Pneumatic device and method for monitoring pneumatic device

Abstract

According to an exemplary embodiment, a method for monitoring a pneumatic device is provided. The method comprises the steps of: collecting a data set of a plurality of operational variables of a pneumatic facility comprising a plurality of pneumatic components; and generating a facility monitoring model, where the facility monitoring model is generated by clustering the data set of the plurality of operational variables based on the density.

Description

Technical Field

[0001] This disclosure relates to pneumatic devices and methods for monitoring pneumatic devices. This application claims the benefit of Korean Patent Application No. 10-2023-0179446, filed on December 12, 2023, the disclosure of which is incorporated herein by reference. Background Technology

[0002] Unlike primary batteries, secondary batteries can be charged and discharged multiple times. They are widely used as power sources for various wireless devices such as mobile phones, laptops, and cordless vacuum cleaners. In recent years, with improved energy density and economies of scale significantly reducing the manufacturing cost per unit capacity of secondary batteries, and with battery electric vehicles (BEVs) achieving driving ranges comparable to fuel-powered vehicles, the primary use of secondary batteries has shifted from mobility devices to transportation.

[0003] In the production of rechargeable batteries, pneumatic equipment is used for handling and transferring materials such as battery cells or electrode semi-finished products. Pneumatic equipment is difficult to control precisely and malfunctions are difficult to predict in advance. Therefore, various studies are underway to predict pneumatic equipment failures in advance to prevent yield drops and increase production output in rechargeable battery manufacturing. Summary of the Invention

[0004] Technical issues

[0005] The problem sought to be solved by the technical concept disclosed herein is to provide a pneumatic device and a method for monitoring the pneumatic device.

[0006] Technical solution

[0007] According to an exemplary embodiment of this disclosure for addressing the aforementioned problems, a method for monitoring pneumatic equipment is provided. The method includes the steps of: collecting a dataset of multiple operating variables of a pneumatic equipment comprising multiple pneumatic components; and generating an equipment monitoring model, wherein the equipment monitoring model is generated by clustering the dataset of multiple operating variables based on density.

[0008] Clustering of datasets with multiple operational variables based on mean-shift clustering.

[0009] The method also includes the following steps: collecting data points of multiple operating variables of the pneumatic equipment; and monitoring the pneumatic equipment based on the data points and the equipment monitoring model.

[0010] The steps for monitoring pneumatic equipment include calculating a score indicating the state of the pneumatic equipment based on the distance of data points from normal clusters, which are determined based on the clustering of datasets of multiple operational variables.

[0011] The method also includes a step of generating alerts based on scores.

[0012] The method also includes a step of preprocessing the dataset with multiple operational variables.

[0013] The preprocessing steps for a dataset with multiple operational variables involve classifying the dataset of multiple operational variables of the pneumatic equipment based on the operating state of the pneumatic equipment.

[0014] Datasets with multiple operands are classified into normal datasets with multiple operands and fault datasets with multiple operands.

[0015] A normal dataset with multiple operands includes the optimal dataset with multiple operands.

[0016] The model is generated based on a normal dataset with multiple operational variables.

[0017] The preprocessing step for the dataset of multiple operational variables standardizes the dataset of multiple operational variables of the pneumatic device.

[0018] The preprocessing steps for datasets with multiple operational variables remove duplicate and missing data from datasets of multiple operational variables for pneumatic equipment.

[0019] The preprocessing step for the dataset of multiple operational variables enables the time synchronization of the datasets of multiple operational variables of the pneumatic device.

[0020] According to an exemplary embodiment, a pneumatic device is provided. The device includes: a plurality of pneumatic components; a sensor configured to detect a plurality of operational variables of the plurality of pneumatic components; and a processor configured to collect a dataset of the plurality of operational variables, wherein the processor includes a device monitoring model generated by clustering the dataset of the plurality of operational variables.

[0021] A device monitoring model is generated by clustering a dataset of multiple operational variables based on density.

[0022] Clustering of datasets with multiple operational variables based on mean-shift clustering.

[0023] The processor is configured to monitor pneumatic equipment based on data points of multiple operating variables of the pneumatic equipment and equipment monitoring models.

[0024] The processor is configured to calculate the distance of data points to normal clusters, which are determined based on the clustering of datasets with multiple operational variables.

[0025] The processor is configured to calculate a score indicating the state of the pneumatic equipment based on the distance of the data point from the normal cluster.

[0026] The processor is configured to generate alarms based on scores.

[0027] Beneficial effects

[0028] According to an exemplary embodiment of this disclosure, a method for monitoring a pneumatic system can cluster normal data of multiple operating variables of the pneumatic device and calculate the distance between data points of the multiple operating variables occurring in real time and the determined clusters. According to an exemplary embodiment, the method for monitoring the pneumatic device can generate alarms or control the operation of the pneumatic device based on changes in scores calculated from the distances, thereby providing a prediction of the state of the pneumatic device (e.g., normal and faulty).

