Sensor system, host unit, prediction device and prediction method

Abstract

The present invention relates to a sensor system, a host unit, a prediction device and a prediction method, which can detect the abnormality or abnormality precursor of a workpiece at an early stage. The sensor system (1) includes: a first sensor (30a) for measuring the workpiece; a second sensor (30b) for measuring the workpiece at a relatively longer period than the first sensor (30a); and a host unit (10), the host unit (10) including: an acquisition unit (11) for acquiring data measured by the first sensor (30a) and data measured by the second sensor (30b); and a generation unit (12) for generating learning data, the learning data being used for machine learning of a learning model, using the acquired data of the first sensor (30a) as input data and the acquired data of the second sensor (30b) as label data representing the nature of the input data.

Description

Technical Field

[0001] The present disclosure relates to a sensor system, a host unit, a prediction device and a prediction method. Background Art

[0002] In the past, multiple sensors were sometimes arranged along a line to detect the presence or absence of workpieces being transported on the line. The data measured by the multiple sensors were sometimes acquired by multiple slave units and forwarded to a master unit, and then collected in a control device such as a programmable logic controller (PLC) connected to the master unit.

[0003] Patent document 1 described below describes a sensor system, which includes a plurality of sensor units and a communication device that sends information received from each sensor unit to a control device. Each sensor unit uses a synchronization signal sent from any sensor unit as a starting point, and after a standby time specified for each sensor unit has passed, sends detection information such as sensor data to the communication device. Here, the standby time of each sensor unit is specified to be different from the standby time of other sensor units. Through the technology described in patent document 1, when data measured by a plurality of sensors are collected in a control device, data can be sent without waiting for a command from the control device, thereby increasing the communication speed.

[0004] Prior art literature

[0005] Patent Literature

[0006] Patent Document 1: Japanese Patent Application Publication No. 2014-96036 Summary of the invention

[0007] Problems to be solved by the invention

[0008] In recent years, research has been conducted on building a sensor system that uses data measured by a plurality of sensors for machine learning of a learning model to generate a learned model and makes a more advanced determination using the learned model.

[0009] Conventionally, a sensor system using a learned model has been proposed which detects an abnormality or a precursor or sign of an abnormality of a workpiece conveyed on the line using a learned model generated based on data measured by a plurality of sensors arranged on the line.

[0010] However, in such a sensor system, it is desired to detect abnormality of a workpiece or a sign or symptom of abnormality as early as possible to suppress the manufacture or production of the workpiece having the abnormality.

[0011] Therefore, one of the objects of the present invention is to provide a sensor system, a host unit, a prediction device, and a prediction method that can detect an abnormality or a sign of an abnormality of a workpiece at an early stage.

[0012] Technical means of solving problems

[0013] A sensor system of one embodiment of the present invention includes: a first sensor for measuring a workpiece; a second sensor for measuring the workpiece at a period longer than that of the first sensor; and a host unit, the host unit including: an acquisition unit for acquiring data measured by the first sensor and data measured by the second sensor; and a generation unit for generating learning data, wherein the learning data is used for machine learning of a learning model, using the data acquired by the first sensor as input data, and using the data acquired by the second sensor as label data representing the nature of the input data.

[0014] According to this scheme, learning data is generated, and the learning data is used for machine learning of a learning model, using the acquired data of the first sensor as input data, and the acquired data of the second sensor as label data indicating the nature of the input data. Thus, the learned model generated using the learning data can output a value (predicted value) using the data of the first sensor whose measurement cycle is shorter than that of the second sensor as input. Therefore, by using the learned model, it is possible to detect abnormalities or signs of abnormalities of a workpiece at an earlier stage than before.

[0015] In the above scheme, the generation unit may generate learning data by associating input data with label data based on a time difference calculated from a moving speed of the workpiece and a distance between the first sensor and the second sensor, a measurement period of the first sensor, and a measurement period of the second sensor.

[0016] According to this scheme, input data and label data are associated with each other based on the time difference, the measurement cycle of the first sensor and the measurement cycle of the second sensor to generate learning data. Thus, data measured on the same or similar workpieces are associated with each other to generate learning data, thereby improving the prediction accuracy of the learned model.

[0017] In the aforementioned aspect, the first sensor may be arranged on the upstream side of the second sensor in the line along which the workpiece moves.

[0018] According to this scheme, the first sensor is arranged upstream of the second sensor in the line where the workpiece moves. Thus, compared with the case where the first sensor is arranged downstream of the second sensor, the generated learned model uses data measured on the workpiece at a relatively early stage on the line as input data, thereby being able to predict abnormalities or signs of abnormalities in the workpiece at an early stage.

[0019] In the aforementioned aspect, the host unit may further include a learning unit that performs machine learning of the learning model using the learning data to generate a learned model.

[0020] According to this aspect, machine learning of the learning model is performed using learning data to generate a learned model. Thus, a learned model that can detect abnormality or abnormality signs of a workpiece at an early stage can be easily generated.

[0021] In the aforementioned solution, the host unit may also include a prediction unit, which inputs the acquired data of the first sensor into the learned model, so that the learned model outputs a predicted value.

[0022] According to this aspect, the acquired data of the first sensor is inputted, and the learned model is made to output a predicted value, thereby making it possible to predict an abnormality or a sign of abnormality of the workpiece at an early stage based on the predicted value.

[0023] The aforementioned scheme may also include a plurality of first sensors, and the host unit may further include a selection unit, which calculates a correlation coefficient between the data acquired from the first sensor and the data acquired from the second sensor with respect to one of the plurality of first sensors.

[0024] According to this aspect, for one of the plurality of first sensors, a correlation coefficient between data acquired by the first sensor and data acquired by the second sensor is calculated, thereby enabling selection of a first sensor among the plurality of first sensors that measures data that has a linear relationship or a nearly linear relationship with data from the second sensor.

[0025] The aforementioned scheme may also include multiple first sensors, and a generating unit generates learning data using data acquired from at least one of the multiple first sensors as input data. The host unit also includes a selecting unit, which calculates a learning progress value representing the proportion of learning progress of the learned model based on the data acquired from the second sensor and a predicted value output by inputting the input data into a learned model generated by performing machine learning of the learning model using the learning data.

[0026] According to this scheme, a learning progress value is calculated based on the acquired data of the second sensor and the predicted value output by inputting the input data to the learned model generated by performing machine learning of the learning model using the learning data. Thus, by selecting at least one from the plurality of first sensors based on the learning progress value, a first sensor can be selected in which the predicted value of the learned model generated by the data of the first sensor is close to the value of the data of the second sensor.

[0027] A host unit of another embodiment of the present invention is used for a sensor system, which includes a first sensor for measuring a workpiece and a second sensor for measuring the workpiece at a period longer than that of the first sensor. The host unit includes: an acquisition unit for acquiring data measured by the first sensor and data measured by the second sensor; and a generation unit for generating learning data, wherein the learning data is used for machine learning of a learning model, using the data acquired by the first sensor as input data and the data acquired by the second sensor as label data representing the nature of the input data.

[0028] According to this scheme, learning data is generated, and the learning data is used for machine learning of a learning model, using the acquired data of the first sensor as input data, and the acquired data of the second sensor as label data indicating the nature of the input data. Thus, the learned model generated using the learning data can output a value (predicted value) using the data of the first sensor whose measurement cycle is shorter than that of the second sensor as input. Therefore, by using the learned model, it is possible to detect abnormalities or signs of abnormalities in a workpiece at an earlier stage than before.

[0029] In the above scheme, the generation unit may generate learning data by associating input data with label data based on a time difference calculated from a moving speed of the workpiece and a distance between the first sensor and the second sensor, a measurement period of the first sensor, and a measurement period of the second sensor.

[0030] According to this scheme, input data and label data are associated based on the time difference, the measurement cycle of the first sensor and the measurement cycle of the second sensor to generate learning data. Thus, data measured on the same or similar workpieces are associated with each other to generate learning data, thereby improving the prediction accuracy of the learned model.

[0031] The aforementioned scheme may further include a learning unit that performs machine learning of the learning model using the learning data to generate a learned model.

[0032] According to this aspect, machine learning of the learning model is performed using learning data to generate a learned model. Thus, a learned model that can detect abnormality or abnormality signs of a workpiece at an early stage can be easily generated.

[0033] The aforementioned scheme may also include a prediction unit, which inputs the acquired data of the first sensor into the learned model, so that the learned model outputs a predicted value.

