Anomaly detection system, molding machine system, anomaly detection device, anomaly detection method, and computer program

Abstract

To provide an anomaly detection system capable of detecting anomalies in a production device without excessively burdening a control device controlling the operation of the production device.SOLUTION: An anomaly detection system comprises: an anomaly detection device for detecting anomalies in a production device; a control device for controlling operation of the production device and sending operation data; and sensors for detecting physical quantities associated with the operation of the production device or a product, and outputting sensor value data. The anomaly detection device is configured to: receive the operation data sent from the control device; acquire the sensor value data output from the sensors; compute a statistical amount of the acquired sensor value data; determine presence or absence of anomalies in the production device on the basis of the computed statistical amount and a threshold value corresponding to the received operation data; and send a determination result and the statistical amount to the production device.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to an abnormality detection system, a molding machine system, an abnormality detection device, an abnormality detection method, and a computer program. [Background technology]

[0002] Patent Document 1 discloses a monitoring method for monitoring the vibrations of each moving part of an injection molding machine using an acceleration sensor, and detecting the state of the molding process and the occurrence of abnormalities in the moving parts. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 5-50480 Summary of the Invention [Problem to be solved by the invention]

[0004] However, Patent Document 1 does not disclose a specific technology for configuring a control device that controls the operation of a manufacturing device such as an injection molding machine and a device that detects abnormalities in the manufacturing device or diagnoses the state of the manufacturing device separately. If the control device is configured to detect and diagnose abnormalities in the manufacturing device, there is a possibility that the operation of the manufacturing device will be adversely affected, such as delays in operation control processing.

[0005] The object of the present disclosure is to provide an anomaly detection system, molding machine system, anomaly detection device, an anomaly detection method, and a computer program that include a control device that controls the operation of a manufacturing device and an anomaly detection device, and that can detect anomalies in the manufacturing device by having each device work together without placing an excessive load on the control device. [Means for solving the problem]

[0006] The anomaly detection system according to the present disclosure is an anomaly detection system equipped with an anomaly detection device that detects anomalies in manufacturing equipment, and includes a control device that controls the operation of the manufacturing equipment and transmits operating data indicating the details of the operating control, and a sensor that detects physical quantities related to the operation or product of the manufacturing equipment and outputs time-series sensor value data indicating the detected physical quantities, and the anomaly detection device includes a communication unit that receives the operating data transmitted from the control device, an acquisition unit that acquires the sensor value data output from the sensor, and a processing unit that calculates statistics of the sensor value data acquired by the acquisition unit and determines whether or not there is an anomaly in the manufacturing equipment based on the calculated statistics and a threshold value corresponding to the received operating data, and the communication unit transmits the determination result and the statistics to the control device.

[0007] A molding machine system according to the present disclosure includes the above-described abnormality detection system and a molding machine, and the abnormality detection system is configured to detect an abnormality in the molding machine.

[0008] The anomaly detection device according to the present disclosure is an anomaly detection device that detects an anomaly in a manufacturing device, and includes a communication unit that receives operating data indicating the content of the operating control transmitted from a control device that controls the operation of the manufacturing device, an acquisition unit that acquires time-series sensor value data output from a sensor that detects physical quantities related to the operation of the manufacturing device or a product, and a processing unit that calculates statistics of the sensor value data acquired by the acquisition unit and determines whether or not there is an anomaly in the manufacturing device based on the calculated statistics and a threshold value corresponding to the acquired operating data, and the communication unit transmits the determination result and the statistics to the control device.

[0009] The anomaly detection method according to the present disclosure is an anomaly detection method in which a computer executes a process for detecting an anomaly in a manufacturing device, in which the computer receives operating data indicating the details of the operating control transmitted from a control device that controls the operation of the manufacturing device, acquires time-series sensor value data output from a sensor that detects physical quantities related to the operation or product of the manufacturing device, calculates statistics of the acquired sensor value data, and determines whether or not there is an anomaly in the manufacturing device based on the calculated statistics and a threshold value corresponding to the acquired operating data, and transmits the determination result to the control device.

[0010] The computer program according to the present disclosure is a computer program for causing a computer to execute a process for detecting an abnormality in a manufacturing apparatus, and causes the computer to execute the following process: receive operating data indicating the content of the operating control transmitted from a control device that controls the operation of the manufacturing apparatus; acquire time-series sensor value data output from a sensor that detects physical quantities related to the operation or product of the manufacturing apparatus; calculate statistics of the acquired sensor value data; determine whether or not there is an abnormality in the manufacturing apparatus based on the calculated statistics and a threshold value corresponding to the acquired operating data; and transmit the determination result to the control device. [Effects of the Invention]

[0011] According to the present disclosure, a control device that controls the operation of a manufacturing device and an abnormality detection device are provided, and by each device working together, abnormalities in the manufacturing device can be detected without placing an excessive load on the control device. [Brief explanation of the drawings]

[0012] [Figure 1] 1 is a block diagram showing an example of the configuration of a molding machine system according to a first embodiment of the present invention. [Figure 2] 1 is a schematic diagram showing a configuration example of a twin-screw kneading extruder according to the first embodiment. [Figure 3] 1 is a block diagram showing an example of the configuration of an anomaly detection device according to a first embodiment of the present invention. [Figure 4]1 is a block diagram showing an example of the configuration of a diagnostic device according to a first embodiment. [Figure 5] FIG. 2 is a conceptual diagram illustrating an example of a record layout of a database according to the first embodiment. [Figure 6] 3 is a flowchart showing a processing procedure for detecting and diagnosing an abnormality in the twin-screw kneading extruder according to the first embodiment. [Figure 7] 3 is a flowchart showing a processing procedure for detecting and diagnosing an abnormality in the twin-screw kneading extruder according to the first embodiment. [Figure 8] 10 is a flowchart showing a procedure for determining an abnormality. [Figure 9] 10 is a flowchart showing a procedure for diagnosing an abnormality. [Figure 10] 10 is a flowchart showing a procedure for diagnosing an abnormality in the screw shaft 11. [Figure 11] This is a time-series data image showing the displacement of two screw shafts. [Figure 12] 10A and 10B are conceptual diagrams illustrating feature amounts and anomaly detection results of time-series data images. [Figure 13] 10 is a flowchart showing a processing procedure for lifespan diagnosis. [Figure 14] FIG. 1 is a conceptual diagram showing an overview of a method for generating a learning model. [Figure 15] FIG. 1 is a conceptual diagram illustrating a learning model in a learning phase. [Figure 16] FIG. 1 is a conceptual diagram illustrating a learning model in a test phase. [Figure 17] 5 is a schematic diagram showing an example of displaying an abnormality determination result etc. in the control device. FIG. [Figure 18] 10A and 10B are schematic diagrams showing examples of displaying abnormality determination results and the like on a terminal device. [Figure 19] 10 is a flowchart showing a feedback process procedure for a diagnosis result according to the second embodiment. [Figure 20] 10 is a flowchart showing a feedback process procedure regarding the state of the twin-screw kneading extruder according to the third embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0013] Anomaly detection system, molding machine system, anomaly detection device, and anomaly detection according to an embodiment of the present invention Specific examples of the method and computer program will be described below with reference to the drawings. Note that the present invention is not limited to these examples, but is defined by the claims, and is intended to include all modifications within the meaning and scope of the claims. In addition, at least some of the embodiments and modifications described below may be combined in any manner.

[0014] 1 is a block diagram showing an example of the configuration of a molding machine system according to Embodiment 1. The molding machine system includes a twin-screw kneading extruder 1, multiple sensors 2, an abnormality detection device 3, a router 4, a diagnostic device 5, a database 6, and a user terminal device 7. The molding machine system is made up of the twin-screw kneading extruder 1 and the abnormality detection system, which is made up of the multiple sensors 2, the abnormality detection device 3, the diagnostic device 5, and the database 6.

