State estimation method, state estimation device, computer program, and machine learning method

Abstract

To provide condition estimation method that can determine the presence or severity of abnormalities in an extruder, regardless of the model, operating method, or raw materials used.SOLUTION: Physical quantity data related to the state of the extruder is acquired, and when the aforementioned physical quantity data or features based on the aforementioned physical quantity data are input, the state of the extruder is determined by inputting the acquired physical quantity data or features based on the physical quantity data into a machine learning model that outputs information indicating the state of the extruder.SELECTED DRAWING: Figure 7

Description

Technical Field

[0001] The present invention relates to a state estimation method, a state estimation device, a computer program, and a machine learning method.

Background Art

[0002] Patent Document 1 discloses a technique in which a load detection device provided in an input shaft portion of a screw shaft detects a load (torque) on the input shaft portion, and when the detected load average value and load amplitude value are not within a predetermined normal range, it is determined that an overload state exists, and when the overload continuous time maintaining the overload state exceeds a set time, an abnormal alarm is notified and / or the rotation of the screw shaft is stopped.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, as a problem peculiar to an extruder, there is a problem that the scale of detection values varies depending on the model. Also, even with the same model, there is a problem that the scale of detection values varies depending on the operation method and raw materials. When the raw materials and operation methods are different, it is difficult to detect abnormalities with simple thresholds. Usually, there is no clear standard for the thresholds for abnormality detection, and they are determined, for example, by skilled operators with operation experience.

[0005] An object of the present disclosure is to provide a state estimation method, a state estimation device, a computer program, and a machine learning method capable of determining the presence or absence or degree of abnormality of an extruder regardless of the model, operation method, and differences in raw materials.

Means for Solving the Problems

[0006] The state estimation method according to one aspect of the present disclosure acquires physical quantity data related to the state of an extruder, and when the physical quantity data or a feature quantity based on the physical quantity data is input, the physical quantity data obtained is input to a machine learning model that outputs information indicating the state of the extruder, thereby determining the state of the extruder.

[0007] The state estimation device according to one aspect of the present disclosure includes a processing unit that acquires physical quantity data related to the state of an extruder, and when the physical quantity data or a feature quantity based on the physical quantity data is input, executes a process of determining the state of the extruder by inputting the obtained physical quantity data or the feature quantity based on the physical quantity data to a machine learning model that outputs information indicating the state of the extruder.

[0008] The computer program according to one aspect of the present disclosure causes a computer to execute a process of determining the state of an extruder by acquiring physical quantity data related to the state of the extruder and inputting the obtained physical quantity data or a feature quantity based on the physical quantity data to a machine learning model that outputs information indicating the state of the extruder when the physical quantity data or the feature quantity based on the physical quantity data is input.

[0009] The machine learning method according to one aspect of the present disclosure acquires physical quantity data related to the state of an extruder, and generates a machine learning model that outputs information indicating the state of the extruder based on the obtained physical quantity data when the physical quantity data or a feature quantity based on the physical quantity data is input.

Advantages of the Invention

[0010] According to the present disclosure, it is possible to determine the presence or absence or degree of abnormality of an extruder regardless of its model, operation method, and difference in raw materials.

Brief Description of the Drawings

[0011]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Figure 15

MODE FOR CARRYING OUT THE INVENTION

[0012] A state estimation method, a state estimation apparatus, a computer program, and a machine learning method according to embodiments of the present disclosure will be described below with reference to the drawings. Note that the present disclosure is not limited to these examples, but is defined by the claims, and is intended to include all modifications within the meaning and scope equivalent to the claims. Also, at least a part of the embodiments described below may be arbitrarily combined.

[0013] FIG. 1 is a block diagram showing a configuration example of an extruder system according to Embodiment 1. The extruder system includes an extruder 1, a plurality of sensors 2, a data collection device 3, a router 4, a state estimation device 5, and a terminal device 6.

[0014] Although one extruder 1 and one data collection device 3 are shown in FIG. 1, a plurality of data collection devices 3 may be configured to be connected to the state estimation device 5 via a network. One or a plurality of extruders 1 may be configured to be connected to the data collection device 3. The state estimation device 5 can collect information on each of the one or a plurality of extruders 1 and estimate the state of each extruder 1.

[0015] The terminal device 6 is a communication terminal having a display unit such as a computer, a tablet terminal, or a smartphone.

[0016] <Extruder 1> FIG. 2 is a schematic diagram showing a configuration example of the extruder 1 according to Embodiment 1. The extruder 1 includes a cylinder 10, two screws 11, and a die 12 (see FIG. 1) provided at the outlet portion of the cylinder 10. The cylinder 10 has an inlet 10a into which a resin raw material is charged. The resin raw material is supplied from the inlet 10a to the cylinder 10 by a feeder 10b. The feeder 10b is a device that supplies the raw material to the cylinder 10 of the extruder while performing weight control. The two screws 11 are arranged substantially parallel to each other in a meshed state and are rotatably inserted into the holes of the cylinder 10, and convey the resin raw material charged into the inlet 10a in the extrusion direction (right direction in FIGS. 1 and 2), and melt and knead it. The melted resin raw material is discharged from the die 12 having a through hole. The screw 11 is configured as a single screw by combining and integrating multiple types of screw pieces. For example, a forward flight piece in the shape of a flight screw for transporting the resin raw material in the forward direction, a reverse flight piece for transporting the resin raw material in the reverse direction, a kneading piece for kneading the resin raw material, etc. are arranged and combined in an order and position according to the characteristics of the resin raw material, thereby constituting the screw 11.