[0029] The effects obtainable from the exemplary embodiments of this disclosure are not limited to those mentioned above, and other effects not mentioned can be clearly derived and understood by those skilled in the art from the following description. That is, those skilled in the art can also derive unintended effects from practicing the exemplary embodiments of this disclosure. Attached Figure Description

[0030] Figure 1 This is a block diagram illustrating a pneumatic device according to an exemplary embodiment.

[0031] Figure 2 This is a flowchart illustrating a method for monitoring a device according to an exemplary embodiment.

[0032] Figures 3 to 8 This is a diagram illustrating an example of multiple operational variables measured by multiple sensors.

[0033] Figure 9 and Figure 10 Example datasets with multiple operational variables are shown.

[0034] Figure 11 This is a diagram illustrating a method for monitoring a device according to an exemplary embodiment.

[0035] Figure 12 The changes in the scores based on the experimental example are illustrated. Detailed Implementation

[0036] In the following, preferred embodiments of the present disclosure will be described in detail with reference to the accompanying drawings. It should be noted that the terms and words used in this specification and claims should not be interpreted in their general or dictionary sense, but rather based on the principle that the inventor can define the concepts of the terms as best suited to describe his disclosure, and to be interpreted in a meaning and concept consistent with the technical concept of the present disclosure.

[0037] Therefore, it should be understood that the embodiments described herein and the configurations shown in the accompanying drawings are merely the most preferred embodiments of this disclosure and are not an exhaustive list of the technical concepts of this disclosure, and various equivalents and modifications may exist that can replace them upon submission.

[0038] Furthermore, in describing this disclosure, specific descriptions of relevant known configurations or features have been omitted where it is believed that a detailed description of the relevant known configurations or features would obscure the essence of this disclosure.

[0039] Because the embodiments of this disclosure are provided to explain the disclosure more fully to those skilled in the art, the shapes and dimensions of the components in the drawings may be shown enlarged, omitted, or schematically for clarity. Therefore, the dimensions or proportions of each component do not necessarily indicate its actual size or proportion.

[0040] (First and Second Embodiments)

[0041] Figure 1 This is a block diagram illustrating a pneumatic device 100 according to an exemplary embodiment.

[0042] Figure 2 This is a flowchart illustrating a method for monitoring a device according to an exemplary embodiment.

[0043] Reference Figure 1 and Figure 2 On page 110, datasets containing multiple operational variables V1, V2, and V3 can be collected.

[0044] The pneumatic device 100 can be configured to perform, for example, a manufacturing process for secondary batteries. The pneumatic device 100 can be configured to perform any of, for example, a mixing process, an electrode process, an assembly process, and an activation process. The pneumatic device 100 may include a plurality of pneumatic components 110, sensors 120, a controller 130, and a processor 140.

[0045] Electrode manufacturing processes can include coating, rolling, and cutting. In the coating process, electrode sheets can be coated with an electrode slurry. The electrode slurry can include active materials, conductors, binders, and solvents. An electrode slurry can be provided by dissolving the active materials, conductors, and binders in a solvent. In the rolling process, electrode sheets coated with electrode slurry can pass between pressure rollers. The rolling process flattens the surface of the electrode sheet and improves the adhesion between the active material and the current collector. A cutting process can separate the electrode sheets into multiple electrode sheets. The cutting process can be selective. Depending on the specifications of the electrode assembly of the final battery cell to be manufactured, the cutting process can be omitted. Finished electrode rolls from the cutting device can be processed by a winding device or a grooving device. Therefore, stacked electrode assemblies or cylindrical electrode assemblies can be provided.

[0046] A grooving process can be performed on the electrode sheet after it has been unwound from the electrode roll by a rewinder. In the grooving process, electrode tabs can be formed on the uncoated portion of the electrode sheet. V-grooves can be further formed on the electrode sheet in the grooving process. In the lamination process, individualized unit electrodes can be attached to the separator. In the lamination process, the separator and unit electrodes can be heat-treated to increase the bonding strength between them. The lamination process can provide half-cells or single-cells. Dual-cells can also be provided through the lamination process. A half-cell can include a separator and a positive electrode, or it can include a separator and a negative electrode. A single-cell can include a positive electrode, a first separator, a negative electrode, and a second separator stacked sequentially. In the stacking process, at least one of the half-cells, single-cells, and dual-cells can be repeatedly stacked. Electrode assemblies can be provided through the stacking process. Electrode assemblies can include strips for securing the positive and negative electrodes. In the folding process, half-cells, single-cells, and dual-cells can be wound around the separator. The packaging process may include inserting the electrode assembly into the housing, filling it with electrolyte, and sealing the housing.