[0034] According to this aspect, the acquired data of the first sensor is inputted, and the learned model is made to output a predicted value, thereby making it possible to predict an abnormality or a sign of abnormality of the workpiece at an early stage based on the predicted value.

[0035] In the aforementioned scheme, the sensor system may also include a plurality of first sensors, and the host unit may further include a selection unit, which calculates a correlation coefficient between the data acquired from the first sensor and the data acquired from the second sensor with respect to one of the plurality of first sensors.

[0036] According to this aspect, for one of the plurality of first sensors, a correlation coefficient between data acquired by the first sensor and data acquired by the second sensor is calculated, thereby enabling selection of a first sensor among the plurality of first sensors that measures data that has a linear relationship or a nearly linear relationship with data from the second sensor.

[0037] In the aforementioned scheme, the sensor system may also include multiple first sensors, and the generating unit generates learning data using data acquired from at least one of the multiple first sensors as input data. The host unit also includes a selecting unit, which calculates a learning progress value representing the proportion of learning progress of the learned model based on the data acquired from the second sensor and a predicted value output by inputting the input data into a learned model generated by performing machine learning of the learning model using the learning data.

[0038] According to this scheme, a learning progress value is calculated based on the acquired data of the second sensor and the predicted value output by inputting the input data to the learned model generated by executing machine learning of the learning model using the learning data. Thus, by selecting at least one from the plurality of first sensors based on the learning progress value, a first sensor can be selected in which the predicted value of the learned model generated by the data of the first sensor is close to the value of the data of the second sensor.

[0039] A prediction device according to another embodiment of the present invention predicts an abnormality or a sign of an abnormality of a workpiece, the prediction device comprising: an acquisition unit, which acquires data measured by a first sensor that measures the workpiece; and a prediction unit, which inputs the acquired data of the first sensor into a learned model, so that the learned model outputs a predicted value, wherein the learned model is generated by performing machine learning of the learning model using learning data, wherein the learning data is generated using the data of the first sensor as input data and using the data of a second sensor that measures the workpiece at a period longer than that of the first sensor as label data representing the nature of the input data.

[0040] According to this scheme, the acquired data of the first sensor is input into the learned model, and the learned model outputs a predicted value. Here, the learned model is generated using the data of the first sensor as input data and the data of the second sensor as label data indicating the nature of the input data, so that the data of the first sensor having a shorter measurement period than the second sensor can be used as input to output a predicted value. Therefore, the abnormality or abnormality sign of the workpiece can be predicted at an early stage based on the predicted value.

[0041] A prediction method according to another embodiment of the present invention predicts an abnormality or a sign of an abnormality of a workpiece, the prediction method comprising the following steps: acquiring data measured by a first sensor that measures the workpiece; and inputting the acquired data of the first sensor into a learned model so that the learned model outputs a predicted value, the learned model being generated by performing machine learning of the learning model using learning data, the learning data being generated by using the data of the first sensor as input data and using the data of the second sensor that measures the workpiece at a period longer than that of the first sensor as label data representing the nature of the input data.

[0042] According to this scheme, the acquired data of the first sensor is input into the learned model, and the learned model outputs a predicted value. Here, the learned model is generated using the data of the first sensor as input data and the data of the second sensor as label data indicating the nature of the input data, so that the data of the first sensor having a shorter measurement period than the second sensor can be used as input to output a predicted value. Therefore, the abnormality or abnormality sign of the workpiece can be predicted at an early stage based on the predicted value.

[0043] Effects of the Invention

[0044] According to the present invention, it is possible to detect an abnormality or a sign of an abnormality of a workpiece at an early stage. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 This is a structural diagram illustrating a schematic structure of an optical measurement device according to one embodiment.

[0046] Figure 2 This is a structural diagram illustrating the physical structure of a master unit and a slave unit in one embodiment.

[0047] Figure 3 This is a structural diagram illustrating the structure of functional blocks of a host unit in one embodiment.

[0048] Figure 4 This is a structural diagram illustrating a schematic structure of a first example of a wire in one embodiment.

[0049] Figure 5 This is a flowchart illustrating a schematic operation of a setting mode process of a master unit in one embodiment.

[0050] Figure 6 This is a flowchart illustrating a schematic operation of a prediction learning process of a host unit in one embodiment.

[0051] Figure 7 This is a conceptual diagram for explaining the association between input data and label data by the generation unit.

[0052] Figure 8 This is a flowchart illustrating a schematic operation of the selection learning process of the master unit in one embodiment.

[0053] Fig. 9 This is a flowchart illustrating a schematic operation of the first sensor selection mode process of the host unit in one embodiment.

[0054] Fig.10 This is a flowchart illustrating the schematic operation of the prediction mode processing of the master unit in one embodiment.

[0055] Fig.11 This is a structural diagram illustrating a schematic structure of a second example of the wire in one embodiment.

[0056] [Explanation of Symbols]

[0057] 1: Sensor system

[0058] 10: Host unit

[0059] 11: Get Department

[0060] 12: Generation Department

[0061] 13: Storage

[0062] 13a: Learning with Data

[0063] 13b: Learning the model

[0064] 14: Learning Department

[0065] 15: Selected Department

[0066] 16: Forecasting Department

[0067] 17: Ministry of Communications

[0068] 18: Display unit

[0069] 20: Slave unit

[0070] 20a: First slave unit

[0071] 20b: Second slave unit

[0072] 30: Sensor

[0073] 30a: First sensor

[0074] 30b: Second sensor

[0075] d: distance

[0076] L, L10, L20: Line

[0077] L11: Hopper

[0078] L12: Heating cylinder

[0079] L13: Screw

[0080] L14: Heater

[0081] L15: Mould

[0082] L16: Cooling device

[0083] L17: Access Device

[0084] L18: Cut-off device

[0085] MA: Materials

[0086] S200: Setting mode processing

[0087] S220: Prediction Learning Process

[0088] S240: Select learning process

[0089] S260: First sensor selection mode processing

[0090] S280: Prediction mode processing

[0091] v: speed

[0092] W, W10, W21, W22: Workpiece

[0093] Δt: time difference DETAILED DESCRIPTION

[0094] The following describes the embodiments of the present invention. In the following description of the drawings, the same or similar parts are represented by the same or similar symbols. However, the drawings are schematic. Therefore, specific dimensions and the like should be judged in accordance with the following description. Moreover, the drawings also include parts with different dimensional relationships or ratios. Furthermore, the technical scope of the present invention should not be limited to the embodiments described.

[0095] First, refer to Figure 1 , while explaining the structure of a sensor system according to one embodiment. Figure 1 It is a structural diagram illustrating a schematic structure of a sensor system 1 in one embodiment.

[0096] like Figure 1 As shown, the sensor system 1 includes, for example, a master unit 10, a first slave unit 20a, a second slave unit 20b, a first sensor 30a, a second sensor 30b, and a PLC 40. The master unit 10 of this embodiment also corresponds to an example of a "prediction device".

[0097] The first sensor 30a and the second sensor 30b are arranged along the line L. On the line L, Figure 1 The workpiece W is transported from left to right (from the front to the back of the paper) in the drawing. The first sensor 30a and the second sensor 30b measure data related to the workpiece W transported on the line L, for example, data indicating the passing condition. The measurement cycles of the first sensor 30a and the second sensor 30b are different from each other, and the second sensor 30b measures the workpiece W in a longer cycle than the first sensor 30a. That is, the first sensor 30a measures the workpiece W in a shorter cycle than the second sensor 30b.

[0098] In addition, the line L is not limited to Figure 1 The example shown. The line L can be any line as long as the workpiece W moves. For example, the line L is a conveying line for conveying the workpiece W, a manufacturing line for manufacturing the workpiece W, a production line for producing the workpiece W, etc., and the type thereof is not limited. Furthermore, the workpiece W is not limited to the final product, and may be, for example, an intermediate product, a semi-finished product, a part, a material, etc.

[0099] The first slave unit 20a is connected to the first sensor 30a, and the second slave unit 20b is connected to the second sensor 30b. Furthermore, the master unit 10 is connected to the first slave unit 20a, the second slave unit 20b, and the PLC 40. In this specification, the first slave unit 20a and the second slave unit 20b are collectively referred to as slave units 20, and the first sensor 30a and the second sensor 30b are collectively referred to as sensors 30.