[0015] <Twin-screw kneading extruder 1> 2 is a schematic diagram showing an example of the configuration of the twin-screw kneading extruder 1 according to the present embodiment 1. The twin-screw kneading extruder 1 includes a cylinder 10 having a hopper 10a into which a resin raw material is introduced, and two screw shafts 11. The two screw shafts 11 are arranged substantially parallel to each other while meshing with each other and are rotatably inserted into holes in the cylinder 10. The resin raw material introduced into the hopper 10a is transported in the extrusion direction (to the right in FIGS. 1 and 2) and melted and kneaded. The screw shaft 11 is configured by combining and integrating multiple types of screw pieces into a single screw shaft 11. For example, the screw shaft 11 is configured by combining forward flight pieces of a flight screw shape that transport the resin raw material in the forward direction, reverse flight pieces that transport the resin raw material in the reverse direction, kneading pieces that knead the resin raw material, etc., in an order and positions according to the properties of the resin raw material.

[0016] The twin-screw kneading extruder 1 also includes a motor 12 that outputs a driving force for rotating the screw shaft 11, a reducer 13 that reduces and transmits the driving force of the motor 12, and a control device 14. The screw shaft 11 is connected to the output shaft of the reducer 13. The screw shaft 11 is rotated by the driving force of the motor 12 that is reduced and transmitted by the reducer 13.

[0017] <Sensor 2> The sensor 2 detects physical quantities related to the operation of the twin-screw kneading extruder 1 or the molded product, and outputs time-series sensor value data indicating the detected physical quantities to the abnormality detection device 3. The physical quantities include temperature, position, speed, acceleration, current, voltage, pressure, time, image data, torque, force, strain, power consumption, weight, etc. These physical quantities can be measured using a thermometer, position sensor, speed sensor, acceleration sensor, ammeter, voltmeter, pressure gauge, timer, camera, torque sensor, wattmeter, weight scale, etc. More specifically, the sensor 2 that detects physical quantities related to the operation of the twin-screw kneading extruder 1 is a vibration sensor such as an acceleration sensor that detects the vibration of the reducer 13, a torque sensor that detects the axial torque applied to the screw shaft 11, a thermometer that detects the temperature of the screw shaft 11, a displacement sensor that detects the displacement of the center of rotation of the screw shaft 11, an optical sensor that detects the dimensions of the molded product or strand, an imaging device that images the molded product or strand, a resin pressure sensor that detects the resin pressure, etc. The sensor 2 for detecting a physical quantity related to the molded product is an optical measuring instrument for detecting the dimensions, chromaticity, brightness, etc. of the molded product, an imaging device, a weighing scale for detecting the weight of the molded product, etc.

[0018] <Control device 14> The control device 14 is a computer that controls the operation of the twin-screw kneading extruder 1, and includes a transmitter / receiver (not shown) that transmits and receives information to and from the abnormality detection device 3, and a display unit. Specifically, the control device 14 transmits operation data indicating the operation state of the twin-screw kneading extruder 1 to the abnormality detection device 3. The operation data includes, for example, the feeder supply amount (supply amount of resin raw material), the rotation speed of the screw shaft 11, the extrusion amount, the cylinder temperature, the resin pressure, the motor current, etc. The control device 14 detects whether or not there is an abnormality in the twin-screw kneading extruder 1, which is transmitted from the abnormality detection device 3. The control device 14 receives the judgment results described below, various statistical quantities of the sensor value data, and diagnostic results relating to abnormalities, lifespan, etc. of the twin-screw kneading extruder 1, and displays the received judgment results, values ​​and graphs of the statistical quantities, and diagnostic results. In addition, the control device 14 monitors abnormalities in the twin-screw kneading extruder 1 using the judgment results, etc., outputs a warning as necessary, and executes processing to stop the operation of the twin-screw kneading extruder 1.

[0019] <Anomaly detection device 3> 3 is a block diagram showing an example of the configuration of the abnormality detection device 3 according to the present embodiment 1. The abnormality detection device 3 is an edge computer, and includes a processing unit 31, a storage unit 32, a communication unit 33, and an acquisition unit 34, and the storage unit 32, the communication unit 33, and the acquisition unit 34 are connected to the processing unit 31. The edge computer constituting the abnormality detection device 3 is a computer with lower hardware specifications than the diagnostic device 5 described below, and executes simple abnormality detection processing for the twin-screw kneading extruder 1.

[0020] The processing unit 31 includes a central processing unit (CPU), a multi-core CPU, an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other processing circuits, a read-only memory (ROM), a random access memory (RAM), etc. The processing unit 31 has internal storage devices such as a memory unit 32, an I / O terminal, etc. The processing unit 31 functions as the anomaly detection device 3 according to the first embodiment by executing a computer program P1 stored in a storage unit 32 described below. Note that each functional unit of the anomaly detection device 3 may be realized by software, or some or all of it may be realized by hardware.

[0021] The storage unit 32 includes a hard disk, an EEPROM (Electrically Erasable Programmable Read Only Memory), The storage unit 32 is a non-volatile memory such as a ROM, a flash memory, etc. The storage unit 32 stores a computer program P1 for causing a computer to execute the anomaly detection method according to the first embodiment.

[0022] The computer program P1 according to the first embodiment may be recorded on a recording medium 30 in a computer-readable manner. The storage unit 32 stores the computer program P1 read from the recording medium 30 by a reading device (not shown). The recording medium 30 is a semiconductor memory such as a flash memory. The recording medium 30 may also be a CD (Compact Disc)-ROM, a DVD (Digital Versatile Disc)-ROM, or a BD (Blu-ray (registered trademark) Disc) The recording medium 30 may be a magnetic disk such as a flexible disk or a hard disk, a magneto-optical disk, etc. Furthermore, the computer program P1 according to the first embodiment may be downloaded from an external server (not shown) connected to a communication network (not shown) and stored in the storage unit 32.

[0023] The communication unit 33 is a communication circuit that transmits and receives information according to a predetermined communication protocol such as Ethernet (registered trademark). The communication unit 33 is connected to the control device 14 via a first communication network such as a LAN, and the processing unit 31 can transmit and receive various information to and from the control device 14 via the communication unit 33. A router 4 is connected to the first network, and the communication unit 33 is connected to a diagnostic device 5 on the cloud, which is a second communication network, via the router 4. The processing unit 31 can send and receive various information to and from the diagnostic device 5 via the communication unit 33 and the router 4.

[0024] The acquisition unit 34 is an input interface for receiving signals. The sensor 2 is connected to the acquisition unit 34, and sensor value data, which is time-series data output from the sensor 2, is input to the acquisition unit 34.

[0025] <Diagnostic device 5 and database 6> FIG. 4 is a block diagram showing an example of the configuration of the diagnostic device 5 according to the first embodiment. The diagnostic device 5 includes: The computer includes a diagnostic processing unit 51, a storage unit 52, and a communication unit 53. The storage unit 52 and the communication unit 53 are connected to the diagnostic processing unit 51.

[0026] The diagnostic processing unit 51 includes an arithmetic processing circuit such as a CPU, a multi-core CPU, a GPU (Graphics Processing Unit), a GPGPU (General-purpose computing on graphics processing units), a TPU (Tensor Processing Unit), an ASIC, an FPGA, or an NPU (Neural Processing Unit), an internal storage device such as a ROM or RAM, an I / O terminal, etc. The diagnostic processing unit 51 functions as the diagnostic device 5 according to the first embodiment by executing a computer program P2 stored in a storage unit 52 described below. Note that each functional unit of the diagnostic device 5 may be realized by software, or some or all of them may be realized by hardware.

[0027] The storage unit 52 is a non-volatile memory such as a hard disk, an EEPROM, a flash memory, etc. The storage unit 52 stores a computer program P2 for causing a computer to execute a process for diagnosing the state of the twin-screw kneading extruder 1, such as abnormalities and the lifespan of the twin-screw kneading extruder 1, and one or more learning models 54.