[0047] The activation process can include charging and discharging, aging, degassing, performance testing, and open-circuit voltage (OCV) based testing. In the activation process, an SEI film can be formed on the negative electrode surface during the first charge. The SEI film is a thin film formed on the surface of the negative electrode material when the battery cell is first charged after manufacturing. When the battery cell is charged, lithium ions migrate from the battery cell to the negative electrode, and during this process, a solid electrolyte interface (SEI) film can be formed on the surface of the negative electrode material due to the chemical reactions that occur when the materials in the electrolyte are first electrolyzed. The SEI can be a type of separator. As lithium ions move from the positive electrode to the negative electrode to charge the battery, the SEI can prevent further degradation reactions in the electrolyte. In the aging process, the charged or discharged battery cell can be stored at room temperature for a predetermined period of time (e.g., 30 minutes to 3 hours, as a non-limiting example) to stabilize the battery cell. The purpose of the aging process is to uniformly distribute the electrolyte within the pouch cell, allowing the electrolyte to permeate both the positive and negative electrodes evenly. The aging process can improve the mobility of lithium ions. Charge-discharge and aging processes can generate gas inside battery cells. Degassing processes remove this gas. Performance testing can include capacity testing and defect screening. OCV-based inspection can screen battery cells for low-voltage defects. Low-voltage defects in battery cells can be caused by metallic foreign objects located inside the cell. For example, if the positive electrode of a battery cell contains metallic foreign objects such as iron or copper, these foreign objects can grow as dendrites on the negative electrode. Dendrites can cause unwanted short circuits inside the battery cell, which can lead to cell failure and fire.

[0048] Each of the plurality of pneumatic components 110 can be configured to operate under pneumatic pressure, or to guide and control pneumatic pressure applied to another component. The plurality of pneumatic components 110 may include actuation devices such as cylinders, rotary actuators, and air chucks, as well as directional control devices such as solenoid valves and mechanical valves. Solenoid valves can redirect compressed air, thereby determining the forward and backward movement of the drive assembly. The plurality of pneumatic components 110 may also include electronic devices such as servo motors. Servo motors can function similarly to cylinders, but can provide more precise control.

[0049] Multiple sensors 120 can be configured to detect the movement of multiple pneumatic components 110. Multiple sensors 120 can be configured to determine multiple operating variables V1, V2, V3 of the multiple pneumatic components 110.

[0050] Controller 130 can be configured to control a plurality of pneumatic components 110. Controller 130 can be configured to receive signals from a plurality of sensors 120 indicating a plurality of operational variables V1, V2, V3. Controller 130 can be configured to determine the state of the plurality of pneumatic components 110. For example, controller 130 can be configured to collect data including error codes indicating whether some of the plurality of pneumatic components 110 are in a normal operating state or a fault state. Controller 130 can be configured to send signals indicating the plurality of operational variables V1, V2, V3 to processor 140.

[0051] Here, controller 130 can be a programmable logic controller (PLC). A PLC is a dedicated form of microprocessor-based controller that uses programmable memory to store instructions and instantiate functions such as logic, sequencing, timing, counting, and arithmetic to control machines and processes. PLCs are easy to operate and program.

[0052] Controller 130 may include a power supply, a CPU, an input interface, an output interface, a communication interface, and a memory device. The power supply may be configured to provide power to other components of controller 130, such as the CPU, input interface, output interface, communication interface, and memory device, for the operation of controller 130. The memory device may include a read-only memory (ROM) configured to store system programs such as an operating system, and a random access memory (RAM) configured to store user programs and data such as status information of input and output devices, timers, counters, and values ​​of other internal devices. The CPU may be configured to control communication between modules that instantiate logic and convert input signals into output motion signals. The CPU may operate based on system programs and user programs stored in the memory device. The CPU may be configured to write check data and measurement data to or read check data and measurement data from the data area of ​​the memory device based on the system programs and user programs. Conditions or data from industrial equipment and production processes may be sent to the CPU via input modules. Results processed by the CPU may be sent to actuators via output modules. The communication interface may be configured to relay the reception and transmission of data between controller 130 and processor 140.

[0053] However, this is not the only possibility; controller 130 may include any of the following: a simple controller, a complex processor such as a microprocessor, CPU, GPU, etc., a software-configurable processor, or dedicated hardware and firmware. Controller 130 may be instantiated by, for example, a general-purpose computer or dedicated hardware such as a digital signal processor (DSP), a field-programmable gate array (FPGA), and an application-specific integrated circuit (ASIC).

[0054] Figures 3 to 8 This is an illustrative drawing used to describe an example of multiple operational variables V1, V2, V3 measured by multiple sensors 120. More specifically, Figures 3 to 8 The example illustrates the reciprocating motion of cylinder 112 in response to a control signal applied from controller 130 to solenoid valve 111.

