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

Abstract

An anomaly detection system comprises an anomaly detection device that detects an anomaly in a manufacturing device, a control device that performs operation control of the manufacturing device and transmits operation data, and a sensor and a diagnostic device that detect physical quantities according to the operation of the manufacturing device or the product and output sensor value data. The anomaly detection device determines whether there is an anomaly in the manufacturing device based on the operation data transmitted from the control device and the sensor value data output from the sensor, and transmits the determination result to the control device. Additionally, the anomaly detection device transmits the operation data and sensor value data to the diagnostic device, and the diagnostic device diagnoses the state of the manufacturing device based on the received data and past sensor value data accumulated in a database, and transmits the diagnosis result to the anomaly detection device. The anomaly detection device transmits the diagnosis result to the control device.

Description

Technology Field

[0001] The present invention relates to an abnormality detection system, a molding equipment 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 vibration of each moving part of an injection molding machine using an acceleration sensor and detecting the molding process state and the occurrence of abnormalities in the moving part. Prior art literature

[0003] Japanese Patent Publication No. JP 5-50480 The problem to be solved

[0004] However, Patent Document 1 does not disclose a specific technology in which a control device for controlling the operation of a manufacturing device, such as an injection molding machine, and a device for detecting abnormalities in the manufacturing device or diagnosing the state of the manufacturing device are configured as separate entities. If the control device is configured to detect and diagnose abnormalities in the manufacturing device, there is a possibility that adverse effects may occur on the operation of the manufacturing device, such as delays in the operation control processing.

[0005] The objective of the present disclosure is to provide an abnormality detection system, a molding machine system, an abnormality detection device, an abnormality detection method, and a computer program, wherein the system comprises a control device for controlling the operation of a manufacturing device, an abnormality detection device, and a diagnostic device, and wherein each device cooperates to detect abnormalities in the manufacturing device and diagnose the state of the manufacturing device without placing an excessive load on the control device. means of solving the problem

[0006] An anomaly detection system according to the present disclosure is an anomaly detection system comprising an anomaly detection device for detecting an anomaly in a manufacturing device and a diagnostic device for diagnosing the state of the manufacturing device, a control device for performing operation control of the manufacturing device and transmitting operation data representing the contents of the operation control to the anomaly detection device, a sensor for detecting a physical quantity according to the operation or product of the manufacturing device and outputting time-series sensor value data representing the detected physical quantity to the anomaly detection device, and a database for accumulating information including the operation data, the sensor value data, and the state of the manufacturing device, wherein the anomaly detection device comprises a first communication unit for receiving the operation data transmitted from the control device, an acquisition unit for acquiring the sensor value data output from the sensor, and a processing unit for calculating a statistical amount of the sensor value data acquired by the acquisition unit and determining whether there is an anomaly in the manufacturing device based on the calculated statistical amount and a threshold value according to the received operation data, wherein the first communication unit transmits the operation data and the sensor value data to the diagnostic device, and the diagnostic device receives the operation data and the sensor value data transmitted from the anomaly detection device The apparatus comprises a second communication unit that receives, and a diagnostic processing unit that diagnoses the state of the manufacturing device based on the driving data and sensor value data received by the second communication unit and information accumulated in the database, wherein the second communication unit transmits the diagnostic result by the diagnostic processing unit to the abnormality detection device, and the first communication unit of the abnormality detection device receives the diagnostic result by the diagnostic processing unit and transmits the judgment result by the processing unit, the statistical amount, and the diagnostic result by the diagnostic processing unit to the control device.

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

[0008] An abnormality detection device according to the present disclosure is an abnormality detection device for detecting an abnormality of a manufacturing device, and comprises a communication unit that receives operation data representing the contents of the operation control transmitted from a control device that performs operation control of the manufacturing device, an acquisition unit that acquires time-series sensor value data output from a sensor that detects physical quantities according to the operation of the manufacturing device or a product, and a processing unit that calculates a statistical amount of the sensor value data acquired by the acquisition unit and determines whether there is an abnormality of the manufacturing device based on the calculated statistical amount and a threshold value according to the acquired operation data. The communication unit transmits the operation data and the sensor value data to an external diagnostic device that diagnoses the state of the manufacturing device, receives a diagnostic result transmitted from the diagnostic device, and transmits a determination result by the processing unit, the statistical amount, and the received diagnostic result to the control device.

[0009] An 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, wherein the computer receives operation data representing the contents of the operation control transmitted from a control device that performs operation control of the manufacturing device, acquires time-series sensor value data output from a sensor that detects physical quantities according to the operation of the manufacturing device or a product, calculates a statistical amount of the acquired sensor value data, determines whether there is an anomaly in the manufacturing device based on the calculated statistical amount and a threshold value based on the acquired operation data, transmits the operation data and the sensor value data to an external diagnostic device that diagnoses the state of the manufacturing device, receives a diagnostic result transmitted from the diagnostic device, and executes a process of transmitting a determination result indicating the presence or absence of an anomaly, the statistical amount, and the received diagnostic result to the control device.

[0010] A computer program according to the present disclosure is a computer program for executing a process of detecting an abnormality in a manufacturing device on a computer, and executes a process of receiving operation data representing the contents of the operation control transmitted from a control device that performs operation control of the manufacturing device, acquiring time-series sensor value data output from a sensor that detects physical quantities according to the operation of the manufacturing device or a product, calculating a statistical amount of the acquired sensor value data, determining whether there is an abnormality in the manufacturing device based on the calculated statistical amount and a threshold value based on the acquired operation data, transmitting the operation data and the sensor value data to an external diagnostic device that diagnoses the state of the manufacturing device, receiving a diagnostic result transmitted from the diagnostic device, and transmitting a determination result indicating the presence or absence of an abnormality, the statistical amount, and the received diagnostic result to the control device. Effects of the invention

[0011] According to the present disclosure, a control device for controlling the operation of a manufacturing device, an abnormality detection device, and a diagnostic device are provided, and by each device cooperating, an abnormality of the manufacturing device can be detected and the condition of the manufacturing device can be diagnosed without placing an excessive load on the control device. Brief explanation of the drawing

[0012] [Fig. 1] This is a block diagram showing an example of the configuration of a molding machine system according to the present embodiment 1. [Fig. 2] This is a schematic diagram showing an example of the configuration of a twin-screw kneading extruder according to the present embodiment 1. [Fig. 3] This is a block diagram showing an example of the configuration of an abnormality detection device according to the present embodiment 1. [Fig. 4] This is a block diagram showing an example of the configuration of a diagnostic device according to Embodiment 1. [Fig. 5] This is a conceptual diagram showing an example of a record layout of a database according to Embodiment 1. [Fig. 6] This is a flowchart showing the processing sequence according to the detection and diagnosis of abnormalities in a twin-screw kneading extruder according to Embodiment 1. [Fig. 7] This is a flowchart showing the processing sequence according to the detection and diagnosis of abnormalities in a twin-screw mixing extruder according to Embodiment 1. [Fig. 8] This is a flowchart showing the processing order of abnormal judgments. [Fig. 9] This is a flowchart showing the processing sequence of abnormal diagnosis. [Fig. 10] This is a flowchart showing the processing sequence for diagnosing abnormalities in a screw shaft (11). [Fig. 11] This is a time-series data image showing the variation of two screw axes. [Fig. 12] This is a conceptual diagram showing the feature quantity and anomaly detection results of a time series data image. [Fig. 13] This is a flowchart showing the processing sequence of life diagnosis. [Fig. 14] This is a conceptual diagram showing an overview of the method for creating a learning model. [Fig. 15] This is a conceptual diagram showing the learning model in the learning phase. [Fig. 16] This is a conceptual diagram showing the learning model in the test phase. [Fig. 17] This is a schematic diagram showing an example of displaying abnormality judgment results, etc. in a control device. [Fig. 18] This is a schematic diagram showing an example of displaying abnormality judgment results, etc. in a terminal device. [Fig. 19] This is a flowchart showing the feedback processing sequence regarding the diagnostic results according to Embodiment 2. [Fig. 20] This is a flowchart showing the feedback processing sequence regarding the state of the twin-screw kneading extruder according to Embodiment 3. Specific details for implementing the invention

[0013] Specific examples of an abnormality detection system, a molding machine system, an abnormality detection device, an abnormality detection method, and a computer program according to embodiments of the present invention will be described below with reference to the drawings. Furthermore, the present invention is not limited to these examples but is defined by the claims, and it is intended that all modifications within the meaning and scope of equivalents to the claims are included. In addition, at least some of the embodiments and variations described below may be arbitrarily combined.