[0028] The computer program P2 and the learning model 54 may be recorded in a computer-readable manner on a recording medium 50. The storage unit 52 stores the computer program P2 and the learning model 54 read from the recording medium 50 by a reading device (not shown). The recording medium 50 is a semiconductor memory such as a flash memory. The recording medium 50 may also be an optical disc such as a CD (Compact Disc)-ROM, a DVD (Digital Versatile Disc)-ROM, or a BD (Blu-ray (registered trademark) Disc). Furthermore, the recording medium 50 may also be a flexible disc. The storage device 52 may be a magnetic disk such as a disk or a hard disk, a magnetic optical disk, etc. Furthermore, the computer program P2 and the learning model 54 may be downloaded from an external server (not shown) connected to a communication network (not shown) and stored in the storage device 52.

[0029] The communication unit 53 is a communication circuit that transmits and receives information according to a predetermined communication protocol such as Ethernet (registered trademark). The communication unit 53 is connected to the anomaly detection device 3, the database 6, and the terminal device 7 via a second communication network, and the processing unit 31 can transmit and receive various types of information to and from the anomaly detection device 3, the database 6, and the terminal device 7 via the communication unit 53.

[0030] 5 is a conceptual diagram showing an example of a record layout of the database 6 according to the embodiment 1. The database 6 includes a hard disk and a DBMS (Database Management System), and stores various information related to the operation of the twin-screw kneading extruder 1. For example, a table of the database 6 has a "No." (record number) column, an "operation date and time" column, an "operation data" column, a "screw configuration" column, a "resin type" column, a "sensor value" column, a "status" column, a "quality evaluation" column, and a "lifespan" column.

[0031] The "Driving Date and Time" column stores information indicating the year, month, date and time when the various data stored as records were obtained. The "operation data" column stores information indicating the operation state of the twin-screw kneading extruder 1, such as the feeder supply amount (supply amount of resin raw material), the rotation speed of the screw shaft 11, the extrusion amount, the cylinder temperature, the resin pressure, the motor current, etc. The "screw configuration" column stores information regarding the type, arrangement, or combination of various screw pieces that make up the screw shaft 11. The "resin type" column stores information indicating the type of resin raw material supplied to the twin-screw kneading extruder 1. The "sensor value" column stores sensor value data. The "sensor value" column may store numerical data as is, or may display the sensor value data, which is time-series data, as an image, as will be described later. The time series data may be stored as an image. The "Status" column stores information indicating the state of the twin-screw kneading extruder 1 when the operation control indicated by the operation data is performed and the sensor value data is obtained. For example, information indicating normality, information indicating an abnormal state where the screw shaft 11 is worn, information indicating an abnormal state where a resin raw material incompatible with the screw shaft 11 is being used, information indicating an overload error, etc. are stored. The "quality evaluation" column stores information indicating the quality evaluation of the molded product produced by the twin-screw kneading extruder 1 when the operation control indicated by the above operation data is performed and the above sensor value data is obtained. For example, information indicating whether the molded product is normal or not, dimensional abnormalities, etc. is stored. The "Lifespan" column stores information indicating the lifespan of the twin-screw kneading extruder 1 used to obtain the above information. The lifespan may be the total product lifespan or the remaining lifespan from the time the above data is obtained.

[0032] The various data stored in the database 6 can be obtained, for example, by trial operating the twin-screw kneading extruder 1, checking the operating state of the twin-screw kneading extruder 1, and evaluating the molded product. Alternatively, the operating state of the twin-screw kneading extruder 1 can be simulated using a simulator to prepare the various data. In the first embodiment, the database 6 stores the data obtained by trial or experimentally operating the twin-screw kneading extruder 1 or the data obtained by simulation.

[0033] <Example of anomaly> Abnormalities that can be detected or diagnosed by the abnormality detection device 3 and the diagnosis device 5 include, for example, abnormal vibration of the reducer 13, overload, abnormal dimensions of the molded product, scratches, cracks, wear, corrosion on the screw shaft 11, performance deterioration or oil abnormalities due to wear of the reducer 13 and its parts, inefficient operating conditions (increased energy consumption, etc.), operating conditions that lead to quality abnormalities (unreasonable start-up, etc.), strand abnormalities (defects in dimensions, foreign matter, color, coils (twists), etc.), resin viscosity (quality) abnormalities, etc. In the first embodiment, the abnormality detection device 3 detects abnormal vibrations of the reducer 13, overloads, dimensional abnormalities of the molded product, strand abnormalities, resin viscosity abnormalities, etc. based on the sensor value data. Specifically, the abnormality detection device 3 calculates statistics such as the average, variance, and root mean square of the sensor values, and detects abnormal vibrations of the reducer 13, overloads, dimensional abnormalities of the molded product, strand abnormalities, resin viscosity abnormalities, etc. through low-load processing based on the calculated statistics. On the other hand, the diagnostic device 5 uses the operation data and sensor value data acquired from the control device 14 via the abnormality detection device 3, as well as past sensor value data stored in the database 6, to diagnose the presence or absence of abnormalities such as scratches, cracks, wear, and corrosion on the screw shaft 11, performance deterioration or oil abnormalities in the reducer 13, inefficient operating conditions, and operating conditions that lead to quality abnormalities. The diagnostic device 5 also performs processing to identify faulty parts of the reducer 13, such as bearings and gears. The division of the anomaly detection process performed by the anomaly detection device 3 and the diagnosis process performed by the diagnosis device 5 is an example, and the anomaly detection device 3 may be configured to perform the process if it is capable of performing the process.

[0034] <Overall flow of anomaly detection processing> 6 and 7 are flowcharts showing the processing procedure for detecting and diagnosing an abnormality in the twin-screw kneading extruder 1 according to embodiment 1. The control device 14 transmits operation data indicating the state of operation control of the twin-screw kneading extruder 1 to the abnormality detection device 3 (step S11).

[0035] The processing unit 31 of the abnormality detection device 3 receives the operation data transmitted from the control device 14 at the communication unit 33 (step S12).

[0036] The processing unit 31 acquires time-series sensor value data output from the sensor 2 (step S13). Then, the processing unit 31 calculates statistics such as the average, variance, and root mean square of the sensor values ​​based on the sensor value data (step S14), and executes an abnormality determination process for the twin-screw kneading extruder 1 based on the calculated statistics (step S15).

[0037] Next, the processing unit 31 transmits the determination result indicating whether or not there is an abnormality in the twin-screw kneading extruder 1 and the statistics calculated in step S14 to the control device 14 via the communication unit 33 (step S16).

[0038] The control device 14 receives the determination result and statistics transmitted from the abnormality detection device 3 (step S17), and monitors the operation of the twin-screw kneading extruder 1 based on the received determination result or statistics (step S18). For example, if the determination result indicates a predetermined abnormality, the control device 14 stops the operation control of the twin-screw kneading extruder 1.

[0039] Next, the control device 14 causes the display unit to display the received determination results, statistics, etc. (step S19).

[0040] Next, we will explain the processing of the anomaly detection device 3 that has transmitted the determination result and statistics. After completing the processing of step S16, the processing unit 31 of the anomaly detection device 3 transmits the operating data received in step S12 and the sensor value data acquired in step S13 to the diagnostic device 5 via the communication unit 33 (step S20).

[0041] The diagnostic device 5 receives the operation data and sensor value data transmitted from the abnormality detection device 3 (step S21), and the diagnostic processing unit 51 executes abnormality diagnosis processing for the twin-screw kneading extruder 1 using the received operation data and sensor value data, as well as past sensor value data stored in the database 6 (step S22). The diagnostic processing unit 51 also executes life diagnosis processing for the twin-screw kneading extruder 1 (step S23). The diagnostic processing unit 51 evaluates material fatigue from the waveform of shaft torque, for example, and calculates the remaining life.

[0042] After completing the processes of steps S22 and S23, the diagnostic processing unit 51 transmits the diagnostic result to the abnormality detection device 3 via the communication unit 53 (step S24).

[0043] The processing unit 31 of the abnormality detection device 3 receives the diagnosis result transmitted from the diagnosis device 5 (step S25), and transmits the received diagnosis result to the control device 14 via the communication unit 33 (step S26).

[0044] The control device 14 receives the diagnosis result transmitted from the abnormality detection device 3 (step S27), and displays the received diagnosis result on the display unit (step S28).