[0017] Further, the extruder 1 includes a motor 13 that outputs a driving force for rotating the screw 11, a speed reducer 14 that decelerates and transmits the driving force of the motor 13, and a control device 15. The screw 11 is connected to the output shaft of the speed reducer 14. The screw 11 rotates by the driving force of the motor 13 that is decelerated and transmitted by the speed reducer 14.

[0018] <Sensor 2> The sensor 2 detects a physical quantity related to the state of the members constituting the extruder 1, and directly or indirectly outputs the physical quantity data obtained by the detection to the data collection device 3. The physical quantity data is data of sensor values in a time series indicating the detected physical quantity. The sensor 2 includes those provided in the extruder 1 as necessary for the operation control of the extruder 1 and those provided separately for estimating the state of the extruder 1. A part of the plurality of sensors 2 is connected to the data collection device 3, and the data collection device 3 acquires the physical quantity data from the sensor 2. A part of the plurality of sensors 2 is connected to the control device 15, and the data collection device 3 acquires the physical quantity data from the sensor 2 via the control device 15.

[0019] 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, a position sensor, a speed sensor, an acceleration sensor, an ammeter, a voltmeter, a pressure gauge, a timer, a camera, a torque sensor, a power meter, a weighing scale, etc.

[0020] The plurality of sensors 2 include, for example, a raw material supply amount detector, a screw speed detector, an extruder power detector, an extruder torque detector, an extruder current detector, an extruder load detector, a resin pressure detector, a resin temperature detector, a cylinder or barrel temperature detector, a cylinder or barrel output detector, an imaging unit that captures a still image or a moving image including the opening of the extruder 1, and a pressure sensor that detects the pressure of the vent of the extruder 1. The raw material supply amount detector is a sensor that detects the supply amount of the resin raw material input by the feeder. The extruder power detector is a sensor that detects the power of the motor 13. The extruder torque detector is a sensor that detects the torque acting on the motor 13 or the screw 11. Further, the sensor 2 also includes any detector that can detect a physical quantity contributing to the estimation of the throughput, screw speed, and load of the extruder 1 which is the extruder.

[0021] In addition, the physical quantity data output from the control device 15 to the data collection device 3 may include a set value of the screw speed, a set value of the torque of the extruder 1, a set value of the current or load, a set value of the raw material supply amount, a set value of the temperature or output of the cylinder or barrel. Further, the physical quantity data may include static information (data indicating static characteristics), for example, the diameter of the extruder 1, the screw length, the load limit, the maximum speed, information indicating the configuration of the screw, and index information related to the screw configuration.

[0022] <Control device 15> The control device 15 is a computer that controls the operation of the extruder 1, and includes a transmission / reception unit (not shown) that transmits and receives information to and from the data collection device 3, and a display unit. Specifically, the control device 15 transmits operation data indicating the operation state of the extruder 1 to the data collection device 3. The operation data includes, for example, the raw material supply amount, the screw speed, the extruder power, the extruder torque, the extruder current, the resin pressure, the resin temperature, etc. The control device 15 receives the state estimation result of the extruder 1 transmitted from the data collection device 3, and displays the received state estimation result. The state estimation result is information indicating the presence or absence of abnormalities and the degree of abnormality of the extruder 1 and its components.

[0023] <Data collection device 3> FIG. 3 is a block diagram showing a configuration example of the data collection device 3 according to Embodiment 1. The data collection device 3 is a computer and includes a control unit 31, a storage unit 32, a communication unit 33, and a data input unit 34, and the storage unit 32, the communication unit 33, and the data input unit 34 are connected to the control unit 31. The data collection device 3 is, for example, a PLC (Programmable Logic Controller).

[0024] The control unit 31 includes an arithmetic processing circuit such as a CPU (Central Processing Unit), a multi-core CPU, an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array), an internal storage device such as a ROM (Read Only Memory) and a RAM (Random Access Memory), and I / O terminals. By executing a control program stored in the storage unit 32 described later, the control unit 31 executes a process of collecting physical quantity data and transmitting it to the state estimation device 5. Each functional unit of the data collection device 3 may be realized software-wise or partially or entirely hardware-wise.

[0025] The storage unit 32 is a non-volatile memory such as a hard disk, an EEPROM (Electrically Erasable Programmable ROM), or a flash memory. The storage unit 32 stores a control program for causing a computer to perform the process of collecting physical quantity data.

[0026] 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 15 via a first communication network such as a LAN, and the control unit 31 can transmit and receive various information to and from the control device 15 via the communication unit 33. The control unit 31 acquires physical quantity data via the communication unit 33. The first network is connected to router 4, and the communication unit 33 is connected to the state estimation device 5 on the cloud, which is the second communication network, via router 4. The control unit 31 can transmit and receive various information to and from the state estimation device 5 via the communication unit 33 and router 4.

[0027] The data input unit 34 is an input interface to which the signal output from the sensor 2 is input. The sensor 2 is connected to the data input unit 34, and the control unit 31 acquires physical quantity data via the data input unit 34.