[0100] In addition, in this embodiment, an example is shown in which the sensor system 1 includes a first sensor 30a, a second sensor 30b, and two slave units, but the present invention is not limited thereto. The number of first sensors, the number of second sensors, and the number of slave units included in the sensor system 1 are arbitrary and may be changed as appropriate. Furthermore, the sensor system 1 may not necessarily include the PLC 40.

[0101] The host unit 10 is connected to the PLC 40 via a communication network such as a local area network (LAN). The slave unit 20 is physically and electrically connected to the host unit 10. In the present embodiment, the host unit 10 stores information received from the slave unit 20 in a storage unit, and sends the stored information to the PLC 40. Therefore, the data acquired by the slave unit 20 is collected and transmitted to the PLC 40 through the host unit 10.

[0102] Specifically, the slave unit 20 transmits a determination signal and detection information to the master unit 10. The so-called determination signal refers to, for example, a signal indicating a determination result related to the workpiece determined by the second slave unit 20b based on the data measured by the second sensor 30b. For example, when the second sensor 30b is a photoelectric sensor, the determination signal is an on signal or an off signal obtained by the second slave unit 20b comparing the amount of light received measured by the second sensor 30b with a threshold value. The detection information is, for example, a detection value obtained by the detection action of the first slave unit 20a. For example, when the first sensor 30a is a photoelectric sensor, the detection action is the action of projecting light and receiving light, and the detection information is the amount of light received.

[0103] The slave unit 20 is installed on the side of the master unit 10. For the communication between the master unit 10 and the slave unit 20, parallel communication or serial communication is used. That is, the master unit 10 and the slave unit 20 are physically connected using a serial transmission path and a parallel transmission path. For example, a determination signal may be sent from the slave unit 20 to the master unit 10 on the parallel transmission path, and detection information may be sent from the slave unit 20 to the master unit 10 on the serial transmission path. In addition, the master unit 10 and the slave unit 20 may be connected using either a serial transmission path or a parallel transmission path.

[0104] Next, refer to Figure 2 , while describing the physical structure of a master unit and a slave unit according to one embodiment. Figure 2 1 is a structural diagram illustrating the physical structure of the master unit 10 and the slave unit 20 in one embodiment.

[0105] like Figure 2 As shown, the master unit 10 includes: input / output connectors 101, 102 used for connection with the PLC 40; a connector 106 used for connection with the slave unit 20; and a power input connector (not shown).

[0106] Furthermore, the host unit 10 includes a microprocessor (MPU) 110, an application specific integrated circuit (ASIC) 112, a parallel communication circuit 116, a serial communication circuit 118, a flash read only memory (Flash ROM) 120, a display device 122, and a power supply circuit (not shown).

[0107] The MPU 110 operates to collectively execute all processing in the host unit 10. The communication ASIC 112 manages communication with the PLC 40. The parallel communication circuit 116 is used for parallel communication between the host unit 10 and the slave unit 20. Similarly, the serial communication circuit 118 is used for serial communication between the host unit 10 and the slave unit 20. The flash ROM 120 is a non-volatile memory that stores a learning model. For example, in the case where the learning model is a neural network, the flash ROM 120 can store weight parameters or a network structure of the neural network. Moreover, in the case where the learning model is a regression model or a decision tree, the flash ROM 120 can store regression parameters or hyperparameters of the decision tree. The display device 122 is a display such as an organic electroluminescence (EL) that displays text information or status.

[0108] The slave unit 20 is provided with connectors 304, 306 for connecting to the master unit 10 or other slave units 20 on both side walls. A plurality of slave units 20 can be connected in a row relative to the master unit 10. Signals from a plurality of slave units 20 are transmitted to adjacent slave units 20 and transmitted to the master unit 10.

[0109] Windows for optical communication based on infrared rays are provided on both sides of the slave unit 20. When a plurality of slave units 20 are connected one by one and arranged in a row using connectors 304 and 306, bidirectional optical communication using infrared rays can be realized between adjacent slave units 20 through the windows for optical communication facing each other.

[0110] The slave unit 20 includes various processing functions implemented by a central processing unit (CPU) 400 and various processing functions implemented by dedicated circuits.

[0111] The CPU 400 controls the light projection control unit 403 to emit infrared rays from the light emitting element (light emitting diode (LED)) 401. The signal generated by the light receiving element (photodiode (PD)) 402 receiving light is amplified by the amplifier circuit 404, converted into a digital signal by the analog / digital (A / D) converter 405, and introduced into the CPU 400. In the CPU 400, the light receiving data, that is, the amount of light received, is directly sent to the host unit 10 as detection information. In addition, in the CPU 400, the on signal or the off signal obtained by judging whether the amount of light received is greater than a preset threshold value is sent to the host unit 10 as a judgment signal.

[0112] Furthermore, the CPU 400 controls the left and right light-emitting circuits 411 and 413, thereby emitting infrared rays from the left and right communication light-emitting elements (LEDs) 407 and 409 to the adjacent slave units 20. The infrared rays coming from the adjacent left and right slave units 20 are received by the left and right light-receiving elements (PD) 406 and 408, and then reach the CPU 400 via the light-receiving circuits 410 and 412. In the CPU 400, the transmission and reception of signals are controlled based on a predetermined protocol, thereby performing optical communication between the adjacent left and right slave units 20.

[0113] The light receiving element 406, the communication light emitting element 409, the light receiving circuit 410, and the light projecting circuit 413 are used to send and receive synchronization signals for preventing mutual interference between the slave units 20. Specifically, in each slave unit 20, the light receiving circuit 410 is directly connected to the light projecting circuit 413. With this structure, the received synchronization signal is quickly sent from the communication light emitting element 409 to another adjacent slave unit 20 via the light projecting circuit 413 without being delayed by the CPU 400.

[0114] The CPU 400 further controls the lighting of the display 414. In addition, the CPU 400 processes the signal from the setting switch 415. Various data required for the operation of the CPU 400 are stored in a recording medium such as an electrically erasable programmable read only memory (EEPROM) 416. The signal obtained from the reset unit 417 is sent to the CPU 400 to reset the measurement control. The reference clock is input from the oscillator (OSC) 418 to the CPU 400.

[0115] The output circuit 419 performs a process of transmitting a determination signal obtained by comparing the amount of received light with a threshold value. As described above, in this embodiment, the determination signal is transmitted to the host unit 10 by parallel communication.

[0116] The transmission path for parallel communication is a transmission path that connects the master unit 10 and each slave unit 20 individually. That is, the plurality of slave units 20 are connected to the master unit 10 through their own parallel communication lines. However, the parallel communication line that connects the slave units 20 other than the slave units 20 adjacent to the master unit 10 to the master unit 10 may pass through other slave units 20.

[0117] The serial communication driver 420 performs reception processing of commands and the like sent from the host unit 10 and transmission processing of detection information (received light amount). In this embodiment, the RS-422 protocol is used for serial communication. The RS-485 protocol may also be used for serial communication.

[0118] The transmission path for serial communication is a transmission path connecting the master unit 10 and all the slave units 20. That is, all the slave units 20 are connected to the master unit 10 via serial communication lines so that signals can be transmitted in a bus format.

[0119] Next, refer to Figure 3 , while explaining the structure of the functional blocks of the host unit according to one embodiment. Figure 3 1 is a block diagram illustrating a configuration of functional blocks of the host unit 10 in one embodiment.

[0120] like Figure 3 As shown, the host unit 10 includes an acquisition section 11 , a generation section 12 , a storage section 13 , a learning section 14 , a selection section 15 , a prediction section 16 , a communication section 17 , and a display section 18 as functional blocks.

[0121] The acquisition unit 11 is configured to acquire data measured by the first sensor 30a and data measured by the second sensor 30b via the slave unit 20. Specifically, the acquisition unit 11 acquires detection information measured by the plurality of sensors 30 from the slave unit 20 through a serial transmission path.

[0122] The generating unit 12 is configured to generate learning data 13a used for machine learning of the learning model. The learning data 13a is data used for teaching learning of the learning model, and includes input data and label data. Here, the input data refers to data input to the learning model during machine learning of the learning model. The input data may be numerical data, but may also be data in other formats. The label data indicates the properties of the input data. The so-called properties of the input data refer to the properties predicted from the input data, such as the presence or absence of an abnormality or a sign of an abnormality of the workpiece W conveyed on the line L, or the type of workpiece W, or the size of the workpiece W, or the position deviation of the workpiece W. The label data is the data that the learning model should output during the machine learning of the learning model, and is the data set as the learning target. The label data may be numerical data, but may also be data in other formats.