[0014] FIG. 1 is a block diagram showing an example of the configuration of a molding machine system according to the present embodiment 1. The molding machine system comprises a twin-screw mixing extruder (1), a plurality of sensors (2), an abnormality detection device (3), a router (4), a diagnostic device (5), a database (6), and a user terminal device (7). Additionally, the molding machine system is composed of a twin-screw mixing extruder (1) and an abnormality detection system, and the abnormality detection system is composed of a plurality of sensors (2), an abnormality detection device (3), a diagnostic device (5), and a database (6).

[0015] <2-screw mixing extruder (1)>

[0016] FIG. 2 is a schematic diagram showing an example of the configuration of a twin-screw mixing extruder (1) according to the present embodiment 1. The twin-screw mixing extruder (1) comprises a cylinder (10) having a hopper (10a) into which resin raw materials are fed, and two screw shafts (11). The two screw shafts (11) are arranged approximately parallel to each other in a meshed state and are rotatably inserted into the hole of the cylinder (10), and convey the resin raw materials fed into the hopper (10a) in the extrusion direction (right direction in FIG. 1 and FIG. 2) and melt and mix.

[0017] The screw shaft (11) is configured as a single screw shaft (11) by combining and integrating multiple types of screw pieces. For example, the screw shaft (11) is configured by arranging and combining a forward flight piece in the shape of a flight screw that transports the resin raw material in the forward direction, a reverse flight piece that transports the resin raw material in the reverse direction, a kneading piece that mixes the resin raw material, etc., in a sequence and position according to the characteristics of the resin raw material.

[0018] Additionally, the two-axis mixing extruder (1) is equipped with a motor (12) that outputs a driving force to rotate a screw shaft (11), a reduction gear (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 reduction gear (13). The screw shaft (11) rotates by the driving force of the motor (12) that is reduced and transmitted by the reduction gear (13).

[0019] <Sensor (2)>

[0020] The sensor (2) detects physical quantities according to the operation of the twin-screw kneading extruder (1) or the molded product, and outputs time-series sensor value data representing the detected physical quantities to the abnormality detection device (3). Physical quantities include temperature, position, speed, acceleration, current, voltage, pressure, time, image data, torque, force, distortion, 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, power meter, weight meter, etc.

[0021] More specifically, the sensor (2) for detecting physical quantities according to the operation of the two-screw mixing extruder (1) is a vibration sensor such as an acceleration sensor that detects vibration of the reduction gear (13), a torque sensor that detects shaft torque applied to the screw shaft (11), a thermometer that detects the temperature of the screw shaft (11), a displacement sensor that detects 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 captures the molded product or strand, a resin pressure sensor that detects resin pressure, etc.

[0022] The sensor (2) for detecting physical quantities according to the molded product is an optical measuring instrument that detects the dimensions, color, brightness, etc. of the molded product, an imaging device, a weighing scale that detects the weight of the molded product, etc.

[0023] <Control device (14)>

[0024] The control device (14) is a computer that performs operation control of the twin-screw mixing extruder (1) and is equipped with a transmitting and receiving unit not shown and a display unit that transmits and receives information with an abnormality detection device (3).

[0025] Specifically, the control device (14) transmits operating data indicating the operating status of the twin-screw mixing extruder (1) to the abnormal detection device (3). The operating data is, for example, the feeder supply amount (supply amount of resin raw material), the rotational speed of the screw shaft (11), the extrusion amount, the cylinder temperature, the resin pressure, the motor current, etc.

[0026] The control device (14) receives the judgment result described later indicating whether there is an abnormality in the twin-screw mixing extruder (1) transmitted from the abnormality detection device (3), various statistical quantities of sensor data, and diagnostic results regarding the abnormality and lifespan of the twin-screw mixing extruder (1), and displays the received judgment result, the value and graph of the statistical quantity, and the diagnostic results. In addition, the control device (14) monitors the abnormality of the twin-screw mixing extruder (1) using the judgment result, etc., outputs a warning as needed, and performs a process to stop the operation of the twin-screw mixing extruder (1).

[0027] <Abnormal detection device (3)>

[0028] FIG. 3 is a block diagram showing an example of the configuration of an abnormality detection device (3) according to the present embodiment 1. The abnormality detection device (3) is an edge computer and is equipped with a processing unit (31), a memory unit (32), a communication unit (33), and an acquisition unit (34), and the memory 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 compared to the diagnostic device (5) described later, and performs simple and easy abnormality detection processing of the twin-screw mixing extruder (1).

[0029] The processing unit (31) has a computational processing circuit such as a CPU (Central Processing Unit), a multi-core CPU, an ASIC (Application Specific Integrated Circuit), or an FPGA (Field-Programmable Gate Array), an internal memory device such as a ROM (Read Only Memory) or RAM (Random Access Memory), and I / O terminals. The processing unit (31) functions as an abnormality detection device (3) according to the present embodiment 1 by executing a computer program (program product) P1 stored by the memory unit (32) described later. In addition, each functional part of the abnormality detection device (3) may be implemented in software, or it may be implemented in part or in whole in hardware.

[0030] The memory unit (32) is a non-volatile memory such as a hard disk, EEPROM (Electrically Erasable Programmable ROM), or flash memory. The memory unit (32) stores a computer program P1 for implementing the abnormal detection method according to the present embodiment 1 on a computer.

[0031] The computer program P1 according to the present embodiment 1 may be in a form that is readable by a computer on a recording medium (30). The memory unit (32) stores the computer program P1 read out from the recording medium (30) by a reading device not shown. The recording medium (30) is a semiconductor memory such as a flash memory. In addition, the recording medium (30) may be an optical disc such as a CD (Compact Disc)-ROM, DVD (Digital Versatile Disc)-ROM, or BD (Blu-ray Disc). In addition, the recording medium (30) may be a magnetic disc such as a flexible disc or a hard disc, or a magnetic optical disc. Furthermore, the computer program P1 according to the present embodiment 1 may be downloaded from an external server not shown that is connected to a communication network not shown and stored in the memory unit (32).

[0032] 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 a control device (14) through a first communication network such as a LAN, and the processing unit (31) can transmit and receive various information between itself and the control device (14) through the communication unit (33).

[0033] A router (4) is connected to the first network, and the communication unit (33) is connected to the diagnostic device (5) on the warehouse cloud, which is the second communication network, through the router (4). The processing unit (31) can transmit and receive various information with the diagnostic device (5) through the communication unit (33) and the router (4).

[0034] The acquisition unit (34) is an input interface that receives a signal. A 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.

[0035] <Diagnostic device (5) and database (6)>

[0036] FIG. 4 is a block diagram showing an example of the configuration of a diagnostic device (5) according to embodiment 1. The diagnostic device (5) is a computer and is equipped with a diagnostic processing unit (51), a memory unit (52), and a communication unit (53). The memory unit (52) and the communication unit (53) are connected to the diagnostic processing unit (51).

[0037] The diagnostic processing unit (51) has computational processing circuits such as a CPU, multi-core CPU, GPU (Graphics Processing Unit), GPGPU (General-purpose computing on graphics processing units), TPU (Tensor Processing Unit), ASIC, FPGA, NPU (Neural Processing Unit), internal memory devices such as ROM and RAM, and I / O terminals. The diagnostic processing unit (51) functions as a diagnostic device (5) according to the present embodiment 1 by executing a computer program (program product) P2 stored by the memory unit (52) described later. In addition, each functional unit of the diagnostic device (5) may be realized in software, and may be partially or entirely realized in hardware.

[0038] The memory unit (52) is a non-volatile memory such as a hard disk, EEPROM, or flash memory. The memory unit (52) stores a computer program P2 for performing a process to diagnose the condition of the twin-screw mixing extruder (1), such as abnormalities or lifespan of the twin-screw mixing extruder (1), and one or more learning models (54).

