JP7317931B2 - Selection device, selection method and selection program - Google Patents

Description

The present invention relates to a selection device, selection method and selection program.

2. Description of the Related Art Conventionally, there has been known a technique for assisting plant operation by reproducing the operation of a plant operator by machine learning. In addition to the normal mode, the operation of the plant includes an abnormal mode such as during sudden rain or during maintenance. Therefore, when performing driving support in the normal mode, it is necessary to remove the operation data in the non-normal mode.

Japanese Patent No. 6649112

However, it is difficult to reproduce the operations of plant operators in normal mode by machine learning. For example, the machine learning model developer cannot determine whether the operation is in the normal mode or the non-normal mode, and it is difficult to eliminate the operation in the non-normal mode. In addition, if the system learns without distinguishing between the correct operation that the system should learn and the incorrect operation that the system does not want to learn, which are mixed in the operation in the actual driving, there is a concern that the accuracy will decrease.

The present invention has been made in view of the above, and an object of the present invention is to reproduce the operation of plant operators in the normal mode by machine learning.

In order to solve the above-described problems and achieve the object, the selection device according to the present invention includes an acquisition unit that acquires operation history data, and gives a predetermined flag to the operation history data according to an input designation. and a selection unit that selects the operation history data to which the predetermined flag is added as teacher data for machine learning.

Advantageous Effects of Invention According to the present invention, machine learning can be used to reproduce the operation of a plant operator in a normal mode.

FIG. 1 is a schematic diagram illustrating a schematic configuration of a selection device. FIG. 2 is a diagram illustrating a data configuration of operation history data. FIG. 3 is a diagram for explaining the processing of the granting unit. FIG. 4 is a diagram for explaining the processing of the granting unit. FIG. 5 is a flowchart illustrating a selection processing procedure. FIG. 6 is a flowchart illustrating a selection processing procedure. FIG. 7 is a diagram illustrating a computer that executes the selection program.

An embodiment of the present invention will be described in detail below with reference to the drawings. It should be noted that the present invention is not limited by this embodiment. Moreover, in the description of the drawings, the same parts are denoted by the same reference numerals.

[Configuration of selected device]
FIG. 1 is a schematic diagram illustrating a schematic configuration of a selection device. As illustrated in FIG. 1 , the selection device 10 of this embodiment is implemented by a general-purpose computer such as a personal computer, and includes an input unit 11 , an output unit 12 , a communication control unit 13 , a storage unit 14 and a control unit 15 .

The input unit 11 is implemented using an input device such as a keyboard and a mouse, and inputs various instruction information such as processing start to the control unit 15 in response to input operations by the operator. The output unit 12 is implemented by a display device such as a liquid crystal display, a printing device such as a printer, or the like.

The communication control unit 13 is implemented by a NIC (Network Interface Card) or the like, and controls communication between an external device and the control unit 15 via a network. For example, the communication control unit 13 controls communication between the control unit 15 and an operation terminal that outputs operation history data to be processed in a selection process to be described later, a management device that manages operation history data, or the like.

The storage unit 14 is implemented by a semiconductor memory device such as a RAM (Random Access Memory) or a flash memory, or a storage device such as a hard disk or an optical disk. In the storage unit 14, a processing program for operating the selection device 10, data used during execution of the processing program, and the like are stored in advance, or are temporarily stored each time processing is performed.

In this embodiment, the storage unit 14 stores operation history data 14a. The storage unit 14 also stores a model generated in a selection process, which will be described later, parameters of the model, and the like. Note that the storage unit 14 may be configured to communicate with the control unit 15 via the communication control unit 13 .

FIG. 2 is a diagram illustrating the data configuration of the operation history data. As illustrated in FIG. 2, the operation history data 14a includes the time and operation history of the operator's operation. The operation history includes values of explanatory variables and objective variables of the model learned by the learning unit 15d, which will be described later. The operation history data 14a also includes flags and weights that are given in selection processing, which will be described later. A weight is a weight in teacher data when a learning unit 15d, which will be described later, uses operation history data as teacher data to learn a model. Further, the operation history data 14a may include operator's comments, as will be described later.