[0123] More specifically, the generator 12 is configured to set the acquired data of the first sensor 30a as input data of the learning model, and set the acquired data of the second sensor 30b as label data used in the teaching learning of the learning model, and generate the learning data 13a including the input data and the label data. By generating the learning data 13a using the acquired data of the first sensor 30a as input data and the acquired data of the second sensor 30b as label data in this way, the learned model generated using the learning data 13a can output a value (predicted value) using the data of the first sensor 30a having a relatively short measurement cycle as input. Therefore, by using the learned model, it is possible to detect an abnormality or a sign of an abnormality of the workpiece W at an earlier stage than before.

[0124] The storage unit 13 stores the learning data 13 a and the learned model 13 b generated by the generation unit 12 .

[0125] The learning unit 14 is configured to perform machine learning of the learning model using the learning data 13a to generate a learned model 13b. For example, when the learning model is a neural network, the learning unit 14 may input the input data of the learning data 13a to the neural network, and update the weight of the neural network by the error back propagation method based on the difference between its output and the label data. In addition, the learning model is not limited to a neural network, and may be a regression model or a decision tree. The learning unit 14 may perform machine learning of the learning model using any algorithm. By using the learning data 13a to perform machine learning of the learning model to generate the learned model 13b, it is possible to easily generate the learned model 13b that detects abnormalities or abnormal signs of the workpiece W at an early stage.

[0126] The selection unit 15 is used to select one or more first sensors 30a from the plurality of first sensors 30a. Here, when a plurality of first sensors 30a are provided on the line L, there is a situation where the data to be used as input data is limited in order to improve the prediction accuracy of the learned model. Therefore, the selection unit 15 calculates a value that serves as an index when selecting the data of the first sensor 30a, and selects one or more first sensors 30a based on the value, or notifies the user of the value to select one or more first sensors 30a.

[0127] In more detail, the selection unit 15 is configured to calculate the correlation coefficient between the data of the first sensor 30a obtained and the data of the second sensor 30b obtained for one of the plurality of first sensors 30a. Generally speaking, there is a correlation between the data measured by the two sensors 30. Therefore, when the absolute value of the correlation coefficient between the data of the first sensor 30a and the data of the second sensor 30b is greater than a predetermined value, the data of the first sensor 30a can be used as input data. In addition, the data of the first sensor 30a whose absolute value among the correlation coefficients between the data of the first sensor 30a and the data of the second sensor 30b is the largest can be used as input data. By calculating the correlation coefficient between the data of the first sensor 30a obtained and the data of the second sensor 30b obtained in this way, it is possible to select the first sensor 30a among the plurality of first sensors 30a that measures data that has a linear relationship or a nearly linear relationship with the data of the second sensor 30b.

[0128] Furthermore, the selection unit 15 is configured to calculate a learning progress value based on the data of the second sensor 30b obtained and the predicted value output by inputting the input data to the learning completion model 13b, wherein the learning completion model 13b is generated by performing machine learning of the learning model using the learning data 13a. Here, the learning data 13a used by the selection unit 15 to calculate the learning progress value is data obtained from at least one of the plurality of first sensors 30a as input data and generated by the generation unit 12. The selection unit 15 generates the learning completion model 13b using the learning data 13a, and calculates a learning progress value indicating the proportion of the learning progress of the learning completion model 13b based on the predicted value output by inputting the aforementioned input data to the generated learning completion model 13b. In addition, the details of the learning progress value will be described later.

[0129] The prediction unit 16 is configured to input the acquired data of the first sensor 30a into the learned model 13b, so that the learned model 13b outputs a predicted value. In addition, the prediction unit 16 is not limited to the case where the output of the learned model 13b is directly used as a predicted value. For example, the prediction unit 16 may perform any post-processing on the output of the learned model 13b to use it as a predicted value. By inputting the data of the first sensor 30a in this way and causing the learned model 13b to output a predicted value, it is possible to predict an abnormality or a sign of an abnormality of the workpiece W at an early stage based on the predicted value.

[0130] The communication unit 17 is an interface for communicating with the PLC 40. The communication unit 17 may be an interface for communicating with an external device other than the PLC 40.

[0131] The display unit 18 is used to display text information or status to notify the user. The display unit 18 may display objects such as numerical data such as predicted values ​​and learning progress rates and their definitions, determination results, predictable notifications, current modes and other statuses, and other settings of the host unit 10.

[0132] In this embodiment, the host unit 10 includes Figure 3 The example of the functional block shown is explained, but it is not limited to this. For example, when the host unit 10 functions as a prediction device for predicting an abnormality or abnormal sign of the workpiece W, the host unit 10 includes: an acquisition unit 11, which acquires data measured by the first sensor 30a; and a prediction unit 16, which inputs the acquired data of the first sensor 30a into the learned model 13b, so that the learned model 13b outputs a predicted value. Thus, the acquired data of the first sensor 30a is input into the learned model 13b, so that the learned model 13b outputs a predicted value. Here, the learned model 13b is generated using the learning data 13a generated by using the data of the first sensor 30a as input data and the data of the second sensor 30b as label data indicating the nature of the input data, so the data of the first sensor 30a having a shorter measurement period than the second sensor 30b can be used as input to output a predicted value. Therefore, the abnormality or abnormal sign of the workpiece W can be predicted at an early stage based on the predicted value.

[0133] In addition, when the host unit 10 is a prediction device that predicts an abnormality or a sign of an abnormality of the workpiece W, the learned model 13b used by the prediction unit 16 and the learning data 13a used for generating the learned model 13b may also be generated by other devices such as external devices. In addition, the sensor unit 10 does not need to include the storage unit 13 that stores the learned model 13b. For example, the learned model 13b may also be stored in other devices such as external devices. The prediction unit 16 sends the acquired data of the first sensor 30a to the other device via the communication unit 17, and receives the predicted value from the other device via the communication unit 17.

[0134] Next, refer to Figure 4 , while describing a first example of a line provided with a first sensor and a second sensor according to one embodiment. Figure 4 This is a structural diagram illustrating a schematic structure of a first example of the line L in one embodiment.

[0135] like Figure 4 As shown, the line L10 provided with the first sensor 30a and the second sensor 30b is used, for example, to heat the material MA while extruding the material MA at a controlled speed to form a workpiece W10. The line L10 includes a hopper L11, a heating cylinder L12, a die L15, a cooling device L16, a receiving device L17, and a cutting device L18.

[0136] The hopper L11 is a container for accommodating the material MA of the workpiece W10. The material MA is supplied from the discharge port to the interior of the heating cylinder L12. The material MA is, for example, a resin. The heating cylinder L12 includes a screw L13 and a heater L14. The heating cylinder L12 extrude the material MA supplied thereto while stirring it with the screw L13 so that the heat of the heater 14 is evenly applied to the material MA. In addition, the extrusion speed of the screw L13 and the temperature of the heater L14 can be fixed or variable.

[0137] The material MA extruded from the heating cylinder L12 is discharged as a workpiece W10 of a specified thickness (thickness) through the die L15. The workpiece W10 is then supplied to the cooling device L16. The cooling device L16 removes the heat from the heater 14 from the workpiece W10 and cools the workpiece W10 to a specified temperature. In addition, the cooling device L16 can be used to cool the workpiece W10 in any manner, for example, it can be air-cooled or water-cooled.

[0138] The workpiece W10 discharged from the cooling device L16 is supplied to the receiving device L17 and then to the cutting device L18. The cutting device L18 cuts the workpiece W10 at a controlled timing, thereby forming the workpiece W10 of a predetermined thickness (thickness) and a predetermined length.

[0139] In such a line L10, for example, the first sensor 30a is arranged at a position between the mold L15 and the cooling device L16, and the second sensor 30b is arranged at a position between the receiving device L17 and the cutting device L18.

[0140] The first sensor 30a on the line L10 is, for example, a transmission type photoelectric sensor, and a light projector and a light receiver are arranged at positions facing each other with the workpiece W10 sandwiched therebetween. The light emitted from the light projector is blocked according to the thickness (thickness) of the workpiece W10, and the light quantity of the unblocked light is measured by the light receiver. The first sensor 30a outputs the measured light receiving quantity as the light receiving quantity data of the workpiece W10. The first sensor 30a can measure the light receiving quantity in a relatively short period, for example, outputting the light receiving quantity data of the workpiece W10 every 10 [μs].

