[0088] The stroke state determination module, connected to the pressure sensing module, is used to determine stroke abnormalities by comparing the pressure characteristics in the pressure signal with the normal pressure reference, and outputs an abnormality type identifier by combining the deviation direction of the pressure characteristics when an abnormal stroke is determined.

[0089] The marking application module, connected to the stroke state determination module, is used to apply readable markings to a preset non-functional area on the material web when an abnormal stroke is determined.

[0090] The mark recognition module is located at the entrance of the vision inspection station. It is used to identify whether there are readable marks on the material surface and generate inspection mode switching instructions accordingly.

[0091] The imaging control module, connected to the marker recognition module and the spatial partition index module, is used to perform image acquisition operations in response to detection mode switching commands and in conjunction with the spatial partition index.

[0092] The defect identification module, connected to the imaging control module, is used to identify defects in the acquired images and generate defect detection results.

[0093] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0094] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0095] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0096] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0097] Finally, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for online defect detection in a die-cutting machine, characterized in that, include: Before the die-cutting machine is run, the dense area of ​​the blade lines and the bending transition area are delineated according to the direction of the steel blade cutting edge line and the indentation steel line in the die-cutting plate design drawing, and the spatial partition index of the material sheet is established. During the operation of the die-cutting machine, the state of the die-cutting reaction force borne by the die-cutting plate in each die-cutting stroke is sensed and pressure signals are obtained. The stroke anomaly is determined by comparing the pressure characteristics in the pressure signal with the normal pressure reference, and when an abnormal stroke is determined, the deviation direction of the pressure characteristics is used to output an anomaly type identifier. When an abnormal stroke is determined, a readable mark is applied to the preset non-functional area on the material sheet corresponding to the abnormal stroke; When materials enter the visual inspection station, the presence of readable markers is identified, and an inspection mode switching command is generated. In response to the detection mode switching command, if it is in local detection mode, the region of interest is determined from the spatial partition index according to the anomaly type identifier, and the camera is controlled to perform local image acquisition; if it is in normal detection mode, the camera is controlled to perform global image acquisition on the material area. The acquired images are used to identify defects and generate material defect detection results.

2. The online defect detection method for a die-cutting machine as described in claim 1, characterized in that: The establishment of the spatial partition index for the material sheet includes the following: Obtain the die-cutting pattern design drawing for the current batch of products, and extract the position of each steel blade edge line and the position and extension direction of each indentation steel line on the die-cutting plate. Using the conveying direction of the material width as the longitudinal reference and the direction perpendicular to the conveying direction as the transverse reference, calculate the spacing between each pair of adjacent steel blade cutting edges one by one, and mark the area with a spacing less than the set spacing threshold as the blade line dense area. Calculate the directional deflection of each local section of each indentation steel line relative to the overall extension direction of the indentation steel line, and mark the local sections with directional deflection greater than the allowable deflection limit as bending transition areas. A two-dimensional coordinate reference system is established with the conveying direction and the width direction as the reference. The start and end positions of the dense blade line area and the bending transition area in the conveying direction and the offset start and end positions in the width direction are determined in the reference system. Each region is assigned a spatial identifier, which includes the region's sequential position in the conveying direction and its offset range in the width direction, forming a spatial partition index for the material sheet.

3. The online defect detection method for a die-cutting machine as described in claim 1, characterized in that: The acquisition of the pressure signal includes the following: A pressure sensing element is installed on the load-bearing surface of the main power transmission path of the die-cutting machine; When the spindle of the die-cutting machine rotates to the pressing start phase of each stroke, the pressure sensing element is triggered to collect the pressure signal; When the spindle of the die-cutting machine rotates to the pressing end phase of the stroke, the pressure sensing element is triggered to stop the pressure signal acquisition, and the continuous waveform of the acquired pressure over time is used as the pressure signal of the stroke.

4. The online defect detection method for a die-cutting machine as described in claim 1, characterized in that: The normal pressure reference includes the following establishment process: After the die-cutting machine runs continuously at its rated operating speed and the current batch of products is confirmed to be qualified, select multiple consecutive die-cutting strokes as the benchmark calibration strokes. Perform the following pressure feature extraction operation on the pressure signal of each reference calibration stroke: The maximum value of the waveform amplitude is located on the continuous waveform from the start phase of pressing to the end phase of pressing in the pressure signal, and is used as the reference to calibrate the peak pressure value of the stroke. The rate of change of pressure rise edge of the continuous waveform is calibrated by taking the amplitude change per unit time within the rising section of the waveform amplitude from the initial value to the peak pressure as the reference calibration stroke. After extracting the characteristic values ​​of all benchmark calibration strokes, the values ​​of the same pressure characteristic value in multiple benchmark calibration strokes are arranged in order of magnitude, and the minimum and maximum values ​​are used as the lower and upper limits of the pressure characteristic value fluctuation boundary, respectively. The normal pressure reference is formed by the fluctuation boundaries of each pressure characteristic value.

