KR20220028065A - Machine learning devices, data processing systems, inference devices, and machine learning methods - Google Patents

Abstract

The present invention provides a machine learning device or the like that enables to accurately grasp abnormalities and abnormalities in a fluid pressure actuated valve without depending on the operator's experience or the like.

Description

Machine learning devices, data processing systems, inference devices, and machine learning methods

The present invention relates to a machine learning apparatus, a data processing system, an inference apparatus, and a machine learning method for performing abnormal diagnosis of a valve system.

BACKGROUND ART Conventionally, a fluid pressure actuating valve that opens and closes a main valve by controlling a driving fluid by an electromagnetic valve is known. For example, in patent document 1, it is a fluid pressure actuation valve used for piping of a plant facility, In the case of an emergency like an abnormality occurring in a facility, a drive fluid is controlled with a solenoid valve, and a ball valve (main valve) is closed. Accordingly, an emergency shutoff valve device for shutting off a fluid flowing through a pipe is disclosed.

Patent Document 1: Japanese Patent Application Laid-Open No. 2009-97539

In fluid pressure drive valves, such as an emergency shut-off valve used for plant equipment as shown in patent document 1, in order to improve the operation rate and reliability of the whole plant equipment, it is preferable that unexpected abnormality does not generate|occur|produce. For this reason, in the fluid pressure actuation valve, in order to operate reliably in an emergency, for example, as a periodic maintenance check, the influence on plant equipment is suppressed as much as possible while opening/closing operation of the fluid pressure actuation valve in a predetermined period is performed. It is confirmed that the fluid pressure actuated valve operates normally. Therefore, in such a fluid pressure actuation valve, it is desired to realize not only post maintenance for catching abnormalities generated in the fluid pressure actuation valves post hoc, but also forefront maintenance for catching abnormal signs.

Here, the abnormal signs of the hydraulically driven valve may be expressed as various events by, for example, opening and closing of the hydraulically driven valve in a predetermined period, but events that may occur in the hydraulically driven valve and The causal relationship of anomalies has not been clearly specified. As a result, rolling maintenance in a fluid pressure drive valve is performed based on judgment dependent on the operator's experience (including tacit knowledge), and there is a problem that the accuracy varies depending on the operator in charge.

In view of the above-mentioned problems, the present invention is for accurately grasping abnormalities and abnormal signs (hereinafter, in the present invention, these are collectively referred to as "abnormalities") in a fluid pressure actuated valve. An object of the present invention is to provide a machine learning apparatus, a data processing system, an inference apparatus, and a machine learning method.

In order to achieve the above object, the machine learning apparatus according to the first aspect of the present invention provides a main valve 11 and a drive device ( 12) and a fluid pressure actuating valve 10 including at least a solenoid valve 1 for controlling supply/exhaust of the actuating fluid A to and from the drive device 12, wherein the main Input data including time series data of the valve opening degree of the valve 11 and time series data of the pressure on the solenoid valve output side of the driving fluid A supplied to and discharged from the drive device 12 during a predetermined period, and corresponding input data a learning data set storage unit 202 for storing a plurality of sets of learning data sets composed of output data composed of diagnostic information of the built-in fluid pressure actuation valve 10; a learning unit (203) for learning a learning model for inferring the correlation of the output data; and a learned model storage unit (204) for storing the learning model learned by the learning unit (203).

ADVANTAGE OF THE INVENTION According to the machine learning apparatus of this invention, based on the various information etc. which can be acquired by the fluid pressure drive valve in a predetermined period, it becomes possible to provide the learning completion model which can estimate the presence or absence of occurrence of abnormality of a fluid pressure drive valve. Therefore, by using this learned model, it becomes possible to implement|achieve to estimate the abnormality which generate|occur|produces in a fluid pressure drive valve with high precision, without relying on an operator's experience.