[0045] In response to a request from the terminal device 7, the anomaly detection device 3 can transmit the anomaly determination result by the processing unit 31, the calculated statistics of the sensor value data, the diagnosis result by the diagnosis processing unit 51, etc. to the terminal device 7. The terminal device 7 displays the received anomaly determination result, statistics, and diagnosis result.

[0046] <Abnormality detection process by abnormality detection device 3> 8 is a flowchart showing the procedure for determining an abnormality. Since the abnormality detection device 3 cannot store a large amount of sensor value data, it executes a process for determining whether or not there is an abnormality in the twin-screw kneading extruder 1 using the operation data and sensor value data of the day.

[0047] The processing unit 31 specifies a threshold value of a determination criterion for determining whether or not there is an abnormality in the twin-screw kneading extruder 1 based on the operation data received in step S12 (step S51). Since torque and the rotation speed of the screw shaft 11 are inversely proportional to each other, the threshold value for detecting torque overload depends on the rotation speed of the screw shaft 11, which is one of the operation data. In this way, the processing unit 31 can identify the threshold value for detecting an abnormality based on the operation data. Specifically, the anomaly detection device 3 stores a table that associates operation data, types of anomalies, and threshold values ​​of judgment criteria in the storage unit 32. The processing unit 31 reads the threshold values ​​by referring to the table using the operation data, and can thereby identify the threshold values ​​that are the judgment criteria for various types of anomalies according to the operation data.

[0048] Next, the processing unit 31 determines whether or not there is abnormal vibration in the reducer 13 based on the statistics of the sensor value data indicating the vibration of the reducer 13 and the threshold value identified in step S51 (step S52). Note that the processing unit 31 may be configured to convert the sensor value data indicating the vibration of the reducer 13 into time frequency components, and determine whether or not there is an abnormality in the reducer 13 based on the data of the time frequency components.

[0049] In addition, the processing unit 31 determines whether or not an overload exists based on the statistics of the sensor value data indicating the shaft torque, the temperature of the screw shaft 11, and the displacement of the center of rotation of the screw shaft 11, and the threshold value identified in step S51 (step S53).

[0050] Furthermore, the processing unit 31 determines whether or not there is a dimensional abnormality in the molded product based on the statistics indicating the dimensions of the molded product and the threshold value specified in step S51 (step S54).

[0051] The processing unit 31 determines whether or not there is an abnormality in the strand based on the statistical amount indicating the difference between the image obtained by capturing the strand and a normal image, and the threshold value specified in step S51 (step S55). The processing unit 31 can measure the strand diameter based on the pixel size (number of pixels) of the image of the strand included in the captured image. The processing unit 31 can also determine the presence or absence of a foreign object by determining the color of the image portion included in the captured image.

[0052] The processing unit 31 determines whether or not there is an abnormality in the resin viscosity based on the time average of the resin pressure detected by the sensor 2, for example, the resin pressure sensor, and the threshold value identified in step S51 (step S56), and ends the abnormality determination process.

[0053] <Abnormality diagnosis process by diagnostic device 5> 9 is a flowchart showing the procedure for diagnosing an abnormality. The diagnosis processing unit 51 performs an abnormality diagnosis of the screw shaft 11 by inputting sensor value data indicating the displacement of the rotation center of the screw shaft 11 or time-frequency data obtained by time-frequency analysis of the sensor value data into the learning model 54 (step S61). The learning model 54 used in step S61 is a convolutional neural network model that, when a time-series data image (described later) representing sensor value data, which is time-series data, is input, extracts features from the time-series data image and outputs the extracted feature quantities. The learning model 54 is a model having a feature extraction layer of a convolutional neural network (CNN), for example, a one-class classification model. The diagnosis processing unit 51 can diagnose the presence or absence of an abnormality in the screw shaft 11 by comparing feature quantities obtained from sensor value data associated with operation data that is the same as or similar to the operation data received in step S12 and obtained when the twin-screw kneading extruder 1 is operating normally with feature quantities of the diagnosis target. A method for generating the learning model 54 and its configuration will be described in detail below.

[0054] Furthermore, the diagnostic processing unit 51 inputs vibration sensor data obtained from the sensor 2, for example, a vibration sensor, installed on the reducer 13, or time-frequency data obtained by time-frequency analysis of the sensor value data into the learning model 54, thereby determining the wear of the reducer 13 and each part thereof. The presence or absence of performance degradation or oil abnormality due to the above is diagnosed (step S62). The learning model 54 used in step S62 is, for example, a support vector machine. The support vector machine is a classifier that learns time frequency components obtained by time-frequency analysis of vibration sensor data related to a normal reducer 13 and time frequency components obtained by time-frequency analysis of vibration sensor data related to an abnormal reducer 13, and classifies the data into a normal class related to a normal reducer 13 and an abnormal class related to an abnormal reducer 13. The frequency components to be analyzed may be specific frequency components, such as main frequency components obtained from the reducer 13 or fault frequency components that appear during a fault. The diagnostic processing unit 51 identifies, for example, the main frequency components, fault frequency components, etc. obtained by time-frequency analysis of the vibration sensor data. The diagnostic processing unit 51 then diagnoses whether the reducer 13 is normal by inputting the time frequency components of the vibration sensor data, such as the main frequency components and fault frequency components, into the support vector machine. The learning model 54 may be a classifier using a convolutional neural network model. For example, when vibration sensor data or time-frequency data obtained by time-frequency analysis of the sensor value data is input, the learning model 54 may be trained to output data indicating an abnormality location and abnormality level of the reducer 13. The abnormality location may be, for example, a bearing, gear, etc. of the reducer 13.

[0055] In addition, the diagnostic processing unit 51 diagnoses whether or not the system is in an inefficient operating state by inputting sensor value data obtained from the sensor 2, indicating the axial torque, the displacement of the center of rotation of the screw shaft 11, vibration, etc., or time-frequency data obtained by time-frequency analysis of the sensor value data, into the learning model 54 (step S63). The learning model 54 used in step S63 is a convolutional neural network model that, when a time-series data image (described later) representing sensor value data, which is time-series data, is input, extracts features from the time-series data image and outputs the extracted feature quantities. The learning model 54 is a model having a feature extraction layer of CNN, such as a one-class classification model. The diagnostic processing unit 51 can diagnose whether the twin-screw kneading extruder 1 is in an inefficient operating state by comparing the feature values ​​obtained from the sensor value data associated with the operating data received in step S12 and when the twin-screw kneading extruder 1 is in an efficient operating state with the feature values ​​of the object to be diagnosed.

[0056] In addition, the diagnostic processing unit 51 diagnoses whether the operating state is one that could lead to quality abnormalities by inputting sensor value data obtained from the sensor 2, indicating the displacement of the center of rotation of the screw shaft 11, vibration, etc., or time-frequency data obtained by time-frequency analysis of the sensor value data, into the learning model 54 (step S64). The learning model 54 used in step S64 is a convolutional neural network model that, when a time-series data image (described later) representing sensor value data, which is time-series data, is input, extracts features from the time-series data image and outputs the extracted feature quantities. The learning model 54 is a model having a feature extraction layer of CNN, such as a one-class classification model. The diagnostic processing unit 51 can diagnose whether the twin-screw kneading extruder 1 is in an operating state that could lead to a quality abnormality by comparing the feature values ​​obtained from the sensor value data associated with the operating data that is the same as or similar to the operating data received in step S12 and when the twin-screw kneading extruder 1 is in a normal operating state that is not in an operating state that could lead to a quality abnormality with the feature values ​​of the object to be diagnosed.

[0057] The diagnostic processing unit 51 may be configured to monitor the operation rate by monitoring the current flow data obtained from the control device 14 via the abnormality detection device 3 and calculating the operation rate for a certain period of time. The diagnostic processing unit 51 may be configured to diagnose whether or not there is an abnormality in the twin-screw kneading extruder 1 based on the operation rate.