[0028] <State estimation device 5> FIG. 4 is a block diagram showing a configuration example of the state estimation device 5 according to Embodiment 1. The state estimation device 5 is a computer and includes a 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 processing unit 51.

[0029] The processing unit 51 is a processor and has 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, an NPU (Neural Processing Unit), an internal storage device such as a ROM and a RAM, and I / O terminals. The processing unit 51 functions as the state estimation device 5 according to the present embodiment by executing a computer program P stored in the storage unit 52 described later. Note that each functional unit of the state estimation device 5 may be realized software-wise or partially or entirely may be realized hardware-wise.

[0030] 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 data collection device 3 and the terminal device 6 via the second communication network, and the processing unit 51 can transmit and receive various information to and from the data collection device 3 and the terminal device 6 via the communication unit 53.

[0031] The storage unit 52 is a non-volatile memory such as a hard disk, EEPROM, or flash memory. The storage unit 52 stores a computer program P for causing a computer to execute a process of estimating the presence or absence and degree of abnormality of the extruder 1 and its constituent members, a first autoencoder 54, a second autoencoder 55, a third autoencoder 56, and a collected data DB57.

[0032] The computer program P and the like may be recorded on a recording medium 50 in a computer-readable manner. The storage unit 52 stores the computer program P and the like 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. 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 (registered trademark) Disc). Further, the recording medium 50 may be a magnetic disc such as a flexible disc or hard disk, a magneto-optical disc, or the like. Furthermore, the computer program P and the like may be downloaded from an external server (not shown) connected to a communication network (not shown) and stored in the storage unit 52.

[0033] When the physical quantity data detected by the sensor 2 is input, the first autoencoder 54 is an autoencoder that reproduces and outputs the physical quantity data that would be obtained from the normal extruder 1 regardless of the state at the time of detection of the physical quantity data.

[0034] FIG. 5 is a conceptual diagram showing the first autoencoder 54 in the learning phase, and FIG. 6 is a conceptual diagram showing the first autoencoder 54 in the detection phase. When physical quantity data is input to the first autoencoder 54, the first autoencoder 54 extracts the features included in the physical quantity data and outputs physical quantity data that reproduces the features of the physical quantity obtained during normal operation. As shown in FIG. 6, when the physical quantity data obtained by detecting an abnormality in the extruder 1 is input to the first autoencoder 54, the first autoencoder 54 outputs physical quantity data that reproduces the features of the physical quantity that would be obtained during normal operation of the extruder 1.

[0035] The first autoencoder 54 has an input layer 54a, an intermediate layer 54b, and an output layer 54c. The intermediate layer 54b has an encoder layer and a decoder layer, and has a network structure that is symmetric on the input side and the output side. The input layer 54a is a layer to which physical quantity data is input. As the physical quantity data, it is preferable to input the detected value detected at the current time and the detected values at a plurality of time points in the past relative to the current time to the input layer 54a. The encoder layer of the intermediate layer 54b is a layer that compresses the dimension of the physical quantity data. By dimension compression, the feature amount of the physical quantity data is extracted. The decoder layer of the intermediate layer 54b is a layer that restores the data (latent variable) dimensionally compressed by the encoder layer to the original dimension. By this restoration, the physical quantity data representing the original features of the physical quantity data, that is, the features of the physical quantity that would be detected when the extruder 1 is normal, is restored. The output layer 54c is a layer that extracts the features of the physical quantity by the encoder layer and the decoder layer and outputs the physical quantity data in which the features during normal operation are reproduced.

[0036] As shown in FIG. 6, the processing unit 51 machine-learns the neural network of the first autoencoder 54 so that the normal physical quantity data input to the first autoencoder 54 and the output physical quantity data are the same. Specifically, various parameters such as the weight coefficients (coupling coefficients) between neurons constituting the neural network are optimized using the error backpropagation method, the steepest descent method, etc. so that the input physical quantity data during normal operation and the reproduced physical quantity data are the same.

[0037] When normal physical quantity data is input to the first autoencoder 54 learned in this way, physical quantity data having substantially the same features as the input physical quantity data is output. When physical quantity data during an abnormal time is input to the first autoencoder 54, as shown in FIG. 6, normal physical quantity data is output.

[0038] FIG. 7 is a conceptual diagram showing the second autoencoder 55 in the detection phase, and FIG. 8 is a conceptual diagram showing the third autoencoder 56 in the detection phase.

[0039] As shown in FIG. 7, the second autoencoder 55 is an autoencoder that, when a feature amount calculated based on a plurality of physical quantity data is input, reproduces and outputs the feature amount that would be obtained from the normal extruder 1. Although various combinations of the plurality of physical quantity data are conceivable, combinations of the power or current of the motor 13 for the extruder and torque, and combinations of the torque of the motor 13 or the screw 11 and the rotational speed are suitable.

[0040] As shown in FIG. 8, the third autoencoder 56 is an autoencoder that, when a feature amount calculated based on physical quantity data and a set value of the extruder 1 related to the physical quantity data is input, reproduces and outputs the feature amount that would be obtained from the normal extruder 1. Although various combinations of the physical quantity data and the set value are conceivable, combinations of the raw material supply amount and the raw material supply amount set value, and combinations of the screw speed and the screw speed set value are suitable.