[0039] The computer program P2 and the learning model (54) may be in a form that is computer-readable on the recording medium (50). The memory 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 flash memory. Also, the recording medium (50) may be an optical disc such as a CD (Compact Disc)-ROM, DVD (Digital Versatile Disc)-ROM, or BD (Blu-ray Disc). Additionally, the recording medium (50) may be a magnetic disc such as a flexible disc or hard disc, or a magnetic optical disc. Furthermore, the computer program P2 and the learning model (54) may be downloaded from an external server not shown that is connected to a communication network not shown and stored in the memory unit (52).

[0040] 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 abnormal detection device (3), the database (6), and the terminal device (7) through a second communication network, and the processing unit (31) can transmit and receive various information between the abnormal detection device (3), the database (6), and the terminal device (7) through the communication unit (53).

[0041] FIG. 5 is a conceptual diagram showing an example of a record layout of a database (6) according to embodiment 1. The database (6) is equipped with a hard disk and a DBMS (Database Management System) and stores various information regarding the operation of a twin-screw mixing extruder (1). For example, the 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 “lifetime” column.

[0042] The "Operation Date and Time" column stores information indicating the year, month, date, etc., from which various data stored as records can be obtained.

[0043] The “operation data” column stores information indicating the operating status of the twin-screw mixing extruder (1), such as the feeder supply amount (supply amount of resin raw material), the rotational speed of the screw shaft (11), the extrusion amount, the cylinder temperature, the resin pressure, the motor current, etc.

[0044] The “screw configuration” column stores information regarding the types, arrangement, or combination of various screw pieces that constitute the screw shaft (11).

[0045] The “Resin Type” column stores information indicating the type of resin raw material supplied to the twin-screw mixing extruder (1).

[0046] The "Sensor" column stores sensor data. In the "Sensor" column, numerical data may be stored as is, or, as described below, sensor data, which is time-series data, may be stored as a time-series image representing the sensor data.

[0047] The “Status” column stores information indicating the state of the twin-screw compounding 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 a normal state, information indicating an abnormal state where wear is occurring on the screw shaft (11), information indicating an abnormal state where a resin raw material unsuitable for the screw shaft (11) is being used, information indicating an overload abnormality, etc. are stored.

[0048] The “Quality Evaluation” column stores information indicating the quality evaluation of a molded product manufactured by a twin-screw kneading extruder (1) when the above-mentioned operating data is used to perform the operating control indicated by the above-mentioned operating data and the above-mentioned sensor value data is obtained. For example, it stores information indicating whether the molded product is normal or not, dimensional abnormalities, etc.

[0049] The “life” column stores information indicating the life of the twin-screw mixing extruder (1) used to obtain the above information. The life may be the total product life, or the remaining life from the time when the above data is obtained.

[0050] The various data stored in the database (6) can be obtained, for example, by operating the twin-screw mixing extruder (1) experimentally, checking the operating state of the twin-screw mixing extruder (1), and evaluating the molded product. Alternatively, the operating state of the twin-screw mixing extruder (1) may be simulated using a simulator to prepare the various data. In this embodiment 1, the database (6) stores the data obtained by operating the twin-screw mixing extruder (1) experimentally or experimentally, or the data obtained through simulation.

[0051] Examples above

[0052] The abnormalities that are detected or diagnosed by the abnormality detection device (3) and the diagnostic device (5) include, for example, abnormal vibration of the reduction gear (13), overload, dimensional abnormality of the molded product, scratches, cracks, wear, corrosion of the screw shaft (11), performance deterioration due to wear of the reduction gear (13) and its parts or abnormality of the oil, inefficient operating conditions (increase in energy consumption, etc.), operating conditions that cause quality abnormalities (forced start, etc.), abnormality of the strand (defects such as dimensions, foreign matter, color, winding (twisting), etc.), and abnormality of resin viscosity (quality).

[0053] In the present embodiment 1, the abnormal detection device (3) detects abnormal vibration of the reduction gear (13), overload, dimensional abnormality of the molded product, strand abnormality, resin viscosity abnormality, etc. based on sensor value data. Specifically, the abnormal detection device (3) calculates statistical quantities such as the average, variance, and root mean square of the sensor values, and detects abnormal vibration of the reduction gear (13), overload, dimensional abnormality of the molded product, strand abnormality, resin viscosity abnormality, etc. through a process with a small load based on the calculated statistical quantities.

[0054] Meanwhile, the diagnostic device (5) uses operating data and sensor data obtained from the control device (14) through the abnormality detection device (3), and past sensor data accumulated in the database (6), to diagnose the presence or absence of abnormalities such as damage, cracks, wear, corrosion of the screw shaft (11), performance deterioration of the reducer (13), oil abnormalities, inefficient operating conditions, and operating conditions causing quality abnormalities. In addition, the diagnostic device (5) also performs processes such as identifying fault locations of the reducer (13), such as bearings, gears, etc.

[0055] In addition, the division of the abnormal detection processing performed by the abnormal detection device (3) and the diagnostic processing performed by the diagnostic device (5) is an example, and if the abnormal detection device (3) can handle it, it may be configured to be executed on the side of the abnormal detection device (3).

[0056] <Overall flow of anomaly detection processing>

[0057] FIGS. 6 and 7 are flowcharts showing the processing sequence for abnormal detection and diagnosis of a twin-screw mixing extruder (1) according to embodiment 1. The control device (14) transmits operation data indicating the operation control status of the twin-screw mixing extruder (1) to the abnormal detection device (3) (step S11).

[0058] The processing unit (31) of the abnormal detection device (3) receives the operation data transmitted from the control device (14) at the communication unit (33) (step S12).

[0059] The processing unit (31) acquires time series sensor value data output from the sensor (2) (step S13). Then, the processing unit (31) calculates statistical quantities such as the mean, variance, and root mean square of the sensor values ​​based on the sensor value data (step S14), and performs an abnormality determination process of the twin-screw mixing extruder (1) based on the calculated statistical quantities (step S15).

[0060] Next, the processing unit (31) transmits the determination result indicating whether there is an abnormality in the twin-screw mixing extruder (1) and the statistical amount calculated in step S14 from the communication unit (33) to the control device (14) (step S16).

[0061] The control device (14) receives a judgment result and statistics transmitted from the abnormality detection device (3) (step S17), and monitors the operation of the twin-screw mixing extruder (1) based on the received judgment result or statistics (step S18). For example, if the judgment result indicates a predetermined abnormality, the control device (14) stops the operation control of the twin-screw mixing extruder (1).

[0062] Next, the control device (14) displays the received judgment result and statistics, etc. on the display unit (step S19).

[0063] Next, the processing of the abnormality detection device (3) that transmitted the judgment result and statistics is described. The processing unit (31) of the abnormality detection device (3), which has finished processing in step S16, transmits the operation data received in step S12 and the sensor value data acquired in step S13 from the communication unit (33) to the diagnostic device (5) (step S20).

[0064] The diagnostic device (5) receives operating data and sensor data transmitted from the abnormal detection device (3) (step S21), and the diagnostic processing unit (51) performs an abnormality diagnosis process of the twin-screw mixing extruder (1) using the received operating data and sensor data, and past sensor data accumulated in the database (6) (step S22). In addition, the diagnostic processing unit (51) performs a lifespan diagnosis process of the twin-screw mixing extruder (1) (step S23). The diagnostic processing unit (51) evaluates material fatigue from the waveform of the shaft torque, for example, and calculates the remaining lifespan.

[0065] The diagnostic processing unit (51), having finished processing steps S22 and S23, transmits the diagnostic result from the communication unit (53) to the abnormal detection device (3) (step S24).

[0066] The processing unit (31) of the abnormality detection device (3) receives a diagnosis result transmitted from the diagnosis device (5) (step S25), and transmits the received diagnosis result from the communication unit (33) to the control device (14) (step S26).

[0067] The control device (14) receives a diagnosis result transmitted from the abnormality detection device (3) (step S27) and displays the received diagnosis result on the display unit (step S28).

