Computer-implemented system and method for automated detection of defects in composite package seals

Abstract

A system (1) for automatically detecting defects in a seal band (2) of a composite package is described. The seal band (2) is formed from a first band portion (3), a second band portion (4), and a layer of heat seal material (5) interposed between the first band portion (3) and the second band portion (4). The system (1) includes a vision system (10) that captures images of the seal band (2), and an analysis unit (12) operatively connected to the vision system (10) that analyzes one or more images acquired by the vision system (10) and determines the quality of the seal band (2) based on the images. The analysis unit (12) includes an artificial intelligence module (24) that uses a machine learning model to analyze the one or more images and detect the presence or absence of defects in the seal band (2).

Description

[Technical Field]

[0001] The present invention relates to a computer-implemented system and method for automated detection of defects in the closure of composite packages filled with certain pourable products, such as certain pourable food products. [Background technology]

[0002] As is well known, many liquid or pourable foods, such as fruit juice, UHT (ultra-high temperature processed) milk, wine, tomato sauce, etc., are packaged in sealed composite packages made from sterilized multi-layer packaging materials.

[0003] A typical example is the Tetra Brik Aseptic®, a parallelepiped-shaped package for pourable food products, which is formed by sealing and folding a laminated strip of packaging material. The packaging material has a multi-layer structure, with a base layer, such as paper or paperboard, covered on both sides with layers of heat-sealable plastic material, such as polyethylene. In the case of aseptic packaging for long-term storage products, the packaging material comprises a layer of oxygen barrier material, such as aluminum foil, which is laminated with a layer of heat-sealable plastic material and covered with a further layer of heat-sealable plastic material that ultimately defines the inner surface of the package that contacts the food.

[0004] Such packages are typically produced on fully automatic packaging machines. In some machines, a web of packaging material is passed through a sterilization unit of the packaging machine to sterilize the web. The sterilized web of packaging material is then held in an isolation chamber, folded lengthwise and sealed to form a tube, and advanced further. The tube is filled with the pourable product, sealed transversely along equally spaced cross sections, and cut as the tube is conveyed through the package-forming device of the packaging machine.

[0005] More particularly, the package forming apparatus includes a plurality of form and seal assemblies which, in use, form, transversely seal and cut the tube to form a single package.

[0006] The forming and sealing assembly comprises a sealing unit for transversely sealing the tube and for obtaining the transverse seal by locally compressing and heating corresponding portions of the layer of heat-sealable plastic material, the heating being performed, for example, by the sealing unit generating ultrasonic vibrations.

[0007] The form-and-seal assembly further includes a cutting device for severing the transverse seals and separating the sealed composite package from the tube to form a first transverse seal band and a second transverse seal band, the first seal band belonging to the immediately preceding formed and sealed composite package and the second seal band belonging to the immediately succeeding formed and sealed composite package.

[0008] The quality of the seal, especially the transverse seal band, plays a vital role in the overall quality of the package, as damage to the packaging material or an incorrect seal can lead to a breach of the protective environment. Therefore, it is important to monitor the quality of the seal during the packaging process carried out on the packaging machine.

[0009] Currently, to perform test measurements for evaluation of seal quality, sample composite packages are taken from a plurality of formed composite packages and analyzed at defined time intervals.

[0010] In particular, one or more samples are prepared from the sample composite package, each having a transverse seal band.

[0011] After preparing each sample, the technician uses stretch pliers or other suitable tool to tear the seal band to expose the seal pattern. The technician then inspects the exposed seal for defects such as channels (i.e., leaks), burn marks, or plastic chunks.

[0012] Test results are highly dependent on the experience of the technician performing the test, and repeating the same test takes time and must be performed on a limited number of samples.

[0013] Although known seal quality control methods generally provide satisfactory results, a need is felt in the industry for further improvements in composite package seal quality control measures and improved methods for detecting defects. Summary of the Invention [Problem to be solved by the invention]

[0014] It is therefore an object of the present invention to provide a solution that allows reliable and reproducible automatic defect detection in the seals of composite packages. [Means for solving the problem]

[0015] According to the present invention, there is provided a system, method and computer program as defined in the accompanying claims. [Brief explanation of the drawings]

[0016] Non-limiting embodiments of the present invention will now be described, by way of example only, with reference to the accompanying drawings, in which:

[0017] [Figure 1] 1 is a schematic diagram illustrating an automatic defect detection system according to an embodiment of the present invention. [Figure 2] 2 is a schematic top view of a seal band inspected by the system of FIG. 1; [Figure 3] FIG. 10 shows a processed image of the seal band at the end of the peeling process, showing possible defects. [Figure 4] FIG. 10 shows a processed image of the seal band at the end of the peeling process, showing possible defects. [Figure 5] FIG. 2 is a schematic block diagram of an analysis unit of the system of FIG. 1. [Figure 6]FIG. 10 is an example of an output from the analysis unit showing an image of a seal band with detected defects and associated confidence scores displayed. [Figure 7] 1 is a top perspective view, with portions removed for clarity, of an automatic test equipment system in accordance with one embodiment of the present invention; [Figure 8A] 1A-1C are schematic diagrams of a sealing band at different times during a peeling procedure. [Figure 8B] 1A-1C are schematic diagrams of a sealing band at different times during a peeling procedure. [Figure 8C] 1A-1C are schematic diagrams of a sealing band at different times during a peeling procedure. [Figure 9] FIG. 1 shows force curves measured during a peeling procedure. DETAILED DESCRIPTION OF THE INVENTION