BRIEF DESCRIPTION OF THE DRAWINGS It is a schematic diagram which shows an example of the fluid pressure drive valve to which the machine learning apparatus etc. which concern on one Embodiment of this invention are applied.
It is a schematic diagram which shows an example of the solenoid valve to which the machine learning apparatus etc. which concern on one Embodiment of this invention are applied.
3 is a block diagram showing an example of a solenoid valve to which a machine learning apparatus or the like according to an embodiment of the present invention is applied.
4 is a schematic block diagram of a machine learning apparatus according to an embodiment of the present invention.
5 is a diagram showing a configuration example (learning with a teacher) data used in a machine learning apparatus or the like according to an embodiment of the present invention.
6 is a diagram showing a configuration example (learning without a teacher) used in a machine learning apparatus or the like according to an embodiment of the present invention.
Fig. 7 is a diagram showing an example of a neural network model for learning with a teacher implemented in the machine learning apparatus according to an embodiment of the present invention.
8 is a flowchart showing an example of a machine learning method according to an embodiment of the present invention.
9 is a schematic block diagram showing a data processing system according to an embodiment of the present invention.
It is a flowchart which shows the example of the data processing process by the data processing system which concerns on one Embodiment of this invention.

EMBODIMENT OF THE INVENTION Hereinafter, each embodiment for implementing this invention with reference to drawings is demonstrated. In addition, in the following, the range necessary for the description for achieving the object of the present invention is schematically shown, and the range necessary for the description of the corresponding part of the present invention will be mainly described, and the point where the description is omitted is based on the known technology. make it as

Before describing the machine learning apparatus, the data processing system, the reasoning apparatus, and the machine learning method related to the embodiment of the present invention, first, a fluid pressure actuation valve to which the machine learning apparatus or the like is applied will be described.

(Hydraulic actuated valve)

1 is a schematic diagram showing an example of a fluid pressure actuation valve 10 according to an embodiment of the present invention. As the fluid pressure actuation valve 10 in this embodiment, for example, it is installed in the pipe 100 through which various gases, petroleum, etc. flow in plant equipment, and at the time of an emergency stop when abnormality etc. generate|occur|produced in plant equipment, It can be used as an emergency shut-off valve for blocking the flow of the pipe 100 . In addition, the installation place and use of the fluid pressure actuation valve 10 are not limited to the said example.

The fluid pressure actuating valve 10 shown in FIG. 1 drives the main valve 11 disposed in the middle of the pipe 100 and the valve shaft 13 connected to the main valve 11 according to the fluid pressure of the driving fluid, whereby the main A fluid pressure type driving device 12 for opening/closing the valve 11 is provided, and a solenoid valve 1 having a function of controlling supply/exhaust of the driving fluid to the driving device 12 .

Instrumentation air (hereinafter, simply referred to as “air”) A is employed as the driving fluid used for the fluid pressure driving valve 10 . This air A is supplied from the air supply source 14 to the solenoid valve 1 through the first air pipe 140 , and is also supplied to the drive device 12 through the second air pipe 141 . In addition, the fluid pressure drive valve 10 includes a communication cable 150 for transmitting and receiving various data between the external device 15 and the solenoid valve 1 , and an external power supply 16 to the solenoid valve 1 . A power cable 160 for supplying power is connected. In addition, as a driving fluid, it is not limited to the said air A, Another gas may be sufficient, and a liquid (for example, oil) may be sufficient as it.

The external device 15 is a device for transmitting and receiving various types of information to and from the fluid pressure actuation valve 10, for example, a computer for plant management (including a local server and a cloud server), an operator (maintenance inspector) It consists of a diagnostic computer used by the company, or an external storage unit such as a USB memory or external HDD. This external device 15 can also be connected to a machine learning device 200 described later to transmit various data constituting a data set for learning. In addition, this external device 15 is provided with a notification means consisting of a GUI (Graphical User Interface), etc. for notifying an operator or the like that an abnormality has occurred or the contents when an abnormality occurs in the fluid pressure actuation valve 10, there is. In addition, wireless communication may be used for communication between the external device 15 and the solenoid valve 1 .