[0068] Additionally, the abnormality detection device (3) can transmit the abnormality determination result by the processing unit (31), the statistical amount of the calculated sensor value data, the diagnosis result by the diagnosis processing unit (51), etc., to the terminal device (7) in response to a request from the terminal device (7). The terminal device (7) displays the received abnormality determination result, statistical amount, and diagnosis result.

[0069] <Processing of abnormal detection by abnormal detection device (3)>

[0070] FIG. 8 is a flowchart showing the processing sequence of abnormality determination. Since the abnormality detection device (3) cannot store a large amount of sensor value data, it performs a process to determine whether there is an abnormality in the twin-screw mixing extruder (1) using the operation data and sensor value data of the day.

[0071] The processing unit (31) specifies a threshold value for a determination criterion to determine whether there is an abnormality in the two-axis mixing extruder (1) based on the operation data received in step S12 (step S51). For example, since the shaft torque and the rotational speed of the screw shaft (11) are inversely proportional, the threshold value for detecting an overload of torque depends on the rotational speed of the screw shaft (11), which is one of the operation data. In this way, the processing unit (31) can specify a threshold value for detecting an abnormality based on the operation data.

[0072] Specifically, the abnormality detection device (3) stores a table in the memory unit (32) that corresponds the operating data, the type of abnormality, and the threshold of the judgment criteria. The processing unit (31) can determine the threshold, which is the judgment criterion for various abnormalities according to the operating data, by reading out the threshold by referring to the table using the operating data.

[0073] Next, the processing unit (31) determines whether there is abnormal vibration of the reduction gear (13) based on the statistical amount of sensor value data indicating vibration of the reduction gear (13) and the threshold value specified in step S51 (step S52). Additionally, the processing unit (31) may be configured to convert the sensor value data indicating vibration of the reduction gear (13) into a time frequency component and determine whether there is abnormality of the reduction gear (13) based on the data of the time frequency component.

[0074] Additionally, the processing unit (31) determines whether there is an overload based on the statistical amount of sensor value data representing shaft torque, the temperature of the screw shaft (11), and the displacement of the center of rotation of the screw shaft (11), and a specific threshold value in step S51 (step S53).

[0075] Additionally, the processing unit (31) determines whether there is an abnormality in the dimensions of the molded product based on a statistical amount representing the dimensions of the molded product and a specific threshold value in step S51 (step S54).

[0076] The processing unit (31) determines whether there is an abnormality in the strand based on a statistical amount representing the difference between the image obtained by capturing the strand and the normal image, and a specific threshold value in step S51 (step S55). Additionally, the processing unit (31) can measure the diameter of the strand by the pixel size (number of pixels) of the image of the strand included in the captured image. Also, the processing unit (31) can determine whether there is a foreign substance by determining the color of the image portion included in the captured image.

[0077] The processing unit (31) determines whether there is an abnormality in resin viscosity based on the time average of the resin pressure detected by the sensor (2), for example, the resin pressure sensor, and a specific threshold value in step S51 (step S56), and terminates the abnormality determination process.

[0078] <Processing of abnormal diagnosis by the diagnostic device (5)>

[0079] FIG. 9 is a flowchart showing the processing sequence of an abnormality diagnosis. The diagnosis processing unit (51) performs an abnormality diagnosis of the screw shaft (11) by inputting sensor data representing the displacement of the rotation center of the screw shaft (11), or time-frequency data obtainable by time-frequency analysis of the sensor data, into the learning model (54) (step S61).

[0080] The learning model (54) used in step S61 is a convolutional neural network model that, for example, when a time-series data image described later is input, which represents sensor data as a time-series data image, extracts features of said time-series data image and outputs the extracted feature quantity. The learning model (54) is a model having a feature extraction layer of a CNN (Convolutional Neural Network), for example, a 1-class classification model. The diagnostic processing unit (51) can correspond to operating data identical or similar to the operating data received in step S12, and can diagnose whether there is an abnormality in the screw shaft (11) by comparing the feature quantity obtained from sensor data when the two-axis mixing extruder (1) is operating normally with the feature quantity of the diagnostic target. Details of the method of creating and the configuration of the learning model (54) will be described later.

[0081] Additionally, the diagnostic processing unit (51) diagnoses whether there is performance deterioration due to wear of the reduction gear (13) and each part thereof or an abnormality of the oil by inputting vibration sensor data obtained from a sensor (2) installed in the reduction gear (13), for example, a vibration sensor, or time frequency data obtained by time frequency analysis of the sensor value data into a learning model (54) (step S62).

[0082] The learning model (54) used in step S62 is, for example, a support vector machine. The support vector machine is a classifier that learns the time frequency components obtained by time frequency analysis of vibration sensor data according to a normal reducer (13) and the time frequency components obtained by time frequency analysis of vibration sensor data according to an abnormal reducer (13), and classifies them into a normal class according to a normal reducer (13) and an abnormal class according to an abnormal reducer (13). In addition, the frequency components to be analyzed may be specific frequency components, such as, for example, major frequency components obtained from the reducer (13) and fault frequency components that appear during a failure. The diagnostic processing unit (51) specifies, for example, major frequency components and fault frequency components obtained by time frequency analysis of vibration sensor data. And, the diagnostic processing unit (51) diagnoses whether the reduction gear (13) is normal or not by inputting the time frequency components of the vibration sensor data, for example, the main frequency component and the fault frequency component, into the support vector machine.

[0083] The learning model (54) may be a classifier using a convolutional neural network model. The learning model (54) may be trained to output data indicating an abnormal location or abnormal level of the reduction gear (13) when, for example, vibration sensor data or time-frequency data obtained by time-frequency analysis of the sensor data is input. An abnormal location is, for example, a bearing or gear of the reduction gear (13).

[0084] Additionally, the diagnostic processing unit (51) diagnoses whether there is an inefficient operating state by inputting sensor data, such as shaft torque, displacement of the center of rotation of the screw shaft (11), and vibration, which can be obtained from the sensor (2), or time-frequency data obtained by time-frequency analysis of the sensor data, into the learning model (54) (step S63).

[0085] The learning model (54) used in step S63 is a convolutional neural network model that, for example, when a time series data image described later is input, which represents sensor data as a time series data image, extracts features of said time series data image and outputs the extracted features. The learning model (54) is a model having a feature extraction layer of a CNN, for example, a 1-class classification model.

[0086] The diagnostic processing unit (51) can correspond to the same or similar operating data as the operating data received in step S12, and can diagnose whether the 2-axis mixing extruder (1) is in an inefficient operating state by comparing the characteristic amount obtained from the sensor value data when the 2-axis mixing extruder (1) is in an efficient operating state with the characteristic amount of the diagnostic target.

[0087] Additionally, the diagnostic processing unit (51) diagnoses whether there is an operating condition that causes quality abnormalities by inputting sensor data, which indicates the shaft torque, displacement of the center of rotation of the screw shaft (11), vibration, etc., which can be obtained from the sensor (2), or time-frequency data obtained by time-frequency analysis of the sensor data, into the learning model (54) (step S64).

[0088] The learning model (54) used in step S64 is a convolutional neural network model that, for example, when a time series data image described later is input, which represents sensor data as a time series data image, extracts features of said time series data image and outputs the extracted features. The learning model (54) is a model having a feature extraction layer of a CNN, for example, a 1-class classification model.

[0089] The diagnostic processing unit (51) can correspond to operating data identical or similar to the operating data received in step S12, and can diagnose whether there is an operating state causing quality abnormalities by comparing the characteristic quantity obtained from sensor value data when the twin-screw mixing extruder (1) is in a normal operating state without an operating state causing quality abnormalities with the characteristic quantity of the diagnostic target.

[0090] Additionally, the diagnostic processing unit (51) may be configured to monitor the operating rate by monitoring the power supply data obtained from the control unit (14) through the abnormal detection device (3) and calculating the operating rate for a certain period of time. The diagnostic processing unit (51) may also be configured to diagnose whether there is an abnormality in the twin-screw mixing extruder (1) based on the operating rate.

[0091] Below, screw shaft abnormality diagnosis is described in detail as an example of a Class 1 classification model.

