A carton sealing abnormality detection method and system based on mutual inductance

Abstract

The present application relates to the technical field of intelligent packaging detection, and more particularly to a carton sealing abnormality detection method and system based on reflection sensing, which comprises constructing an adjustable physical detection channel, using a reflection type photoelectric sensing array to collect carton contour signals, and identifying sealing abnormality through adaptive threshold segmentation and an improved DTW algorithm; a confidence assessment model is constructed in combination with multi-modal data, and weighted exponential smoothing filtering is used to confirm effective abnormality; a PLC controller accurately separates abnormal cartons based on a predictive trajectory mapping algorithm, and device fault early warning is realized through edge computing cluster analysis. The present application can improve detection accuracy, reduce false positive rate and support intelligent maintenance.

Description

Technical Field

[0001] This invention belongs to the field of industrial automation testing technology, specifically a method and system for detecting abnormalities in carton sealing based on through-beam induction. Background Technology

[0002] In automated packaging production lines, the stability of carton sealing quality directly affects the safety and efficiency of subsequent logistics and transportation. To ensure correct application of sealing tape, existing technologies generally employ machine vision-based inspection solutions: industrial cameras are installed below the conveyor line to capture images of the sealing area of ​​the passing cartons, and specialized vision inspection software is used to compare the captured images with a standard template to determine whether there are any abnormalities such as missing, misaligned, or uncompressed tape. Once an abnormality is detected, the system outputs an alarm signal or triggers a shutdown mechanism. However, this type of vision inspection solution has significant drawbacks: on the one hand, its hardware is complex, typically requiring high-resolution cameras, industrial control computers, and professional image processing software, resulting in high overall equipment costs, with a single system often costing tens of thousands of yuan, making it unsuitable for widespread application in small and medium-sized production lines; on the other hand, since the camera must be installed below the carton's running path to obtain an effective viewing angle, dust generated by friction at the bottom of the carton easily adheres to the lens surface during actual operation, causing image blurring or even misjudgment, requiring frequent manual cleaning and maintenance, which not only increases the maintenance burden but also affects the long-term stability and reliability of the inspection system. Furthermore, vision systems are highly sensitive to ambient light, surface reflections on cartons, and color changes, further limiting their adaptability in complex industrial environments. Therefore, there is an urgent need for a simple, low-cost, interference-resistant, and easy-to-maintain method and system for detecting abnormal carton sealing, to overcome the shortcomings of existing vision inspection technologies in terms of cost, stability, and applicability. Summary of the Invention

[0003] This invention provides a method and system for detecting abnormalities in carton sealing based on photoelectric induction, which can at least solve some of the problems existing in the prior art.

[0004] A first aspect of this invention provides a method for detecting abnormal carton sealing based on through-beam induction, comprising: constructing an adjustable physical detection channel based on a perforated conveyor line; forming a dual-dimensional dimensional constraint benchmark by means of an adjustable height upper sealing detection sensor symmetrically arranged vertically and an adjustable width lower sealing perforated structure; when the carton is running along the conveyor line, using a through-beam photoelectric induction array to collect spatial penetration signals of the carton outline edge in real time; constructing a dynamic occlusion feature vector based on the spatial penetration signal; performing binarization processing on the occlusion feature vector using an adaptive threshold segmentation algorithm to obtain an abnormal protrusion area identification sequence; combining a standard size template in a carton specification parameter library, using an improved dynamic time warping (DTW) distance metric algorithm to calculate the deformation deviation between the current carton outline and the standard template; if the deformation deviation exceeds a preset safety threshold, it is determined to be a sealing abnormality event;

[0005] Based on the sealing anomaly event trigger signal, a multimodal state fusion module integrates conveyor speed, carton position coding, and historical false alarm rate data to construct an anomaly confidence assessment model. This model uses a weighted exponential smoothing formula to filter instantaneous signal jitter; its expression is: ,