As the drive method of the fluid pressure actuation valve 10 of this embodiment, the airless closing method is employ|adopted. Therefore, during normal operation, by supplying (supplying) air A from the air supply source 14 to the drive device 12 through the solenoid valve 1 (supply air), the main valve 11 is operated to open, and during an emergency stop or test During operation, the main valve 11 is operated to close by discharging (exhausting) the air A from the driving device 12 through the solenoid valve 1 . In addition, the fluid pressure actuation valve 10 may employ|adopt an airless open system, and in that case, it is closed by supplying air A to the drive device 12, and air A is removed from the drive device 12. By discharging, the main valve 11 is operated to close.

A ball valve is employed as the main valve 11 . As a specific structure of this main valve 11, the valve box 110 arrange|positioned in the middle of the piping 100, and the ball-shaped valve body 111 formed rotatably in the valve box 110 are provided. In addition, the first end 130A of the valve shaft 13 is connected to the upper portion of the valve body 111 . As the valve shaft 13 is rotationally driven by 0 degrees to 90 degrees, the valve body 111 rotates in the valve box 110, and the fully open state (the state shown in FIG. 1) and the fully closed state of the main valve 11 state is switched In addition, the valve used as the main valve 11 is not limited to a ball valve, For example, a butterfly valve or other on-off valve may be sufficient.

The drive device 12 employs a single-acting air cylinder mechanism disposed between the main valve 11 and the solenoid valve 1 . As a specific configuration of the driving device 12 , a cylindrical cylinder 120 and a pair of pistons 122A and 122B formed to be reciprocally linearly movable in the cylinder 120 and connected via a piston rod 121 . ), a coil spring 123 formed on the side of the first piston 122A, an air supply/exhaust port 124 formed on the side of the second piston 122B, and a valve disposed to pass through the cylinder 120 in a radial direction The transmission mechanism 125 formed in the part where the shaft 13 and the piston rod 121 orthogonally crosses is provided. Further, the drive device 12 is not limited to a single-actuation type, and may be configured in other types such as, for example, a dual-actuation type.

The first piston 122A is elastically supported by the coil spring 123 to operate the main valve 11 in the closing direction. Further, the second piston 122B operates the main valve 11 in the open direction by the air A (supply air) supplied from the air supply/exhaust port 124 (resisting the elastic force of the coil spring 123). is to pressurize In addition, the transmission mechanism 125 is comprised of a rack and pinion mechanism, a link mechanism, a cam mechanism, etc., converts the reciprocating linear motion of the piston rod 121 into rotational motion, and transmits it to the valve shaft 13 .

The valve shaft 13 is formed on a shaft, and is disposed so as to pass through the driving device 12 in a rotatable state. The first end 130A of the valve shaft 13 is connected to the main valve 11 , and the second end 130B of the valve shaft 13 is pivotally supported by the solenoid valve 1 . In addition, the valve shaft 13 may be one in which a plurality of shafts are connected via a coupling or the like.

The solenoid valve 1 has a function of controlling the supply and exhaust of air A to and from the drive device 12, for example, a normally closed type in two positions (“open” when energized, “closed” when de-energized) It is configured as a three-way solenoid valve of This solenoid valve 1 has a spool part 2 for switching a flow path through which air A flows inside a housing part 6 functioning as a housing of an indoor type or explosion-proof type solenoid valve 1, and an energized state. The solenoid part 3 for displacing the spool part 2 according to (when energized or de-energized) is provided. In addition, this solenoid valve 1 is not limited to the 3-way solenoid valve of the type mentioned above, 3-position may be sufficient, a normally open type may be sufficient, a 4-way solenoid valve etc. may be sufficient, Based on these arbitrary combinations It can be composed of various formations. In addition, in this embodiment, although the solenoid valve 1 is used as a pilot valve in the fluid pressure drive valve 10, the use of the solenoid valve 1 is not limited to this.

