Virtual Detection System for Condition Monitoring of Container Packaging Machines

Abstract

A virtual detection system (1) for monitoring the state of a container packaging machine (2) for packaging a container filled with a pourable food, the system (1) being configured to reconstruct a target state monitoring signal based on input data from the container packaging machine (2), the system (1) comprising: an input module (6) configured to receive from the container packaging machine (2) input data indicating the target state monitoring signal to be reconstructed; an artificial intelligence (AI) module (8) configured to execute a machine learning algorithm that generates an output state monitoring signal that is a reconstruction of the target state monitoring signal based on the input data; an output module (7) configured to provide the output state monitoring signal generated by the AI module (8); and a state monitoring module (9) designed to evaluate and / or predict the state of the container packaging machine (2) based on the output state monitoring signal.

Description

Technical Field

[0001] The present invention relates to a container packaging machine, in particular to a virtual sensing system for monitoring the manufacturing state of a composite package filled with a pourable food product.

Background Art

[0002] As is well known, many liquids or pourable food products, such as fruit juice, UHT (ultra-high temperature treated) milk, wine, tomato sauce, etc., are distributed and sold in composite packages made of multilayer composite packaging materials.

[0003] As a typical example, a parallelepiped package for a pourable food product known as Tetra Brik Aseptic (registered trademark) can be mentioned, which is made by sealing and folding a laminated strip-shaped packaging material. This packaging material has a multilayer structure covered on both sides with heat-sealing plastic materials, for example layers of polyethylene, on a base layer of carton and / or paper. In the case of aseptic packaging for long-term storable products, the packaging material includes a layer of an oxygen barrier material, for example aluminum foil, which layer overlaps with a layer of heat-sealing plastic material and is further covered with another layer of heat-sealing plastic material to finally form the inner surface of the package that contacts the food product.

[0004] This type of composite package is usually manufactured in a fully automated packaging line (or factory), where the composite package is formed starting from a web of multilayer composite packaging material and filled with a pourable food product.

[0005] A typical packaging line at least includes a filling machine that starts from a multi-layer composite packaging material to form a composite packaging material and fills the composite packaging material with food that can be poured into it. Further, the packaging line may include additional container packaging machines upstream and / or downstream. Downstream packaging machines may include, for example, a buffer unit for temporarily buffering the composite package, an application unit for applying a straw or other elements onto the composite package, a grouping unit for grouping a plurality of composite packages together within a storage unit (such as a pallet), for example, one or more of a palletizer unit.

[0006] As is well known, monitoring the state of container packaging machines and related components is also important for evaluating the current operating state and the possibility of abnormalities in related components, and predicting the occurrence of failures and damages.

[0007] Condition monitoring is usually implemented by installing appropriate monitoring sensors in the packaging machine, acquiring corresponding detection signals, processing the same detection signals with appropriate evaluation and prediction algorithms, and providing a display of the current and predicted operating states of the same packaging machine.

[0008] In particular, vibration analysis constitutes a relevant part of such condition monitoring. Vibration analysis monitors the level and pattern of vibration signals within the packaging machine, detects abnormal vibrations of key components (such as bearings, gears, servo motors, etc.), and is designed to evaluate the overall state of the same machine. Vibration analysis realizes, for example, real-time response to state changes, remote condition monitoring, and predictive maintenance.

[0009] Therefore, vibration sensors, especially acceleration sensors, are coupled to the main parts of the packaging machine to monitor the corresponding operations. For example, the vibration sensor is coupled to the servo motor via a support flange and provides a detection signal indicating the vibration generated during operation by the same servo motor.

[0010] However, installing and maintaining state monitoring sensors on packaging machines is costly and time-consuming. Furthermore, there may be locations on the packaging machine where it is difficult or impossible to install monitoring sensors (for example, there are access problems to a part of the machine, or it operates in a harsh environment unsuitable for installing monitoring sensors, etc.).

[0011] In such cases, it may be difficult or impossible to achieve the conventional monitoring of the operating state of the corresponding parts of the packaging machine with the desired level of reliability and accuracy.

Summary of the Invention

Problems to be Solved by the Invention

[0012] An object of the present disclosure is to provide an improved solution that can at least partially overcome the above problems of known state monitoring solutions.

[0013] Therefore, according to the present disclosure, a virtual sensing system for state monitoring of a container packaging machine is provided as defined in the appended claims.