[0055] Reference Figures 3 to 8 The cylinder 112 may include a first contact sensor 112S1 and a second contact sensor 112S2. For example... Figure 3 and Figure 8 As shown, when the pistol 112P of cylinder 112 is in the retracted position, the first contact sensor 112S1 can be in the open state and the second contact sensor 112S2 can be in the closed state. Figure 5 and Figure 6As shown, when the pistol 112P of cylinder 112 is in the forward position, the first contact sensor 112S1 can be in the closed state and the second contact sensor 112S2 can be in the open state. Figure 4 and Figure 7 As shown, when the pistol 112P of cylinder 112 moves between the forward and backward positions, the first contact sensor 112S1 and the second contact sensor 112S2 can be in the off state.

[0056] exist Figure 3 In this configuration, solenoid valve 111 can switch from a closed state to an open state. Therefore, as... Figure 4 As shown, the transition of solenoid valve 111 to the open state causes the pistol 112P of cylinder 112 to advance, thereby causing the first contact sensor 112S1 to turn to the closed state. The forward reaction time can be Figure 3 Configuration and Figure 4 The time between configurations. More specifically, the forward reaction time can be the time it takes for the pistol 112P of cylinder 112 to begin moving forward after solenoid valve 111 switches to the open state.

[0057] The time when solenoid valve 111 switches to the open state can be determined as the time when the control signal from controller 130 controlling solenoid valve 111 is sent to solenoid valve 111. Therefore, the forward reaction time can be the difference between the time when the control signal from controller 130 controlling solenoid valve 111 is sent to solenoid valve 111 and the time when the first sensor 111S1 switches from the open state to the closed state.

[0058] exist Figure 5 In the middle, the pistol 112P of cylinder 112 can reach the forward position, thereby switching the second contact sensor 112S2 to the on state. The forward movement time can be Figure 4 Configuration and Figure 5 The forward movement time can be the time taken for the pistol 112P to move to the forward position after the movement of the pistol 112P has begun. The forward movement time can also be the difference between the time it takes for the first sensor 111S1 to transition from the on state to the off state and the time it takes for the second sensor 111S2 to transition from the off state to the on state.

[0059] exist Figure 6 In this configuration, solenoid valve 111 can switch from the open state to the closed state. Figure 7 In the middle, when the pistol 112P of cylinder 112 is retracted, the solenoid valve 111 switches to the closed state, which can cause the second contact sensor 112S2 to switch to the closed state. The retraction reaction time can be Figure 6 configuration and Figure 7The time between configurations. More specifically, the retraction reaction time can be the time it takes for the pistol 112P of cylinder 112 to begin retraction after solenoid valve 111 switches to the closed state.

[0060] The time when solenoid valve 111 switches to the closed state can be determined as the time when the control signal from controller 130 controlling solenoid valve 111 is sent to solenoid valve 111. Therefore, the back-off reaction time can be the difference between the time when the control signal from controller 130 controlling solenoid valve 111 is sent to solenoid valve 111 and the time when the second contact sensor 112S2 switches from the open state to the closed state.

[0061] exist Figure 8 In the middle, the pistol 112P of cylinder 112 can reach the backward position, thereby causing the first contact sensor 112S1 to switch from the on state to the off state. The backward travel time can be Figure 7 Configuration and Figure 8 The time between the configurations. More specifically, the backward travel time can be the time it takes for the pistol 112P to move to the backward position after the movement of the pistol 112P has begun. The backward travel time can also be the difference between the time it takes for the second sensor 111S2 to transition from the on state to the off state and the time it takes for the first sensor 111S1 to transition from the off state to the on state.

[0062] Although Figures 3 to 8 This describes a dataset of four or more operational variables collected from two pneumatic components; however, the dimensionality of the collected data can vary depending on the number of sensors applied to the pneumatic components. For example, if one of the solenoid valve 111 and cylinder 112 is equipped with a flow sensor and a pressure sensor, the flow sensor can sense the flow rate of one of the solenoid valve 111 and cylinder 112, and the pressure sensor can sense the air pressure of one of the solenoid valve 111 and cylinder 112. Therefore, two additional datasets of multiple operational variables can be collected.

[0063] In another example, pneumatic component 110 may include a vacuum chuck and a solenoid valve configured to regulate the pressure applied to the vacuum chuck. Multiple sensors 120 may be configured to calculate the working response time, working, non-working, and non-working response time of the vacuum chuck from the vacuum chuck and the solenoid valve. The working and non-working status of the vacuum chuck may be determined based on pressure sensors. When the vacuum chuck is working, a high pressure may be applied to the vacuum chuck, and when the vacuum chuck is not working, a low pressure may be applied to the vacuum chuck. The working and non-working response times of the vacuum chuck may be the time from when the controller 130 sends a control signal to the solenoid valve to the change in pressure applied to the vacuum chuck.