[0058] Below, screw shaft abnormality diagnosis will be described in detail as an example of a one-class classification model. 10 is a flowchart showing the procedure for diagnosing an abnormality in the screw shaft 11. The diagnosis processing unit 51 converts a plurality of sensor value data into one or more time-series data images represented by images (step S71), and if there are a plurality of converted time-series data images, combines the plurality of time-series data images into a single time-series data image (step S72). For example, if the sensor value data is data indicating displacement in two directions of each of the two screw shafts 11, the diagnosis processing unit 51 combines a first time-series data image representing, as an image, the displacement in two directions about the rotation center axis of the first screw shaft 11 and a second time-series data image representing, as an image, the displacement in two directions about the rotation center axis of the second screw shaft 11.

[0059] FIG. 11 is a time series data image showing the displacement of the two screw shafts 11. The first time series data image shows the displacement of the first screw shaft 11. The first time series data image is a substantially square image, and the displacement amounts in the X-axis and Y-axis directions perpendicular to the rotation center of the screw shaft 11 are plotted as X-axis and Y-axis coordinate values. The first time series data image is a plot of displacement for three rotations. The second time series data image can be created in the same way. The combined time series data image is a substantially square image twice the size (four times the area) of the first time series data image. The time series data image contains the information contained in the original images without changing the aspect ratio of the first and second time series data images. Specifically, the time series data image has the first and second time series data images arranged side by side on the upper left and right sides in FIG. 11, and a blank image arranged below to make the time series data image a substantially square. Note that the arrangement of the first and second time series data images and the blank image is not particularly limited, and is not particularly limited as long as the first and second time series data images are included as they are.

[0060] Note that other multiple sensor value data can also be converted into time-series data images and synthesized in the same manner. In the example shown in Fig. 11, a total of eight sensor value data can be represented by images, but when nine time-series data images of 3x3 are synthesized, 18 sensor value data can be represented by one time-series data image.

[0061] Next, the diagnostic processing unit 51 inputs the time-series data image synthesized in step S72 into the learning model 54, thereby calculating the feature amount of the time-series data image (step S73).

[0062] Furthermore, the diagnostic processing unit 51 reads out from the database 6 the sensor value data that is associated with the same or similar operating data as the operating data received in step S12 and that is obtained when the twin-screw kneading extruder 1 is operating normally (step S74), converts it into a time series data image in the same manner (step S75), synthesizes it (step S76), and inputs the synthesized time series data image into the learning model 54 to calculate the feature quantities of the time series data image during normal operation (step S77).

[0063] Next, the diagnostic processing unit 51 calculates an outlier score for the feature quantity of the time-series data image calculated in step S73 (step S78). The outlier score is a numerical value obtained by evaluating the degree of outlier of the feature quantity of the time-series data image calculated in step S77 relative to the feature quantity of the time-series data image obtained when the twin-screw kneading extruder 1 is operating normally (hereinafter referred to as sample feature quantity). The outlier score is, for example, LOF (Local Outlier Factor). LOF is the local density ld(P) of the feature quantity to be detected as an anomaly and the local density average of the neighborhood group of the feature quantity (the k nearest sample feature quantities to the feature quantity).<ld(Q)> Relative to<ld(Q)> / ld(P). The greater the LOF, the more likely it is to be an outlier. The degree of abnormality increases. P indicates the feature quantity that is the target of anomaly detection, and Q indicates the sample feature quantity.

[0064] Then, the processing unit 31 determines whether or not the calculated outlier score is equal to or greater than a predetermined threshold value, thereby determining whether or not there is an abnormality in the twin-screw kneading extruder 1 (step S79). When the outlier score is the above-mentioned LOF, the threshold value is a value equal to or greater than 1. When the LOF is equal to or greater than the threshold value, the diagnostic processing unit 51 determines that there is an abnormality, and when the LOF is less than the threshold value, it determines that there is a normality.

[0065] The abnormality determination using the LOF described above is one example, and it may be determined by the k-nearest neighbor method or SVM whether or not the twin-screw kneading extruder 1 is abnormal. Also, it may be configured to determine whether or not the twin-screw kneading extruder 1 is abnormal by determining whether or not the feature amount of the abnormality detection target is significantly different from the sample feature amount in normal times using the Hotelling method.

[0066] FIG. 12 is a conceptual diagram showing feature quantities and anomaly detection results of a time-series data image. FIG. 12 shows high-dimensional feature quantities of the time-series data image reduced to two-dimensional feature quantities and plotted on a two-dimensional plane. The horizontal and vertical axes of the graph shown in FIG. 12 represent the first and second feature quantities of the dimension-reduced time-series data image. The dimension reduction of the feature quantities can be performed using a dimension reduction algorithm such as t-SNE. As shown in FIG. 12, the learning model 54 is trained so that the local density of the feature quantity output from the learning model 54 in a normal state is high. The dashed circle shows an image of the threshold value. If the statistical distance between the feature value of the time-series data image that is the target of anomaly detection (for example, a star-shaped hexagonal plot) and the group of feature values ​​of the time-series data image in a normal state is short (local density is relatively high) and the LOF value is small, the twin-screw kneading extruder 1 is estimated to be in a normal state. If the statistical distance between the feature value of the time-series data image that is the target of anomaly detection (for example, an X-mark plot) and the group of feature values ​​of the time-series data image in a normal state is long (local density is relatively low) and the LOF value is large, the twin-screw kneading extruder 1 is estimated to be in an abnormal state. By diagnosing an abnormality using the displacement of the rotational center axis of the screw shaft 11, it is possible to diagnose whether or not there is an abnormality such as wear of the screw shaft 11.

[0067] Although the above description has mainly focused on abnormality detection based on the displacement of the screw shaft 11, the present invention is not limited to this. For example, a frequency spectrum can be calculated based on sensor value data indicating the vibration of the reducer 13, a time-series data image representing the frequency spectrum can be generated, and the generated time-series data image can be used to diagnose abnormal vibration of the reducer 13. Furthermore, a time-series data image is generated that represents the sensor value data indicating the shaft torque applied to the screw shaft 11, and the presence or absence of an overload can be diagnosed using the generated time-series data image.

[0068] 13 is a flowchart showing the processing procedure for lifespan diagnosis. The diagnostic processing unit 51 converts a plurality of sensor value data into one or a plurality of time-series data images represented by images (step S91), and if there are a plurality of converted time-series data images, synthesizes the plurality of time-series data images into a single time-series data image (step S92). The sensor value data is, for example, data indicating the shaft torque or the displacement of the center of rotation of the screw shaft 11. Next, the diagnostic processing unit 51 inputs the time-series data image synthesized in step S92 into the learning model 54, thereby calculating the feature quantities of the time-series data image (step S93).

[0069] Furthermore, the diagnostic processing unit 51 reads out from the database 6 specific sensor value data, for example, sensor value data of the twin-screw kneading extruder 1 that is associated with the same or similar operation data as the operation data received in step S12 and has a normal lifespan (step S94), and similarly The time series data image is converted into a time series data image (step S95), synthesized (step S96), and the synthesized time series data image is input to the learning model 54 to calculate the feature quantities of the time series data image related to the twin-screw kneading extruder 1 having a normal lifespan (step S97).

[0070] Next, the diagnostic processing unit 51 calculates an outlier score for the feature amount of the time-series data image calculated in step S97 (step S98). Then, the processing unit 31 determines whether the calculated outlier score is equal to or greater than a predetermined threshold, thereby determining whether the life of the twin-screw kneading extruder 1 to be diagnosed is normal (step S99). When the outlier score is the above-mentioned LOF, the threshold is a value of 1 or greater. When the LOF is equal to or greater than the threshold, the diagnostic processing unit 51 determines that the life is short, and when the LOF is less than the threshold, it determines that the life is normal.

[0071] Although an example of diagnosing whether the lifespan is normal or not has been described here, it is possible to determine whether the twin-screw mixer / extruder 1 has a specific lifespan by comparing the features of the sensor value data obtained from the twin-screw mixer / extruder 1 having a specific lifespan with the features of the current sensor value data.

[0072] <Learning model generation method> A method for generating the learning model 54 used in the diagnostic process by the diagnostic device 5 will be described. 14 is a conceptual diagram showing an outline of a method for generating the learning model 54. A method for generating the learning model 54 for determining whether the operation of the twin-screw kneading extruder 1 is normal or not will be described. An example in which the diagnosis processing unit 51 of the diagnosis device 5 performs machine learning to generate the learning model 54 will be described below.