5. The online defect detection method for a die-cutting machine as described in claim 4, characterized in that: The output anomaly type identifier Includes the following processes: Extract the peak pressure value and the rate of change of the pressure rise edge from the pressure signal of the current stroke. Compare the peak pressure value and the rate of change of pressure rise with their respective normal pressure references; If the eigenvalue is within the fluctuation boundary, then no deviation has occurred; If the eigenvalue is higher than the upper limit of the fluctuation boundary, a positive deviation is determined to have occurred; If the eigenvalue is below the lower limit of the fluctuation boundary, a negative deviation is determined to have occurred; When neither the peak pressure value nor the rate of change of the pressure rise edge deviates, it is determined to be a normal stroke; When only a single feature deviates, it is determined to be an abnormal stroke, and the abnormality type is marked as a single offset abnormality; When both the peak pressure value and the rate of change of the pressure rise deviate, compare the directions of deviation: if the directions of deviation are the same, mark the anomaly type as a pressure continuous deflection anomaly; if the directions of deviation are opposite, mark the anomaly type as a pressure sudden change anomaly.

6. The online defect detection method for a die-cutting machine as described in claim 1, characterized in that: The application of readable tags includes the following: A marker is set on the die-cutting station. When the current stroke is determined to be an abnormal stroke, a trigger signal carrying an abnormality type identifier is sent to the marker. After receiving the trigger signal, the marker applies a readable mark to the preset non-functional area on the material web corresponding to the abnormal stroke.

7. The online defect detection method for a die-cutting machine as described in claim 1, characterized in that: The generation detection mode switching instruction includes the following: A marker recognizer is set up at the entrance of the vision inspection station to scan the preset non-functional area on the material surface entering the sensing area to determine whether there is a readable mark. If it exists, the presence signal and its corresponding anomaly type identifier are transmitted to the imaging control module; if it does not exist, the absence signal is transmitted. The imaging control module completes image acquisition configuration before the material enters the imaging field of view based on the received signal: it is configured to local detection mode when a marker presence signal is received, and otherwise configured to regular detection mode.

8. The online defect detection method for a die-cutting machine as described in claim 7, characterized in that: The specific operation in response to the detection mode switching command is as follows: In the normal inspection mode, the camera is controlled to perform global image acquisition on the material area; In local detection mode, the area of ​​interest is determined according to the anomaly type identifier: the area of ​​interest for pressure sudden change anomaly is the area with dense blade lines, the area of ​​interest for pressure continuous deflection anomaly is the bending transition area, and the area of ​​interest for single offset anomaly is both the area with dense blade lines and the bending transition area. Read the spatial identifier corresponding to the area of ​​interest from the spatial partition index, and determine the spatial coordinates of the area of ​​interest on the material sheet based on the conveying direction sequence position and the offset range in the width direction contained in the spatial identifier; The camera is controlled to move sequentially above each area of ​​interest according to the sequence position in the transport direction, and the horizontal framing range is adjusted according to the offset range in the width direction to perform local image acquisition.

9. The online defect detection method for a die-cutting machine as described in claim 1, characterized in that: The defect identification process is as follows: In the normal detection mode, defect features are extracted from the global image and compared with the material sheet design features defined according to the pattern design drawing to identify defects within the sheet. In local detection mode, defect features are extracted from local images of each region of interest and compared with the material sheet design features defined according to the pattern design drawing to identify defects existing in the region of interest.

10. An online defect detection system for a die-cutting machine, characterized in that, Includes the following modules: The spatial partition index module is used to establish a spatial partition index for the material sheet based on the pattern design drawing of the current batch of die-cut products before the die-cutting machine runs. The pressure sensing module is installed on the load-bearing surface of the die-cutting machine to sense the pressure state of each die-cutting stroke and output a pressure signal. The stroke state determination module is used to determine stroke abnormalities by comparing the pressure characteristics in the pressure signal with the normal pressure reference, and outputs an abnormality type identifier by combining the deviation direction of the pressure characteristics when an abnormal stroke is determined. The marking application module is used to apply readable markings to a preset non-functional area on the material web when an abnormal stroke is detected. The mark recognition module is located at the entrance of the vision inspection station and is used to identify whether there are readable marks on the material surface and generate inspection mode switching instructions accordingly. The imaging control module is used to perform image acquisition operations in response to detection mode switching commands and in conjunction with spatial partition indexes; The defect identification module is used to identify defects in the acquired images and generate defect detection results.

Images