[0018] Figure 1 shows a schematic diagram, generally designated 1, of an automatic defect detection system for determining the quality of the seal, in particular the transverse sealing band 2, of a pourable product filled in a composite package (not shown), for example filled with a pourable food product, such as (pasteurized) milk, fruit juice, wine, tomato sauce, salt, sugar, etc.

[0019] The composite package may be formed from a multi-layer packaging material, which may comprise a layer of fibrous material, such as paper or paperboard, and at least two layers of heat-sealable plastic material (e.g., polyethylene) sandwiching the fibrous layer, one of the two layers of heat-sealable plastic material forming the inner surface of the composite package that contacts the pourable product.

[0020] The packaging material may further comprise a layer of gas and light barrier material (e.g., aluminum foil or ethylene vinyl alcohol (EVOH) film), particularly configured to be disposed between the layer of heat-sealable plastic material and the layer of fibrous material. Preferably, the packaging material may comprise a further layer of heat-sealable plastic material sandwiched between the gas and light barrier layer and the fibrous layer.

[0021] The sealed composite package may comprise at least one longitudinal sealing band and in particular two transverse sealing bands.

[0022] In some non-limiting embodiments, the sealed composite package may be obtained from a single blank formed into a sleeve having a longitudinal seal band, and the sleeve is then formed into the sealed composite package, requiring the formation of at least one, and particularly two, transverse seal bands.

[0023] In another non-limiting embodiment, the multi-layer packaging material may be provided in the form of a web. In particular, the web may include a plurality of repeating patterns, each pattern defining a blank for forming a respective sealed composite package. In such an embodiment, the web is formed into a tube, sealed longitudinally, filled with a pourable product, sealed transversely, and cut along equally spaced cross sections to form respective seal bands.

[0024] In some possible non-limiting examples, the sealed composite package includes a longitudinal seal band and a pair of transverse seal bands disposed on opposite sides of the sealed composite package, in particular, one transverse seal band defining an upper transverse seal band and the other transverse seal band defining a lower transverse seal band.

[0025] As shown schematically in FIG. 2, each sealing band 2 (particularly the lateral sealing bands) may be formed from a first band portion 3, a second band portion 4, and a heat-sealable material layer 5 interposed between the first band portion 3 and the second band portion 4.

[0026] The sealing band 2 can be formed by overlapping two portions of a multi-layer packaging material and sealing the two overlapping portions.

[0027] The first band portion 3 and the second band portion 4 may each comprise a layer of textile material.

[0028] The heat sealable material layer 5 is formed by sealing two overlapping portions together and fusing the respective heat sealable plastic material layers together. In particular, the heat sealable material layer 5 may be formed from at least a first portion of heat sealable material 6 and a second portion of heat sealable material 7.

[0029] Returning to Figure 1, the automatic defect detection system 1 comprises a vision system 10 having an imaging device 11 (e.g. a camera) configured to acquire images of the sealing band 2 (e.g. positioned at a distance above the sealing band 2), and an analysis unit 12 operatively connected to the vision system 10 and configured to analyze one or more images acquired by the imaging device 11 to determine the quality (and assess possible defects) of the sealing band 2. In particular, the analysis unit 12 may be configured to receive and at least temporarily store these images.

[0030] The analysis unit 12 may be implemented by a digital processing unit configured to execute a computer program consisting of a plurality of software instructions, such as a microprocessor, a microcontroller, a PLC (Programmable Logic Controller) or any suitable processing and computing unit.

[0031] In a possible embodiment, as described below, the imaging device 11 of the vision system 10 may be configured to take a series of images capturing the changes over time in the peeling of the sealing band 2 during the break test, in particular the changes over time in the shape of the heat sealable material layer 5.

[0032] The break test can be carried out by an automatic test apparatus (described in detail below) comprising a first gripping device for gripping the first band portion 3, a second gripping device for gripping the second band portion 4, and an actuation device for causing relative movement between the first gripping device and the second gripping device to detect peeling of the sealing band 2.

[0033] In this specification, "peeling" refers to the controlled separation of specific portions of the sealing band 2 from each other, specifically, peeling occurring between the first band portion 3 and the second band portion 4.

[0034] The analysis unit 12 may be configured to analyze one or more images taken during the peeling of the sealing band 2 in order to determine the quality of the same sealing band 2, in particular to detect defects that may affect the sealing quality (e.g. channels, burns, plastic chunks, etc. that compromise the sterility of the composite package).