[0092] FIG. 10 is a flowchart showing the processing sequence for diagnosing an abnormality of a screw shaft (11). The diagnostic processing unit (51) converts a plurality of sensor data into one or more time-series data images (step S71), and if there are multiple converted time-series data images, it synthesizes the plurality of time-series data images into one time-series data image (step S72). For example, if the sensor data is data representing displacement in two directions of each of the two screw shafts (11), the diagnostic processing unit (51) synthesizes a first time-series data image representing displacement in two directions of the rotation center axis of the first screw shaft (11) and a second time-series data image representing displacement in two directions of the rotation center axis of the second screw shaft (11).

[0093] FIG. 11 is a time series data image showing the displacement of 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 roughly square image, and the displacement amounts in the X-axis and Y-axis directions, which are orthogonal to the rotation center of the screw shaft (11), are plotted as coordinate values ​​of the X-axis and Y-axis. The first time series data image plots the displacement of three rotations. The second time series data image can also be created in this way.

[0094] The synthesized time series data image is an approximately square image having twice the size (four times the area) of the first time series data image. The time series data image contains the information of the original image as is, 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 of FIG. 11, and a blank image arranged on the lower side to make the time series data image approximately square. Furthermore, the arrangement method of the first and second time series data images and the blank image is not particularly limited, and as long as the first and second time series data images are included as is, the arrangement method is not particularly limited.

[0095] In addition, multiple other sensor data can be converted into time-series data images and combined in the same way. In the example shown in FIG. 11, a total of 8 sensor data can be displayed as images, but when combining 9 time-series data images of a 3x3 grid, 18 sensor data can be displayed as a single time-series data image.

[0096] Next, the diagnostic processing unit (51) calculates the feature quantity of the time series data image by inputting the time series data image synthesized in step S72 into the learning model (54) (step S73).

[0097] Additionally, the diagnostic processing unit (51) can correspond to the same or similar operating data received in step S12 from the database (6), and reads out sensor value data when the twin-axis mixing extruder (1) is operating normally (step S74), converts it into a time-series data image by the same method (step S75), synthesizes it (step S76), and calculates the feature quantity of the time-series data image during normal operation by inputting the synthesized time-series data image into the learning model (54) (step S77).

[0098] 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 outliers of the feature quantity of the time-series data image calculated in step S77 against the feature quantity of the time-series data image obtained during normal operation of the twin-screw mixing extruder (1) (hereinafter referred to as the sample feature quantity). The outlier score is, for example, a Local Outlier Factor (LOF). LOF is the average of the local density ld (P) of the feature quantity to be detected as an anomaly and the local density of the neighboring group of the feature quantity (k sample feature quantities closest to the feature quantity).<ld(Q)> The difference between the two<ld(Q)> It can be expressed as / ld(P). As LOF is greater than 1, the degree of outlier value increases. Also, P represents the feature quantity targeted for outlier detection, and Q represents the sample feature quantity.

[0099] Then, the processing unit (31) determines whether there is an abnormality in the twin-screw mixing extruder (1) by determining whether the calculated abnormality score is greater than or equal to a predetermined threshold (step S79). When the abnormality score is the LOF above, the threshold is a value greater than or equal to 1. When the LOF is greater than or equal to the threshold, the diagnostic processing unit (51) determines that there is an abnormality, and when the LOF is less than the threshold, it determines that it is normal.

[0100] In addition, the abnormality determination using the LOF described above is merely an example, and it is acceptable to determine whether the twin-screw mixing extruder (1) is abnormal or not by the k-proximate method or SVM. Also, it is acceptable to configure the system to determine whether the twin-screw mixing extruder (1) is abnormal or not by determining whether the characteristic quantity of the abnormality detection target deviates from the normal sample characteristic quantity by the Hotelling method.

[0101] FIG. 12 is a conceptual diagram showing the feature quantity and anomaly detection results of a time-series data image. FIG. 12 shows the high-dimensional feature quantity of a time-series data image reduced in dimensionality to a two-dimensional feature quantity and plotted on a two-dimensional plane. The horizontal and vertical axes of the graph shown in FIG. 12 represent the first feature quantity and the second feature quantity of the dimensionally reduced time-series data image. In addition, the dimensionality reduction of the feature quantity can be performed by a dimensionality reduction algorithm such as t-SNE, for example.

[0102] As shown in FIG. 12, the learning model (54) is trained so that the local density of normal feature quantities output from the learning model (54) increases.

[0103] The dashed circle represents the image of the threshold. If the statistical distance between the feature quantity of the time series data image subject to abnormal detection (e.g., a formed hexagon plot) and the group of feature quantities of the time series data image in normal conditions is short (local density is relatively high) and the LOF value is small, it is estimated that the twin-screw mixing extruder (1) is in a normal state. If the statistical distance between the feature quantity of the time series data image subject to abnormal detection (e.g., an X-mark plot) and the group of feature quantities of the time series data image in normal conditions is long (local density is relatively low) and the LOF value is large, it is estimated that the twin-screw mixing extruder (1) is in an abnormal state.

[0104] By diagnosing abnormalities using the displacement of the rotational center axis of the screw shaft (11), it is possible to diagnose whether there is an abnormality such as wear on the screw shaft (11).

[0105] Additionally, although abnormal detection based on the displacement of the screw shaft (11) has been mainly described, it is not limited to this. For example, based on sensor data indicating vibration of the reduction gear (13), a frequency spectrum can be calculated, a time-series data image representing the frequency spectrum can be generated, and the vibration abnormality of the reduction gear (13) can be diagnosed using the generated time-series data image.

[0106] In addition, a time series data image is generated that represents sensor value data indicating shaft torque applied to the screw shaft (11), and the presence or absence of overload can be diagnosed using the generated time series data image.

[0107] FIG. 13 is a flowchart showing the processing sequence of life diagnosis. The diagnosis processing unit (51) converts multiple sensor data into one or more time-series data images that represent multiple sensor data (step S91), and if there are multiple converted time-series data images, synthesizes the multiple time-series data images into one time-series data image (step S92). The sensor data is, for example, data representing shaft torque or the displacement of the center of rotation of the screw shaft (11). Next, the diagnosis processing unit (51) calculates the feature quantity of the time-series data image by inputting the time-series data image synthesized in step S92 into a learning model (54) (step S93).

[0108] Additionally, the diagnostic processing unit (51) reads out sensor data of a twin-screw mixing extruder (1) that has a normal lifespan and can correspond to specific sensor data, for example, operating data received in step S12, from the database (6) (step S94), converts it into a time-series data image by making it identical (step S95), synthesizes it (step S96), and calculates the feature quantity of the time-series data image according to the twin-screw mixing extruder (1) that has a normal lifespan by inputting the synthesized time-series data image into the learning model (54) (step S97).

[0109] Next, the diagnostic processing unit (51) calculates an outlier score of the feature quantity of the time series data image calculated in step S97 (step S98). Then, the processing unit (31) determines whether the lifespan of the twin-screw mixing extruder (1) to be diagnosed has a normal lifespan by determining whether the calculated outlier score is greater than or equal to a predetermined threshold (step S99). When the outlier score is the LOF above, the threshold is a value greater than or equal to 1. If the LOF is greater than or equal to the threshold, the diagnostic processing unit (51) determines that the lifespan is short, and if the LOF is less than the threshold, it determines that the lifespan is normal.

[0110] In addition, 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 kneading extruder (1) has a specific lifespan or not by comparing the characteristic amount of sensor data obtained from the twin-screw kneading extruder (1) having a specific lifespan with the characteristic amount of current sensor data.

[0111] <How to Create a Learning Model>

[0112] A method for generating a learning model (54) used for diagnostic processing by a diagnostic device (5) is explained.

[0113] FIG. 14 is a conceptual diagram showing an overview of the method for generating a learning model (54). A method for generating a learning model (54) to determine whether the operation of the twin-screw mixing extruder (1) is normal or not is described. Below, an example of machine learning the learning model (54) by the diagnostic processing unit (51) of the diagnostic device (5) is described.