[0014] According to a second aspect, a method for monitoring the state of a container packaging machine as defined in the appended claims is provided.

[0015] According to a third aspect, a computer program product including instructions for causing a computing unit to execute the method is provided when the program is executed by the computing unit.

Means for Solving the Problems

[0016] Hereinafter, embodiments of the present invention will be exemplarily described with reference to the accompanying drawings.

Brief Description of the Drawings

[0017]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

[0018] FIG. 1 is a schematic block diagram of a computer-implemented virtual sensing system for monitoring the operation of a container packaging machine 2 configured to manufacture packaging containers filled with pourable products, particularly foods such as milk, cream, fruit juice, wine, tomato sauce, sugar, salt, emulsions, solutions containing solid particles (e.g., beans), etc., and particularly composite packages formed from multi-layer composite packaging materials, generally indicated by 1.

[0019] The container packaging machine 2 is part of a container packaging line 2' that includes further processing machines (not shown here), and the container packaging machine 2 is, for example, a filling machine.

[0020] The multi-layer composite packaging material may comprise a layer of fibrous material such as paper or cardboard, and at least two layers of heat-sealing plastic material such as polyethylene interposed between the layers of fibrous material. One of these two layers of heat-sealing plastic material defines the inner aspect of the packaging container that ultimately contacts the pourable food packaged within the same packaging container.

[0021] The multilayer composite packaging material may also comprise a layer of gas-barrier and light-barrier material, such as aluminum foil or ethylene vinyl alcohol (EVOH) film, disposed in particular between one of the layers of heat-sealing plastic material and the layer of fibrous material. Preferably, the multilayer composite packaging material may comprise a further layer of heat-sealing plastic material interposed between the layer of gas-barrier and light-barrier material and the layer of fibrous material.

[0022] Each packaging container may extend along a longitudinal axis and may have a longitudinal seam portion (extending along the respective longitudinal axis) and a pair of transverse seal bands, in particular a transverse upper seal band and a transverse lower seal band. In particular, each packaging container may have a substantially parallelepiped structure.

[0023] Furthermore, the packaging container may comprise at least two transverse walls (transverse to the respective longitudinal axis) and a plurality of transverse walls disposed on opposite sides of the packaging container and extending between the transverse walls.

[0024] More specifically, each respective transverse wall of each packaging container may define a bottom wall and the other respective transverse walls may define a top wall. In particular, the bottom wall may define a support surface adapted to be placed on a (horizontal) plane, such as a shelf within a distribution point, and the top wall is opposite the bottom wall.

[0025] Although not shown in detail, the container packaging machine 2 comprises at least - a conveying device for advancing a web of multilayer packaging material along a forward path, - a tube forming device for forming a tube from the advancing web of multilayer packaging material, - a sealing device for sealing the tube longitudinally, - a filling device connected to an injection duct for filling the tube with pourable food products, - a package forming unit for forming a composite package from the tube filled with pourable food products by forming the tube and sealing the tube transversely, and may be provided with.

[0026] More specifically, the package forming unit may at least include - a forming device configured to form a tube, seal it laterally, and cut it laterally to obtain a composite package, particularly a pillow package; - a final folder configured to receive the composite package from the forming device and form a packaging container from the composite package, particularly a pillow package. It may be provided with.

[0027] The container packaging machine 2 further includes a control device 4 for controlling the packaging operation. The control device 4 includes, for example, a PLC (Programable Logic Controller) unit or any suitable processing and computing unit, and is configured to execute a computer program designed to control the operation of the container packaging machine 2.

[0028] According to one aspect of the present solution, the virtual sensing system 1 reconstructs a target signal, particularly a state monitoring target signal useful for state monitoring of the container packaging machine 2, such as a vibration signal, by data fusion of other signals, measured values, parameters, and / or information (generally referred to as input data) available from the same environment as the target signal, that is, the same container packaging machine 2.

[0029] The virtual sensing system 1 is configured to learn how to interpret the relationships between the input data and combine them so that it can reconstruct the target signal with sufficient accuracy to replace an actual physical sensor, particularly a vibration sensor (such as an acceleration sensor). Therefore, for the purpose of state monitoring of the container packaging machine 2, there may be no vibration sensor.