In addition, in the example shown in FIG. 2 , the “non-subject flag” of learning is given as a flag, but the present invention is not limited to this. For example, as will be described later, the flag may be a "target flag" for learning, or a plurality of flags including a "non-target flag", a "target flag", or other flags may be added.

Returning to the description of FIG. The control unit 15 is implemented using a CPU (Central Processing Unit) or the like, and executes a processing program stored in a memory. Thereby, the control unit 15 functions as an acquisition unit 15a, a provision unit 15b, a selection unit 15c, and a learning unit 15d, as illustrated in FIG. Each or part of these functional units may be implemented in different hardware. For example, the learning unit 15d may be implemented in a device separate from other functional units. Also, the control unit 15 may include other functional units.

The acquisition unit 15a acquires operation history data. For example, the acquisition unit 15a acquires operation history data indicating the operation history of the operation terminal by the operator, which is the target of the selection process described later, via the input unit 11, from the operation terminal, or at the operation terminal. Acquired via the communication control unit 13 from a management device or the like that manages the operation history. The acquisition unit 15a acquires operation history data in advance and stores it in the storage unit 14 . Note that the acquiring unit 15a may transfer the acquired operation history data to the attaching unit 15b described below in real time without storing the acquired operation history data in the storage unit 14. FIG.

The adding unit 15b adds a predetermined flag to the acquired operation history data according to the input designation. Specifically, the adding unit 15b receives an input instructing to add a predetermined flag, and adds the predetermined flag.

Here, FIGS. 3 and 4 are diagrams for explaining the processing of the granting unit. For example, as illustrated in FIG. 3, the assigning unit 15b presents the operation history data to the operator and accepts an input instructing to assign a flag. Then, as illustrated in FIG. 2, the adding unit 15b adds a flag to the data corresponding to the specified portion in the operation history data 14a.

In FIG. 3, the values of explanatory variables such as raw materials, pressure, and temperature in the operation history are displayed in chronological order in the lower part (b). Also, in the upper part (a), the values of the objective variables of the operation history are displayed in chronological order.

Further, as indicated by vertical hatching in FIG. Accept designation. For example, an operator can specify which maneuvers went well and which they want the machine learning model to actively learn.

As indicated by horizontal hatching in the upper part (a) of FIG. Accepts designation by the operator. For example, when an operator recognizes an operation that has failed, the operator explicitly designates that portion as an operation not targeted at that time.

In addition, as shown in the upper part (a) of FIG. 3, the operator can also add a comment to a portion specified as a flagged portion. In this case, for example, a comment item is added to the operation history data 14a shown in FIG.

For example, it is possible to leave as a comment the reason why the operation was successful at the place where the learning object flag is given. In the example shown in FIG. 3, a comment such as "Successfully stabilized quality by finely manipulating in accordance with quality change of raw materials" is added.

Alternatively, it is possible to leave a comment as a reason for excluding the object of learning at the place where the flag not to be studied is given. In the example shown in FIG. 3, a comment such as "a temperature drop was predicted due to a thunderstorm, but no thunderstorm occurred, so it is not subject to learning" is added. As a result, when a plurality of operators input the object/non-object of the learning data, it becomes possible to match the line of sight with other operators, and it becomes a reference when improving accuracy later.

Also, in FIG. 3, a "collection stop" button is displayed. By clicking this button, the operator can stop data collection on the UI when he/she does not want to collect operation history data as teaching data for learning under unusual conditions, such as during maintenance inspections. . Thus, the operator can specify the learning exclusion period.

The adding unit 15b can further add the weight in the teacher data to the operation history data according to the input designation. Specifically, the assigning unit 15b further receives an input designating a weight, and assigns the weight to the operation history data as illustrated in FIG. For example, the operator assigns a greater weight to a location considered to be particularly important among a plurality of locations designated to be given a "target flag". As a result, it is possible to more accurately reflect the judgment of the on-site operator, so that it is possible to learn a model that more accurately reproduces the operation in the normal mode.