[0114] First, as a training dataset required to create a 1-class classification model, a plurality of time-series data images that can be detected and obtained when the twin-screw kneading extruder (1) is operating normally, and arbitrary multiple reference images A, B, and C unrelated to the time-series data images are prepared. Reference images A, B, and C are image data of multiple classes selected from a training dataset such as Image Net, for example. In the example shown in FIG. 14, reference images A, B, and C with labels 0, 1, and 2 attached are prepared.

[0115] Then, machine learning is performed on a learning model (54) having a feature extraction layer composed of a plurality of convolutional layers and pooling layers that extract features of images, with the time-series data images during normal operation and reference images A, B, and C. Specifically, the diagnostic processing unit (51) trains the learning model (54) to output feature quantities capable of distinguishing the features of each image, that is, feature quantities in which the local density of the feature quantities of the time-series data images is high as shown in the right diagram of FIG. 14, and furthermore, the distinguishability with the feature quantities of reference images A, B, and C is high.

[0116] Below, the details of the method for creating a 1-class classification model are explained.

[0117] FIG. 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.

[0118] 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 repeating structure of multiple convolutional layers and pooling layers. The classification layer has, for example, one or multiple layers of fully connected layers. The first neural network (54a) is pre-trained using a training dataset such as Image Net.

[0119] The second neural network (54a) has the same network configuration as the first neural network (54a), and the various parameters (weight coefficients) characterizing the feature extraction layer and the classification layer are also the same. Hereinafter, in the learning phase of the first and second neural networks (54a, 54a), the various parameters of the feature extraction layer and the classification layer share the same values.

[0120] Then, the diagnostic processing unit (51) inputs reference image A into the first neural network (54a) and calculates a loss function, which is a descriptiveness loss. Meanwhile, the diagnostic processing unit (51) inputs a normal time-series data image into the second neural network (54a) and calculates a loss function, which is a compactness loss. The descriptiveness loss is a loss function generally used for the training of a classifier, for example, a cross-entropy error. The compactness loss is expressed by the following equations (1), (2), and (3). The compactness loss is a value corresponding to the variance of the output of the second neural network (54a) within a batch, that is, in the training data set.

[0121]

[0122] And, the diagnostic processing unit (51) calculates the total loss based on the technical loss and the compact loss. The total loss is expressed, for example, by Equation (4) as follows.

[0123]

[0124] The diagnostic processing unit (51) trains the first and second neural networks (54a, 54a) by optimizing various parameters of the first and second neural networks (54a, 54a) using backpropagation or the like so that the total loss indicated in (4) is reduced. Additionally, when machine learning, it is acceptable to fix the various parameters of the front layer and adjust the parameters of the multiple layers of the rear layer.

[0125] In this way, the feature output from the feature extraction layer of the first and second neural networks (54a, 54a) learned in this manner, as shown in FIG. 14, increases the local density of the feature of the time-series data image that can be obtained when the twin-screw mixing extruder (1) is in normal operation, and lowers the local density of the feature of another arbitrary image.

[0126] FIG. 16 is a conceptual diagram showing a learning model (54) in the test phase. As shown in FIG. 16, the learning model (54) according to the present embodiment 1 can be constructed using the input layer and the feature extraction layer of the second neural network (54a) learned as described above. Therefore, the classification layer shown in FIG. 15 is not an essential component of the learning model (54) according to the present embodiment 1. However, the second neural network (54a) shown in FIG. 15 may be used as the learning model (54) as is, and the feature amount output from the feature extraction layer may be used.

[0127] Additionally, FIG. 16 illustrates two learning models (54), but this conceptually represents the state in which a time series data image to be detected and a time series data image of normal time that is a sample are input, and does not indicate that two models exist.

[0128] In the test phase, the diagnostic processing unit (51) outputs the feature quantity of the time series data image in normal conditions by inputting the time series data image obtained in normal conditions into the learning model (54) as sample data. Meanwhile, the feature quantity of the time series data image of the abnormality detection target is output by inputting the time series data image of the abnormality detection target into the learning model (54). Then, the diagnostic processing unit (51) calculates the outlier score of the feature quantity of the abnormality detection target with respect to the feature quantity of the sample data, and determines whether the time series data image of the abnormality detection target is abnormal or not, that is, whether the two-screw mixing extruder (1) is abnormal or not, by comparing the calculated outlier score with a threshold value.

[0129] In addition, FIG. 16 shows an example in which sample data is input into a learning model (54) to calculate the feature quantity of the sample data, but the feature quantity of a plurality of sample data that has been calculated in advance is stored in a memory unit (32), and in the abnormality detection process, the abnormality score may be calculated using the feature quantity of the sample data stored in the memory unit (32).

[0130] Additionally, although a learning model (54) for diagnosing whether there is an abnormality in the twin-screw mixing extruder (1) has been described, a learning model (54) for diagnosing whether the twin-screw mixing extruder (1) has a normal lifespan can be generated by training the learning model (54) using sensor data obtained from the twin-screw mixing extruder (1) which has a normal lifespan.

[0131] Additionally, although the above description describes an example of classifying normal and abnormal conditions, it is acceptable to configure the device to distinguish a typical abnormal time series data image from other abnormal time series data images by using a learning model (54) which is a plurality of 1-class classification models. For example, the abnormal detection device (3) is equipped with a first learning model (54) that has been trained on a normal time series data image as in the above embodiment 1. Also, the abnormal detection device (3) is equipped with a second learning model (54) that has been trained on a normal time series data image and a first abnormal time series data image. By using the first learning model (54) and the second learning model (54), the abnormal detection device (3) can distinguish between a normal time series data image, a first abnormal time series data image, and other abnormal time series data images. In the same way, by equipping the abnormality detection device (3) with three or more learning models (54), it can be configured to be able to distinguish two or more time-series data images of abnormalities.

[0132] Additionally, although the diagnosis of screw shaft abnormalities has been described in detail, a learning model (54) that performs a diagnosis of whether there is an inefficient operating state in steps S63 and S64, and a diagnosis of whether there is an operating state that causes quality abnormalities, can also be configured and created in the same way.

[0133] In addition, although a 1-class classification model was described as an example of a learning model (54), general neural networks such as CNN, U-Net, RNN (Recurrent Neural Network), other SVM (Support Vector Machine), Bayesian network, or regression tree may also be used.

[0134] For example, when sensor data in normal conditions and sensor data in various abnormal conditions are accumulated, the state of the twin-screw mixing extruder (1) is used as teaching data, a learning data set is created by labeling the sensor data or time-series data images, and a learning model (54) composed of CNN, etc. is trained. The state of the twin-screw mixing extruder (1) includes a normal state indicating that the twin-screw mixing extruder (1) is normal, an abnormal state where the screw shaft (11) is worn, an abnormal state where resin raw material unsuitable for the screw shaft (11) is used, an overload abnormal state where an overload occurs, an abnormal vibration of the reduction gear (13), an abnormal state where the lifespan is short, and an abnormal state such as the dimensions of the molded product.

[0135] The learning model (54) has an input layer, an intermediate layer, and an output layer. The intermediate layer has a plurality of convolutional layers and pooling layers that extract features of an image. The output layer has a plurality of nodes corresponding to a plurality of states of the 2-axis mixing extruder (1) and outputs a certainty that it is in the said state. When sensor data or a time-series data image corresponding to the sensor data is input, the learning model (54) optimizes the weight coefficients of the intermediate layer so that the state of the 2-axis mixing extruder (1) output from the learning model (54) becomes close to the state indicated by the teacher data. The said weight coefficients are, for example, weights between neurons (coupling coefficients), etc. The method of optimizing parameters is not particularly limited, but optimization of various parameters is performed using, for example, steepest descent method, backpropagation method, etc.

[0136] According to the learning model (54) generated in this way, the state of the two-screw mixing extruder (1) can be diagnosed by inputting the time series data image of the sensor value data obtained from the two-screw mixing extruder (1) to the learning model (54). For example, the diagnosis processing unit (51) determines that the two-screw mixing extruder (1) is in a state corresponding to the node that outputs the highest confidence level.

[0137] Additionally, the learning model (54) may be configured to output a more optimal molding condition or an adjustment amount for the molding condition when a time-series data image is input. The diagnostic device (5) transmits the molding condition output from the learning model (54) to the abnormality detection device (3), and the abnormality detection device (3) transmits the received molding condition to the control device (14).