[0030] More specifically, the virtual sensing system 1 is implemented in the processing computing unit 1', an input module 6 configured to receive input data from the container packaging machine 2 indicating (or related to) the target state monitoring signal to be reconstructed; Based on the input data received from the input module 6, an artificial intelligence (AI) module 8 is configured to implement an appropriate neural network algorithm, particularly an appropriate neural network algorithm with recursiveness, in order to generate an output state monitoring signal that is an accurate approximation of the target state monitoring signal. An output module 7 is configured to provide the output state monitoring signal generated by the AI module 8 for further processing by a state monitoring module 9 that is designed to evaluate the actual state of the container packaging machine 2 and / or predict a future state based on the same output state monitoring signal. That is, the virtual sensing system 1 includes a state monitoring unit 9.

[0031] The state monitoring module 9 may be implemented in the same processing computing unit 1' and configured to perform state monitoring of the container packaging machine 2, generally the container packaging line 2'. For example, the processing operation unit 1' may be configured to perform vibration analysis when the output state monitoring signal is configured to reconstruct a target vibration signal.

[0032] In a possible embodiment, the virtual sensing system 1 is remotely implemented from the container packaging machine 2 in a central processing unit 100 (including the processing computing unit 1' in this case) located in a remote server (e.g., within a "cloud"), where operation data and information related to the container packaging line 2' are received and processed for the control and management of its operation.

[0033] In another possible embodiment, the virtual sensing system 1 may be locally implemented in the container packaging machine 2, for example, in the control device 4 of the same container packaging machine 2.

[0034] There are many advantages to the virtualization of the actual monitoring sensors performed by the virtual sensing system 1, such as: Reducing the number of physical sensors on-site, resulting in reduced installation and maintenance costs; For example, in cases where there is no solution other than installing expensive wireless sensors for very small parts, moving parts, etc., virtual sensorization of important points where actual sensor installation is very difficult or impossible; By collecting data and adapting the neural network algorithm implemented by the AI module 8 to those specific anomalies, there is a possibility of detecting new anomalies that cannot be detected by current condition monitoring techniques, improving the anomaly detection of the on-site fault monitoring solution.

[0035] In order to accurately replace the physical monitoring sensors of the container packaging machine 2, the AI module 8 of the virtual sensing system 1 needs to be appropriately trained so that it can train the neural network algorithm based on the collection of appropriate input data. For example, the weighting of the neural network is adjusted through the training process.

[0036] As shown in FIG. 1, such training may be performed by the training module 10, and the training module 10 may be included in (or alternatively, outside of) the above-described processing computing unit 1' and / or the central processing unit 100.

[0037] The training may be performed with reference to a number of training container packaging machines 2 in which the physical monitoring sensors are (at least initially) held for the purpose of comparing the corresponding detection signals with the output condition monitoring signals generated by the AI module 8 and adjusting the weights (generally, parameters) of the neural network implemented in the same AI module 8 accordingly.

[0038] As schematically shown in FIG. 2, during the training or "development" of the AI module 8 (herein referred to as the "virtual sensor") of the virtual sensing system 1, the input module 6 receives both the target condition monitoring signal (in this case, the actual vibration signal provided by the physical actual sensor 15 (in this case, an acceleration sensor)) and the above-described input data indicating the target condition monitoring signal to be reconstructed.

[0039] In the illustrated example where the physical monitoring sensor removed from the container packaging machine 2 is a vibration sensor coupled to a machine servo motor (for example, a rotary servo motor configured to drive a jaw designed to cut a tube to obtain a composite package), the input data may include operation signals related to the same servo motor, such as torque signals and speed signals detected by an appropriate sensor coupled to the servo motor.

[0040] To increase the pool of input data, the first and second derivatives of the same torque and speed signals can also be considered.

[0041] During the development stage, the output state monitoring signal provided by the virtual sensing system 1 is compared with the target state monitoring signal, and the parameters of the AI module 8 (for example, the weights of the corresponding neural network) are adjusted based on the comparison (for example, the difference) between the same output state monitoring signal and the target state monitoring signal.

[0042] After appropriate training of the AI module 8, for the state monitoring of the container packaging machine 2 at the site where the physical sensor 15 is no longer installed or used (in other words, the container packaging machine 2 has no physical sensor 15 used for state monitoring purposes), the deployment of the virtual sensing system 1 can be implemented. The vibration signal is estimated by the virtual sensing system 1 and provides the output state monitoring signal generated by the AI module 8.