The spool part 2 has an input port 20 connected to the air supply source 14 through a first air pipe 140 and an output port 20 connected to the drive device 12 through a second air pipe 141 ( 21 ) and an exhaust port 22 for discharging exhaust gas from the drive device 12 .

The solenoid portion 3 displaces the spool portion 2 to communicate between the input port 20 and the output port 21 when energized, and between the output port 21 and the exhaust port 22 when de-energized. The spool part 2 is displaced to communicate with the

With the above-described series of configurations, when the solenoid valve 1 is energized, the air A (supply air) from the air supply source 14 flows through the first air pipe 140 , the input port 20 , Flowing through the output port 21 and the second air pipe 141 in this order, and being supplied to the air supply/exhaust port 124 , the second piston 122B is pressed and the coil spring 123 is compressed. And, when the valve shaft 13 is rotationally driven through the piston rod 121 and the transmission mechanism 125 as much as the piston rod 121 moves according to the compression of the coil spring 123 , the valve in the valve box 110 is The sieve 111 rotates, and the main valve 11 is operated in a fully opened state.

On the other hand, when the solenoid valve 1 is in a non-energized state, the air A (exhaust) in the cylinder 120 flows from the air supply/exhaust port 124 to the second air pipe 141 , the output port 21 and It flows in the order of the exhaust port 22, and is discharged|emitted to the outside air, so that the urging force of the 2nd piston 122B falls, and the coil spring 123 restores|restores from a compressed state. And, when the valve shaft 13 is rotationally driven through the transmission mechanism 125 as much as the piston rod 121 moves according to the restoration of the coil spring 123 , the valve body 111 rotates in the valve box 110 . Thus, the main valve 11 is operated in a fully closed state.

2 : is sectional drawing which shows an example of the solenoid valve 1 which concerns on one Embodiment of this invention. As shown in Fig. 2, the solenoid valve 1 according to the present embodiment includes a plurality of sensors ( 4), a substrate 5 on which at least one of the plurality of sensors 4 is mounted, a spool portion 2, a solenoid portion 3, a accommodating portion for accommodating the plurality of sensors 4 and the substrate 5 (6) is provided.

The accommodating part 6 includes a first accommodating part 60 for accommodating the spool part 2, and adjacent to the first accommodating part 60, as well as a solenoid part 3, a plurality of sensors 4 and a substrate ( 5) and a second accommodating part 61 for accommodating it, and a terminal box 62 to which the communication cable 150 and the power cable 160 are connected. The first accommodating portion 60 and the second accommodating portion 61 are made of, for example, a metal material such as aluminum.

The first accommodating portion 60 has openings (not shown) that each function as an input port 20 , an output port 21 , and an exhaust port 22 .

The second accommodating part 61 includes a cylindrical housing 610 having both ends (a first housing end 610a and a second housing end 610b) open, and a body 611 disposed inside the housing 610 . ), a solenoid cover 612 that covers the solenoid part 3 fixed to the first housing end 610a from outside air, and a terminal box 62 fixed to the second housing end 610b from outside air. A cover 613 is provided.

The housing 610 is formed at the lower portion of the shaft insertion hole 610c into which the second end 130B of the valve shaft 13 is inserted, and the body insertion hole 610d formed on the upper portion and into which the body 611 is inserted; , is formed on the side of the second housing end 610b and has a cable insertion hole 610e into which the communication cable 150 and the power cable 160 are inserted.

In the first accommodating part 60 and the second accommodating part 61 , the body 611 is branched from the input-side flow path 26 to communicate between the input-side flow path 26 and the first pressure sensor 40 . a first flow path 63 to be used, a second flow path 64 branching from the output side flow path 27 and communicating between the output side flow path 27 and the second pressure sensor 41, the spool part 2 and the solenoid part A spool flow path 65 through which air A for interlocking (3) flows is formed.

The spool part 2 includes a spool hole 23 formed in a second accommodating part 61 functioning as a spool case, a spool valve 24 movably disposed in the spool hole 23 , and a spool valve 24 . The spool spring 25 for elastically supporting the 27 and an exhaust passage 28 communicating between the exhaust port 22 and the spool hole 23 .