[0064] In another example, the pneumatic component 110 may include a servo motor, and multiple operating variables V1, V2, V3 may include the load factor of the servo motor. Here, the load factor of the servo motor may represent the ratio of the servo motor's output to its maximum output. The load factor of the servo motor may be determined based on either the servo motor's drive current or its drive force.

[0065] Multiple sensors 120 can be configured to send multiple operational variables V1, V2, V3 to processor 140. Processor 140 can be configured to collect a dataset of multiple operational variables V1, V2, V3. The multiple operational variables V1, V2, V3 can be collected by processor 140 via controller 130. In another example, processor 140 can be configured to load a dataset including multiple operational variables V1, V2, V3 stored on a server such as MES, data warehouse, and network attached storage.

[0066] Each data point in a dataset of multiple operands V1, V2, and V3 can be a multidimensional quantity. Each dimension of a data point in the dataset of multiple operands V1, V2, and V3 can correspond to different operands V1, V2, and V3. For example, to collect a dataset of multiple operands V1, V2, and V3, M operands can be extracted from N components. In this case, each data point in the dataset of multiple operands V1, V2, and V3 can have an M-dimensional value. Here, N and M can each be integers. M can be greater than N, but is not limited to this.

[0067] Figure 9 and Figure 10 Example datasets with multiple operands V1, V2, and V3 are provided. More specifically, Figure 9 The forward reaction time and forward travel time of the electrode transfer device are illustrated, and Figure 10 The forward reaction time and forward travel time of the electrode tab cutting device are illustrated. Figure 9 and Figure 10 In the diagram, the horizontal axis represents the data collection date, and the vertical axis represents the forward reaction time and forward progress time. Figure 9 and Figure 10 In the diagram, circular dots represent the forward travel time, and triangular dots represent the forward reaction time.

[0068] Reference Figure 9 and Figure 10 It was observed that the dispersion (i.e., stratification) of the electrode tab cutting device with a relatively large number of parts was greater than that of the electrode transfer device with a relatively small number of parts.

[0069] Refer again Figure 1and Figure 2 In P120, a dataset with multiple operands V1, V2, and V3 can be preprocessed. Processor 140 can be configured to preprocess the dataset with multiple operands V1, V2, and V3. Preprocessing the dataset with multiple operands V1, V2, and V3 may include: classifying the dataset with multiple operands V1, V2, and V3; synchronizing the dataset with multiple operands V1, V2, and V3 in time; normalizing the dataset with multiple operands V1, V2, and V3; and removing duplicate and missing data from the dataset with multiple operands V1, V2, and V3.

[0070] Classifying a dataset of multiple operational variables V1, V2, V3 may include determining the state of the pneumatic device 100 when the dataset of multiple operational variables V1, V2, V3 was collected. According to an exemplary embodiment, the processor 140 may be configured to generate a device monitoring model based on a subset of the datasets of multiple operational variables V1, V2, V3 selected based on the classification of the datasets of multiple operational variables V1, V2, V3.

[0071] For example, the operating state of pneumatic equipment 100 may include normal and faulty states. If controller 130 generates error codes for some of the multiple pneumatic components 110, the operating state of pneumatic equipment 100 may be faulty. Conversely, if controller 130 does not generate error codes, the operating state of pneumatic equipment 100 may be normal. Here, controller 130 may record and collect error codes when the value of sensor 120 exceeds an upper threshold or falls below a lower threshold. As a non-limiting example, error codes may be sent by controller 130 to a server such as a manufacturing execution system (MES) and may be stored in the MES. The dataset of multiple operating variables V1, V2, V3 may include a portion collected when the secondary battery manufacturing equipment is in a normal operating state, and the dataset of multiple operating variables V1, V2, V3 may include a portion collected when the secondary battery manufacturing equipment is in a faulty operating state.

[0072] The portion of the dataset containing multiple operating variables V1, V2, and V3 collected when the secondary battery manufacturing equipment is in normal operating condition can be referred to as the normal dataset of multiple operating variables V1, V2, and V3. The portion of the dataset containing multiple operating variables V1, V2, and V3 collected when the secondary battery manufacturing equipment is in faulty operating condition can be referred to as the fault dataset of multiple operating variables V1, V2, and V3. In some cases, the optimal dataset for multiple operating variables V1, V2, and V3 can be selected from the normal dataset. The normal dataset for multiple operating variables V1, V2, and V3 can include the optimal dataset for multiple operating variables V1, V2, and V3. By selecting the optimal dataset for multiple operating variables V1, V2, and V3, the normal dataset can exclude the deteriorated portion caused by the operation of multiple pneumatic components 110.