[0073] First, as a training dataset required to generate a one-class classification model, a plurality of time-series data images obtained by detection when the twin-screw kneading extruder 1 is operating normally and a plurality of arbitrary reference images A, B, and C unrelated to the time-series data images are prepared. The reference images A, B, and C are image data of multiple classes selected from a training dataset such as Image Net. In the example shown in FIG. 14, reference images A, B, and C labeled with labels 0, 1, and 2 are prepared.

[0074] Then, a learning model 54 having a feature extraction layer made up of multiple convolutional layers and pooling layers that extract image features is trained by machine learning on the time-series data images during normal operation and the reference images A, B, and C. Specifically, the diagnosis processing unit 51 trains the learning model 54 so as to output feature amounts that can distinguish the features of each image, that is, feature amounts that have a high local density of feature amounts of the time-series data images and are highly distinguishable from the feature amounts of the reference images A, B, and C, as shown in the right diagram of Fig. 14 . The method for generating a one-class classification model will be explained in detail below.

[0075] 15 is a conceptual diagram showing a learning model 54 in the learning phase. First, a first neural network (reference network) 54a and a second neural network (secondary network) 54a are prepared.

[0076] The first neural network 54a is a CNN having an input layer, a feature extraction layer, and a classification layer. The feature extraction layer has a repetitive structure of multiple convolutional layers and pooling layers. The classification layer has, for example, one or more fully connected layers. The first neural network 54a has been pre-trained using a training dataset such as ImageNet. The second neural network 54a has the same network configuration as the first neural network 54a, and the various parameters (weighting coefficients) that characterize the feature extraction layer and classification layer are also the same. Hereinafter, during the learning phase of the first and second neural networks 54a, 54a, the various parameters of the feature extraction layer and classification layer share the same values.

[0077] Then, the diagnostic processing unit 51 inputs the reference image A to the first neural network 54a and calculates a descriptive loss, which is a loss function. Meanwhile, the diagnostic processing unit 51 inputs a normal time-series data image to the second neural network 54a and calculates a compactness loss, which is a loss function. The descriptive loss is a loss function commonly used in classifier training, such as cross-entropy error. The compact loss is expressed by the following formulas (1), (2), and (3). The compact loss is a value corresponding to the variance of the output of the second neural network 54a within the batch, i.e., the training dataset.

[0078]

number

[0079] Then, the diagnostic processing unit 51 calculates a total loss based on the descriptive loss and the compact loss. The total loss is expressed, for example, by the following formula (4).

[0080]

number

[0081] The diagnostic processing unit 51 optimizes various parameters of the first and second neural networks 54a, 54a using backpropagation or the like so as to reduce the total loss expressed in (4) above, thereby training the first and second neural networks 54a, 54a. Note that, during machine learning, it is advisable to fix various parameters in the earlier stages and adjust parameters in multiple layers in the later stages.

[0082] The features output from the feature extraction layers of the first and second neural networks 54a, 54a trained in this manner have a high local density of features of the time-series data images obtained when the twin-screw kneading extruder 1 is operating normally, and a low local density with features of other arbitrary images, as shown in FIG.

[0083] FIG. 16 is a conceptual diagram showing the learning model 54 in the test phase. As shown in FIG. 16, the learning model 54 according to the first embodiment can be configured using the input layer and feature extraction layer of the second neural network 54a trained as described above. Therefore, the classification layer shown in FIG. 15 is not an essential component of the learning model 54 according to the first embodiment. However, it is also possible to use the second neural network 54a shown in FIG. 15 as the learning model 54 as it is, and use the features output from the feature extraction layer. Although two learning models 54 are illustrated in Figure 16, this is a conceptual illustration of the state in which a time series data image to be detected and a normal time series data image as a sample are input, and does not indicate the existence of two models.

[0084] In the test phase, the diagnostic processing unit 51 inputs time-series data images obtained under normal conditions as sample data to the learning model 54, thereby outputting feature quantities of the time-series data images under normal conditions. On the other hand, by inputting time-series data images of an abnormality detection target to the learning model 54, the diagnostic processing unit 51 outputs feature quantities of the time-series data images. Then, the diagnostic processing unit 51 calculates outlier scores of the feature quantities of the abnormality detection target with respect to the feature quantities of the sample data, and compares the calculated outlier score with a threshold value to determine whether the time-series data image of the abnormality detection target is abnormal, i.e., whether the twin-screw kneading extruder 1 is abnormal.

[0085] Note that Figure 16 shows an example in which sample data is input into the learning model 54 and the features of the sample data are calculated, but it is also possible to store the features of multiple sample data calculated in advance in the memory unit 32, and in the anomaly detection process, calculate the outlier score using the features of the sample data stored in the memory unit 32.

[0086] Furthermore, a learning model 54 for diagnosing whether or not there is an abnormality in the twin-screw kneading extruder 1 has been described, but by training the learning model 54 using sensor value data obtained from a twin-screw kneading extruder 1 having a normal lifespan, a learning model 54 for diagnosing whether or not the twin-screw kneading extruder 1 has a normal lifespan can be generated.

[0087] Although the above description has been given of an example of classifying normal and abnormal conditions, the anomaly detection device 3 may be configured to distinguish time-series image data at a typical abnormal time from time-series image data at other abnormal times using multiple learning models 54, which are one-class classification models. For example, similar to the first embodiment, the anomaly detection device 3 includes a first learning model 54 that has been trained on time-series image data at normal times. The anomaly detection device 3 also includes a second learning model 54 that has been trained on time-series image data at normal times and time-series image data at a first abnormal time. By using the first learning model 54 and the second learning model 54, the anomaly detection device 3 can distinguish time-series image data at normal times, time-series image data at a first abnormal time, and time-series image data at other abnormal times. Similarly, by providing the anomaly detection device 3 with three or more learning models 54, the device can be configured to be able to distinguish time-series image data at two or more abnormal times.

[0088] Furthermore, although the screw shaft abnormality diagnosis has been described in detail, a learning model 54 that diagnoses whether or not the operation is inefficient in steps S63 and S64, and whether or not the operation is in a state that could lead to quality abnormalities, can also be constructed and generated in the same way.

[0089] Furthermore, although a one-class classification model has been described as an example of the learning model 54, it is also possible to use general neural networks such as CNN, U-Net, RNN (Recurrent Neural Network), other SVMs (Support Vector Machines), Bayesian networks, regression trees, etc. For example, when sensor value data in normal times and sensor value data in various abnormal times are accumulated, the state of the twin-screw kneading extruder 1 is used as training data, and a learning data set is created in which the sensor value data or time-series data images are labeled, and a learning model 54 composed of CNN or the like is trained. The states of the twin-screw kneading extruder 1 include a normal state indicating that the twin-screw kneading extruder 1 is operating normally, an abnormal state in which the screw shaft 11 is worn, an abnormal state in which a resin raw material that is not suitable for the screw shaft 11 is being used, an abnormal overload state in which an overload is occurring, an abnormal vibration of the reducer 13, an abnormal state in which the lifespan is short, an abnormal state in which the dimensions of the molded product are abnormal, and the like. The learning model 54 has an input layer, an intermediate layer, and an output layer. The intermediate layer has multiple convolution layers and a pooling layer that extract image features. The output layer has multiple nodes corresponding to multiple states of the twin-screw kneading extruder 1 and outputs the confidence level of the state. When sensor value data or a time-series data image related to the sensor value data is input, the learning model 54 optimizes the weight coefficients of the intermediate layer so that the state of the twin-screw kneading extruder 1 output from the learning model 54 approaches the state indicated by the training data. The weight coefficients are, for example, weights (coupling coefficients) between neurons. The parameter optimization method is not particularly limited, and various parameters are optimized using, for example, the steepest descent method, the backpropagation method, etc. According to the learning model 54 generated in this manner, the state of the twin-screw kneading extruder 1 can be diagnosed by inputting time-series data images of sensor value data obtained from the twin-screw kneading extruder 1 to be diagnosed into the learning model 54. For example, the diagnosis processing unit 51 determines that the twin-screw kneading extruder 1 is in a state corresponding to the node that outputs the highest certainty.