[0035] In particular, as will be described below, the analysis unit 12 may be configured to analyze the images using artificial intelligence algorithms.

[0036] The same analysis unit 12 may be connected to a control unit 100 of a packaging machine (not shown) that produces the sealed composite packages and is configured to adjust production, and in particular the formation of the seal, based on the results of the analysis performed.

[0037] As shown in FIG. 1, the automatic defect detection system 1 may further comprise an illumination device 14 configured to illuminate (using polarized or unpolarized light) the seal band 2 to be inspected.

[0038] 1 shows only one illumination device 14. However, preferably, a pair of illumination devices may be provided on opposite sides of the imaging device 11, which may improve the quality of the images captured by the imaging device.

[0039] In a possible embodiment, the automatic defect detection system 1 may further comprise a polarizing filter 15 interposed between the imaging device 11 of the vision system 10 and the sealing band 2, which is positioned at a distance (H) from the same sealing band 2. In the example shown, light is irradiated onto the polarizing filter 15 at an angle of incidence α of less than 90 degrees (preferably less than 60 degrees or less than 45 degrees) and reflected onto the imaging device 11.

[0040] In another embodiment, the light may pass through the sample under test (along with the corresponding sealing band) and then be received by the imaging device 11 .

[0041] In yet another embodiment, the polarizing filter 15 may be integrated into the optics of the imaging device 11 (in which case no additional polarizing filter outside the imaging device 11 is required).

[0042] In this regard, the applicant has recognised that implementing a technology known as Polarsens® in the imaging device 11 will improve image quality and thereby improve seal inspection and characterisation by the analysis unit 12.

[0043] PolarSense is a CMOS image sensor technology in which multiple polarizers with different angles are formed on the chip (specifically, underneath the on-chip lens) during the semiconductor manufacturing process, enabling highly accurate pixel alignment. Using this technology, a single camera device equipped with a polarization CMOS sensor can capture four differently polarized images in a single shot, thereby increasing the obtainable information about the sample.

[0044] The four images can be combined using appropriate software recombination algorithms. Combining the four images generates additional digital data content, which can facilitate analysis and improve results.

[0045] Applicants have shown that this technique can be applied to quality control applications, particularly seal inspection (including break testing and analysis of plastic bags), to improve analytical results.

[0046] In particular, tests carried out by the applicant have shown that the use of certain recombination algorithms allows for increased accuracy in defect detection and improved analysis of stress and material properties (all information that can be used to assess the quality of the sealing process). These recombination algorithms are Linear Polarization Degree, Linear Polarization Angle, and Glare Reduction.

[0047] In particular, when using glare reduction algorithms, different glosses of materials can highlight areas that exhibit different sealing properties, identifying areas with particularly sealed inner layers, while areas with non-uniform gloss / reflectivity can indicate a sealing defect.

[0048] In this regard, Figure 3 shows an image taken with a camera equipped with integrated PolarSense technology and processed with a glare reduction algorithm, taken at the end of the fracture test after the sealing band 2 had been subjected to the peel process in a case where the sealing band 2 had defects (specifically the unsealed areas).

[0049] Figure 4 shows an image of the same seal band 2, but processed with a linear polarization algorithm. Again, the defects (unsealed areas) are highlighted by a clearly different reflection pattern.

[0050] The combination of PolarSense technology with appropriate image processing algorithms (which can be implemented in the analysis unit 12) improves seal characterization (both in break testing and plastic bag inspection) and increases the efficiency of machine learning algorithms for automatic defect detection and classification.

[0051] With reference to FIG. 5, a possible embodiment of the analysis unit 12 of the automatic defect detection system 1 will be described in more detail.

[0052] The analysis unit 12 comprises an input module 20 configured to receive image data relating to images acquired by the vision system 10 of the same automatic defect detection system 1 .

[0053] In a preferred embodiment, the input module 20 may be configured to receive sensor data from multiple sensors 21. In this case, the sensors 21 may be part of the automatic defect detection system 1.

[0054] The sensor 21 may be configured to provide sensor data related to the inspection of the sealing band 2 , in particular sensor data related to the sealing test procedure and / or the operation of the packaging machine in forming the sealing band 2 .

[0055] For example, during production of a packaging machine, the sensor data acquired by the sensor 21 may relate to one or more of: the forming and sealing device; the product level, pressure and / or flow rate; and the machine parameters and settings.

[0056] Furthermore, as described above, when a break test is performed by an automated test device to inspect the sealing quality, the sensor data relates to the break force applied by the automated test device, and the sensor may include a force sensor (e.g., a force cell and / or an encoder connected to a drive device of the test device, etc.).

[0057] In this case, the sensor data may include a force profile, such as a time-dependent force curve and / or a position-dependent force curve, and in particular the position may indicate the relative distance between the first clamping device and the second clamping device and may be related to the force acting during peeling of the sealing band 2 and / or the force acting on the sealing band 2 by the drive device as the drive device moves the first clamping device and the second clamping device away from each other.