Note that the example shown in FIG. 3 illustrates a case where the provision unit 15b presents the operation history data acquired by the acquisition unit 15a in real time. In this case, the operation history data is stored in the operation history data 14a of the storage unit 14 after being given flags and weights.

On the other hand, the adding unit 15b can present the operation history data 14a pre-stored in the storage unit 14, add a flag, and update the operation history data 14a. FIG. 4 illustrates a case where the operation history data 14a is updated. In this case as well, the operator can specify the locations to which the flags are attached, as in the example shown in FIG. In the example shown in FIG. 4 as well, it is possible to specify the learning exclusion period as illustrated by hatching with oblique lines.

Returning to the description of FIG. The selection unit 15c selects operation history data to which a predetermined flag is added as teacher data for machine learning. For example, the selection unit 15c extracts the operation history data to which the “target flag” for learning is added from the operation history data 14a, and uses it as teacher data.

In the example shown in FIG. 2, the "non-target flag" is given, but for example, the "weight" can be interpreted as the "target flag". In that case, the selection unit 15c can also select only data exceeding a predetermined threshold as teacher data. Alternatively, the selection unit 15c may directly use the weight value as a machine learning parameter.

This makes it possible, for example, to add a flag indicating a learning target to the operation history data to be learned, and to specify the operation history data to be used as teacher data by the operator on site. Therefore, it is possible to reproduce the operation of the plant operator in the normal mode by machine learning. Therefore, in the operation of the plant in the normal mode, it is possible to perform operation support with high accuracy.

The selection unit 15c selects teacher data by excluding, from the operation history data 14a, the operation history data to which the learning "non-target flag" is given, which is determined by the on-site operator to be an unsteady operation. may In this way, by explicitly excluding the operation history data not to be learned from the training data for machine learning, it is possible to reproduce the operation in the normal mode with high accuracy.

Further, the adding unit 15b can also add a plurality of types of predetermined flags to the operation history data according to the input designation. In this case, the selection unit 15c selects operation history data for each predetermined flag.

For example, when the operation history data 14a is given a learning "non-target flag" or other flags in addition to the learning "target flag", the selection unit 15c selects the operation history data 14a for each flag. converge on In that case, the selection unit 15c may use the operation history data to which the “target flag” is assigned as the teacher data, or may exclude the operation history data to which the “non-target flag” is assigned and use it as the teacher data. .

The learning unit 15d learns a machine learning model using the selected teacher data. As a result, it becomes possible to accurately reproduce the operation of the plant operator in the normal mode by machine learning.

[Selection procedure]
Next, an example of selection processing by the selection device 10 according to the present embodiment will be described with reference to FIGS. 5 and 6. FIG. 5 and 6 are flowcharts illustrating selection processing procedures. First, the flowchart of FIG. 5 is started, for example, when an input instructing the start of selection processing is received.

First, the acquisition unit 15a acquires operation history data (step S1). For example, the acquisition unit 15a acquires operation history data indicating the operation history of the operation terminal by the operator, which is the target of the selection process, via the input unit 11, from the operation terminal, or from the operation history at the operation terminal. is acquired via the communication control unit 13 from a management device or the like that manages the .

Next, the assigning unit 15b presents the operation history data to the operator, and when an input designating attachment of a flag is received (step S2, Yes), the flag is added to the specified location in the operation history data 14a. A flag is attached to the corresponding data (step S3).

Further, when there is no input instructing to add a flag (step S2, No), the adding unit 15b advances the process to step S4.

Similarly, when receiving an input specifying a weight in the teacher data from the operator (step S4, Yes), the assigning unit 15b assigns a weight to the operation history data 14a (step S5). This completes a series of processes.

If there is no input designating the weight in the teacher data (step S4, No), the imparting unit 15b terminates the series of processes.

Next, the flowchart of FIG. 6 is started at the timing when an input instructing the start of processing is received, for example. Note that the process of step S11 may be started following the process of step S5 shown in FIG.