[0090] Furthermore, the learning model 54 may be configured to output more optimal molding conditions or adjustment amounts for the molding conditions when a time-series data image is input. The diagnosis device 5 transmits the molding conditions output from the learning model 54 to the anomaly detection device 3, and the anomaly detection device 3 transmits the received molding conditions to the control device 14.

[0091] <Display process of diagnosis results and judgment results> An example of displaying the determination results and statistics in step S19 will be described. 17 is a schematic diagram showing an example of the display of abnormality determination results, etc. in the control device 14. The determination result, etc. display screen 141 displayed on the display unit has a graph display section 141a, a warning display section 141b, an AI diagnosis display section 141c, and a processed value display section 141d. The graph display section 141a displays, for example, a graph of the time change in statistical quantities of shaft torque, vibration, and displacement of the screw shaft 11. The warning display section 141b displays the received determination result indicating whether or not there is an abnormality in the twin-screw kneading extruder 1, and if there is an abnormality, displays the content of the warning. The AI ​​diagnosis display section 141c displays the diagnosis result by the diagnostic device 5, which will be described later. The processed value display section 141d displays the value of the received statistical quantity.

[0092] An example of displaying the diagnosis result in step S28 will be described. FIG. 18 is a schematic diagram showing an example of display of abnormality determination results, etc., on the terminal device 7. The determination result display screen 71 includes a graph display section 71a, a warning history display section 71b, an AI diagnosis history display section 71c, a processed value display section 71d, and a data selection section 71e. The data selection section 71e accepts selection of data to be displayed. For example, the data selection section 71e accepts selection of the operation date and time of the twin-screw kneading extruder 1, the type of statistics to display, and the diagnosis items. When multiple twin-screw kneading extruders 1 are connected to the first network, the data selection section 71e may be configured to accept selection of the twin-screw kneading extruder 1. The terminal device 7 transmits the operation date and time selected by the data selection section 71e, the identification information of the twin-screw kneading extruder 1, etc., to the abnormality detection device 3, and requests the history of abnormality determination results, the statistics of sensor value data, and the history of diagnosis results for the selected twin-screw kneading extruder 1 and operation date and time. The terminal device 7 receives this information transmitted from the diagnosis device 5 in response to the request and displays the received information. The graph display unit 71a displays a graph of the time change in the statistical quantity of the sensor value data. The warning history display unit 71b displays the history of received judgment results indicating the presence or absence of an abnormality in the twin-screw kneading extruder 1. The AI ​​diagnosis history display unit 71c displays the history of diagnosis results related to abnormalities or the lifespan of the twin-screw kneading extruder 1. The processed value display unit 71d displays the values ​​of the received statistical quantities.

[0093] <Operations and Effects of the Molding Machine System According to the First Embodiment> As described above, the molding machine system according to the present embodiment 1 is equipped with a control device 14 that controls the operation of the twin-screw kneading extruder 1, an abnormality detection device 3, and a diagnostic device 5, and these devices work together to detect abnormalities in the twin-screw kneading extruder 1 and diagnose detailed conditions such as abnormalities and the lifespan of the twin-screw kneading extruder 1 without placing an excessive load on the control device 14.

[0094] Furthermore, the abnormality detection device 3 can easily detect an abnormality in the twin-screw kneading extruder 1 using the sensor value data and operation data, and transmit the abnormality determination result to the control device 14. The control device 14 can receive the abnormality determination result of the twin-screw kneading extruder 1 simply by transmitting the operation data to the abnormality detection device 3, and can monitor the operation of the twin-screw kneading extruder 1 based on the abnormality determination result.

[0095] Furthermore, the abnormality detection device 3 can calculate statistics of the sensor value data and transmit them to the control device 14, and can display graphs of the statistics and values ​​of the statistics on the display unit.

[0096] Furthermore, the abnormality detection device 3 can detect abnormal vibrations in the reducer 13, overload on the screw shaft 11, and dimensional abnormalities in the molded product, and transmit such abnormalities to the control device .

[0097] Furthermore, the diagnostic device 5 can diagnose in detail the abnormalities and lifespan of the twin-screw kneading extruder 1 based on the current operation data and sensor value data of the twin-screw kneading extruder 1 and the past sensor value data accumulated in the database 6. The diagnostic device 5 can perform condition diagnosis of the twin-screw kneading extruder 1 that the abnormality detection device 3 cannot perform. The abnormality detection device 3 can acquire the diagnostic results by transmitting the current operation data and sensor value data of the twin-screw kneading extruder 1 to the diagnostic device 5 and requesting a diagnosis, and can then transmit the acquired diagnostic results to the control device 14. The control device 14 can receive the diagnostic results from the diagnostic device 5 and display the diagnostic results on the display unit simply by transmitting the operation data to the abnormality detection device 3.

[0098] Furthermore, the diagnostic device 5 can diagnose the state of the twin-screw kneading extruder 1, such as abnormality and end of life, using the learning model 54.

[0099] Furthermore, by configuring the learning model 54 as a one-class classification model, the learning model 54 can be trained by machine learning using time series data images obtained when the twin-screw kneading extruder 1 is operating normally, without using time series data images of the twin-screw kneading extruder 1 when an abnormality occurs.

[0100] Furthermore, by expressing the sensor value data output from the multiple sensors 2 as a single time-series data image and calculating the feature quantities of the time-series data image using the learning model 54, it is possible to easily obtain feature quantities that accurately represent the state of the twin-screw kneading extruder 1. By using the feature quantities obtained in this manner, it is possible to accurately diagnose the state of the twin-screw kneading extruder 1.

[0101] Furthermore, in the first embodiment, an example has been described in which the diagnostic device 5 performs machine learning to generate the learning model 54, but the learning model 54 may be generated by machine learning using another external computer or server.

[0102] Furthermore, although this embodiment 1 has mainly described abnormality detection and diagnosis of the twin-screw kneading extruder 1, it may also be configured to perform abnormality detection and diagnosis of molding machines such as injection molding machines, film molding machines, and other manufacturing equipment.

[0103] (Variation) The diagnostic device 5 is configured to execute a billing process for the diagnostic process using the learning model 54. For example, the diagnostic device 5 may calculate a diagnostic fee depending on the number of twin-screw kneading extruders 1, the number of sensors 2, the types of sensors 2, the amount of data to be diagnosed, the number of learning models 54 to be used, and the like, and may store the calculated diagnostic fee, the user ID of the abnormality diagnosis system, the identification ID of the twin-screw kneading extruder 1, the date and time of use, the amount of data, and the diagnostic items in the storage unit 52 in association with each other.

[0104] (Embodiment 2) The molding machine system according to the second embodiment differs from the second embodiment in that it receives user feedback on the diagnostic results and updates the learning model 54 or the diagnostic criteria. Since the other configurations of the molding machine system are the same as those of the molding machine system according to the first embodiment, the same reference numerals are used for the same parts and detailed descriptions are omitted.

[0105] 19 is a flowchart showing the procedure for feedback processing related to the diagnosis result according to embodiment 2. The abnormality detection device 3 receives, via the control device 14, the appropriateness of the diagnosis result or the correct state of the twin-screw kneading extruder 1 (step S211). The processing unit 31 of the abnormality detection device 3 transmits, to the diagnosis device 5, feedback information including the operation data and sensor value data that are the basis for obtaining the diagnosis result to be evaluated, and data indicating the correct state of the twin-screw kneading extruder 1 (step S212). The above-mentioned content of the feedback information is an example. For example, when the diagnosis ID, operation data, sensor value data, and diagnosis result are stored in the database 6, the abnormality detection device 3 may transmit feedback information including the diagnosis ID attached to the received diagnosis result and the correct state of the twin-screw kneading extruder 1 to the diagnosis device 5. The abnormality detection device 3 may also be configured to transmit feedback information including the diagnosis ID and whether the diagnosis result is appropriate to the diagnosis device 5.