[0058] The above sensor data may include a stress / strain displacement profile of the material of the sealing band 2. Additionally, the difference in force applied by the first and second clamping devices of the automated test equipment may be included in the same sensor data.

[0059] The above sensor data may be provided automatically to the analysis unit 12 by the same automated test equipment.

[0060] The sensors may be associated with a packaging machine and the sensor data may include, for example, processed data relating to the operation of a sealing device that seals the composite package.

[0061] In particular, the input module 20 may be configured to correlate the timing of detection of images by the vision system 10 with sensor data (e.g., force profiles) so that correlations (e.g., time correlations) between image data and sensor data may be determined.

[0062] The analysis unit 12 further comprises a pre-processing module 22 connected to the input module 20 and adapted to pre-process the data received from the input module 20 (image data and, in a preferred embodiment, sensor data).

[0063] In particular, the pre-processing module 22 may perform one or more pre-processing steps on the image data, such as filtering (to remove unwanted data), reshaping, normalizing, re-coloring, and / or cropping (to divide the image into regions of interest). Pre-processing of the sensor data may include operations such as normalization, amplification, and filtering.

[0064] As explained above, pre-processing of the image data may include performing a suitable recombination algorithm on the images acquired by the imaging device 11 (in this case provided with an integrated polarizing filter).

[0065] The analysis unit 12 further comprises an artificial intelligence (AI) module 24 connected to the pre-processing module 22 and configured to receive the pre-processed input data.

[0066] The artificial intelligence module 24 is configured to analyze the pre-processed input data using, for example, a machine learning model based on a neural network, to automatically determine the quality of the seal band 2, and in particular to detect the presence or absence of channels (i.e., leaky seals), burn marks, plastic lumps, and other defects.

[0067] In a possible embodiment, the artificial intelligence module 24 is configured to process the image data and the sensor data jointly or in conjunction to assess the quality of the seal band 2 using a multi-modal approach (i.e. based on data of different nature). The sensor data may comprise other data in addition to the image data (i.e. images of the seal band).

[0068] The machine learning model may be a neural network. The machine learning model may be a deep learning model. A deep learning model should be understood as a model based on deep learning. Deep learning should be understood as machine learning in which hypotheses take the form of complex algebraic circuits with adjustable connection strengths. The term "deep" means that the circuits are typically organized into many layers, and the computational path from input to output has multiple layers.

[0069] The machine learning model may be implemented as a neural network, such as a convolutional neural network (CNN), that performs deep learning analysis for the detection of labeled defects. The machine learning model may be trained to output quality measures based on input data. The machine learning model may be trained using supervised learning, i.e., the machine learning model may be trained based on training input data (e.g., image data or sensor data associated with quality measures, or image data associated with quality measures). In particular, the quality measures may indicate the presence or absence of one or more defects in the input data (e.g., image data). Optionally, the quality measures may indicate the type of defect.

[0070] The quality measure may be a range of continuous values, such as ranging from 0 to 1, or the quality measure may be a set of discrete values.

[0071] For defects or imperfections in the seal band 2, the value of the quality measure may indicate the severity of the detected defect or imperfection. Alternatively, the quality measure may be a binary value such as "OK" or "not-OK".

[0072] The quality measurements may further comprise information indicative of the type of defect or imperfection, if any, present in the sealing band 2. Information indicative of the type of defect may, for example, be that the seal is not properly formed or that the packaging material is damaged, either of which may result in an unsterile package.

[0073] The quality metric may indicate a confidence score for the output results from the machine learning model, i.e., the degree of confidence that the defects or malfunctions detected by the machine learning model are in fact defects or malfunctions.

[0074] In particular, the artificial intelligence module 24 is configured to interpret and aggregate input data (including data of different nature, such as image data and sensor data) and generate output results relevant to quality control based on certain parameters such as confidence level, defect location, defect type (defect class), etc.

[0075] In particular, any seal defects or imperfections that may occur in the seal portion of the composite package may be displayed in the output results.

[0076] Each defect may be characterized through the final layer of the neural network with a bounding box containing the corresponding spatial information (defect location) and a score, allowing it to be classified as a specific defect among various cases present in the field. The optimal threshold that maximizes the accuracy of the algorithm is determined based on the score, which indicates the confidence in the defect prediction.

[0077] In a possible embodiment, the supervised neural network is trained on a training dataset comprising example data generated directly on the packaging machine, which example data is annotated and labeled with the help of an expert operator, in particular in terms of variations, defects, warnings, objects of interest and suspicious areas.

[0078] As explained above, in a possible embodiment, a supervised neural network is developed to process image data from the visual domain and sensor (or process) data, such as stress / strain displacement curves, using a multi-domain architecture that is able to combine the results of two AI architectures operating in different data domains.

[0079] Thus, the training data set may comprise both image data and sensor data, both of which may be independently labeled based on possible defects or anomalies.

[0080] As shown in FIG. 5, the automatic defect detection system 1 may further include a deviation monitoring module 26 connected to the artificial intelligence module 24 and configured to monitor potential deviations in the output results from the same artificial intelligence module 24.