[0006] in, The current level of confidence is the level of abnormality. This represents the current initial trigger state of the through-beam induction sensor (0 or 1), where α is an adaptive smoothing coefficient whose value is dynamically adjusted based on the vibration frequency of the conveyor line. f is the real-time vibration frequency. The reference stable frequency is given by k, which is the sensitivity adjustment constant; when When θ is the confidence threshold, the anomaly is confirmed to be valid;

[0007] Upon confirming the anomaly, the PLC controller generates a diversion control command based on the real-time pose encoding of the carton. An improved predictive trajectory mapping algorithm precisely guides the abnormal carton to the resealing station. This algorithm incorporates a conveyor delay compensation factor. Where L is the distance from the sensing point to the shunt point. For conveying speed, To mitigate the response delay of the actuator and ensure strict synchronization between the diversion action and the carton position, an audible and visual alarm is triggered, and the anomaly type, timestamp, and carton batch information are recorded. The edge computing unit performs cluster analysis on continuous anomaly patterns and uses an improved DBSCAN density clustering algorithm to identify potential sealing equipment failure trends and provide early warnings of maintenance needs.

[0008] A second aspect of this invention provides a carton sealing anomaly detection system based on through-beam induction, comprising: a first unit for constructing an adjustable physical detection channel based on a perforated conveyor line; forming a dual-dimensional dimensional constraint reference through an adjustable-height upper sealing detection sensor symmetrically arranged vertically and an adjustable-width lower sealing perforated structure; when the carton runs along the conveyor line, using a through-beam photoelectric induction array to collect spatial penetration signals of the carton outline edge in real time; constructing a dynamic occlusion feature vector based on the spatial penetration signal; performing binarization processing on the occlusion feature vector using an adaptive threshold segmentation algorithm to obtain an abnormal protrusion area identification sequence; combining a standard size template in the carton specification parameter library, using an improved dynamic time warping (DTW) distance metric algorithm to calculate the deformation deviation between the current carton outline and the standard template; if the deformation deviation exceeds a preset safety threshold, it is determined to be a sealing anomaly event;

[0009] The second unit is used to construct an anomaly confidence assessment model based on the sealing anomaly event trigger signal, by integrating conveyor speed, carton position code, and historical false alarm rate data through a multimodal state fusion module. The model uses a weighted exponential smoothing formula to filter instantaneous signal jitter, and its expression is: ,

[0010] in, The current level of confidence is the level of abnormality. This represents the current initial trigger state of the through-beam induction sensor (0 or 1), where α is an adaptive smoothing coefficient whose value is dynamically adjusted based on the vibration frequency of the conveyor line. f is the real-time vibration frequency. The reference stable frequency is given by k, which is the sensitivity adjustment constant; when When θ is the confidence threshold, the anomaly is confirmed to be valid;

[0011] The third unit, upon confirming an anomaly, uses the PLC controller to generate diversion control commands based on the real-time pose encoding of the cartons. An improved predictive trajectory mapping algorithm precisely guides the abnormal cartons to the resealing station. This algorithm incorporates a conveyor delay compensation factor. Where L is the distance from the sensing point to the shunt point. For conveying speed, To mitigate the response delay of the actuator and ensure strict synchronization between the diversion action and the carton position, an audible and visual alarm is triggered, and the anomaly type, timestamp, and carton batch information are recorded. The edge computing unit performs cluster analysis on continuous anomaly patterns and uses an improved DBSCAN density clustering algorithm to identify potential sealing equipment failure trends and provide early warnings of maintenance needs.

[0012] A third aspect of the present invention provides an electronic device, comprising: a processor; and a memory for storing processor-executable instructions; wherein the processor is configured to invoke the instructions stored in the memory to execute the aforementioned method. A fourth aspect of the present invention provides a computer-readable storage medium having computer program instructions stored thereon, which, when executed by a processor, implement the aforementioned method.