[0043] It should be noted that the development stage of the AI module 8 may be implemented considering a selected number of container packaging machines 2 for training, for example, the number of machines within the site of the corresponding manufacturing company. In this case, the container packaging machine 2 installed at the customer's site may not be equipped with the physical sensor 15 used for state monitoring purposes, at least in the deployment stage described above.

[0044] As an alternative, a selected number of container packaging machines 2 provided to selected customers may also be involved in the development of the AI module 8 as training container packaging machines.

[0045] According to one aspect of this solution, for the training of the AI module 8 of the virtual sensing system 1, a continuous learning approach is implemented by the training module 10.

[0046] Continuous machine learning, also known as sequential learning or online learning, is a machine learning approach in which an AI model continuously learns and evolves based on increasing input data while retaining previously learned knowledge. Therefore, this approach provides the ability for the AI model to autonomously learn and adapt when new and different input data comes in.

[0047] In the context of this solution, the continuous machine learning applied to the AI module 8 provides for continuously repeating training phases based on different input data, and continuously re-training the neural network while gradually adjusting the weights and parameters of the corresponding neural network algorithm. This approach can continuously improve the performance of the AI module 8, and can improve the performance of the AI module 8 until the point where it is considered industrially deployable (at which point all the advantages of the virtual sensing solution can be utilized on-site).

[0048] As schematically shown in FIG. 3, the continuous learning approach provides for collecting input data from the selected training container packaging machines 2 as shown in step 20. It should be noted that these training container packaging machines 2 may, for collecting input data, in this case, be training machines within the manufacturer's premises and / or on-site training machines installed within the customer's premises where physical monitoring sensors 15 are installed.

[0049] All of the collected data is received, for example, by the central processing unit 100 and stored in the data storage device 19.

[0050] The input data is then used, as shown in step 21, for the training of the neural network algorithm in the AI module 8 of the virtual sensing system 1, in particular for the adjustment of relevant parameters (such as weights).

[0051] In this regard, it should be noted that monitoring algorithms generally used by the same central processing unit 100 to monitor the operation and performance of the related container packaging machine 2, such as a predictive maintenance model, a performance monitoring model, a quality control model, and a process control model, are utilized to provide data labels for the training of the AI module 8, and basically the virtual sensing system 1 can select the input data that must be retrained.

[0052] Next, as shown in step 22, the performance of the AI module 8 in the reconstruction of the target state monitoring signal is evaluated with any appropriate measurement criteria by comparing the same target state monitoring signal with the output state monitoring signal provided by the virtual sensing system 1.

[0053] If the performance is evaluated as good or satisfactory (e.g., meeting or exceeding a certain quality threshold), as shown in step 24, the virtual sensing system 1 is considered ready for on-site deployment, and the same virtual sensing system 1 is used for the state monitoring of the container packaging machine 2 at the site where no physical or actual state monitoring sensors are installed (or the physical sensors are no longer used for this purpose).

[0054] As shown in the same Figure 3, the training of the AI module 8 is continued by further collecting input data and further possible adjustment of the parameters of the same AI module 8. By this continuous training, it is possible to adapt, for example, to changes or modifications that may occur in the container packaging machine 2.

[0055] Next, with reference to FIG. 4, a further aspect of the solution regarding a specific training pipeline for the AI module 8 of the virtual sensing system 1 will be described.

[0056] It should be noted that, in particular, the AI module 8 is configured to implement a regression algorithm so as to output the actual estimated value of the same signal when arranged to reconstruct an approximation of the target state monitoring signal.

[0057] However, the applicant has discovered that it may not be possible to obtain the best possible training results even when training the AI module 8 configured as a regressor.

[0058] Therefore, in one aspect of the solution, a training pipeline for the AI module 8 is envisioned, according to which the training is split into two consecutive phases: " A first training phase 26 in which the AI module 8 is configured as a classifier and thus assigns an output class (which is provided as the output of the AI module 8) among a certain number of classes to the input data; and A second training phase 28 in which the same AI module 8 is actually configured as a regressor and provides the actual output value of the reconstructed signal.

[0059] More specifically, as shown in FIG. 4, in the first training phase 26 of the pipeline, the AI module 8 is configured to implement a classification algorithm (of any suitable nature depending on the situation) as shown in step 30 and provide a classification result.

[0060] The result of the classification is used to adjust the parameters of the AI module 8, for example to measure the performance of the neural network via a confusion matrix, and in particular to set the weights of the neural network as shown in step 32.