[0141] The second sensor 30b on line L10 is, for example, a laser-type length measuring sensor, and a light projector and a light receiver are arranged at positions facing each other with the workpiece W10 sandwiched therebetween. The laser emitted from the light projector is blocked corresponding to the thickness (roughness) of the workpiece W10, and the thickness (roughness) of the workpiece W10 is measured based on the laser that is not blocked and incident on the light receiver. The second sensor 30b outputs the thickness (roughness) data of the workpiece W10. The resolution of the thickness (roughness) data output by the second sensor 30b is, for example, 10 [μm]. The second sensor 30b can measure the thickness (roughness) of the workpiece W10 at a relatively long period, for example, outputting the thickness (roughness) data of the workpiece W10 every 500 [μs].

[0142] The first sensor 30a is arranged on the upstream side ( Figure 4 Thus, the sensor 30a is arranged on the downstream side ( Figure 4 Compared with the case on the right side of FIG. 1 ), the generated learned model 13b uses data measured on the workpiece W10 at a relatively early stage on the line L10 as input data, so it is possible to predict an abnormality or a sign of an abnormality of the workpiece W10 at an early stage.

[0143] Next, refer to Figures 5 to 10 , while explaining an example of the operation of the host unit according to one embodiment. Figure 5 This is a flowchart illustrating a schematic operation of the setting mode process S200 of the host unit 10 in one embodiment. Figure 6This is a flowchart illustrating the schematic operation of the prediction learning process S220 of the host unit 10 in one embodiment. Figure 7 This is a conceptual diagram for explaining the association between input data and label data by the generating unit 12 . Figure 8 This is a flowchart illustrating the schematic operation of the selection learning process S240 of the host unit 10 in one embodiment. Fig. 9 This is a flowchart illustrating the schematic operation of the first sensor selection mode process S260 of the host unit 10 in one embodiment. Fig.10 This is a flowchart illustrating the schematic operation of the prediction mode processing S280 of the host unit 10 in one embodiment.

[0144] The host unit 10 includes a plurality of modes, for example, a setting mode for performing settings required for executing each mode, a learning mode for generating a learned model, and a prediction mode for performing prediction using a learned model. When a plurality of first sensors 30a are provided on the line L, the host unit 10 may also include a first sensor selection mode. The user can select a mode included in the host unit 10 through operation.

[0145] When the mode is changed by the operation of the user, for example, the host unit 10 executes Figure 5 In the following, except for the cases specifically described, the first sensor 30a and the second sensor 30b are used to set the Figure 4 This is explained using an example on line L10 shown.

[0146] <Setting mode processing>

[0147] like Figure 5 As shown, initially, the host unit 10 determines whether various setting values ​​input by the user's operation have been changed from the current values ​​(S201). The various setting values ​​are, for example, the setting value for the first sensor 30a, the setting value for the second sensor 30b, and the time difference Δt between sensors described later used by the host unit 10, the determination value, the upper threshold, the lower threshold, and the setting of whether to perform additional learning when the learning completion model is generated.

[0148] If the result of determination in step S201 is that any of the various setting values ​​has been changed from the current value, the host unit 10 reflects the changed content of the setting value (S202). After step S202, the host unit 10 determines whether the learning condition has been changed (S203). For example, when at least one of the plurality of first sensors 30a and second sensors 30b is provided, and when a certain first sensor 30a is changed to another first sensor 30a and / or a certain second sensor 30b is changed to another second sensor 30b by setting, it is determined that the learning condition has been changed.

[0149] If the result of determination in step S203 is that the learning condition has changed, the host unit 10 deletes the learned model 13b stored in the storage unit 13 (S204). In addition, the host unit 10 may also send the learned model 13b stored in the storage unit 13 to an external device such as the PLC 40, or write it out to other storage devices for temporary storage, while deleting the learned model 13b, or instead of deleting it.

[0150] If the result of determination in step S201 is that the setting value has not been changed, if the result of determination in step S203 is that the learning condition has not been changed, or after step S204, the host unit 10 determines whether the current mode is the learning mode (S205).

[0151] If the result of determination in step S205 is that the current mode is the learning mode, the host unit 10 performs prediction learning processing S220 and selection learning processing S240 described below. After the prediction learning processing S220 and selection learning processing S240, the host unit 10 ends the setting mode processing S200.

[0152] In addition, the execution of the selection learning process S240 is not limited to the case where it is performed after the prediction learning process S220. The selection learning process S240 may be performed before the prediction learning process S220, or may be performed in parallel with the prediction learning process S220. Furthermore, when there is only one first sensor 30a, or when there are multiple first sensors 30a and all data of the multiple first sensors 30a are used, the host unit 10 may not perform the selection learning process S240.

[0153] On the other hand, if the result of determination in step S205 is that the current mode is not the learning mode, the host unit 10 determines whether the current mode is the first sensor selection mode ( S206 ).

[0154] If the result of determination in step S206 is that the current mode is the first sensor selection mode, the host unit 10 performs the first sensor selection mode process S260 described below. After the first sensor selection mode process S260, the host unit 10 ends the setting mode process S200.

[0155] In addition, when there are multiple first sensors 30a and at least one of the multiple first sensors 30a must be selected, the host unit 10 may also perform at least one of the selected learning process S240 and the first sensor selection mode process S260. In addition, an arbitrary first sensor 30a may be selected from the multiple first sensors 30a by the user's operation. At this time, if the user selects a first sensor 30a different from the previous first sensor 30a, the host unit 10 determines that the learning condition has changed in the determination of step S203.

[0156] On the other hand, if the result of determination in step S206 is that the current mode is not the first sensor selection mode, the host unit 10 determines whether the current mode is the prediction mode ( S207 ).

[0157] If the result of determination in step S207 is that the current mode is the prediction mode, the host unit 10 refers to the storage unit 13 to determine whether there is a learned model 13 b ( S208 ).

[0158] If the result of determination in step S208 is that there is a learned model 13b, the host unit 10 performs a prediction mode process S280 described later. After the prediction mode process S280, the host unit 10 ends the setting mode process S200.

[0159] On the other hand, if the result of determination in step S208 is that there is no learned model 13b, the host unit 10 sends an error signal to the PLC 40 or an external device via the communication unit 17, and displays an error on the display unit 18 to notify the user (user) of the error (S209). After step S209, the host unit 10 ends the setting mode processing S200.

[0160] <Prediction learning processing>

[0161] When the prediction learning process S220 is started, Figure 6 As shown, the acquisition unit 11 acquires data from the sensor 30 via the slave unit 20 ( S221 ).

[0162] Next, the generator 12 determines whether any of the acquired data has been updated ( S222 ).

[0163] If the result of determination in step S222 is that any of the acquired data is updated, the generator 12 generates learning data 13a (S223). The generated learning data 13a is stored in the storage unit 13. Next, the learning unit 14 uses the learning data 13a to perform machine learning of the learning model and generates a learned model 13b (S224). The generated learned model 13b outputs a predicted value when input data is input. If a learned model already exists, the learning unit 14 uses the learning data 13a to perform additional learning and generates an updated learned model 13b.

[0164] In addition, the generated learning model 13b is not limited to the case where one input data is used to output one prediction value. For example, the learning model 13b can also use multiple input data with different timings to output prediction values. In this case, if the measurement cycle is short enough, the effect of rapid prediction in the early stage can also be maintained.

[0165] On the other hand, if the result of determination in step S222 is that none of the acquired data has been updated, the host unit 10 repeats steps S221 and S222 until any of the acquired data has been updated.

[0166] After step S224, the prediction unit 16 inputs the acquired data of the first sensor 30a as input data to the learned model 13b, so that it outputs a predicted value (S225). Next, the learning unit 14 calculates the learning progress value of the learned model 13b based on the output predicted value (S226). The learning progress value is an indicator indicating the progress state of the machine learning of the learning model, for example, a ratio (%) indicating the learning progress of the learned model 13b. The learning progress value is expressed as the following formula (1) using the data of the second sensor 30b, i.e., the measured value A and the predicted value A' of the learned model 13b.