[0081] In particular, the deviation monitoring module 26 is configured to evaluate deviations in the performance of the machine learning model over time and determine the need for retraining of the machine learning model if the deviations exceed a predetermined threshold.

[0082] The deviation monitoring module 26 may be configured to evaluate the actual performance of the machine learning model relative to its past performance (e.g., performance in the time interval one hour prior to the current time) using a suitable performance metric (e.g., a metric indicating how well the machine learning model performs on the input data).

[0083] For example, deviation monitoring module 26 may detect a change in the reliability or accuracy of the output results (eg, a deviation of the output score from a range of 98%-99% to a significantly different range of 90%-91%).

[0084] Additionally or alternatively, deviation monitoring module 26 may be configured to evaluate deviations in performance relative to output results associated with a training dataset (e.g., deviations associated with quality measures that increase or decrease exponentially over time).

[0085] The deviation monitoring module 26 may be configured to identify possible deviations in the nature or characteristics of the input data, for example, compared to a current batch of input data based on a comparison with a previous batch of input data (received in an earlier time interval compared to the current batch) and / or a comparison with a training data set.

[0086] For example, changes in the distribution of a particular property (eg, color) of the image data are evaluated by the deviation monitoring module 26 .

[0087] Possible causes of significant deviations could be, for example, changes in the materials used to manufacture the sealed composite package, or changes in defects or imperfections that may occur in the seal band 2 being inspected (including new types of defects that the machine learning model did not learn about during training).

[0088] It is emphasized again that appropriate criteria and statistical methods can be implemented to evaluate the above deviations.

[0089] If the deviation monitoring module 26 detects significant deviations (deviations above a certain threshold) in the output results or input data (either or both), it may be determined that retraining of the machine learning model is necessary.

[0090] The machine learning model may be retrained using different training data, and this process may be repeated until the same machine learning model reaches a certain level of performance (e.g., greater than 95% accuracy), after which it can be used again to test seal quality resulting from the manufacturing process.

[0091] As shown in FIG. 5, the automatic defect detection system 1 may further comprise an output module 28 configured to output the results of the analysis performed by the artificial intelligence module 24 to provide feedback on the seal quality.

[0092] According to a possible embodiment, the output module 28 may be configured to interface with a human-machine interface to display the results of the evaluation performed by the artificial intelligence module 24 to a human operator.

[0093] For example, Figure 6 shows an output that can be provided by the output module 28, in particular an image of the inspected seal band 2, where areas where defects were detected by the machine learning model are highlighted along with the corresponding accuracy level.

[0094] The same output module 28 may be operably connected to the control unit 100 of the packaging machine (or directly to the sealing device of the same packaging machine) and configured to provide feedback data for directly controlling the operation of the same packaging machine and / or the sealing device, in particular depending on the determined quality of the inspected sealing band 2.

[0095] In particular, the values ​​and / or settings of the parameters used in the sealing procedure performed by the sealing device may be suggested to the operator by the output module 28 or may be sent directly to the control unit 100 of the packaging machine depending on the results of the seal quality evaluation.

[0096] For example, the parameters or settings may include the value of the heating temperature or electrical current used in forming the sealing band 2 by the sealing device of the packaging machine.

[0097] The same output module 28 may further be configured to generate alarms or warnings based on the assessment of the seal quality, and to directly or indirectly cause the packaging machine to shut down in case of serious malfunctions.

[0098] With reference to FIG. 7, the automatic test device 30 for assessing the quality of the sealing band 2 will now be described in more detail.

[0099] In particular, the automatic defect detection system 1 may form a component of this automatic test equipment 30, as shown diagrammatically in the same FIG.

[0100] Alternatively, the same automatic defect detection system 1 may be located externally and communicatively connected to the automatic test equipment 30 .

[0101] As shown in FIG. 7, the automatic test equipment 30 - a first clamping device 32 and a second clamping device 33 configured to clamp a sample including a sealing band 2 (not shown) to be inspected between the first clamping device 32 and the second clamping device 33; an actuation device 35 configured to drive a relative movement of the first clamping device 32 and the second clamping device 33 so as to cause peeling of the same sealing band 2; Equipped with.

[0102] In particular, the actuation device 35 is configured to change the relative position of the first clamping device 32 and the second clamping device 33 in a direction D that is laterally, in particular (substantially) perpendicular to the sealing band 2. That is to say, the actuation device 35 is configured to apply a force to the sealing band 2 in a laterally, in particular perpendicular, direction.

[0103] Thereby, the actuation device 35 is configured to cause peeling so as to separate the first band portion 3 and the second band portion 4 from one another, which in particular makes it possible to observe the layer of thermoadhesive material 5 and to inspect the seal quality.

[0104] The actuation device 35 is capable of repeatedly performing measurements on different samples under controlled and defined conditions. In particular, the actuation device 35 may be configured to apply defined force profiles (force per time) and / or motion profiles (velocity per time) to the different sealing bands 2. Such defined force profiles and / or motion profiles can be repeatedly applied.