[0041] Since the configurations of the second autoencoder 55 and the third autoencoder 56 are the same as that of the first autoencoder 54, detailed descriptions thereof are omitted.

[0042] In addition, the first autoencoder 54, the second autoencoder 55, and the third autoencoder 56 may be a CNN autoencoder having a CNN (Convolutional Neural Network), an RNN encoder having an RNN (Recurrent Neural Network), or an LSTM autoencoder having an LSTM (Long Short Term Memory). In the case of a CNN autoencoder, it is preferable to input, as physical quantity data, image data representing a detected value waveform for a predetermined period including the current time.

[0043] Note that a plurality of the first autoencoder 54, the second autoencoder 55, and the third autoencoder 56 may be prepared respectively according to the type of physical quantity data. [[ID=⑦]] In addition, in the present embodiment, for convenience of understanding the invention, the first autoencoder 54, the second autoencoder 55, and the third autoencoder 56 will be separately described as different autoencoders. However, if possible, they may be configured with one or two autoencoders.

[0044] FIG. 9 is a conceptual diagram showing an example of a record layout of the collected data DB57. The collected data DB57 includes a hard disk and a DBMS (DataBase Management System), and stores various physical quantity data collected from the extruder 1. For example, the collected data DB57 has a column of "No." (record number), a column of "equipment ID", a column of "operation date and time", a column of "set value", and a column of "physical quantity data".

[0045] The "equipment ID" column stores the equipment identifier of the extruder 1. The "operation date and time" column stores information indicating the year, month, date and time when various data stored as records were obtained. The "set value" column stores the set values set in the extruder 1. The set values include, for example, a raw material supply amount set value and a screw speed set value. The "physical quantity data" column stores time-series physical quantities indicating the operating state of the extruder 1, such as raw material supply amount, screw speed, extruder power, extruder torque, extruder current, resin pressure, resin temperature, etc.

[0046] <Machine learning process> FIG. 10 is a flowchart showing the learning process procedure of the autoencoder. The processing unit 51 of the state estimation device 5 acquires physical quantity data and set values related to the state of the extruder 1 from the data collection device 3 (step S11). The processing unit 51 stores the acquired physical quantity data in the collection data DB57. When the physical quantity data necessary for machine learning is collected, the following processing is executed.

[0047] Next, the processing unit 51 executes the machine learning process of the first autoencoder 54 based on the acquired physical quantity data (step S12).

[0048] Also, the processing unit 51 calculates a feature amount based on the acquired plurality of physical quantity data (step S13), and executes the machine learning process of the second autoencoder 55 based on the calculated feature amount (step S14).

[0049] Furthermore, the processing unit 51 calculates a feature amount based on the acquired physical quantity data and the set value related to the physical quantity data (step S15), and executes the machine learning process of the third autoencoder 56 based on the calculated feature amount (step S16). The method for calculating the feature amount in step S16 is not particularly limited, but it is preferable to perform scaling conversion of the physical quantity data based on the set value.

[0050] The processing unit 51 that has completed the learning of the first to third autoencoders 54, 55, 56 stores the data for configuring the learned first to third autoencoders 54, 55, 56 in the storage unit 52 (step S17).

[0051] Although the method of collecting physical quantity data and generating the first to third autoencoders 54, 55, 56 has been described, the first to third autoencoders 54, 55, 56 may be machine-learned while collecting physical quantity data during the operation of the extruder 1. Also, based on the physical quantity data acquired during the operation of the extruder 1, the first to third autoencoders 54, 55, 56 may be configured to be additionally learned at an appropriate timing.

[0052] <State determination process> FIG. 11 is a flowchart showing a processing procedure for state estimation of the extruder 1 according to Embodiment 1. The processing unit 51 of the state estimation device 5 acquires physical quantity data and set values related to the state of the extruder 1 from the data collection device 3 (step S31).

[0053] Next, the processing unit 51 inputs the acquired physical quantity data into the first autoencoder 54 to reproduce the physical quantity data in a normal state (step S32). Then, the processing unit 51 determines the presence or absence of an abnormality in the extruder 1 by comparing the physical quantity data input to the first autoencoder 54 with the physical quantity data output from the first autoencoder 54 (step S33).

[0054] Further, the processing unit 51 may obtain the degree of abnormality of the extruder 1 by calculating the difference between the physical quantity data input to the first autoencoder 54 and the physical quantity data output from the first autoencoder 54. The processing unit 51 may use the difference between the value of the physical quantity data input to the first autoencoder 54 and the value of the physical quantity data output from the first autoencoder 54 at each time point in the time series as a residual, and calculate the integrated value of the residuals as the degree of abnormality of the extruder 1. Further, the processing unit 51 may calculate the sum of squared residuals and the square root of the sum of squares of residuals based on the residuals as the degree of abnormality of the extruder 1.

[0055] Next, the processing unit 51 calculates a feature amount based on the acquired plurality of physical quantity data (step S34), and inputs the calculated feature amount into the second autoencoder 55 to reproduce the normal feature amount (step S35). Then, the processing unit 51 determines the presence or absence of an abnormality in the extruder 1 by comparing the feature amount input to the second autoencoder 55 with the feature amount output from the second autoencoder 55 (step S36). Further, the processing unit 51 may obtain the degree of abnormality of the extruder 1 by calculating the difference between the feature amount input to the second autoencoder 55 and the feature amount output from the second autoencoder 55, similar to step S33.