The solenoid unit 3 includes a solenoid case 30 , a solenoid coil 31 accommodated in the solenoid case 30 , a movable iron core 32 movably disposed in the solenoid coil 31 , and a solenoid coil 31 . A fixed iron core 33 disposed in a fixed state therein, and a solenoid spring 34 for elastically supporting the movable iron core 32 are provided.

When the solenoid valve 1 is switched from the non-energized state to the energized state, in the solenoid part 3, a coil current flows through the solenoid coil 31 so that the solenoid coil 31 generates an electromagnetic force, and the electromagnetic force As a result, the movable iron core 32 resists the elastic force of the solenoid spring 34 and is attracted to the fixed iron core 33 , whereby the circulation state of the air A flowing through the spool flow path 65 is switched. Then, in the spool part 2 , when the circulation state of the air A flowing through the spool flow path 65 is switched, the spool valve 24 is moved against the elastic force of the spool spring 25 , so that the input The state of communication between the port 20 and the exhaust port 22 is switched to the state of communication between the input port 20 and the output port 21 .

The substrate 5 includes a first substrate 50 arranged so that the substrate surfaces 500A and 500B are along the valve axis 13 inserted from the shaft insertion hole 610c, and a second substrate arranged close to the terminal box 62 . A second substrate (51) and a third substrate (52) arranged adjacent to the solenoid portion (3) are provided.

Among the substrate surfaces 500A and 500B of the first substrate 50 , the body 611 , the solenoid unit 3 , and the third substrate 52 are disposed on the first substrate surface 500A side. The second substrate 51 and the terminal box 62 are disposed on the side of the second substrate surface 500B opposite to the side of the first substrate surface 500A.

A sensor 4 is disposed at positions of the first substrate 50 , the second substrate 51 , and the third substrate 52 . The sensor 4 includes, for example, a first pressure sensor 40 that measures the fluid pressure of the air A flowing through the input-side flow path 26 and the first flow path 63 , the output-side flow path 27 , and The second pressure sensor 41 for measuring the fluid pressure of the air A flowing through the second flow path 64 and the rotation angle when the valve shaft 13 is rotationally driven are measured, and according to the rotation angle, the main and a main valve opening degree sensor 42 for acquiring valve opening degree information of the valve 11 .

The main valve opening degree sensor 42 is, for example, constituted by a magnetic sensor, and the permanent magnet 131 mounted on the second end 130B of the valve shaft 13 measures the magnetic strength generated, The valve opening degree information of the main valve 11 is acquired according to the magnetic strength. This main valve opening degree sensor 42 is located around the axis of the valve shaft 13 among the first substrate surfaces 500A of the first substrate 5 arranged so as to be along the valve axis 13 inserted from the shaft insertion port 610c. It is preferable to be placed in a position opposite to the outer periphery of the. Thereby, in the accommodating part 6, the main valve opening degree sensor 42 mounted on the 1st board|substrate 50 and the 2nd end 130B of the valve shaft 13 are brought close to each other without wasting an arrangement space. It becomes possible to arrange|position, and valve opening degree information can be acquired accurately.

3 : is a block diagram which shows an example of the solenoid valve 1 which concerns on one Embodiment of this invention. As shown in FIG. 3 , the solenoid valve 1 is an example of an electrical configuration. In addition to the substrate 3 and the sensor 4 described above, a control unit 7 for controlling the solenoid valve 1 , and an external device 15 . ) and a communication unit (8) for communicating with, and a power supply circuit unit (9) connected to an external power source (16).

The plurality of sensors 4 is a sensor group for measuring the physical quantity of each part, and in addition to the first pressure sensor 40 , the second pressure sensor 41 , and the main valve opening degree sensor 42 , the solenoid part 3 is A voltage sensor 43 for measuring the supply voltage to the air conditioner, a current/resistance sensor 44 for measuring a current value during energization and a resistance value when not energized in the solenoid part 3, and the internal temperature of the accommodating part 6 A temperature sensor 45 for measuring , and a magnetic sensor 46 for measuring the intensity of magnetism generated by the solenoid unit 3 are provided.