[0013] This invention achieves significant technical advantages in terms of sensitivity, accuracy, anti-interference capability, and adaptability in detecting abnormal carton sealing through a full-link design that incorporates an adjustable physical detection channel, a through-beam induction array, improved DTW deformation analysis, adaptive confidence assessment, and predictive shunt control. Attached Figure Description

[0014] Figure 1 This is a schematic diagram of a carton sealing anomaly detection system based on through-beam induction provided in an embodiment of the present invention. It shows the spatial layout and signal connection relationship between the hollow conveyor line, the adjustable height upper sealing detection sensor, the adjustable width lower sealing hollow structure, the through-beam photoelectric sensor array, the PLC controller, the diversion mechanism, and the resealing station.

[0015] Figure 2 This is a logical block diagram of the anomaly detection and processing flow in an embodiment of the present invention. It sequentially shows the functional modules and their data flow for spatial penetration signal acquisition, dynamic occlusion feature vector construction, adaptive threshold segmentation, improved DTW deformation deviation calculation, anomaly confidence assessment, diversion control command generation, and edge computing clustering analysis. Detailed Implementation

[0016] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0017] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Figure 1As shown, the present invention provides a carton sealing anomaly detection system based on through-beam induction, comprising a perforated conveyor line (labeled 1), an adjustable-height upper sealing detection sensor (labeled 2), an adjustable-width lower sealing perforated structure (labeled 3), a through-beam photoelectric sensor array (labeled 4), a PLC controller (labeled 5), a diversion mechanism (labeled 6), and a resealing station (labeled 7). The components interact and transmit commands via signal lines and a control bus, forming a complete closed-loop detection and response system. In practical applications, the carton to be inspected moves at a constant speed along the perforated conveyor line 1, with its top sealing area and bottom sealing area passing through a two-dimensional physical constraint channel formed by the upper sealing detection sensor 2 and the lower sealing perforated structure 3, respectively. This channel can be flexibly adjusted according to different carton specifications: the upper sealing detection sensor 2 is height-adjustable via an electric lifting bracket to accommodate the upper sealing limit of cartons of different heights; the lower sealing perforated structure 3 uses a sliding rail-type lateral adjustment mechanism to change the spacing between the perforated edges on both sides, thereby matching the bottom contour of cartons of different widths.

[0018] Once the cardboard box enters the detection area, the through-beam photoelectric sensor array 4 begins operation. This array consists of multiple sets of symmetrically arranged infrared emitters and receivers, densely packed along the conveying direction to form a high-resolution spatial penetration sensing surface. The edges of the cardboard box partially block the light beam, and the receiver generates a continuous spatial penetration signal sequence accordingly. This signal, after analog-to-digital conversion, is input to the edge computing unit to construct a dynamic occlusion feature map. Each element This indicates whether the i-th sensing point is occluded (1 for occlusion, 0 for no occlusion). Subsequently, the system calls an adaptive threshold segmentation algorithm to binarize this feature vector. This algorithm dynamically sets the segmentation threshold T based on the historical average contour data of the current batch of cartons. If the value is greater than T, it is marked as an abnormally prominent region, and the final output is an abnormally prominent region identifier sequence. ,in ∈{0,1}.

[0019] To further determine if there are any sealing abnormalities, the system retrieves the standard size template for the corresponding model from the carton specification parameter database. The template is an ideally shaped cardboard box outline occlusion vector without tape lifting or seal misalignment. Next, an improved Dynamic Time Warping (DTW) distance metric algorithm is used to calculate the deformation deviation between the current cardboard box outline VV and the standard template M. Traditional DTW algorithms are susceptible to local noise interference during path search. Therefore, this invention introduces a weighted local constraint mechanism to limit the path slope within a reasonable range and assign higher weights to boundary regions. Its mathematical expression is as follows: ,in, For the set of all legal aligned paths, The location-related weighting function is defined as follows:

[0020] Here, β>0 is the edge sensitivity coefficient, and δ is the boundary region length threshold. If the calculated... Exceeding the preset safety threshold If so, it is initially determined to be an abnormal sealing event, and the subsequent confidence assessment process is triggered.