[0167] Learning progress value = 100-|A-A'| / A×100…(1)

[0168] The expression of the learning progress value is not limited to formula (1). For example, the learning progress value can also be the absolute value of the difference between the measured value and the predicted value, such as |A-A'|. In this case, the smaller the learning progress value, the better the progress, that is, it means that the prediction is accurate.

[0169] Next, the learning unit 14 compares the calculated learning progress value with a predetermined determination value, and determines whether the learning progress value is larger than the determination value ( S227 ).

[0170] If the result of determination in step S227 is that the learning progress value is greater than the determination value, the learning unit 14 sends a signal to the PLC 40 or an external device via the communication unit 17, and displays this on the display unit 18, notifying the user (user) that the prediction can be made by the prediction model (S228). At this time, the learning unit 14 may also notify the learning progress value together with the prediction. Thus, the user (user) can know that the learning completed model 13b that can predict the state of the workpiece W10 has been generated.

[0171] On the other hand, if the result of determination in step S227 is that the learning progress value is equal to or less than the determination value, or after step S228 , the learning unit 14 determines whether to complete the learning based on the operation of the user ( S229 ).

[0172] If the result of determination in step S229 is that the learning is completed, the learning unit 14 stores the learned model generated in step S224 in the storage unit 13 ( S230 ), and ends the prediction learning process S220 .

[0173] On the other hand, if the result of determination in step S229 is that the learning is not to be completed, the host unit 10 repeats steps S221 to S229 until the learning is completed.

[0174] When generating the learning data 13a in step S223, various combinations of input data and label data can be considered. Figure 7 As shown in FIG. 1 , the following case is considered: the measurement cycle of the first sensor 30a is 100 [μs], the measurement cycle of the second sensor 30b is 500 [μs], the first sensor 30a and the second sensor 30b are separated by a distance d, and the workpiece W10 is moving at a speed v. The measurement cycle of the second sensor 30b is five times the measurement cycle of the first sensor 30a, and the second sensor 30b continuously outputs data ak from the measurement of data ak until the measurement of the next data ak+1 ( Figure 7 Indicated in brackets).

[0175] In addition, the distance d is not limited to the distance between the installation position of the first sensor 30a and the installation position of the second sensor 30b. For example, if it is assumed that the workpiece W10 moves in the X-axis direction, the optical axis of the first sensor 30a is parallel to the Y-axis, and the optical axis of the second sensor 30b is parallel to the Z-axis, then the distance between the measurement points on the workpiece W10 is meaningful, not the distance between the installation positions of each sensor 30. In this case, the distance d is the distance between the measurement point of the first sensor 30a and the measurement point of the second sensor 30b.

[0176] For example, when the time difference Δt (= distance d / speed v) is 700 [μs], the generator 12 associates the data ak of the second sensor 30b as label data and the data bk-7 of the first sensor 30a as input data. Similarly, the generator 12 associates the data ak+1 of the second sensor 30b as label data and the data bk-2 of the first sensor 30a as input data. By generating the learning data 13a by associating the input data with the label data based on the time difference Δt, the measurement cycle of the first sensor 30a, and the measurement cycle of the second sensor 30b, the data measured on the same or similar workpiece W10 are associated with each other to generate the learning data 13a, so that the prediction accuracy of the learned model 13b can be improved.

[0177] The speed v may be a preset value or may be acquired by a rotary encoder installed in a feeding mechanism of the device, such as a motor, etc. In particular, when the speed v is not constant, the generator 12 can associate the input data with the label data with high accuracy.

[0178] Figure 7 In the example shown, the example is shown in which the generator 12 uses the data of the second sensor 30b as label data when the data of the second sensor 30b is updated, and associates the corresponding data of the first sensor 30a as input data based on the time difference Δt, the measurement cycle of the first sensor 30a, and the measurement cycle of the second sensor 30b, thereby generating the learning data 13a, but the present invention is not limited thereto. For example, the generator 12 may use the data of the first sensor 30a as input data when the data of the first sensor 30a is updated, and associates the corresponding data of the second sensor 30b as label data based on the time difference Δt, the measurement cycle of the first sensor 30a, and the measurement cycle of the second sensor 30b, thereby generating the learning data 13a.

[0179] The measurement period of at least one of the first sensor 30a and the second sensor 30b may not be fixed. In this case, the measurement time stamp of the sensor 30 is recorded in association with the measurement result, and the generator 12 compares the measurement time of the first sensor 30a and the second sensor 30b in consideration of the time difference Δt, thereby associating the input data with the tag data.

[0180] In the following, when referring to the plurality of first sensors 30 a , description will be given using an example in which n (n is an integer equal to or greater than 2) first sensors 30 a are provided at the same or substantially the same position on the line L10 .

[0181] <Selection of learning process>

[0182] When the selection learning process S240 is started, Figure 8 As shown, the acquisition unit 11 acquires data from the sensor 30 via the slave unit 20 (S241). Next, the selection unit 15 sets "1" to the suffix i (S242). The suffix i indicates the number of the n first sensors 30a and takes an integer value from "1" to "n".

[0183] Next, the generator 12 determines whether the data of the second sensor 30 b among the acquired data is updated ( S243 ).

[0184] If the result of determination in step S243 is that the data of the second sensor 30b is updated, the generating unit 12 generates learning data (S244). The generated learning data is stored in the storage unit 13. Next, the selecting unit 15 generates a learned model of the i-th first sensor 30a by machine learning using the learning data 13a (S245). In this way, a learned model is generated for each first sensor 30a. When the data of the i-th first sensor 30a is input as input data, the generated learned model outputs a predicted value. If the learned model of the i-th first sensor 30a already exists, the selecting unit 15 performs additional learning using the learning data 13a to generate an updated learned model.

[0185] On the other hand, if the result of determination in step S243 is that the data of the second sensor 30 b is not updated, the host unit 10 repeats steps S241 to S243 until the data of the second sensor 30 b is updated.

[0186] After step S245, the selection unit 15 inputs the data obtained from the i-th first sensor 30a as input data into the learning completion model of the i-th first sensor 30a, so that it outputs a predicted value (S246). Next, the selection unit 15 calculates the learning progress value of the learning completion model of the i-th first sensor 30a based on the output predicted value (S247). The learning progress value can be calculated using formula (1) in the same way as the aforementioned learning progress value. By calculating the learning progress value based on the data obtained from the second sensor 30b and the predicted value output by inputting the data obtained from the i-th first sensor 30a to the learning completion model of the i-th first sensor 30a, at least one is selected from the plurality of first sensors 30a based on the learning progress value, thereby selecting a first sensor whose predicted value of the learning completion model generated by the data of the first sensor 30a is close to the value of the data of the second sensor 30b.

[0187] Next, the selection unit 15 determines whether the value of the suffix i is equal to the number n of the first sensors 30 a ( S248 ).

[0188] If the result of determination in step S248 is that the value of the suffix i is equal to the number n of the first sensors 30a, the selection unit 15 sends a signal to the PLC 40 or an external device via the communication unit 17 to notify the user (user) of the learning progress values ​​of the learned models in all the first sensors 30a (S249). Thus, the user (user) can know the learning progress values ​​of the learned models of each first sensor 30a.

[0189] On the other hand, if the result of the determination in step S249 is that the value of the suffix i is not equal to the number n of the first sensors 30a, the selection unit 15 adds "1" to the suffix i (S250). Then, the host unit 10 repeats steps S244 to S248 and S250 until the value of the suffix i is equal to the number n of the first sensors 30a.

[0190] After step S249 , the selection unit 15 determines whether to complete the learning based on the operation of the user (S251 ).

[0191] If the result of the determination in step S251 is that the learning is completed, the selection unit 15 selects at least one first sensor 30a from the plurality of first sensors 30a based on the operation of the user (S252). At this time, the first sensor with the largest learning progress value of the learning completed model among all the first sensors 30a may be notified to the user (user) for selection, or the first sensor with a learning progress value of the learning completed model of a predetermined value, such as 80 [%] or more may be notified to the user (user) for selection.

[0192] Next, the selection unit 15 stores the selected learned model of the first sensor 30a in the storage unit 13 for storage (S253), and ends the selected learning process S240. In addition, the learned model of the unselected first sensor 30a can be stored in the storage unit 13, deleted, or stored in another storage device.

[0193] On the other hand, if the result of determination in step S251 is not to complete the learning, the host unit 10 repeats steps S241 to S251 until the learning is completed.