First, the selection unit 15c selects the operation history data 14a to which a predetermined flag is added as teacher data for machine learning (step S11). For example, the selection unit 15c extracts the operation history data to which the “target flag” for learning is added from the operation history data 14a, and uses it as teacher data. Alternatively, the selection unit 15c may select teacher data by excluding operation history data to which the “non-target flag” is added from the operation history data 14a.

Next, the learning unit 15d learns a machine learning model using the selected teacher data (step S12), and outputs a model generated by learning (step S13). For example, the learning unit 15d applies the model to a site that reproduces the operations of plant operators. This completes a series of selection processes.

[effect]
As described above, in the selection device 10 of the above embodiment, the acquisition unit 15a acquires operation history data. The adding unit 15b adds a predetermined flag to the operation history data according to the input designation. The selection unit 15c selects operation history data to which a predetermined flag is added as teacher data for machine learning.

This makes it possible, for example, to add a flag indicating a learning target to the operation history data to be learned, and to specify the operation history data to be used as teacher data by the operator on site. Therefore, it is possible to reproduce the operation of the plant operator in the normal mode by machine learning. Therefore, in the operation of the plant in the normal mode, it is possible to perform operation support with high accuracy.

In addition, the assigning unit 15b further assigns the weight in the teacher data to the operation history data in accordance with the input designation. As a result, it is possible to more accurately reflect the judgment of the operator at the site, so that the operation in the normal mode can be reproduced with higher accuracy.

Further, the adding unit 15b adds a plurality of predetermined flags to the operation history data according to the input designation, and the selecting unit 15c selects the operation history data for each predetermined flag. As a result, for example, it is possible to attach a flag indicating that the operation is not subject to learning to the operation history data determined by the operator on site to be an abnormal operation. Therefore, it is possible to clearly reproduce operation history data that is not to be learned and exclude it from machine learning teacher data, thereby reproducing operations in the normal mode with high accuracy.

Also, the learning unit 15d learns a machine learning model using the selected teacher data. As a result, it becomes possible to reproduce the operation of plant operators in the normal mode by machine learning.

[System configuration, etc.]
Each component of each device illustrated is functionally conceptual, and does not necessarily need to be physically configured as illustrated. In other words, the specific form of distribution and integration of each device is not limited to the one shown in the figure, and all or part of them can be functionally or physically distributed and integrated in arbitrary units according to various loads and usage conditions. Can be integrated and configured. Furthermore, all or any part of each processing function performed by each device is realized by a CPU or GPU and a program analyzed and executed by the CPU or GPU, or as hardware by wired logic can be realized.

Further, among the processes described in this embodiment, all or part of the processes described as being automatically performed can be performed manually, or the processes described as being performed manually can be performed manually. All or part of this can also be done automatically by known methods. In addition, information including processing procedures, control procedures, specific names, and various data and parameters shown in the above documents and drawings can be arbitrarily changed unless otherwise specified.

[program]
It is also possible to create a program in which the processing executed by the selection device described in the above embodiment is described in a computer-executable language. For example, it is possible to create a program in which the processing executed by the selection device 10 according to the embodiment is described in a computer-executable language. In this case, the same effects as those of the above embodiments can be obtained by having the computer execute the program. Further, such a program may be recorded in a computer-readable recording medium, and the program recorded in this recording medium may be read by a computer and executed to realize the same processing as in the above embodiments.

FIG. 7 is a diagram showing an example of a computer that executes a selection program. Computer 1000 includes, for example, memory 1010 , CPU 1020 , hard disk drive interface 1030 , disk drive interface 1040 , serial port interface 1050 , video adapter 1060 and network interface 1070 . These units are connected by a bus 1080 .