[0021] like Figure 2 As shown, after an abnormal event is triggered, the system enters the second stage—anomaly confidence assessment. This stage is executed by the multimodal state fusion module, integrating multi-source information from the conveyor speed sensor, carton position encoder, and historical false alarm database. The core model uses a weighted exponential smoothing formula to filter instantaneous signal jitter, the expression of which is:

[0022] ,

[0023] in, The current level of confidence is the level of abnormality. This represents the current initial trigger state of the through-beam induction sensor (0 or 1), where α is an adaptive smoothing coefficient whose value is dynamically adjusted based on the vibration frequency of the conveyor line. f is the real-time vibration frequency. The reference stable frequency (usually set to the average vibration frequency of the equipment during no-load operation, such as 2 Hz), and k is the sensitivity adjustment constant, with an empirical value of 35. When the conveyor line experiences increased vibration due to mechanical wear or sudden load changes (i.e., f>), As α increases, the system becomes more reliant on the current instantaneous signal, improving response speed; conversely, when operating smoothly, α decreases, enhancing the smoothing effect of historical data and suppressing false alarms. If the current confidence level... If the confidence threshold θ (usually set to 0.85) is exceeded, the anomaly is confirmed to be valid, and the process proceeds to the third stage.

[0024] Upon confirming the anomaly, PLC controller 5 immediately initiates the shunt control logic. First, it uses a high-precision encoder installed on the conveyor line to acquire the real-time pose code of the carton, determining its current position and speed. Subsequently, an improved predictive trajectory mapping algorithm is invoked to calculate the timing of the diversion execution. This algorithm introduces a transport delay compensation factor Δt, the expression of which is:

[0025] Where L is the physical distance from the sensing point (i.e., the center position of the through-beam photoelectric sensor array 4) to the diversion point (i.e., the operating position of the diversion mechanism 6). The current conveying speed, The inherent response delay of the actuator (such as a pneumatic push rod or servo baffle) is determined by factory calibration, typically 50~150 ms. Based on this, the PLC controller 5 issues control commands Δt time in advance to ensure that the diversion mechanism 6 operates accurately when abnormal cartons arrive at the designated position, guiding them to the resealing station 7 and preventing them from being mixed into the qualified product flow.

[0026] Simultaneously, the system triggers an audible and visual alarm (not marked in the diagram, typically integrated into the PLC control cabinet), emitting a visual flashing and buzzer alert to remind operators to pay attention to the abnormal situation. The system automatically records complete information about this abnormal event, including the abnormality type (such as "top sealing tape lifted" or "bottom seal misalignment," identified by the distribution pattern of the abnormality protrusion area identifier sequence B), a precise timestamp (accurate to the millisecond level), and the carton batch number (obtained via RFID or barcode scanning), and uploads it to the central database.

[0027] To further enhance the intelligence level of the equipment, the edge computing unit continuously monitors continuous anomaly patterns. When multiple similar anomalies occur within the same time period (e.g., the top sealing tape of five consecutive cartons is lifted at the same location), the system activates the improved DBSCAN density clustering algorithm for fault trend analysis. This algorithm introduces a three-dimensional feature vector F=(t,x,y,c) based on the traditional DBSCAN, where t is the timestamp, (x,y) are the coordinates of the anomaly on the carton outline, and c is the anomaly category code. This is achieved by setting a minimum number of neighborhood points. Based on the spatiotemporal radius ϵ, high-density abnormal clusters are identified. If a cluster continues to grow and is concentrated in a specific equipment station (such as the belt roller area of ​​a carton sealing machine), it is determined to be a potential carton sealing equipment failure (such as insufficient belt tension, wear of pressure rollers, etc.), and a maintenance early warning work order is automatically generated and pushed to the equipment management system to achieve predictive maintenance.