[0056] Furthermore, the processing unit 51 calculates a feature amount based on the acquired physical quantity data and the set value related to the physical quantity data (step S37), and reproduces the feature amount in a normal state by inputting the calculated feature amount to the third autoencoder 56 (step S38). Then, the processing unit 51 determines the presence or absence of an abnormality in the extruder 1 by comparing the feature amount input to the third autoencoder 56 with the feature amount output from the third autoencoder 56 (step S39). Also, similar to step S33, the processing unit 51 may obtain the degree of abnormality of the extruder 1 by calculating the difference between the feature amount input to the third autoencoder 56 and the feature amount output from the third autoencoder 56.

[0057] Next, the processing unit 51 comprehensively calculates the presence or absence or the degree of abnormality of the extruder 1 based on the determination results of step S33, step S36, and step S39 (step S40), and ends the processing. For example, when the presence or absence of an abnormality is determined in each step, the processing unit 51 may determine the presence or absence of an abnormality in the extruder 1 based on the most frequent determination result. When the degree of abnormality is calculated in each step, the processing unit 51 may calculate the average value, the maximum value, etc. of the degree of abnormality as the comprehensively determined degree of abnormality.

[0058] As described above, according to the state estimation method and the like according to the first embodiment, it is possible to determine the presence or absence or the degree of abnormality of the extruder 1 regardless of its model, operation method, and difference in raw materials. Generally, due to differences in the model, operation method, and resin raw material of the extruder, characteristics such as the scale of the physical quantity data to be detected are different, so it is impossible to cope with abnormal detection using a threshold value. The state estimation method according to the first embodiment is configured to perform an abnormality determination by reproducing and comparing the physical quantity data in a normal state using an autoencoder. Therefore, it is possible to detect an abnormality regardless of the model, operation method, and difference in resin raw material of the extruder.

[0059] In particular, in the first embodiment, by using the torque of the motor 13 or the screw 11 and the feature amount calculated based on the rotational speed, it is possible to more accurately calculate the presence or absence or the degree of abnormality of the extruder 1. Also, by using the feature amount calculated based on the power or current and the torque of the motor 13 for the extruder as a plurality of physical quantity data, it is possible to more accurately calculate the presence or absence or the degree of abnormality of the extruder 1.

[0060] Furthermore, in the first embodiment, by using the feature amount calculated based on the raw material supply amount and the raw material supply amount set value, it is possible to more accurately calculate the presence or absence or the degree of abnormality of the extruder 1. Also, by using the feature amount calculated based on the screw speed and the screw speed set value, it is possible to more accurately calculate the presence or absence or the degree of abnormality of the extruder 1.

[0061] Note that in the first embodiment, an example of detecting the abnormality of the extruder 1 using an autoencoder has been described. However, other arbitrary neural networks, learning models configured using a multi-layer perceptron (MLP), etc., algorithms such as decision trees, random forests, and SVM (Support Vector Machine) may be used to configure the determination of the abnormality of the extruder 1.

[0062] Also, in the first embodiment, an example in which the state estimation device 5, which is a server on the cloud, executes the state estimation process has been described. However, it may be configured such that the control device 15, the data collection device 3 such as a PLC, or a local computer connected to the data collection device 3 executes the state estimation process. When the data collection device 3 executes the state estimation process, the communication unit 33 is unnecessary.

[0063] Furthermore, in the first embodiment, an example of determining the presence or absence of an abnormality in the extruder 1 has been mainly described, but it may be configured to determine any state of the extruder 1. For example, it may be configured to determine an abnormal state of the extruder 1, a normal state of the extruder 1, an operating state of the extruder 1, or a state related to a combination thereof. Specifically, when physical quantity data or feature quantities are input, the autoencoder may be machine-learned to reproduce and output the physical quantity data or feature quantities obtained from the extruder 1 in a specific abnormal state. By comparing the acquired physical quantity data or feature quantities with the physical quantity data or feature quantities output from the autoencoder, it is possible to determine whether it is in a specific abnormal state. Also, when physical quantity data or feature quantities are input, the autoencoder may be machine-learned to reproduce and output the physical quantity data or feature quantities obtained from the extruder 1 in a specific operating state. By comparing the acquired physical quantity data or feature quantities with the physical quantity data or feature quantities output from the autoencoder, it is possible to determine whether the extruder 1 is in a specific operating state.

[0064] (Embodiment 2) The state estimation system according to the second embodiment differs from the first embodiment in the method of calculating feature quantities based on a plurality of physical quantity data. Since the other configurations of the state estimation system are the same as those of the system according to the first embodiment, the same reference numerals are given to the same parts, and detailed descriptions thereof are omitted.

[0065] FIG. 12 is a conceptual diagram showing a feature quantity calculation model 58 and a second autoencoder 55 according to the second embodiment. The state estimation device 5 according to the second embodiment stores the feature quantity calculation model 58 in the storage unit 52.

[0066] The feature quantity calculation model 58 has an input layer 58a, an intermediate layer 58b, and an output layer 58c. The feature quantity calculation model 58 is, for example, a learning model similar to the first autoencoder 54.