[0105] In particular, the actuation device 35 comprises: - a first carriage 37 holding the first clamping device 32; - a second carriage 38 holding the second clamping device 33; and an actuator 39, in particular an electric motor, operatively connected to the first carriage 37 and / or the second carriage 38 and configured to control the relative position of the first carriage 37 and the second carriage 38 and thereby the relative position of the first clamping device 32 and the second clamping device 33; may also be provided.

[0106] The actuation device 35 may further comprise one or more tracks 40, in particular having a linear shape and configured to guide the movement of the first carriage 37 and the second carriage 38.

[0107] In particular, the first carriage 37 and the second carriage 38 comprise slide elements 41 that movably couple the first carriage 37 and the second carriage 38 to one or more tracks 40 .

[0108] 7, the imaging device 11 of the vision system 10 of the automatic defect detection system 1 is coupled to the housing 36 of the test fixture 30 and is configured to capture images of the sealing band 2 as it peels off. In particular, it is configured to capture a series of images that represent the changes over time in the peeling of the sealing band 2, in particular the changes over time in the shape of the thermoadhesive material layer 5.

[0109] In this regard, FIGS. 8A, 8B and 8C show schematic diagrams of exemplary shapes of the sample together with the sealing band 2 at different points in the peeling process.

[0110] The housing 36 of the test fixture 30 is provided with an illumination device 14 (e.g., an LED bar illuminator) configured to illuminate the seal band 2. As shown in FIG. 7, a pair of illumination devices 14 are preferably disposed on opposite sides of the imaging device 11.

[0111] In this embodiment, the analysis unit 12 of the automatic defect detection system 1 is disposed within a housing 36. Although not shown, the analysis unit 12 may be operatively connected to a control unit of the automatic test equipment 30 to control its overall operation.

[0112] The automatic testing equipment 30 comprises a detection device for detecting at least one force acting when peeling the sealing band 2 and / or for detecting a relative position between the first clamping device 32 and the second clamping device 33. The analysis unit 12 may further be configured to receive the at least one force and / or position as input data, and the artificial intelligence module 24 may be configured to process the image data and the at least one force and / or position using a machine learning model.

[0113] In particular, the force detection device (not shown) may be configured to detect the peeling force acting when the sealing band 2 is peeled and / or the force acting when the actuating device 35 moves the first clamping device 32 and the second clamping device 33 away from each other.

[0114] The force sensing device may comprise one or more force sensors and / or encoders connected to the actuator 39 .

[0115] In this embodiment, the detection device is therefore part of a sensor 21 which, as previously described, forms part of the automatic defect detection system 1 .

[0116] For example, the force detection device can detect a force curve as shown in Figure 9. Points I, II, and III on the force curve indicate the instants at which the vision system 10 captures images of the sealing band 2, and at these capture times, the sealing band 2 may have the exemplary shapes shown in Figures 8A, 8B, and 8C, respectively.

[0117] In some preferred non-limiting embodiments, the automated defect detection system 1 (and test apparatus 30) may be part of a packaging plant configured to form sealed composite packages filled with pourable products.

[0118] In particular, the automatic defect detection system 1 (and test device 30) may be configured to determine the quality of one or more sealing bands 2 of a sample composite package taken from the overall production, particularly at predetermined time intervals and / or at specific events.

[0119] The packaging plant - a packaging device having one or more sealing bands 2 (in particular two transverse sealing bands and / or one longitudinal sealing band) for forming a plurality of sealed composite packages filled with a pourable product; and - a sampling unit for sampling one or more sealed composite packages from a plurality of sealed composite packages; may also be provided.

[0120] In particular, the packaging machine produces a plurality of composite packages that advance along an advance path, and the sampling unit is positioned along the advance path downstream of the packaging machine.

[0121] The automatic testing device 30 is used to determine the quality of one or more seal bands 2 of one or more sealed composite packages sampled by the sampling unit. In particular, one or more samples are prepared from the sampled composite packages.

[0122] More particularly, the packaging machine may be configured to produce sealed composite packages from multi-layer packaging material, particularly provided in web form.

[0123] Furthermore, the packaging plant, in particular the packaging machine, may comprise a control unit (the above-mentioned control unit 100) configured to control the operation of the packaging machine.

[0124] Furthermore, the packaging apparatus may comprise a lateral sealing device for generating lateral seal bands of the sealed composite package. In particular, the control unit is operatively connected to the lateral sealing device and configured to control the operation of the lateral sealing device in response to the quality of the one or more seal bands 2 measured by the automatic defect detection system 1.

[0125] More particularly, the packaging device comprises: - a web transport device which transports the web along a web advancement path in a known manner to a forming station where, in use, the web is formed into a tube; - isolation chambers with an internal environment, in particular a sterile internal environment containing sterile gas, separated from the (hostile) external environment; - a tube forming and sealing device that, in use, forms the web into a tube within an internal environment and longitudinally seals the tube within the internal environment; - a filling device for filling the tube with a pourable product; and - a package forming device configured in use to advance the tube and transversely seal and transversely cut the tube to form a composite package; may also be provided.