[0138] <Display Processing of Diagnosis and Judgment Results>

[0139] An example of displaying judgment results and statistics, etc. in Step S19 is explained.

[0140] FIG. 17 is a schematic diagram showing an example of displaying an abnormality determination result, etc. in a control device (14). The display screen (141) for the determination result, etc. displayed in the display unit has a graph display unit (141a), a warning display unit (141b), an AI diagnosis display unit (141c), and a processing value display unit (141d). The graph display unit (141a) displays, for example, the time change of statistics of shaft torque, vibration of the screw shaft (11), and displacement as a graph. The warning display unit (141b) displays a determination result indicating whether there is an abnormality in the received two-axis mixing extruder (1), and if there is an abnormality, displays a warning content. The AI ​​diagnosis display unit (141c) displays the diagnosis result by the diagnosis device (5) described later. The processing value display unit (141d) displays the value of the received statistics.

[0141] Explain an example of the display of the diagnostic result in Step S28.

[0142] FIG. 18 is a schematic diagram showing an example of displaying an abnormality judgment result, etc. in a terminal device (7). The judgment result display screen (71) has a graph display unit (71a), a warning history display unit (71b), an AI diagnosis history display unit (71c), a processing value display unit (71d), and a data selection unit (71e). The data selection unit (71e) accepts the selection of data to be displayed. For example, the data selection unit (71e) accepts the operation time of the twin-screw mixing extruder (1), the type of statistical amount to be displayed, and the selection of a diagnosis item. When a plurality of twin-screw mixing extruders (1) are connected to the first network, the data selection unit (71e) may be configured to accept the selection of the twin-screw mixing extruder (1). The terminal device (7) transmits the operation time selected by the data selection unit (71e), identification information of the twin-screw mixing extruder (1), etc., to the abnormality detection device (3), and requests the history of the abnormality judgment result at the selected twin-screw mixing extruder (1) and operation time, the statistical amount of sensor value data, and the history of the diagnosis result. The terminal device (7) receives such information transmitted from the diagnosis device (5) in response to the request and displays the received information.

[0143] The graph display unit (71a) displays the time change of the statistical amount of sensor data as a graph. The warning history display unit (71b) displays the history of the judgment result indicating whether there is an abnormality in the received twin-screw kneading extruder (1). The AI ​​diagnosis history display unit (71c) displays the history of the diagnosis result regarding the abnormality or lifespan of the twin-screw kneading extruder (1). The processing value display unit (71d) displays the value of the received statistical amount.

[0144] <Effect of operation of the molding machine system according to the present embodiment 1>

[0145] In summary, according to the molding machine system of the present embodiment 1, a control device (14) for controlling the operation of a twin-screw mixing extruder (1), an abnormality detection device (3), and a diagnostic device (5) are provided, and by each device cooperating, an abnormality of the twin-screw mixing extruder (1) can be detected without placing an excessive load on the control device (14), and detailed conditions such as the abnormality and lifespan of the twin-screw mixing extruder (1) can be diagnosed.

[0146] Additionally, the abnormality detection device (3) can simply and easily detect an abnormality of the twin-screw mixing extruder (1) using sensor 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 mixing extruder (1) by simply transmitting the operation data to the abnormality detection device (3), and can monitor the operation of the twin-screw mixing extruder (1) based on the abnormality determination result.

[0147] In addition, the abnormality detection device (3) can calculate the statistical amount of sensor data and transmit it to the control device (14), and display a graph of the statistical amount and the value of the statistical amount on the display unit.

[0148] Furthermore, the abnormal detection device (3) can detect abnormal vibration of the reduction gear (13), overload of the screw shaft (11), and dimensional abnormalities of the molded product, and transmit these abnormalities to the control device (14).

[0149] Furthermore, the diagnostic device (5) can diagnose abnormalities and lifespan of the twin-screw mixing extruder (1) in detail based on the current operating data and sensor data of the twin-screw mixing extruder (1) and past sensor data accumulated in the database (6). The diagnostic device (5) can perform a diagnosis of the condition of the twin-screw mixing extruder (1) which is impossible with the abnormality detection device (3).

[0150] The abnormality detection device (3) can obtain a diagnosis result by transmitting the current operation data and sensor value data of the twin-screw mixing extruder (1) to the diagnosis device (5) and requesting a diagnosis, and can transmit the obtained diagnosis result to the control device (14). The control device (14) can receive the diagnosis result from the diagnosis device (5) and display the diagnosis result on the display unit simply by transmitting the operation data to the abnormality detection device (3).

[0151] Furthermore, the diagnostic device (5) can diagnose the condition, such as abnormalities and lifespan, of the twin-screw mixing extruder (1) using a learning model (54).

[0152] Furthermore, by configuring the learning model (54) as a 1-class classification model, the learning model (54) can be machine-learned using the time-series data images obtained during normal operation of the 2-screw mixing extruder (1) without using the time-series data images of the 2-screw mixing extruder (1) during abnormal operation.

[0153] Furthermore, by expressing sensor data output from multiple sensors (2) as a single time-series data image and calculating the feature quantity of the time-series data image using a learning model (54), the feature quantity indicating the state of the twin-screw mixing extruder (1) with high precision can be obtained simply and easily. By using the feature quantity obtained in this way, the state of the twin-screw mixing extruder (1) can be diagnosed with high precision.

[0154] Furthermore, although the first embodiment described an example in which the diagnostic device (5) machine learns the learning model (54), it is acceptable to machine learn the learning model (54) using another external computer or server.

[0155] Furthermore, although the above embodiment 1 mainly describes the detection and diagnosis of abnormalities in the twin-screw mixing extruder (1), it may also be configured to detect and diagnose abnormalities in molding machines such as injection molding machines and film molding machines, and other manufacturing devices.

[0156] (Variation Example)

[0157] The diagnostic device (5) may be configured to perform billing processing for diagnostic processing using a learning model (54). For example, the diagnostic device (5) may calculate a diagnostic fee based on the number of twin-screw mixing extruders (1), the number of sensors (2), the type of sensors (2), the amount of data of the diagnostic processing target, the number of learning models (54) used, etc., and may store the calculated diagnostic fee, the user ID of the abnormal diagnosis system, the identification ID of the twin-screw mixing extruder (1), the date and time of use, the amount of data, and the diagnostic item in the memory unit (52).

[0158] (Embodiment 2)

[0159] The molding machine system according to Embodiment 2 differs from Embodiment 1 in that it accepts user feedback regarding the diagnosis results and updates the learning model (54) or the diagnosis criteria. Since the other components of the molding machine system are the same as those of the molding machine system according to Embodiment 1, the same reference numerals are assigned to the same locations and detailed descriptions are omitted.

[0160] FIG. 19 is a flowchart showing the feedback processing sequence regarding the diagnosis result according to Embodiment 2. The abnormality detection device (3) receives the suitability of the diagnosis result or the correct state of the twin-screw mixing extruder (1) through the control device (14) (Step S211). The processing unit (31) of the abnormality detection device (3) transmits feedback information to the diagnosis device (5), including the operating data and sensor value data that caused the diagnosis result to be evaluated, and data indicating the correct state of the twin-screw mixing extruder (1) (Step S212).

[0161] In addition, the content of the above feedback information is an example. For example, when a diagnosis ID, operating data, sensor value data, and diagnosis result are stored in a 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 mixing extruder (1) to the diagnosis device (5). In addition, the abnormality detection device (3) may be configured to transmit feedback information including the diagnosis ID and the suitability of the diagnosis result to the diagnosis device (5).

[0162] The diagnostic device (5) receives 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, it stores the operating data, sensor values, and the correct state of the twin-screw mixing extruder (1). Then, the diagnostic processing device further trains the learning model (54) based on the feedback information, or updates the diagnostic criteria (step S215).

[0163] For example, if there is an error in the above diagnosis result and the twin-screw mixing extruder (1) is normal, the sensor value data with the diagnosis result error is added to the learning data set as normal data, and the learning model (54) is further trained, thereby improving the precision of feature extraction by the learning model (54).