[0067] For example, the processing unit 51 performs machine learning on the neural network of the feature quantity calculation model 58 so that the first physical quantity data in a normal state input to the feature quantity calculation model 58 and the output first physical quantity data are the same. Further, the processing unit 51 performs machine learning on the neural network of the feature quantity calculation model 58 so that the second physical quantity data in a normal state input to the feature quantity calculation model 58 and the output second physical quantity data are the same. The first physical quantity data is, for example, torque data acting on the screw 11. The second physical quantity data is, for example, the rotational speed of the screw 11.

[0068] The feature quantity calculation model 58 thus learned is learned so as to be able to extract the features of each of the first physical quantity data and the second physical quantity data. As shown in FIG. 12, when the first physical quantity data and the second physical quantity data are input to the feature quantity calculation model 58, the features of both the first and second physical quantity data are extracted as latent variables. This latent variable corresponds to a feature quantity calculated based on the first physical quantity data and the second physical quantity data.

[0069] The processing unit 51 acquires a latent variable having the features of both the first and second physical quantity data from the intermediate layer 58b of the feature quantity calculation model 58, and inputs the acquired latent variable as a feature quantity to the second autoencoder 55 in the same manner as in the first embodiment, thereby calculating a reproduced latent variable in a normal state. Then, the processing unit 51 determines the presence or absence and the degree of abnormality of the extruder 1 by comparing the latent variable input to the second autoencoder 55 with the latent variable output from the second autoencoder 55. Needless to say, the second autoencoder 55 is learned to output a latent variable that would be detected and calculated in a normal state when a feature quantity that is a latent variable is input.

[0070] As described above, according to the state estimation method and the like according to the second embodiment, compared with the rule-based arithmetic processing, more accurate feature quantities based on a plurality of physical quantity data can be calculated, and the state of the extruder 1 can be determined more accurately.

[0071] In addition, in the second embodiment, torque and rotational speed were described as the first and second physical quantity data. However, the power or current of the motor 13 for the extruder may be used as the first physical quantity data, and torque may be used as the second physical quantity data.

[0072] (Embodiment 3) The state estimation system according to Embodiment 3 differs from that of Embodiment 1 in the method for determining the state of the extruder 1. Since the other configurations of the state estimation system are the same as those of the system according to Embodiment 1, the same reference numerals are given to the same parts, and detailed descriptions thereof are omitted.

[0073] FIG. 13 is a conceptual diagram showing the state estimation method according to Embodiment 3. The state estimation device 5 according to Embodiment 3 stores in the storage unit 52 an autoencoder 354a and a VAE 354b as the first autoencoder 54, and a CNN autoencoder 355a and an LSTM autoencoder 355b as the second autoencoder 55. Note that the autoencoder 354a is an autoencoder that processes physical quantity data which is a time-series numerical sequence, while the CNN autoencoder 355a is an autoencoder that processes physical quantity data represented by an image.

[0074] The processing unit 51 also includes a threshold determination unit 351a, a comparison processing unit 351b, and an integrated determination unit 351c as functional units.

[0075] The threshold determination unit 351a determines the presence or absence of an abnormality in the extruder 1 by comparing the physical quantity data acquired by the communication unit 53 with a predetermined threshold. Note that the threshold determination unit 351a may compare the threshold with the value of the physical quantity data, and calculate the integrated value, sum of squares, and square root of the sum of squares of the amount exceeding the threshold as the degree of abnormality.

[0076] The comparison processing unit 351b determines the presence or absence of an abnormality in the extruder 1 or calculates the degree of abnormality by comparing the physical quantity data or feature quantity input to each type of autoencoder with the physical quantity data or feature quantity output. A machine learning model using a neural network such as an MLP may be used to comprehensively determine the presence or absence of an abnormality or the degree of abnormality in the extruder 1. Further, a comprehensive determination may be made by combining the determination process using the machine learning model and the determination process based on rules.

[0077] The comprehensive determination unit 351c comprehensively determines the presence or absence of an abnormality in the extruder 1 based on the determination result by the threshold determination unit 351a and the determination result by the comparison processing unit 351b. For example, the presence or absence of an abnormality in the extruder 1 is determined based on the most frequent determination result. Also, when the degree of abnormality is calculated by the threshold determination unit 351a and the comparison processing unit 351b, the comprehensive determination unit 351c may be configured to calculate the average value, maximum value, etc. of the degree of abnormality as the comprehensively determined degree of abnormality.

[0078] As described above, according to the state estimation method and the like according to the third embodiment, the state of the extruder 1 can be determined more accurately by the comprehensive determination process based on the determination results obtained using a plurality of types of autoencoders.

[0079] (Embodiment 4) The state estimation system according to the fourth embodiment is different from the first embodiment in that it discriminates the operating state of the extruder 1 and performs an abnormality determination of the extruder 1 using an autoencoder corresponding to the operating state. Since the other configurations of the state estimation system are the same as those of the system according to the first embodiment, the same reference numerals are given to the same parts, and detailed descriptions thereof are omitted.

[0080] FIG. 14 is a block diagram showing a configuration example of the state estimation device 5 according to the fourth embodiment. The state estimation device 5 according to the fourth embodiment stores an operating state discrimination model 59 for discriminating the operating state of the extruder 1 in the storage unit 52. The operating state corresponds to the operating process of the extruder, and the operating state includes, for example, a stop state, a startup operating state, a steady operating state, and a shutdown operating state.