[0061] This constitutes the pre-training phase of the AI module 8, followed by the second training phase 28 of the pipeline, and the AI module 8 is configured to implement a regression algorithm (which can also be of any suitable nature).

[0062] In this second training phase, as shown in step 34, the AI module 8 undergoes fine-tuning to reconstruct the actual value of the target state monitoring signal with high accuracy. Thanks to the parameters (such as weights) pre-set by the classifier in the first training phase, the training of the AI module 8 in this second training phase becomes easier and more accurate results can be obtained.

[0063] In particular, as schematically shown in step 36, the output state monitoring signal provided by the trained virtual detection system 1 closely matches the target state monitoring signal.

[0064] According to a possible embodiment, the AI module 8 is configured to implement a recurrent neural network algorithm, particularly the LSTM (Long Short Term Memory) algorithm, for the reconstruction of the target state monitoring signal.

[0065] As schematically shown in FIG. 5, the AI module 8 includes a plurality (three in the example) of input stages 40, and each input stage receives respective input data (such as torque signal and speed signal, and their first and second derivative values).

[0066] The AI module 8 further includes a plurality of neural network cells or stages 42 (three in the example), and each cell or stage receives respective input data from the corresponding input stage 40 via respective input weighting blocks 43, and also receives the output of the previous cell according to a recurrent architecture.

[0067] The outputs of the various neural network cells 42 are provided to an attention stage 45 that implements a suitable attention algorithm that gives more or less importance (i.e., focuses) to one or more of the outputs received from the neural network cells 42 within the recurrent architecture via their respective output weighting blocks 44.

[0068] The AI module 8 further comprises an output stage 46 following the attention stage 45 and is configured to provide a reconstructed value of the output state monitoring signal.

[0069] The applicant has found that using a recurrent neural network algorithm, in particular the LSTM (Long Short Term Memory) algorithm, for the reconstruction of the target state monitoring signal results in more accurate and reliable results in signal reconstruction and is particularly advantageous.

[0070] The advantages of the solution discussed will be apparent from the foregoing description.

[0071] In any case, it is again emphasized that the solution discussed enables the following: reducing the number of actual sensors placed on site, and as a result, reducing the associated installation and maintenance costs; virtual sensing of critical points where the installation of actual sensors is very difficult or impossible; improving anomaly detection for fault monitoring on site.

[0072] In particular, using the continuous learning approach discussed for the training and development of the artificial intelligence module of the virtual sensing system, the performance of signal reconstruction can be continuously improved thanks to the gradually growing input data pool until the same virtual sensing system can be deployed in an industrial environment.

[0073] With a similar continuous learning approach, by repeating the training over a long period of time, it is possible to adapt to changes in the container packaging machine and its operation, such as changes in materials and production parameters.

[0074] The discussed training pipeline with two different consecutive training phases, where the AI module is first trained as a classifier and then as a regressor, has proven to be advantageous for improving the performance of signal reconstruction.

[0075] Furthermore, by using a recurrent neural network architecture, the performance of signal reconstruction in the virtual sensing system can be further improved.

[0076] However, it is obvious that modifications may be made to the content described herein without departing from the scope of protection defined in the appended claims.

[0077] In particular, it is emphasized that the AI module 8 of the virtual sensing system 1 can implement different types of neural network algorithms suitable for the reconstruction of the target state monitoring signal.

[0078] Furthermore, the input data includes additional signals and / or higher-order derivatives of the same signal.

[0079] Also, it is emphasized that the discussed solution can be applied to any packaging line 2' including any packaging machine 2 and to any type of pourable food product.

[0080] The systems and methods disclosed herein can be implemented as software, firmware, hardware, or any combination thereof. In a hardware implementation, the division of tasks between the functional units or modules mentioned in the above description does not necessarily correspond to the division into physical units. On the contrary, one physical module can execute multiple functions, and one task can be jointly executed by multiple physical modules.

[0081] Certain modules or all modules may be implemented as software executed by a digital signal processor or a microprocessor, or may be implemented as hardware or an application - specific integrated circuit. Such software can be distributed on a computer - readable medium that includes a computer storage medium (or non - transitory medium) and a communication medium (or transitory medium). As is well known to those skilled in the art, the term computer - readable medium includes both volatile and non - volatile, removable and non - removable media implemented in any method or technology for storing information such as computer - readable instructions, data structures, program modules or other data. Computer storage media includes RAM, ROM, EEPROM, flash memory or other memory technologies, CD - ROM, digital versatile disks (DVD) or other optical disk storage devices, magnetic cassettes, magnetic tapes, magnetic disk storage devices or other magnetic storage devices, or other media that can be used to store desired information and can be accessed by a computer, but is not limited thereto.