[0028] In actual production line deployment, this system supports rapid switching between multiple carton specifications. Operators select the carton model corresponding to the current production order via the HMI interface. The system automatically loads the corresponding standard template M from the parameter library and simultaneously adjusts the height of the upper sealing detection sensor 2 and the width of the lower sealing cutout structure 3. The entire process can be completed within 30 seconds without downtime for debugging. Furthermore, the system has self-learning capabilities: for newly introduced carton models, if no pre-stored template exists, it can automatically construct an initial template by collecting 10-20 qualified samples, and continuously optimize the template accuracy through an online update mechanism during subsequent operation.

[0029] In summary, this invention constructs a two-dimensional physical constraint channel based on a perforated conveyor line and through-beam induction, combined with dynamic occlusion feature extraction, improved DTW deformation analysis, adaptive confidence assessment, and predictive diversion control, achieving high-precision, low-false-alarm, and fast-response detection and processing of carton sealing anomalies. Furthermore, relying on edge computing and density clustering technologies, it extends to equipment health monitoring and predictive maintenance, significantly improving the automation and intelligence level of packaging production lines.

Claims

1. A method for detecting abnormal sealant application in cardboard boxes based on through-beam induction, characterized in that, include: An adjustable physical detection channel is constructed based on the hollow conveyor line (1). A dual-dimensional size constraint benchmark is formed by the adjustable height upper sealing detection sensor (2) arranged symmetrically on the top and bottom and the adjustable width lower sealing hollow structure (3). When the carton runs along the conveyor line, the spatial penetration signal of the carton outline edge is collected in real time using the through-beam photoelectric sensor array (4). A dynamic occlusion feature vector is constructed based on the spatial penetration signal. The occlusion feature vector is binarized by the adaptive threshold segmentation algorithm to obtain the abnormal protrusion area identification sequence. Combined with the standard size template in the carton specification parameter library, the improved dynamic time warping (DTW) distance measurement algorithm is used to calculate the deformation deviation between the current carton outline and the standard template. If the deformation deviation exceeds the preset safety threshold, it is determined to be a sealing abnormal event. Based on the sealing anomaly event trigger signal, an anomaly confidence assessment model is constructed by integrating conveyor speed, carton position code, and historical false alarm rate data through a multimodal state fusion module. This model uses a weighted exponential smoothing formula to filter instantaneous signal jitter, and its expression is as follows: ,in, The current level of confidence is the level of abnormality. This represents the current initial trigger state of the through-beam induction sensor (0 or 1), where α is an adaptive smoothing coefficient whose value is dynamically adjusted based on the vibration frequency of the conveyor line. f is the real-time vibration frequency. The reference stable frequency is given by k, which is the sensitivity adjustment constant; when When θ is the confidence threshold, the abnormality is confirmed to be valid; after the abnormality is confirmed, the PLC (5) controller generates a diversion control command based on the real-time pose encoding of the carton, and accurately guides the abnormal carton to the resealing station (7) through an improved predictive trajectory mapping algorithm. The algorithm introduces a conveying delay compensation factor. Where L is the distance from the sensing point to the shunt point. For conveying speed, To mitigate the response delay of the actuator and ensure strict synchronization between the diversion action and the carton position, an audible and visual alarm is triggered, and the anomaly type, timestamp, and carton batch information are recorded. The edge computing unit performs cluster analysis on continuous anomaly patterns, and the improved DBSCAN density clustering algorithm is used to identify potential sealing equipment failure trends.

2. The method according to claim 1, characterized in that, The through-beam photoelectric sensor array (4) consists of multiple sets of infrared transmitters and receivers arranged symmetrically vertically, densely arranged along the transmission direction, used to generate a continuous spatial penetration signal sequence, which is then input into the edge computing unit after analog-to-digital conversion to construct a dynamic occlusion feature vector. Each element This indicates whether the i-th sensing point is blocked.

3. The method according to claim 1, characterized in that, The adaptive threshold segmentation algorithm dynamically sets the segmentation threshold T based on the historical average contour data of the current batch of cartons. If the value is greater than T, it is marked as an abnormally prominent region, and the final output is an abnormally prominent region identifier sequence. ,in ∈{0,1}.