[0081] The operation status discrimination model 59 has an input layer (not shown), an intermediate layer, and an output layer. The operation status discrimination model 59 is a learning model that outputs data indicating the operation status of the extruder 1 at the time when the physical quantity data is detected, for example, when the physical quantity data is input.

[0082] For example, prepare learning data indicating the operation status, and use the learning data to perform machine learning on the operation status discrimination model 59. The data indicating the operation status is, for example, data used for unsupervised learning in a normal state, semi-supervised learning in which only abnormal states are extracted, or supervised learning with labels. For example, in the case of supervised learning, prepare learning data in which data indicating the operation status of the extruder 1 when the physical quantity data is obtained is added as teacher data to the physical quantity data. Then, when the physical quantity data of the learning data is input to the operation status discrimination model 59, the neural network of the operation status discrimination model 59 is machine-learned so that the operation status indicated by the data output from the operation status discrimination model 59 matches the operation status indicated by the teacher data. The operation status discrimination model 59 may be unsupervised learned using physical quantity data obtained in a plurality of operation statuses such as a stop state, a startup operation state, a steady operation state, and a shutdown operation state as learning data. For example, the operation status discrimination model 59 is a model that clusters physical quantity data and outputs data indicating the class (operation status) to which the input physical quantity data belongs. Note that the operation status discrimination model 59 may be unsupervised learned using physical quantity data obtained in one type of operation status as learning data. The operation status discrimination model 59 may be semi-supervised learned using learning data including physical quantity data with label data indicating the operation status and physical quantity data without label data. For example, the operation status discrimination model 59 may be semi-supervised learned using physical quantity data with label data indicating an abnormal state and physical quantity data without label data for other states.

[0083] In addition, the storage unit 52 stores the first to third autoencoders 54, 55, and 56 learned for different driving situations. For example, the storage unit 52 stores the first autoencoder 54 for determining an abnormality during startup operation, the first autoencoder 54 for determining an abnormality during steady operation, and the first autoencoder 54 for determining an abnormality during shutdown operation. The same applies to the second autoencoder 55 and the third autoencoder 56.

[0084] FIG. 15 is a flowchart showing a processing procedure for state estimation of the extruder 1 according to Embodiment 4. The processing unit 51 of the state estimation device 5 acquires physical quantity data and set values related to the state of the extruder 1 from the data collection device 3 (step S431).

[0085] Next, the processing unit 51 discriminates the driving situation of the extruder 1 using the driving situation discrimination model 59 (step S432). Specifically, the processing unit 51 inputs the acquired physical quantity data into the driving situation discrimination model 59, and discriminates the current driving situation of the extruder 1 based on the data output from the driving situation discrimination model 59.

[0086] Then, the processing unit 51 selects the first to third autoencoders 54, 55, and 56 according to the driving situation discriminated in step S432, and uses the selected autoencoder to determine the presence or absence of an abnormality in the extruder 1 or calculate the degree of abnormality (step S433). The method for abnormality determination using the first to third autoencoders 54, 55, and 56 is the same as that in Embodiment 1. Then, the processing unit 51, in the same manner as in Embodiment 1, synthesizes the determination results obtained using each autoencoder to calculate the presence or absence of an abnormality or the degree of abnormality of the extruder 1 (step S434), and ends the processing.

[0087] As described above, according to the state estimation method and the like according to the present Embodiment 4, an autoencoder corresponding to the driving situation of the extruder 1 can be selected, and the presence or absence of an abnormality in the extruder 1 can be determined, or the degree of abnormality can be calculated.

[0088] Means for solving the problems of the present disclosure are appended. (Appendix 1) A state estimation method for determining the state of an extruder by obtaining physical quantity data related to the state of the extruder and inputting the obtained physical quantity data or a feature quantity based on the physical quantity data into a machine learning model that outputs information indicating the state of the extruder when the physical quantity data or the feature quantity based on the physical quantity data is input. (Appendix 2) The machine learning model is a model that outputs information indicating the state of the extruder when the physical quantity data and data indicating the set value of the extruder or the static characteristics of the extruder are input, or when a feature quantity based on the physical quantity data and data indicating the set value of the extruder or the static characteristics of the extruder is input. The physical quantity data related to the state of the extruder and data indicating the set value of the extruder or the static characteristics of the extruder are obtained, and the obtained physical quantity data and data indicating the set value of the extruder or the static characteristics of the extruder are input, or a feature quantity based on the obtained physical quantity data and data indicating the set value of the extruder or the static characteristics of the extruder is input into the machine learning model, so as to determine the state of the extruder according to the state estimation method described in Appendix 1. (Appendix 3) The machine learning model includes an autoencoder that reproduces and outputs the physical quantity data or the feature quantity obtained from the extruder in a predetermined state when the physical quantity data or a feature quantity based on the physical quantity data is input. The obtained physical quantity data or the feature quantity is input into the autoencoder, and it is determined whether the extruder is in the predetermined state by comparing the obtained physical quantity data or the feature quantity with the physical quantity data or the feature quantity output from the autoencoder according to the state estimation method described in Appendix 1 or Appendix 2. (Appendix 4) The state of the extruder includes an abnormal state of the extruder, a normal state of the extruder, or an operating state of the extruder according to the state estimation method described in any one of Appendix 1 to Appendix 3. (Appendix 5) The feature quantity is calculated based on multiple types of the physical quantity data according to the state estimation method described in any one of Appendix 1 to Appendix 4. (Appendix 6) The extruder has a motor and a screw, and calculates the characteristic quantity based on the rotational speed and torque of the motor or the screw. The state estimation method according to any one of Appendices 1 to 5. (Appendix 7) The extruder has a motor and a screw, and calculates the characteristic quantity based on the power or current of the motor and the torque. The state estimation method according to any one of Appendices 1 to 6. (Appendix 8) Obtain the set value of the extruder related to the physical quantity data, and calculate the characteristic quantity based on the physical quantity data and the set value related to the physical quantity data. The state estimation method according to any one of Appendices 1 to 7. (Appendix 9) The extruder has a screw, and the related physical quantity data and set value include a combination of a raw material supply amount and a raw material supply amount set value, or a combination of a screw speed and a screw speed set value. The state estimation method according to any one of Appendices 1 to 8. (Appendix 10) Use a plurality of types of the machine learning models to respectively determine the state of the extruder, and comprehensively determine the presence or absence or degree of abnormality of the extruder by comprehensively combining each determination result. The state estimation method according to any one of Appendices 1 to 9.