[0082] The state - monitoring signal of interest can be a vibration signal. Advantageously, this system can replace a physical vibration sensor that may not be present in the container packaging machine (2) for state - monitoring purposes, for example, due to the presence of a harsh environment or physical constraints.

[0083] In particular, the target monitoring signal may be a vibration signal related to the servo - motor of the container packaging machine 2. The AI module may be configured to generate a number of output state - monitoring signals that are reconstructions of a number of target state - monitoring signals corresponding to vibration signals related to a number of servo - motors.

[0084] The input data may include a number of operation signals related to components of the packaging machine 2 that cause vibration, such as servo - motors. Such operation signals may include torque signals and / or speed signals, and optionally corresponding derivatives.

[0085] The machine learning algorithm may be trained based on a training data set including reference input data indicating a reference target state monitoring signal to be reconstructed and the reference target state monitoring signal. The input data may be collected simultaneously with respect to the reference target state monitoring signal. That is, the acquired input data may indicate the acquired target state monitoring signal.

[0086] The container packaging machine 2 may be operated in a training phase to collect the reference input data and the reference target state monitoring signal. The container packaging machine 2 may be the same as the one disclosed above or may be a different container packaging machine 2.

[0087] One or more embodiments may relate to a method for monitoring the state of a container packaging machine (2) for packaging a container filled with a pourable food product, the method comprising reconstructing a target state monitoring signal, such as a vibration signal associated with a servo motor within the container packaging machine, based on input data from the container packaging machine (2), such as operating signals associated with a number of servo motors within the container packaging machine (2), preferably operating signals including torque signals and speed signals, more preferably operating signals including corresponding derivatives of the torque signals and speed signals.

[0088] The step of reconstructing comprises receiving, from the container packaging machine (2), input data indicating the target state monitoring signal to be reconstructed; implementing a machine learning algorithm that generates an output state monitoring signal that is a reconstruction of the target state monitoring signal based on the input data; providing the output state monitoring signal generated by the AI module (8); processing the output state monitoring signal to evaluate and / or predict the state of the container packaging machine (2) based on the output state monitoring signal; and comprising

[0089] This method may include using an algorithm designed to monitor the operation and / or performance of the training container packaging machine (2) as a function of input data, including one or more of a predictive maintenance model, a performance monitoring model, a quality control model, and a process control model, to provide data labels for training the AI module (8).

[0090] One or more embodiments may relate to a method for training the aforementioned AI module 8 and machine learning algorithm, the method comprising: - obtaining input data from the container packaging machine 2, such as a reference packaging machine 2 or the same previously disclosed packaging machine used during normal operation; - obtaining a status monitoring signal from a physical status monitoring sensor (15) disposed on the container packaging machine (2); - training the AI module 8 and the machine learning algorithm based on the input data and the status monitoring signal. comprising.

[0091] The method may include implementing a training pipeline for training the AI module (8), according to which the training is divided into the following two consecutive phases: A first training phase in which the AI module (8) is configured as a classifier; A second training phase in which the AI module (8) is configured as a regressor and provides the actual output value of the reconstructed signal.

[0092] The method comprises In the first training phase of the pipeline, during the pre-training stage of the AI module (8), implementing a classification algorithm and providing classification results used to adjust the parameters of the machine learning algorithm; In the second training phase of the pipeline, implementing a regression algorithm for fine-tuning the parameters of the machine learning algorithm. may be provided.

[0093] One or more embodiments may also relate to an AI module 8 and a method for verifying the aforementioned machine learning algorithm, the method including the step of evaluating the performance of the machine learning algorithm by checking whether a state monitoring signal from a state monitoring sensor substantially corresponds to an output state monitoring signal generated by the AI module (8).

[0094] During verification of the machine learning algorithm, the acquired physical signal, i.e., the state monitoring signal from the state monitoring sensor, may be compared with the output state monitoring signal generated by the AI module. If the output state monitoring signal differs by a predetermined threshold for a predetermined period, the machine learning algorithm may be re-trained.

[0095] One or more embodiments may also relate to a computer program product including instructions that cause a computing unit to execute at least one of a method for monitoring the state of a container packaging machine, a method for training an AI module 8 and a machine learning algorithm, and / or a method for verifying an AI module 8 and a machine learning algorithm when the program is executed by the computing unit.