[0194] Figure 8In the example shown, the following example is shown, that is, the selection unit 15 generates a learning completion model and calculates the learning progress value for each first sensor 30a, but it is not limited to this. For example, the selection unit 15 can also set any m (m is an integer greater than 2 and less than n) of the n first sensors 30a as a group, generate a learning completion model for each group, and calculate the learning progress value of the learning completion model of the group. In this case, the input data is the data of all the first sensors 30a included in the group. Moreover, the first sensors 30a selected are not each first sensor 30a, but are selected in groups.

[0195] <First sensor selection mode processing>

[0196] When the first sensor selection mode process S260 is started, as shown in FIG. Fig. 9 As shown, the acquisition unit 11 acquires a predetermined amount of data from the sensor 30 via the slave unit 20 (S261). The predetermined amount of data is, for example, 255 sets of data sets. Next, the selection unit 15 sets the suffix j to "1" (S262). In addition, the suffix j represents the number of the n first sensors 30a, and takes an integer value from "1" to "n".

[0197] Next, the selection unit 15 calculates the correlation coefficient between the j-th first sensor 30a and the second sensor 30b using the data group of the j-th first sensor 30a and the data group of the second sensor 30b (S263).

[0198] Next, the selection unit 15 determines whether the value of the suffix j is equal to the number n of the first sensors 30 a ( S264 ).

[0199] If the judgment result of step S264 is that the value of the suffix j is equal to the number n of first sensors 30a, the selection unit 15 sends a signal to the PLC 40 or an external device via the communication unit 17 to notify the user (user) of the correlation coefficient of all the data of the first sensor 30a relative to the data of the second sensor 30b (S265).

[0200] On the other hand, if the result of determination in step S264 is that the value of the suffix j is not equal to the number n of the first sensors 30a, the selection unit 15 adds "1" to the suffix j (S266). Then, the host unit 10 repeats steps S263, S264, and S266 until the value of the suffix j is equal to the number n of the first sensors 30a.

[0201] After step S267 , the selection unit 15 determines whether the selection of the first sensor 30 a is completed based on the operation of the user ( S267 ).

[0202] If the result of determination in step S267 is that the selection of the first sensor 30a is completed, the selection unit 15 selects at least one first sensor 30a from the plurality of first sensors 30a based on the operation of the user (S268), and ends the first sensor selection mode processing S260. At this time, the first sensor having the largest absolute value of the correlation coefficient with the data of the second sensor 30b among all the first sensors 30a may be notified to the user (user) for selection, or the first sensor having the absolute value of the correlation coefficient with the data of the second sensor 30b greater than a predetermined value may be notified to the user (user) for selection.

[0203] On the other hand, if the result of determination in step S267 is that the selection of the first sensor 30a is not to be completed, the host unit 10 repeats steps S261 to S267 until the selection of the first sensor 30a is completed.

[0204] <Prediction model processing>

[0205] When the prediction mode processing S280 is started, Fig.10 As shown, the acquisition unit 11 acquires data from the first sensor 30a via the slave unit 20 (S281).

[0206] Next, the prediction unit 16 reads out the learned model 13 b stored in the storage unit 13 , and inputs the acquired data of the first sensor 30 a as input data to the learned model 13 b to cause the learned model 13 b to output a predicted value ( S282 ).

[0207] Next, the prediction unit 16 determines whether the output prediction value is greater than the upper threshold or less than the lower threshold (S283). For example, when the specified value of the thickness (roughness) of the workpiece W10 is 20 [mm] and the allowable range is ± 1 [mm], the upper threshold is set to 21 [mm] and the lower threshold is set to 19 [mm] respectively.

[0208] If the result of determination in step S283 is that the predicted value is greater than the upper threshold or less than the lower threshold, the prediction unit 16 sets the determination result to "ON" (S284). On the other hand, if the result of determination in step S283 is that the predicted value is less than the upper threshold and greater than the lower threshold, the prediction unit 16 sets the determination result to "OFF" (S285).

[0209] After step S284 or after step S285, the prediction unit 16 sends a signal to the PLC 40 or an external device via the communication unit 17, and displays it on the display unit 18 to notify the user (user) of the predicted value and the judgment result (S286). Thus, the user (user) can know the thickness (thickness) of the workpiece W10 predicted based on the data of the first sensor 30a, and whether the predicted thickness (thickness) of the workpiece W10 is within the allowable range of the specified value or outside the allowable range.

[0210] After step S286, the prediction unit 16 determines whether to interrupt the prediction based on the operation of the user (S287).

[0211] If the result of determination in step S287 is to interrupt the prediction, the prediction mode processing S280 is terminated.

[0212] On the other hand, if the result of determination in step S287 is not to interrupt the prediction, the host unit 10 repeats steps S281 to S287 until the prediction is interrupted.

[0213] In this embodiment, the sensor system 1 and the host unit 10 are applied to Figure 4 The sensor system 1 and the host unit 10 are also applicable to other forms of lines and other arrangements of the first sensors and the second sensors.

[0214] Next, refer to Fig.11 , while describing a second example of a line provided with a first sensor and a second sensor according to one embodiment. Fig.11 This is a structural diagram illustrating a schematic structure of a second example of the line L in one embodiment.

[0215] like Fig.11 As shown, line L20 from Fig.11 A plurality of workpieces W21 and W22 are transported in a direction from the upper right to the lower left (from the back side to the front side of the paper).

[0216] Three first sensors 30a and one second sensor 30b are arranged at the same or substantially the same position in the conveying direction of the line L20. The three first sensors 30a are arranged along the width direction ( Figure 1 They are arranged at a predetermined interval (in the left and right directions).

[0217] Each first sensor 30a is, for example, a reflective photoelectric sensor, and a light projector and a light receiver are formed integrally. The light emitted from the light projector is reflected by the workpiece W21, W22 or the background, and the light receiver measures the amount of light reflected. Each first sensor 30a outputs the measured amount of light received as the light received amount data of the workpiece W21, W22. Figure 4 In the illustrated example, the light receiving amount is measured in a shorter cycle than that of the second sensor 30 b.

[0218] The second sensor 30b is, for example, a displacement sensor, in which a light projector and a light receiver are formed integrally. When light emitted from the light projector is reflected by the workpieces W21 and W22, the distance to the workpieces W21 and W22 is measured based on the reflected light incident on the light receiver. The second sensor 30b outputs distance data to the workpieces W21 and W22. Figure 4 In the illustrated example, the distance to the workpieces W21 and W22 is measured at a longer cycle than that of the first sensor 30 a.

[0219] and Figure 4 Similarly, in the example shown Fig.11 In the example shown, the host unit 10 can also generate learning data using the data of the three first sensors 30 a as input data and the data of the second sensor 30 b as label data.

[0220] Fig.11 In the example shown, the first sensor 30a outputs light receiving amount data and the second sensor 30b outputs distance data, and the physical quantities of the two are different. That is, the machine learning of the learning model is performed using the generated learning data, and the generated learned model converts the physical quantity in the prediction.

[0221] The input data of the learning data is not limited to the case where the output data of the first sensor 30a is used directly. For example, data (information) obtained by calculating the measurement values ​​of a plurality of first sensors 30a may be used as the input data of the learning data.

[0222] Furthermore, the label data of the learning data is not limited to the case where the output data of the second sensor 30b is directly used. For example, when a sensor that measures distance (displacement) or three-dimensional position is used as the second sensor 30b, the width or height of the workpieces W21 and W22 can be obtained by using two or more second sensors 30b and performing operations such as subtraction or addition on the respective measured values. In this case, these operation results can also be used as label data for the learning data.

[0223] When the machine learning of the learning model is performed and it is determined that the learning progress is sufficient, the host unit 10 may be operated without the second sensor 30 b , that is, to predict an abnormality or a sign of an abnormality of the workpiece W. In this case, the cost of the equipment can be saved.

[0224] The above is a description of an exemplary embodiment of the present invention. According to the sensor system 1 and the host unit 10 of one embodiment of the present invention, the learning data 13a is generated by using the data of the first sensor 30a acquired as input data and the data of the second sensor 30b acquired as label data. Thus, the learned model 13b generated using the learning data 13a can output a value (predicted value) by using the data of the first sensor 30a having a shorter measurement period than the second sensor 30b as input. Therefore, by using the learned model 13b, it is possible to detect an abnormality or a sign of an abnormality of the workpiece W at an earlier stage than before.