[0073] At least some of the multiple sensors 120 have a temporal resolution on the order of milliseconds, and therefore each of the multiple sensors 120 can sense and store multiple manipulated variables V1, V2, V3 at different times. According to an exemplary embodiment, the data points of each of the multiple manipulated variables V1, V2, and V3 can be segmented at a set time interval (e.g., one second), and a representative value for the segmentation interval can be selected to compensate for differences in the times at which the multiple manipulated variables V1, V2, and V3 are collected. Here, the representative value of the multiple manipulated variables V1, V2, and V3 can be any one of the average, standard deviation, median, maximum, and minimum values ​​within the interval of the multiple manipulated variables V1, V2, and V3.

[0074] Normalization of the dataset of multiple manipulated variables V1, V2, and V3 can be standard normalization. Normalizing the dataset of multiple manipulated variables V1, V2, and V3 can include dividing each of the multiple manipulated variables V1, V2, and V3 by the absolute value of the maximum value of each of the multiple manipulated variables V1, V2, and V3. Each of the multiple manipulated variables V1, V2, and V3 has a different value depending on the operation of an associated pneumatic component among the multiple pneumatic components 110. Therefore, some of the multiple manipulated variables V1, V2, and V3 may have relatively large absolute values ​​compared to other parts of the multiple manipulated variables V1, V2, and V3.

[0075] The method for monitoring device status described in this paper relies on the distance between newly measured data points and clusters of datasets comprising multiple operands V1, V2, V3, such that if the datasets of multiple operands V1, V2, V3 and newly measured data points of multiple operands V1, V2, V3 are not normalized, unintended weights are applied to multiple operands V1, V2, V3.

[0076] For example, without normalization, a cylinder with a relatively long forward travel time among the multiple operands V1, V2, and V3 due to differences in drive configuration will be more important for monitoring device status than a cylinder with a relatively short forward travel time among the multiple operands V1, V2, and V3. According to an exemplary embodiment, the importance of each of the multiple operands V1, V2, and V3 can be balanced by normalizing the datasets of the multiple operands V1, V2, and V3.

[0077] Furthermore, duplicate and missing data in the dataset containing multiple operational variables V1, V2, and V3 may skew the distribution of the dataset. According to an exemplary embodiment, by removing duplicate and missing data from the dataset containing multiple operational variables V1, V2, and V3, distribution distortion of the dataset used to generate the equipment monitoring model can be avoided, and the reliability of the pneumatic device 100 and the method for monitoring the pneumatic device 100 can be improved.

[0078] Next, a device monitoring model can be generated at P130. The device monitoring model can be included in processor 140. The device monitoring model can be generated by processor 140. According to an exemplary embodiment, the device monitoring model can be generated based on a dataset of multiple operational variables V1, V2, and V3. According to an exemplary embodiment, the device monitoring model can be generated based on a normal dataset of data for multiple operational variables V1, V2, and V3. According to an exemplary embodiment, the device monitoring model can be generated based on an optimal dataset of data for multiple operational variables V1, V2, and V3.

[0079] The generated device monitoring model may include classifying any one of the following into multiple clusters based on density: a dataset containing multiple operands V1, V2, and V3; a normal dataset containing multiple operands V1, V2, and V3; and an optimal dataset containing multiple operands V1, V2, and V3. According to an exemplary implementation, the device monitoring model for processor 140 can be generated using mean-shift clustering.

[0080] Mean-shift clustering can involve using kernel density estimation (KDE) to obtain the probability density function of the cluster centroids and shifting the centroids in the direction that increases the probability density function. The centroid shift can be repeated for a specified number of iterations, or it can terminate when the clusters have not shifted. KDE can be defined according to the following formula.

[0081]

[0082] Here, K is the kernel function, x is the value of the random variable, xi is the data point in the normal dataset with multiple operational variables (V1, V2, V3), and h is the bandwidth. Setting the bandwidth is necessary for mean-shift clustering. Mean-shift clustering can provide more rigorous and flexible clustering because it does not require any settings for the shape of the dataset and the number of clusters.

[0083] Examples of kernels include uniform kernels, triangular kernels, double-weighted kernels, Laplace kernels, Cauchy kernels, square kernels, Yepanechnikov kernels, normal kernels, and Gaussian kernels. A kernel is a positive function that satisfies both normalization and symmetry conditions.

[0084] According to an exemplary implementation, the device monitoring model of processor 140 can be generated by either K-means clustering or density-based noisy spatial clustering application (DBSCAN).

[0085] K-means clustering establishes cluster centroids and clusters data points based on the distance between the centroids and the data points. The cluster centroids are then shifted to the mean point of the data, and the data points are re-clustered based on the shifted centroids. In K-means clustering, the number of cluster centroids is preset in the design of the equipment monitoring model, and iterations can be repeated until no more centroids need to be moved.