In addition, the plurality of sensors 4 are a sensor group for acquiring information about the operation history of each part, and as the operation time of the solenoid part 3, at least one of the sum of the energization time for the solenoid part and the current energization interlocking time An operation time meter (timer) 47 to measure, and an operation counter (counter) 48 for counting the number of operations of the solenoid valve 1 , the drive device 12 , and the main valve 11 are provided.

In addition, these sensors 40 to 48 are not limited to those in which each sensor is individually formed as described above, and since a specific sensor also functions as another sensor, the other sensors may not be formed individually. . For example, the magnetic sensor 46 measures the magnetic strength generated by the solenoid unit 3 and, based on the magnetic strength, obtains the current value when the solenoid unit 3 is energized. , the current/resistance sensor 44 does not need to be formed individually. In addition, the microcontroller 70 may incorporate the function of a sensor, or may implement|achieve a part of the function of a sensor, For example, the microcontroller 70 may set the operation time meter 47 and the operation counter 48 to By incorporating it, the operation time meter 47 and the operation counter 48 do not need to be formed separately.

The control unit 7 processes information indicating the state of each part of the solenoid valve 1 acquired by the plurality of sensors 4 and a microcontroller 70 for controlling each part of the solenoid valve 1; A valve test switch 71 is provided which controls the energization state of the solenoid part 3 and performs the opening/closing operation of the main valve 11 at the time of a test operation.

The microcontroller 70 includes a processor (not shown) such as a CPU (Central Processing Unit), and a memory configured by a ROM (Read Only Memory), a RAM (Random Access Memory), or the like. The microcontroller 70 may include a function for realizing the data processing system 300 described later in the present embodiment.

The valve test switch 71 receives a command from the microcontroller 70 when a predetermined test operation condition is satisfied, and executes a stroke test of the fluid pressure driving valve 10 as a test operation.

The stroke test is executed, for example, by any of a full stroke test and a partial stroke test. In the full stroke test, the main valve 11 is operated to the fully closed state by switching from the energized state to the non-energized state in the fully open state, and returns to the fully open state by switching from the non-energized state to the energized state in the fully closed state. is executed as The partial stroke test is performed by switching the main valve 11 from the energized state to the non-energized state in the fully open state without operating the main valve 11 to the fully closed state (that is, without stopping the plant equipment) to a predetermined opening degree. It is executed by returning to the fully open state by switching from the non-energized state to the energized state in the partially closed state.

In addition, as the test driving condition, for example, an execution time according to an execution frequency (eg, once a year) specified as a set value by an administrator or a specific specified date and time arrives, or When an execution command is received or a test execution button (not shown) provided on the solenoid valve 1 is operated by an administrator, the test operation conditions may be satisfied and the test operation (stroke test) may be executed.

(machine learning device)

In the fluid pressure actuation valve 10 provided with the above-mentioned series of structures, by providing the some sensor 4 mentioned above, for example, at the time of normal operation and abnormal operation (for example, opening/closing operation is Various types of information of the fluid pressure actuation valve 10 can be acquired during the test run and emergency stop performed. Therefore, in the following, machine learning for learning an inference model (learning completed model) capable of estimating diagnostic information of the fluid pressure actuation valve 10 based on information (state variable) obtainable from the fluid pressure actuation valve 10 . The device 200 will be described. In addition, the machine learning apparatus 200 referred to herein is not only provided as a device that operates alone, but also a program for causing an arbitrary processor to execute the operation described below, or one to plural for executing the operation. and provided in the form of a non-transitory computer-readable medium having instructions thereon.