4. The method according to claim 1, characterized in that, The improved Dynamic Time Warping (DTW) distance metric algorithm introduces a weighted local constraint mechanism to limit the path slope within a reasonable range and assigns higher weights to boundary regions. Its mathematical expression is as follows: ,in, For standard size templates, For the set of all legal aligned paths, The location-related weighting function is defined as follows: β>0 is the edge sensitivity coefficient, and δ is the boundary region length threshold.

5. The method according to claim 1, characterized in that, The reference stable frequency The average vibration frequency of the equipment during no-load operation is 2 Hz; the sensitivity adjustment constant k ranges from 3 to 5; and the confidence threshold θ ranges from 0.7 to 0.

85.

6. The method according to claim 1, characterized in that, The actuator response delay in the delivery delay compensation factor Δt The value range is 50 ms to 150 ms, obtained through factory calibration; the sensing point is the center position of the through-beam photoelectric sensor array (4), and the diversion point is the operating position of the diversion mechanism (6).

7. The method according to claim 1, characterized in that, The improved DBSCAN density clustering algorithm introduces a time-space-type three-dimensional feature vector F=(t,x,y,c), where t is the timestamp, (x,y) is the coordinate position of the anomaly on the carton outline, and c is the anomaly category code; by setting a minimum number of neighborhood points... Based on the spatiotemporal radius ϵ, high-density anomaly clusters are identified, and maintenance early warning work orders are generated when the anomaly clusters continue to grow and are concentrated in specific equipment workstations.

8. A carton sealing anomaly detection system based on through-beam induction, characterized in that, include: The first unit is used to construct an adjustable physical detection channel based on the hollow conveyor line (1). A two-dimensional size constraint benchmark is formed by the adjustable height upper sealing detection sensor (2) arranged symmetrically on the top and bottom and the adjustable width lower sealing hollow structure (3). When the carton runs along the conveyor line, the spatial penetration signal of the carton outline edge is collected in real time by the through-beam photoelectric sensor array (4). Based on the spatial penetration signal, a dynamic occlusion feature vector is constructed. The occlusion feature vector is binarized by the adaptive threshold segmentation algorithm to obtain the abnormal protrusion area identification sequence. Combined with the standard size template in the carton specification parameter library, the improved dynamic time warping (DTW) distance measurement algorithm is used to calculate the deformation deviation between the current carton outline and the standard template. If the deformation deviation exceeds the preset safety threshold, it is determined to be a sealing abnormal event. The second unit is used to construct an anomaly confidence assessment model based on the sealing anomaly event trigger signal, by integrating conveyor speed, carton position code, and historical false alarm rate data through a multimodal state fusion module. The model uses a weighted exponential smoothing formula to filter instantaneous signal jitter, and its expression is: ,in, The current level of confidence is the level of abnormality. This represents the current initial trigger state of the through-beam induction sensor (0 or 1), where α is an adaptive smoothing coefficient whose value is dynamically adjusted based on the vibration frequency of the conveyor line. f is the real-time vibration frequency. The reference stable frequency is given by k, which is the sensitivity adjustment constant; when At that time, confirm that the exception is valid; The third unit is used to confirm the abnormality. The PLC controller (5) generates a diversion control command based on the real-time pose encoding of the carton. The abnormal carton is accurately guided to the resealing station (7) through an improved predictive trajectory mapping algorithm. The algorithm introduces a conveying delay compensation factor. Where L is the distance from the sensing point to the shunt point. For conveying speed, To mitigate the response delay of the actuator, an audible and visual alarm is triggered, and the anomaly type, timestamp, and carton batch information are recorded. The continuous anomaly pattern is clustered and analyzed by the edge computing unit, and the improved DBSCAN density clustering algorithm is used to identify potential failure trends of the sealing equipment.

9. An electronic device, characterized in that, include: processor; A memory for storing processor-executable instructions; wherein the processor is configured to invoke the instructions stored in the memory to perform the method according to any one of claims 1 to 7.

Images