[0086] DBSCAN involves finding the satisfying... The minimum number of data points within a radius are identified as core points, and if core points are found in their neighboring points, then these core points are connected to each other. Here, neighboring points are those located within the radius of the data points. Other data points within the radius. DBSCAN needs to be configured. Radius and minimum number of data points. In DBSCAN, although the number of neighboring points of a boundary point is less than the minimum number of data points, the neighboring points of a boundary point can include core points. In DBSCAN, the number of neighboring points of a noise point is less than the minimum number of data points, and the neighboring points of a noise point do not include core points.

[0087] Figure 11 This is a diagram illustrating a method for monitoring a device according to an exemplary embodiment.

[0088] Reference Figure 1 , Figure 2 and Figure 11 At P140, pneumatic device 100 can be monitored. Pneumatic device 100 can be monitored by processor 140. At P130, datasets of multiple manipulated variables V1, V2, and V3, one of the normal datasets of multiple manipulated variables V1, V2, and V3, and the optimal dataset of multiple manipulated variables V1, V2, and V3 can be assigned to multiple clusters G1, G2, G3, and G4.

[0089] The following describes an example of determining multiple clusters G1, G2, G3, G4 by assigning multiple operational variables V1, V2, V3 to a normal dataset. In this case, the multiple clusters G1, G2, G3, G4 can represent different stages of normal operation of the pneumatic device 100.

[0090] For example, one of the multiple clusters G1, G2, G3, and G4 can be a non-operating cluster, another of the multiple clusters G1, G2, G3, and G4 can be an initial operating cluster, and another of the multiple clusters G1, G2, G3, and G4 can be a main operating cluster. The non-operating cluster can be a sub-cluster of normal datasets of multiple operating variables V1, V2, and V3 originating from the phase when the pneumatic device 100 is not operating. The initial operating cluster can be a sub-cluster of normal datasets of multiple operating variables V1, V2, and V3 originating from the phase after each element (e.g., multiple pneumatic components 100) has started operating but before it has entered a stable state. The main operating cluster can be a sub-cluster of normal datasets of multiple operating variables V1, V2, and V3 originating from the phase after each element (e.g., multiple pneumatic components 100) has started operating but has entered a stable state.

[0091] The pneumatic device 100 can be monitored by the processor 140. The processor 140 can be configured to monitor the pneumatic device 100 based on data points of the operating variables V1, V2, V3 of the pneumatic device 100 and the device monitoring model.

[0092] Monitoring the pneumatic device 100 may include calculating the distances of new data points (NDPs) collected from multiple sensors 120 to the centroids C1, C2, C3, C4 of each of the clusters G1, G2, G3, G4 in the normal dataset. The processor 140 may be configured to calculate the distances of the data points (NDPs) to the centroids (C1, C2, C3, C4) of each cluster (G1, G2, G3, G4) in the normal dataset.

[0093] exist Figure 11 Although two-dimensional distances are shown here for illustrative purposes, the distances to clusters G1, G2, G3, and G4 can be calculated in the same dimension as the data points in the datasets of the multiple operands V1, V2, and V3. For example, if the data points in the datasets of the multiple operands V1, V2, and V3 are M-dimensional, then the distances to clusters in the normal dataset can be calculated in M-dimensional space.

[0094] According to an exemplary embodiment, monitoring the pneumatic device 100 may further include calculating a score based on the distance of a data point to a cluster of data sets containing multiple operational variables V1, V2, V3. The score may be calculated by the processor 140. The score may be calculated based on a representative value of the distance between a data point in the datasets containing multiple operational variables V1, V2, V3 and a cluster of data points (e.g., the centroid of the cluster). The representative value may be any one of the mean, standard deviation, median, maximum, and minimum values ​​of the distance between the data point and the cluster of data sets containing multiple operational variables V1, V2, V3 (e.g., the centroid of the cluster).

[0095] The method for monitoring the equipment may also include generating alarms based on scores. According to an exemplary embodiment, processor 140 may be configured to generate commands to control the operation of pneumatic equipment 100 based on a comparison of a score and a threshold, or to provide an alarm to an operator. For example, processor 140 may be configured to generate instructions to control the operation of pneumatic equipment 100 or to provide an alarm to an operator if the score changes below or above a threshold.

[0096] Figure 12 The changes in the scores based on the experimental example are illustrated.

[0097] Reference Figure 1 , Figure 2 , Figure 11 and Figure 12 Monitoring pneumatic device 100 at P140 may include monitoring a fraction of data points (NDPs) SC originating from sensor 120 of pneumatic device 100. Figure 12 In this system, new data points (NDP) can be collected in real time, and therefore, scores (SC) can be calculated in near real-time. Figure 12 In this context, representative distance (RD) and score (SC) can be calculated based on four operands (Va, Vb, Vc, Vd). Multiple operands Va, Vb, Vc, and Vd are... Figure 1 Examples of multiple operands V1, V2, V3.