[0164] Additionally, it may be configured to change the threshold for diagnosing abnormalities in the twin-screw mixing extruder (1) based on the calculated score value. For example, if there is an error in the diagnosis result of an abnormality and it is determined that the twin-screw mixing extruder (1) is normal, the diagnosis processing unit (51) increases the threshold. If there is an error in the diagnosis result of normality and it is determined that the twin-screw mixing extruder (1) is abnormal, the diagnosis processing unit (51) decreases the threshold.

[0165] In addition, changes to the learning model (54) and diagnostic criteria may be carried out by receiving and accumulating multiple feedback information.

[0166] According to the molding machine system of embodiment 2, the diagnostic precision of the diagnostic device (5) can be further improved by feedback from the user.

[0167] (Embodiment 3)

[0168] The molding machine system according to Embodiment 3 differs from Embodiment 1 in that it accumulates operational data, sensor data, and data indicating the status of the twin-screw mixing extruder (1) that can be obtained at the site in a database (6), and additionally trains a learning model (54). Since the other components of the molding machine system are the same as those of the molding machine system according to Embodiment 1, the same reference numerals are assigned to the same locations, and detailed descriptions are omitted.

[0169] FIG. 20 is a flowchart showing the feedback processing sequence regarding the state of the twin-screw mixing extruder (1) according to embodiment 3. The abnormality detection device (3) receives the state of the twin-screw mixing extruder (1) through the control device (14) (step S311). That is, the abnormality detection device (3) receives the current state of the twin-screw mixing extruder (1) evaluated by the user. Then, the processing unit (31) of the abnormality detection device (3) transmits information including operating data, sensor value data, and the state of the twin-screw mixing extruder (1) to the diagnostic device (5) (step S312).

[0170] Additionally, the abnormal detection device (3) may receive information such as the configuration of the screw shaft (11) of the twin-screw mixing extruder (1), the type of resin raw material, the quality evaluation of the molded product, and the estimated lifespan, and transmit it to the diagnostic device (5).

[0171] The diagnostic device (5) receives information including operation data, sensor data, the status of the twin-screw mixing extruder (1), and other information transmitted from the abnormal 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 data, the status of the twin-screw mixing extruder (1), and other information. Then, the diagnostic processing unit further trains the learning model (54) using a learning data set including sensor data based on user evaluations accumulated in the database (6) (step S315).

[0172] According to the abnormal detection device (3) of embodiment 3, the diagnostic accuracy of the diagnostic device (5) can be further improved by additionally learning the learning model (54) by reflecting the user's evaluation.

[0173] In addition, a data set for supervised learning can be created based on the information transmitted from the abnormal detection device (3) in this manner. The diagnostic device (5) may be configured to generate a learning model (54) that outputs information according to the state of the two-axis mixing extruder (1) when sensor data or time-series data images are input using the data set for supervised learning, as described in Embodiment 1, or to perform additional learning. Explanation of the symbols

[0174] 1 Twin-screw mixing extruder 2 sensors 3 or more detection devices 4 routers 5 Diagnostic Devices 6 databases 7 terminal device 10 cylinders 11 screw shaft 12 motors 13 Reducer 14 Control Unit 31 Processing unit 33 Communications Department 34 Acquisition Department 51 Diagnostic Processing Unit 53 Communications Department 54 Learning Models 54a Neural Network P1, P2 computer program

Claims

Claim 1 An anomaly detection system comprising an anomaly detection device for detecting an anomaly in a manufacturing device and a diagnostic device for diagnosing the state of the manufacturing device, a control device for performing operation control of the manufacturing device and transmitting operation data representing the contents of the operation control to the anomaly detection device, a sensor for detecting a physical quantity according to the operation or product of the manufacturing device and outputting time-series sensor value data representing the detected physical quantity to the anomaly detection device, and a database for accumulating information including the operation data, the sensor value data, and the state of the manufacturing device, wherein the anomaly detection device comprises a first communication unit for receiving the operation data transmitted from the control device, an acquisition unit for acquiring the sensor value data output from the sensor, and a processing unit for calculating a statistical amount of the sensor value data acquired by the acquisition unit and determining whether there is an anomaly in the manufacturing device based on the calculated statistical amount and a threshold value according to the received operation data, wherein the first communication unit transmits the operation data and the sensor value data to the diagnostic device, and the diagnostic device comprises a second communication unit for receiving the operation data and the sensor value data transmitted from the anomaly detection device, and An abnormality detection system comprising a diagnostic processing unit that diagnoses the state of the manufacturing device based on the driving data and sensor data received from the second communication unit and information accumulated in the database, wherein the second communication unit transmits the diagnostic result by the diagnostic processing unit to the abnormality detection device, and the first communication unit of the abnormality detection device receives the diagnostic result by the diagnostic processing unit and transmits the judgment result by the processing unit, the statistical amount, and the diagnostic result by the diagnostic processing unit to the control device. Claim 2 In claim 1, the status of the manufacturing device accumulated in the database includes information regarding the lifespan of the manufacturing device and information regarding abnormalities of the manufacturing device, and the diagnostic processing unit is an abnormality detection system that diagnoses the lifespan or the presence or absence of abnormalities of the manufacturing device. Claim 3 An anomaly detection system according to claim 1 or 2, wherein the diagnostic device comprises a learning model that outputs a characteristic quantity of the manufacturing device or product when the sensor value data is input, and the diagnostic processing unit diagnoses the state of the manufacturing device by comparing the characteristic quantity obtainable by inputting the received sensor value data into the learning model with the characteristic quantity obtainable by inputting the sensor value data accumulated in the database into the learning model. Claim 4 An anomaly detection system according to paragraph 3, wherein the second communication unit receives feedback information indicating the suitability or correctness of the diagnosis result or the diagnosis result by the diagnosis processing unit, and the diagnosis processing unit further trains the learning model based on the received feedback information or updates the diagnosis criteria based on the feature quantity. Claim 5 An abnormality detection system according to claim 1 or 2, wherein the first communication unit or the second communication unit receives information indicating the state of the manufacturing device corresponding to the operation data and the sensor value data, and the abnormality detection device or the diagnostic device stores the received state in the database corresponding to the operation data and the sensor value data. Claim 6 A molding machine system comprising an abnormality detection system described in claim 1 or 2 and a molding machine, wherein the abnormality detection system detects an abnormality of the molding machine. Claim 7 delete Claim 8 An anomaly detection method implemented by an anomaly detection system comprising an anomaly detection device for detecting an anomaly in a manufacturing device and a diagnostic device for diagnosing the state of the manufacturing device, wherein the control unit of the anomaly detection system performs operation control of the manufacturing device and transmits operation data representing the contents of the operation control to the anomaly detection system; the sensor of the anomaly detection system detects a physical quantity according to the operation of the manufacturing device or a product and outputs time-series sensor value data representing the detected physical quantity to the anomaly detection system; the database of the anomaly detection system performs a process of accumulating information including the operation data, the sensor value data, and the state of the manufacturing device; the first communication unit of the anomaly detection system receives the operation data transmitted from the control unit; the acquisition unit of the anomaly detection system acquires the sensor value data output from the sensor; and the processing unit of the anomaly detection system calculates a statistical amount of the sensor value data acquired by the acquisition unit and determines whether there is an anomaly in the manufacturing device based on the calculated statistical amount and a threshold value according to the received operation data. When the first communication unit transmits the driving data and the sensor value data to the diagnostic device, the second communication unit of the diagnostic device performs a process of receiving the driving data and the sensor value data transmitted from the abnormality detection device, and the diagnostic processing unit of the diagnostic device performs a process of diagnosing the state of the manufacturing device based on the driving data and the sensor value data received from the second communication unit and information accumulated in the database, and when the second communication unit transmits the diagnostic result by the diagnostic processing unit to the abnormality detection device, the first communication unit of the abnormality detection device receives the diagnostic result by the diagnostic processing unit, and the judgment result by the processing unit and the statistical amount, andAn abnormality detection method that performs a process of transmitting the diagnosis result by the diagnosis processing unit to the control device. Claim 9 A computer-readable recording medium having a program for executing the abnormal detection method of paragraph 8.

Images