[0106] The diagnostic device 5 receives the feedback information transmitted from the abnormality detection device 3 (step S213) and stores the received feedback information in the database 6 (step S214). That is, the diagnostic device 5 stores the operation data, the sensor values, and the correct state of the twin-screw kneading extruder 1. Then, the diagnostic processing device additionally trains the learning model 54 based on the feedback information or updates the diagnostic criteria (step S215).

[0107] For example, if the diagnosis result of an abnormality is incorrect and the twin-screw kneading extruder 1 is normal, the sensor value data with the incorrect diagnosis result can be added to the learning dataset as normal data, and the learning model 54 can be further trained, thereby improving the accuracy of feature extraction by the learning model 54.

[0108] Furthermore, the threshold value for diagnosing an abnormality in the twin-screw kneading extruder 1 may be changed based on the calculated score value. For example, if the diagnosis result of an abnormality is incorrect and the twin-screw kneading extruder 1 is normal, the diagnostic processing unit 51 increases the threshold value. If the diagnosis result of a normality is incorrect and the twin-screw kneading extruder 1 is abnormal, the diagnostic processing unit 51 decreases the threshold value.

[0109] It is preferable that the learning model 54 and the diagnostic criteria are changed by receiving and accumulating multiple pieces of feedback information.

[0110] According to the molding machine system of the second embodiment, the diagnostic accuracy of the diagnostic device 5 can be further improved by feedback from the user.

[0111] (Embodiment 3) The molding machine system according to the third embodiment differs from the third embodiment in that the operating data, sensor value data, and data indicating the state of the twin-screw kneading extruder 1 obtained on-site are stored in the database 6, and the learning model 54 is additionally trained. The other configurations of the molding machine system are the same as those according to the first embodiment. Since it is similar to the molding machine system, the same parts are given the same reference numerals and detailed explanations are omitted.

[0112] 20 is a flowchart showing the feedback processing procedure regarding the state of the twin-screw kneading extruder 1 according to embodiment 3. The abnormality detection device 3 receives the state of the twin-screw kneading extruder 1 via the control device 14 (step S311). That is, the abnormality detection device 3 receives the current state of the twin-screw kneading extruder 1 evaluated by the user. Then, the processing unit 31 of the abnormality detection device 3 transmits information including the operation data, sensor value data, and the state of the twin-screw kneading extruder 1 to the diagnosis device 5 (step S312). The abnormality detection device 3 may also receive and transmit to the diagnosis device 5 information such as the configuration of the screw shaft 11 of the twin-screw kneading extruder 1, the type of resin raw material, the quality evaluation of the molded product, and the estimated lifespan.

[0113] The diagnostic device 5 receives information including the operation data, sensor value data, the state of the twin-screw kneading extruder 1, and other information transmitted from the abnormality detection device 3 (step S313), and stores the received information in the database 6 (step S314). That is, the diagnostic processing unit 51 stores the operation data, sensor value data, the state of the twin-screw kneading extruder 1, and other information. Then, the diagnostic processing device additionally trains the learning model 54 using a learning dataset including the sensor value data based on user evaluations accumulated in the database 6 (step S315).

[0114] According to the anomaly detection device 3 of the third embodiment, the diagnosis accuracy of the diagnosis device 5 can be further improved by additionally learning the learning model 54 while reflecting the user's evaluation.

[0115] It is possible to create a data set for supervised learning based on the information transmitted from the anomaly detection device 3 in this manner. As described in the first embodiment, the diagnostic device 5 may be configured to use the data set for supervised learning to generate a learning model 54 that outputs information related to the state of the twin-screw kneading extruder 1, or to perform additional learning when sensor value data or time-series data images are input. [Explanation of symbols]

[0116] 1 Twin-screw kneading extruder 2 sensors 3. Anomaly detection device 4. Router 5 Diagnostic equipment 6 Database 7 Terminal Equipment 10 cylinders 11 Screw shaft 12 motors 13 Reducer 14 Control device 31 Processing section 33 Communications Department 34 Acquisition Department 51 Diagnostic processing unit 53 Communications Department 54 Learning Model 54a Neural Networks P1, P2 computer programs

Claims

1. An anomaly detection system including an anomaly detection device that detects an anomaly in a manufacturing device, a control device that controls the operation of the manufacturing apparatus and transmits operation data indicating the details of the operation control; a sensor that detects a physical quantity related to the operation of the manufacturing apparatus or the product and outputs time-series sensor value data indicating the detected physical quantity; Equipped with The abnormality detection device a communication unit that receives the operation data transmitted from the control device; an acquisition unit that acquires sensor value data output from the sensor; a processing unit that calculates statistics of the sensor value data acquired by the acquisition unit, and determines whether or not there is an abnormality in the manufacturing apparatus based on the calculated statistics and a threshold value according to the received operation data; Equipped with The communication unit transmits the determination result and the statistics to an external terminal device in response to a request from the terminal device. Anomaly detection system.

2. The control device Receives the determination result transmitted from the abnormality detection device and monitors the state of the manufacturing equipment. The anomaly detection system according to claim 1 .

3. The control device Receives the statistics transmitted from the anomaly detection device, and displays a graph or numerical values ​​based on the received statistics. The anomaly detection system according to claim 1 or 2.

4. the manufacturing apparatus is a molding machine having a reducer, the sensor detects vibration of the reducer; The processing unit determines whether or not abnormal vibration occurs in the reducer. The anomaly detection system according to any one of claims 1 to 3.

5. the manufacturing apparatus is a molding machine having a screw shaft, the sensor detects a shaft torque applied to the screw shaft, a temperature of the screw shaft, or a displacement of the screw shaft; The processing unit determines whether or not there is an overload on the molding machine and the screw shaft. The anomaly detection system according to any one of claims 1 to 4.

6. the manufacturing apparatus is a molding machine, the sensor detects dimensions of a molded product produced by the molding machine; The processing unit determines whether or not there is an abnormality in the dimensions of the molded product. The anomaly detection system according to any one of claims 1 to 5.

7. The anomaly detection system according to any one of claims 1 to 6; Molding machine and Equipped with The abnormality detection system is configured to detect an abnormality in the molding machine. Molding machine system.

8. An abnormality detection device that detects an abnormality in a manufacturing device, a communication unit that receives operation data indicating the details of the operation control transmitted from a control device that controls the operation of the manufacturing apparatus; an acquisition unit that acquires time-series sensor value data output from a sensor that detects a physical quantity related to the operation of the manufacturing apparatus or the product; a processing unit that calculates statistics of the sensor value data acquired by the acquisition unit, and determines whether or not there is an abnormality in the manufacturing apparatus based on the calculated statistics and a threshold value according to the acquired operating data; Equipped with The communication unit transmits the determination result and the statistics to an external terminal device in response to a request from the terminal device. Anomaly detection device.

9. An anomaly detection method in which a computer executes a process for detecting an anomaly in a manufacturing device, The computer receiving operation data indicating the details of the operation control transmitted from a control device that controls the operation of the manufacturing apparatus; acquiring time-series sensor value data output from a sensor that detects a physical quantity related to the operation of the manufacturing device or the product; Calculate statistics of the acquired sensor value data, determining whether or not an abnormality exists in the manufacturing equipment based on the calculated statistical amount and a threshold value according to the acquired operating data; The determination result is transmitted to an external terminal device in response to a request from the terminal device. Anomaly detection methods to perform processing.

10. A computer program for causing a computer to execute a process for detecting an abnormality in a manufacturing device, receiving operation data indicating the details of the operation control transmitted from a control device that controls the operation of the manufacturing apparatus; acquiring time-series sensor value data output from a sensor that detects a physical quantity related to the operation of the manufacturing device or the product; Calculate statistics of the acquired sensor value data, determining whether or not an abnormality exists in the manufacturing equipment based on the calculated statistical amount and a threshold value according to the acquired operating data; The determination result is transmitted to an external terminal device in response to a request from the terminal device. A computer program for causing the computer to execute a process.

Images