[0098] At the first time point (t1), an alarm can be generated by changes in the fraction. At the first time point, stratification of multiple manipulated variables Va, Vb, Vc, and Vd is identified. Stratification of multiple manipulated variables Va, Vb, Vc, and Vd indicates that multiple manipulated variables Va, Vb, Vc, and Vd deviate from their normal states. Subsequently, a fault in pneumatic equipment 100 is identified at the second time point (t2), adjacent to the first time point (t1). At the second time point t2, with the occurrence of a fault in pneumatic equipment 100, a peak may appear in the representative distance RD, and a trough may appear in the fraction SC. Figure 12In the experimental examples, it was confirmed that the method for monitoring pneumatic equipment according to the exemplary implementation provides a prediction of the failure of pneumatic equipment 100.

[0099] The present disclosure has been described in more detail above with reference to the accompanying drawings and embodiments. However, it should be understood that the configurations shown in the drawings or the embodiments described herein are merely one embodiment of the present disclosure and do not represent all the technical concepts of the present disclosure, and various equivalents and modifications may exist that can replace them when the present disclosure is submitted.

Claims

1. A method for monitoring pneumatic equipment, the method comprising the following steps: Collect a dataset of multiple operating variables for a pneumatic device that includes multiple pneumatic components; as well as Generate an equipment monitoring model, in which, The pneumatic equipment monitoring method includes generating the equipment monitoring model by clustering the dataset of the plurality of operational variables based on density.

2. The pneumatic equipment monitoring method according to claim 1, wherein, The dataset of the multiple operational variables is clustered based on mean-shift clustering.

3. The pneumatic equipment monitoring method according to claim 1, further comprising the following steps: Collect data points of the multiple operating variables of the pneumatic device; as well as The pneumatic equipment is monitored based on the data points and the equipment monitoring model.

4. The pneumatic equipment monitoring method according to claim 3, wherein, The steps of monitoring the pneumatic equipment include: calculating a score indicating the state of the pneumatic equipment based on the distance of the data point from the normal cluster, the normal cluster being determined based on the clustering of the datasets of the plurality of operational variables.

5. The pneumatic equipment monitoring method according to claim 4, further comprising the following steps: An alert is generated based on the score.

6. The pneumatic equipment monitoring method according to claim 1, further comprising the following steps: The dataset containing the multiple operational variables is preprocessed.

7. The pneumatic equipment monitoring method according to claim 6, wherein, The step of preprocessing the dataset of the multiple operational variables involves classifying the dataset of the multiple operational variables of the pneumatic device based on the operating state of the pneumatic device.

8. The pneumatic equipment monitoring method according to claim 7, wherein, The datasets of the plurality of operands are classified into normal datasets of the plurality of operands and fault datasets of the plurality of operands.

9. The pneumatic equipment monitoring method according to claim 8, wherein, The normal dataset of the plurality of operands includes the optimal dataset of the plurality of operands.

10. The pneumatic equipment monitoring method according to claim 8, wherein, The model is generated based on the normal dataset containing the multiple operational variables.

11. The pneumatic equipment monitoring method according to claim 6, wherein, The step of preprocessing the dataset of the plurality of operational variables standardizes the dataset of the plurality of operational variables of the pneumatic device.

12. The pneumatic equipment monitoring method according to claim 6, wherein, The step of preprocessing the dataset of the plurality of operational variables removes duplicate and missing data from the dataset of the plurality of operational variables of the pneumatic device.

13. The pneumatic equipment monitoring method according to claim 6, wherein, The step of preprocessing the dataset of the plurality of operational variables enables the time synchronization of the dataset of the plurality of operational variables of the pneumatic device.

14. A pneumatic device, the pneumatic device comprising: Multiple pneumatic components; A sensor configured to detect multiple operating variables of the plurality of pneumatic components; as well as A processor configured to collect a dataset of the plurality of operational variables, wherein, The processor includes a device monitoring model generated by clustering the dataset of the plurality of operational variables.

15. The pneumatic device according to claim 14, wherein, The device monitoring model is generated by clustering the dataset of the multiple operational variables based on density.

16. The pneumatic device according to claim 14, wherein, The dataset of the multiple operational variables is clustered based on mean-shift clustering.

17. The pneumatic device according to claim 14, wherein, The processor is configured to monitor the pneumatic device based on data points of the plurality of operating variables of the pneumatic device and the device monitoring model.

18. The pneumatic device according to claim 17, wherein, The processor is configured to calculate the distance of the data point to a normal cluster, which is determined based on the clustering of the dataset with the plurality of operational variables.

19. The pneumatic device according to claim 18, wherein, The processor is configured to calculate a score indicating the state of the pneumatic device based on the distance of the data point from the normal cluster.

20. The pneumatic device according to claim 19, wherein, The processor is configured to generate an alarm based on the score.

Images