[0126] In particular, the package forming apparatus may comprise a lateral sealing device.

[0127] Furthermore, the packaging apparatus may comprise a sterilization unit configured to sterilize the web in use at a sterilization station, which may in particular be located upstream of the forming station along the advancement path of the web.

[0128] Preferably, the tube forming and sealing device may be positioned so that the tube exhibits a vertical orientation.

[0129] More particularly, the tube forming and sealing device may be configured to gradually fold the web into a tube, particularly by overlapping longitudinal edges of the web, thereby forming a longitudinal joint in the tube in use, particularly a section of the longitudinal joint that defines a longitudinal seal band in the sealed composite package.

[0130] More specifically, the tube forming and sealing device may comprise at least two forming ring assemblies, particularly disposed within the isolation chamber, configured to cooperate with each other to gradually fold the web into a tube, particularly by overlapping longitudinal ends of the web with each other.

[0131] Furthermore, the tube forming and sealing device may comprise a sealing head, particularly arranged in the isolation chamber, configured to longitudinally seal the tube, particularly along the longitudinal joint.

[0132] Additionally, the tube forming and sealing device may include a pressure assembly configured to apply a mechanical force to the longitudinal joint.

[0133] The package forming device - a plurality of forming and sealing assemblies, each assembly configured to at least form (form) the tube, transversely seal the tube, and in particular transversely cut the tube; and - a transfer unit configured to advance the forming and sealing assembly; may also be provided.

[0134] More specifically, each molded and sealed assembly: - a molded shell configured to at least partially define the shape of the composite package; and - a sealing unit adapted to compress the tube at least laterally, in particular to flatten and press it, and to laterally seal the tube, in particular during the tube's advancement and between two composite packages; may also be provided.

[0135] In particular, the lateral sealing device may comprise a sealing unit of a forming and sealing assembly. Preferably, the control unit 100 may be configured to control the operation (e.g., power, force, etc.) of the sealing unit depending on the determined quality of the lateral sealing band 2.

[0136] Additionally, the forming and sealing assembly may include a cutting device for transversely cutting the tube.

[0137] Preferably, the sealing device is configured to form the main sealing band, and in particular the cutting device may be configured to cut the main sealing band transversely. More preferably, after cutting, each main sealing band forms a transverse sealing band 2 for the preceding composite package and a transverse sealing band 2 for the following composite package, in particular the latter still being part of the tube.

[0138] More particularly, each molded shell may comprise at least a first half shell and a second half shell configured to at least partially define the shape of the composite package. In particular, the first half shell and the second half shell may be configured to contact the tube from opposite sides of the tube.

[0139] Furthermore, the sealing unit may comprise a first part (e.g., a sonotrode), in particular associated with the first half-shell, and a second part (e.g., an anvil), in particular associated with the second half-shell, which may be configured to cooperate with one another, in particular to laterally compress and laterally seal the tube.

[0140] In particular, the control unit 100 may be configured to control the operation of, for example, the first part of the sonotrode depending on the determined quality of the sealing band 2 .

[0141] In use, the packaging plant produces sealed composite packages filled with pourable products, particularly pourable food products.

[0142] The advantages of this solution will be apparent from the above description.

[0143] In particular, the automatic defect detection system 1 can evaluate the sealing quality of composite packages, in particular the sealing quality of transverse seal bands, with high accuracy and reproducibility, especially without relying on the operator's experience and / or different test conditions.

[0144] Obviously, modifications may be made to the solutions described herein without departing from the scope of protection defined in the appended claims.

[0145] In particular, it is re-emphasized that the present solution is applicable to seal inspections independent of the type of seal (longitudinal or transverse), the method of seal production (e.g. hot air, induction heating, ultrasonic heating, etc.) and the test conditions (e.g. both in the case of break tests and in the case of analysis of plastic bags).

[0146] Furthermore, in a possible embodiment, the automatic defect detection system 1 may comprise a reading (or scanning) device configured to read (or scan) the identification code (or identification mark) applied to the sample and the sealing band 2. The identification code may be a QR code, a bar code, a number, or any other type of code. The reading device may be implemented by the vision system 10 described above (in which case the camera of the vision system 10 reads the identification code) or may comprise a dedicated camera.

[0147] In particular, the analysis unit 12 may be configured to associate the image of the sealing band 2 captured by the vision system 10 with this identification code, thereby enabling information related to the sample (e.g., information about the manufacturing process of the sample) to be associated with the image captured by the vision system 10, thereby enabling improved traceability and quality analysis of the sample.