The memory 1010 includes a ROM (Read Only Memory) 1011 and a RAM 1012 . The ROM 1011 stores a boot program such as BIOS (Basic Input Output System). Hard disk drive interface 1030 is connected to hard disk drive 1031 . Disk drive interface 1040 is connected to disk drive 1041 . A removable storage medium such as a magnetic disk or an optical disk is inserted into the disk drive 1041, for example. A mouse 1051 and a keyboard 1052 are connected to the serial port interface 1050, for example. For example, a display 1061 is connected to the video adapter 1060 .

Here, the hard disk drive 1031 stores an OS 1091, application programs 1092, program modules 1093 and program data 1094, for example. Each piece of information described in the above embodiment is stored in the hard disk drive 1031 or the memory 1010, for example.

Also, the selection program is stored in hard disk drive 1031 as program module 1093 in which commands to be executed by computer 1000 are described, for example. Specifically, the hard disk drive 1031 stores a program module 1093 that describes each process executed by the selection device 10 described in the above embodiment.

Data used for information processing by the selection program is stored as program data 1094 in the hard disk drive 1031, for example. Then, the CPU 1020 reads out the program module 1093 and the program data 1094 stored in the hard disk drive 1031 to the RAM 1012 as necessary, and executes each procedure described above.

Note that the program module 1093 and program data 1094 related to the selection program are not limited to being stored in the hard disk drive 1031. For example, they may be stored in a removable storage medium and read by the CPU 1020 via the disk drive 1041 or the like. may be Alternatively, the program module 1093 and program data 1094 related to the selection program are stored in another computer connected via a network such as a LAN (Local Area Network) or a WAN (Wide Area Network), and are transmitted via the network interface 1070. It may be read by CPU 1020 .

Although the embodiments to which the invention made by the present inventor is applied have been described above, the present invention is not limited by the descriptions and drawings forming a part of the disclosure of the present invention according to the embodiments. That is, other embodiments, examples, operation techniques, etc. made by those skilled in the art based on this embodiment are all included in the scope of the present invention.

10 selection device 11 input unit 12 output unit 13 communication control unit 14 storage unit 14a operation history data 15 control unit 15a acquisition unit 15b addition unit 15c selection unit 15d learning unit

Claims (4)

Based on the instruction to stop collecting the operation history data displayed on the screen that displays the operation history data in real time , by switching between acquiring and stopping the acquisition of the data in real time, the operator to be learned by the machine learning model. an acquisition unit that acquires the operation history data including the operation of
A target flag indicating whether or not the machine learning model is to be learned, or a granting unit that grants an out-of-target flag , a weight in machine learning training data, and a predetermined comment;
Based on the target flag or the non-target flag and the weight for each operation history data to which the target flag or the non-target flag, the weight, and the predetermined comment are assigned a selection unit that selects as training data for the machine learning;
A selection device comprising: 2. The selection device according to claim 1, further comprising a learning unit that learns a machine learning model using the selected teacher data. A selection method executed by a selection device,
Based on the instruction to stop collecting the operation history data displayed on the screen that displays the operation history data in real time , by switching between acquiring and stopping the acquisition of the data in real time, the operator to be learned by the machine learning model. an acquisition step of acquiring the operation history data including the operation of
Target flags indicating whether or not the machine learning model is to be learned, or an assignment step of assigning a non-target flag , a weight in machine learning training data, and a predetermined comment;
Based on the target flag or the non-target flag and the weight for each operation history data to which the target flag or the non-target flag, the weight, and the predetermined comment are assigned A selection step of selecting as training data for the machine learning;
A selection method comprising: Based on the instruction to stop collecting the operation history data displayed on the screen that displays the operation history data in real time , by switching between acquiring and stopping the acquisition of the data in real time, the operator to be learned by the machine learning model. an acquisition step of acquiring the operation history data including the operation of
A target flag indicating whether or not the machine learning model is to be learned, or A granting step of granting a non-target flag , a weight in machine learning teacher data, and a predetermined comment;
Based on the target flag or the non-target flag and the weight for each operation history data to which the target flag or the non-target flag, the weight, and the predetermined comment are assigned A selection step of selecting as training data for the machine learning;
A selection program for executing a computer.

Images