[0225] Furthermore, according to the host unit 10 and the prediction method of one embodiment of the present invention, the acquired data of the first sensor 30a is input into the learned model 13b, so that the learned model 13b outputs a predicted value. Here, the learned model 13b is generated using the learning data 13a generated by using the data of the first sensor 30a as input data and the data of the second sensor 30b as label data indicating the nature of the input data, so that the data of the first sensor 30a having a shorter measurement period than the second sensor 30b can be used as input to output a predicted value. Therefore, the abnormality or abnormality sign of the workpiece W can be predicted at an early stage based on the predicted value.

[0226] The above-described embodiments are intended to facilitate understanding of the present invention and are not intended to limit the present invention. The various elements and their configurations, materials, conditions, shapes, and sizes included in the embodiments are not limited to those illustrated and may be appropriately changed. Furthermore, the structures shown in different embodiments may be partially replaced or combined with each other.

[0227] (Note 1)

[0228] A sensor system 1, comprising:

[0229] The first sensor 30a measures the workpiece;

[0230] The second sensor 30b measures the workpiece at a longer period than the first sensor 30a; and

[0231] Host unit 10,

[0232] The host unit 10 includes:

[0233] The acquisition unit 11 acquires data measured by the first sensor 30a and data measured by the second sensor 30b; and

[0234] The generating unit 12 generates learning data used for machine learning of the learning model, using the data acquired by the first sensor 30a as input data and the data acquired by the second sensor 30b as label data indicating the properties of the input data.

[0235] (Note 8)

[0236] A host unit 10 is used in a sensor system 1, wherein the sensor system 1 includes a first sensor 30a for measuring a workpiece and a second sensor 30b for measuring the workpiece with a longer period than the first sensor 30a. The host unit 10 includes:

[0237] The acquisition unit 11 acquires data measured by the first sensor 30a and data measured by the second sensor 30b; and

[0238] The generating unit 12 generates learning data used for machine learning of the learning model, using the data acquired by the first sensor 30a as input data and the data acquired by the second sensor 30b as label data indicating the properties of the input data.

[0239] (Note 14)

[0240] A prediction device 10 is provided for predicting an abnormality or a sign of an abnormality of a workpiece. The prediction device 10 comprises:

[0241] The acquisition unit 11 acquires data measured by the first sensor 30a for measuring the workpiece; and

[0242] The prediction unit 16 inputs the acquired data of the first sensor 30a into the learned model, so that the learned model outputs a predicted value.

[0243] The learned model is generated by performing machine learning of the learning model using learning data, wherein the learning data is generated using the data of the first sensor 30a as input data and the data of the second sensor 30b that measures the workpiece at a period longer than that of the first sensor 30a as label data representing the nature of the input data.

[0244] (Note 15)

[0245] A prediction method for predicting an abnormality or a sign of an abnormality of a workpiece, the prediction method comprising:

[0246] Step (S281), acquiring data measured by the first sensor 30a for measuring the workpiece; and

[0247] Step (S282), inputting the acquired data of the first sensor 30a into the learned model, so that the learned model outputs a predicted value,

[0248] The learned model is generated by performing machine learning of the learning model using learning data, wherein the learning data is generated using the data of the first sensor 30a as input data and the data of the second sensor 30b that measures the workpiece at a period longer than that of the first sensor 30a as label data representing the nature of the input data.

Claims

1. A sensor system comprising: The first sensor measures the workpiece; a second sensor for measuring the workpiece at a longer period than the first sensor; as well as Main unit, The host unit comprises: an acquisition unit, which acquires data measured by the first sensor and data measured by the second sensor; as well as A generation unit generates learning data, wherein the learning data is used for machine learning of a learning model, and uses the data obtained from the first sensor as input data, and uses the data obtained from the second sensor as label data representing the nature of the input data, wherein the nature of the input data includes whether there is an abnormality or a sign of an abnormality when the workpiece is transported online, the type of the workpiece, the size of the workpiece, or the position deviation of the workpiece.

2. The sensor system according to claim 1, wherein The generating unit generates the learning data by associating the input data with the label data based on a time difference calculated from a moving speed of the workpiece and a distance between the first sensor and the second sensor, a measurement cycle of the first sensor, and a measurement cycle of the second sensor.

3. The sensor system according to claim 1 or 2, wherein The first sensor is arranged on an upstream side with respect to the second sensor in the line along which the workpiece moves.

4. The sensor system according to claim 1 or 2, wherein The host unit further includes a learning unit that performs machine learning of a learning model using the learning data to generate a learned model.

5. The sensor system according to claim 4, wherein The host unit further includes a prediction unit configured to input the acquired data of the first sensor into the learned model so that the learned model outputs a predicted value.

6. The sensor system according to claim 1 or 2, comprising a plurality of said first sensors, The host unit further includes a selection unit that calculates a correlation coefficient between the data acquired by the first sensor and the data acquired by the second sensor, with respect to one of the plurality of first sensors.

7. The sensor system according to claim 1 or 2, comprising a plurality of said first sensors, The generating unit generates learning data using data acquired from at least one of the plurality of first sensors as input data, The host unit also includes a selection unit, which calculates a learning progress value representing the proportion of learning progress of the learned model based on the data obtained from the second sensor and a predicted value output by inputting the input data into a learned model generated by performing machine learning of the learning model using the learning data.

8. A host unit used in a sensor system, the sensor system comprising a first sensor for measuring a workpiece and a second sensor for measuring the workpiece with a longer period than the first sensor, the host unit comprising: an acquisition unit, which acquires data measured by the first sensor and data measured by the second sensor; as well as A generation unit generates learning data, wherein the learning data is used for machine learning of a learning model, and uses the data obtained from the first sensor as input data, and uses the data obtained from the second sensor as label data representing the nature of the input data, wherein the nature of the input data includes whether there is an abnormality or a sign of an abnormality when the workpiece is transported online, the type of the workpiece, the size of the workpiece, or the position deviation of the workpiece.

9. The host unit according to claim 8, wherein The generating unit generates the learning data by associating the input data with the label data based on a time difference calculated from a moving speed of the workpiece and a distance between the first sensor and the second sensor, a measurement cycle of the first sensor, and a measurement cycle of the second sensor.

10. The host unit according to claim 8 or 9, further comprising: The learning unit performs machine learning of the learning model using the learning data to generate a learned model.

11. The host unit according to claim 10, further comprising: The prediction unit inputs the acquired data of the first sensor into the learned model, so that the learned model outputs a predicted value.

12. The host unit according to claim 8 or 9, wherein The sensor system comprises a plurality of the first sensors, The host unit further includes a selection unit that calculates a correlation coefficient between the data acquired by the first sensor and the data acquired by the second sensor, with respect to one of the plurality of first sensors.

13. The host unit according to claim 8 or 9, wherein The sensor system comprises a plurality of the first sensors, The generating unit generates learning data using data acquired from at least one of the plurality of first sensors as input data, The host unit also includes a selection unit, which calculates a learning progress value representing the proportion of learning progress of the learned model based on the data obtained from the second sensor and a predicted value output by inputting the input data into a learned model generated by performing machine learning of the learning model using the learning data.

14. A prediction device for predicting an abnormality or a sign of an abnormality of a workpiece, the prediction device comprising: an acquisition unit that acquires data measured by a first sensor that measures the workpiece; as well as The prediction unit inputs the acquired data of the first sensor into the learned model so that the learned model outputs a predicted value. The learned model is generated by performing machine learning of the learning model using learning data, wherein the learning data is generated using the data of the first sensor as input data and the data of a second sensor that measures the workpiece at a period longer than that of the first sensor as label data representing the properties of the input data, wherein the properties of the input data are used to represent the presence or absence of an abnormality or a sign of an abnormality when the workpiece is transported online based on the predicted value.

15. A prediction method for predicting an abnormality or a sign of an abnormality of a workpiece, the prediction method comprising the following steps: acquiring data measured by a first sensor for measuring the workpiece; and inputting the acquired data of the first sensor into a learned model so that the learned model outputs a predicted value, The learned model is generated by performing machine learning of the learning model using learning data, wherein the learning data is generated using the data of the first sensor as input data and the data of a second sensor that measures the workpiece at a period longer than that of the first sensor as label data representing the properties of the input data, wherein the properties of the input data are used to represent the presence or absence of an abnormality or a sign of an abnormality when the workpiece is transported online based on the predicted value.

Images