Claims

1. A virtual detection system (1) for monitoring the status of a container packaging machine (2) for packaging containers filled with pourable food, The system (1) is configured to reconstruct a target state monitoring signal based on input data from the container packaging machine (2), and the system (1) An input module (6) is configured to receive input data indicating a target state monitoring signal to be reconfigured from the container packaging machine (2), An artificial intelligence (AI) module (8) is configured to execute a machine learning algorithm that generates an output state monitoring signal, which is a reconstruction of the target state monitoring signal, based on the input data, An output module (7) is configured to provide an output state monitoring signal generated by the AI ​​module (8), A condition monitoring module (9) is designed to evaluate and / or predict the state of the container packaging machine (2) based on the outputted condition monitoring signal, Equipped with, System (1).

2. The aforementioned target state monitoring signal is a vibration signal. The system according to claim 1.

3. The machine learning algorithm is trained on a training dataset which includes reference input data showing the reference target state monitoring signal to be reconstructed, and the reference target state monitoring signal. The system according to claim 1.

4. The system further comprises a training module (10) operably coupled to the AI ​​module (8) and configured to train the AI ​​module (8) based on a continuous learning approach. The system according to claim 1.

5. The AI ​​module (8) is a neural network comprising a number of neural network cells (42) arranged according to a recurrent architecture, each receiving its own input data, and an attention stage (45) that executes an attention algorithm on the output received from the neural network cells (42) and provides a reconstructed value of an output state monitoring signal. The system according to claim 1.

6. The monitored signal is a vibration signal related to the servo motor of the container packaging machine (2), The input data includes a number of operating signals related to a number of servo motors in the container packaging machine (2), preferably the number of operating signals includes torque signals and / or speed signals, and more preferably the number of operating signals includes corresponding derivatives of the torque signals and speed signals. The system according to claim 1.

7. The AI ​​module (8) is included in a central processing unit (100) installed on a remote server outside the container packaging machine (2), The system according to claim 1.

8. A packaging line (2') comprising the virtual detection system (1) according to claim 1.

9. A method for monitoring the status of a container packaging machine (2) for packaging containers filled with pourable food, The step includes reconstructing a target state monitoring signal based on input data from the container packaging machine (2), and the reconstruction step includes, The steps include receiving input data indicating a target state monitoring signal to be reconstructed from the container packaging machine (2), The steps include: implementing a machine learning algorithm to generate an output state monitoring signal, which is a reconstruction of the target state monitoring signal, based on the aforementioned input data; The steps include providing an output status monitoring signal generated by the AI ​​module (8), The steps include processing the output status monitoring signal and evaluating and / or predicting the state of the container packaging machine (2) based on the output status monitoring signal, A method that includes [a certain feature].

10. The steps include acquiring input data from the container packaging machine (2), The steps include: acquiring a status monitoring signal from a physical status monitoring sensor (15) located in the container packaging machine (2); A step of training the AI ​​module (8) based on the input data and the status monitoring signal, Furthermore, The method according to claim 9.

11. The step further comprises carrying out a training pipeline for training the AI ​​module (8), wherein the training is The first training phase involves configuring the AI ​​module (8) as a classifier, and The AI ​​module (8) is configured as a regressor and is divided into two consecutive phases: a second training phase which provides the actual output value of the reconstructed signal, The method according to claim 9.

12. In the first training phase of the pipeline, the pre-training stage of the AI ​​module (8) includes the steps of implementing a classification algorithm and providing classification results used to adjust the parameters of the machine learning algorithm, The second training phase of the pipeline includes the step of implementing a regression algorithm for fine-tuning the parameters of the machine learning algorithm, Equipped with, The method according to claim 11.

13. The process includes a step of evaluating the performance of the machine learning algorithm by checking whether the status monitoring signal from the status monitoring sensor substantially corresponds to the output status monitoring signal generated by the AI ​​module (8), The method according to claim 10.

14. To provide data labels for training the AI ​​module (8), the step includes utilizing an algorithm designed to monitor the operation and / or performance of a training container packaging machine (2) as a function of input data, which includes one or more of a predictive maintenance model, a performance monitoring model, a quality control model, and a process control model. The method according to claim 9.

15. A computer program that, when executed by a computing unit, includes instructions causing the computing unit to perform the method according to claim 9.

Images