Explanation of Signs

[0089] 1: Extruder 2: Sensor 3: Data collection device 4: Router 5: State estimation device 6: Terminal device 10: Cylinder 11: Screw 12: Die 13: Motor 14: Reducer 15: Control device 50: Recording medium 51: Processing unit 52: Storage unit 53: Communication unit 54: First autoencoder 55: Second autoencoder 56: Third Autoencoder 57: Collected Data DB P: Computer Program

Claims

1. Obtaining physical quantity data related to the state of an extruder, When the physical quantity data or a feature quantity based on the physical quantity data is input, determining the state of the extruder by inputting the obtained physical quantity data or the feature quantity based on the physical quantity data into a machine learning model that outputs information indicating the state of the extruder State estimation method.

2. The machine learning model is A model that outputs information indicating the state of the extruder when the physical quantity data and data indicating the set value of the extruder or the static characteristics of the extruder are input, or when a feature quantity based on the physical quantity data and data indicating the set value of the extruder or the static characteristics of the extruder is input, Obtaining the physical quantity data related to the state of the extruder and data indicating the set value of the extruder or the static characteristics of the extruder, Inputting the obtained physical quantity data and data indicating the set value of the extruder or the static characteristics of the extruder, or inputting a feature quantity based on the obtained physical quantity data and data indicating the set value of the extruder or the static characteristics of the extruder into the machine learning model to determine the state of the extruder The state estimation method according to claim 1.

3. The machine learning model is Including an autoencoder that, when the physical quantity data or a feature quantity based on the physical quantity data is input, reproduces and outputs the physical quantity data or the feature quantity obtained from the extruder in a predetermined state, Inputting the obtained physical quantity data or the feature quantity into the autoencoder, Determining whether the extruder is in the predetermined state by comparing the obtained physical quantity data or the feature quantity with the physical quantity data or the feature quantity output from the autoencoder The state estimation method according to claim 1.

4. The state of the extruder includes an abnormal state of the extruder, a normal state of the extruder, or an operating state of the extruder The state estimation method according to claim 1.

5. Calculating the feature quantity based on a plurality of types of the physical quantity data The state estimation method according to claim 1.

6. The extruder has a motor and a screw, Calculating the feature quantity based on the rotational speed and torque of the motor or the screw The state estimation method according to claim 5.

7. The extruder has a motor and a screw, Calculating the feature quantity based on the power or current of the motor and torque The state estimation method according to claim 5.

8. Obtain the set value of the extruder related to the physical quantity data, Calculate the feature quantity based on the physical quantity data and the set value related to the physical quantity data The state estimation method according to claim 1.

9. The extruder has a screw, The related physical quantity data and the set value include a combination of a raw material supply amount and a raw material supply amount set value, or a combination of a screw speed and a screw speed set value The state estimation method according to claim 8, including the above.

10. Use a plurality of types of the machine learning models to respectively determine the state of the extruder, and comprehensively determine the presence or absence or degree of abnormality of the extruder by comprehensively combining each determination result The state estimation method according to any one of claims 1 to 9.

11. Obtain physical quantity data related to the state of the extruder, A state estimation device including a processing unit that executes a process of determining the state of the extruder by inputting the obtained physical quantity data or a feature quantity based on the physical quantity data into a machine learning model that outputs information indicating the state of the extruder when the physical quantity data or a feature quantity based on the physical quantity data is input.

12. Obtain physical quantity data related to the state of the extruder, When the physical quantity data or a feature quantity based on the physical quantity data is input, determine the state of the extruder by inputting the obtained physical quantity data or a feature quantity based on the physical quantity data into a machine learning model that outputs information indicating the state of the extruder A computer program for causing a computer to execute the process.

13. Obtain physical quantity data related to the state of the extruder, Generate a machine learning model that outputs information indicating the state of the extruder when the physical quantity data or a feature quantity based on the physical quantity data is input, based on the obtained physical quantity data Machine learning method.

Images