Claims

1. A system (1) for automatically detecting defects in a sealing band (2) of a composite package formed by a packaging machine, the sealing band (2) comprising a first band portion (3), a second band portion (4), and a layer of heat sealable material (5) interposed between the first band portion (3) and the second band portion (4), the system (1) comprising: a vision system (10) configured to acquire images of said sealing band (2) during a seal test procedure; an analysis unit (12) operatively connected to the vision system (10) and configured to analyze one or more images acquired by the vision system (10) and determine a quality of the sealing band (2) based on the one or more images; Equipped with the analysis unit (12) comprises an artificial intelligence module (24) configured to analyze the one or more images and detect defects in the sealing band (2) using a machine learning model, the machine learning model being trained by a training dataset comprising image data related to quality measurements; System (1). A system according to any one of claims 1 to 9.

2. at least one sensor; the sensors generate sensor data indicative of a sealing test procedure and / or operation of a packaging machine during formation of the seal band (2); the analysis unit (12) is configured to receive as input data both image data provided by the vision system (10) and the sensor data; the artificial intelligence module (24) is configured to process the image data and the sensor data using the machine learning model in a multimodal approach to assess the presence or absence of defects in the sealing band (2) and preferably identify the type of the defects. The system of claim 1 .

3. the machine learning model is trained based on a training dataset including sensor data and image data related to the quality measurement; The system of claim 2 .

4. the analysis unit (12) further comprises a deviation monitoring module (26) that evaluates deviations in the performance of the machine learning model over time and determines the need for retraining of the machine learning model if the deviations exceed a predetermined threshold. The system according to any one of claims 1 to 3.

5. the deviation monitoring module (26) is configured to monitor deviations in output results from the artificial intelligence module (24) and / or to monitor deviations in characteristics of the input data compared to previous batches of input data and / or training data sets; The system of claim 4.

6. and an automatic test device (30) configured to cause peeling of the sealing band (2), and the vision system (10) configured to capture one or more images of the sealing band (2) during peeling of the sealing band (2) by the automatic test device (30). The system according to any one of claims 1 to 5.

7. The automatic test equipment (30) a first clamping device (32) and a second clamping device (33), the first clamping device (32) and the second clamping device (33) being configured to clamp the sealing band (2) between the first clamping device (32) and the second clamping device (33); an actuation device (35) configured to drive relative movement of the first clamping device (32) and the second clamping device (33) to cause peeling of the sealing band (2); Equipped with The system of claim 6.

8. the automatic testing device (30) further comprises a detection device for detecting at least one force acting when peeling the sealing band (2) and / or for detecting a relative position between the first clamping device (32) and the second clamping device (33); the analysis unit (12) is configured to receive at least one of force and / or position as input data; the artificial intelligence module (24) is configured to process the image data and the at least one force and / or position using the machine learning model; A system according to any one of claims 1 to 7.

9. The imaging device (11) is provided with at least one polarizing filter (15), and at least one image is configured to capture the sealing band (2) through the polarizing filter (15). A system according to any one of claims 1 to 8.

10. the analysis unit (12) further comprises an output module (28), which is configured to output results relating to sealing defects based on the quality of the inspected sealing bands to an operator via a user interface and / or to a control unit (100) of the packaging machine as feedback data. A system according to any one of claims 1 to 9.

11. 1. A packaging plant for packaging pourable products into sealed composite packages, comprising: The packaging plant comprises: a packaging machine having one or more sealing bands (2) for forming a plurality of sealed composite packages filled with a pourable product; a sampling unit for sampling one or more sealed composite packages from the plurality of sealed composite packages; an automatic defect detection system (1) according to any one of claims 1 to 10 for determining the quality of one or more seal bands (2) of one or more sampled sealed composite packages; Equipped with Packaging plant.

12. The packaging device a lateral sealing device that generates a lateral seal band for the sealed composite package; a control unit (100) operatively connected to the lateral sealing device and configured to control operation of the lateral sealing device in response to a quality of one or more seal bands (2) of one or more sampled sealed composite packages determined by the system (1); Equipped with 12. Packaging plant according to claim 11.

13. A method for automatically detecting defects in a sealing band (2) of a composite package, the sealing band (2) being formed from a first band portion (3), a second band portion (4) and a layer of heat sealable material (5) interposed between the first band portion (3) and the second band portion (4), the method comprising: Taking an image of the sealing band (2), Analysing the quality of the sealing band (2) based on one or more images, the analyzing comprises processing the one or more images with a machine learning model to detect defects in the seal band (2), the machine learning model being trained with a training dataset including image data associated with known quality measures. method.

14. the analyzing step comprises receiving as input both image data associated with the image and sensor data indicative of a sealing test procedure and / or operation of a packaging machine in forming the seal band (2), and using the machine learning model to analyze the image data and the sensor data using a multimodal approach to assess the quality of the seal band (2). The method of claim 13.

15. causing the sealing band (2) to peel off, acquiring said one or more images during peeling; 15. The method of claim 13 or 14.

16. and detecting at least one force acting on the sealing band (2) during the peeling, and wherein the analyzing step comprises receiving the image data and the at least one force as input data.

16. The method of claim 15.

17. comprising computer instructions configured to, when executed by one or more processors of a processing system, perform the method of any one of claims 13 to 16; Computer program.

Images