4 is a schematic block diagram of a machine learning apparatus 200 according to an embodiment of the present invention. The machine learning apparatus 200 which concerns on this embodiment, as shown in FIG. 4, the data set acquisition unit 201 for learning, the data set storage unit 202 for learning, the learning unit 203, and learning completed model memory|storage A unit 204 is provided.

The learning data set acquisition unit 201 is an interface unit for acquiring a plurality of data constituting the learning (training) data set from various devices connected via, for example, a wired or wireless communication line. Here, as various devices connected to the data set acquisition unit 201 for learning, the operator's computer PC1 used by the operator of the external device 15 and the fluid pressure actuation valve 10, etc. are mentioned, for example. . In addition, although the example in which the external device 15 and the computer PC1 were connected separately is shown in FIG. 4, the external device 15 and the computer PC1 for workers may be comprised by the same computer. In this data set acquisition unit 201 for learning, the detection data of the plurality of sensors 4 of the fluid pressure actuation valve 10 are acquired as input data from, for example, the external device 15, and the input data Diagnostic information of the fluid pressure actuation valve 10 associated with it can be acquired as output data, for example from the computer PC1 for an operator. Then, the input data and output data correlated with each other constitute a learning data set of 1, which will be described later.

5 is a diagram showing a configuration example (learning with a teacher) data used in the machine learning apparatus 200 according to an embodiment of the present invention. 6 is a diagram showing a configuration example (learning without a teacher) of data used in the machine learning apparatus 200 according to an embodiment of the present invention. 5 and 6 are also appropriately referred to in the description of the data processing system and the reasoning apparatus.

As shown in Figs. 5 and 6, the training data set is, as input data, at least time series data of the valve opening degree of the main valve 11 in a predetermined period, and the pressure of the solenoid valve output side of the air A Time-series data, as output data, includes diagnostic information of the fluid pressure actuation valve 10, and points to a data set for use in machine learning to be described later. Although an example is demonstrated and shown below about the detail of these various data, this invention is not limited to these.

The valve opening degree of the main valve 11 indicates the value of the open/closed state of the main valve 11 , and can be acquired from the above-described main valve opening degree sensor 42 .

The pressure on the solenoid valve output side of the air A is the pressure of the air A supplied and exhausted from the solenoid valve 1 to the drive device 12 during a predetermined period, and the air (A) from the solenoid valve 1 to the drive device 12 is The pressure of the air A (supply air) when A) is supplied and the air A (exhaust air) when the air A is discharged from the drive device 12 to the outside through the solenoid valve 1 includes pressure. The pressure on the output side of the solenoid valve of the air A can be acquired by the second pressure sensor 41 described above.

The time series data is constituted by a plurality of data each acquired at a plurality of different time points within a predetermined period, and is acquired, for example, at a predetermined sampling period. In this embodiment, the time series data of the valve opening degree of the main valve 11 and the time series data of the pressure on the output side of the solenoid valve of air A are acquired at a plurality of time points in the same sampling period and in the same phase (state without phase difference). However, at least one of the sampling period and the phase may be different.

The predetermined period is a period during which the opening/closing operation of the main valve 11 is performed during abnormal operation (including test operation and emergency stop), for example, a stroke test in the fluid pressure actuation valve 10 is performed It consists of the execution period when it is done. The predetermined period may be the entire period from the start of the test to the end of the test in the execution period of the stroke test, or may be a part of the period. Therefore, the predetermined period may be, for example, the entire period of the full stroke test (full open state → fully closed state → fully open state) or the entire period of the partial stroke test (fully open state → partially closed state), and the full stroke The duration of a part of the test (fully open → fully closed, or fully closed → fully open, etc.) or part of the partial stroke test (fully open → partially closed, partially closed → fully open) state, etc.) may be sufficient, and it is not limited to these.

The diagnostic information of the fluid pressure actuation valve 10 is information indicating whether some kind of abnormality has occurred in the fluid pressure actuation valve 10 when the abnormal diagnosis of the fluid pressure actuation valve 10 is performed, and the data A number of formats can be employed. Abnormalities include not only post-mortem abnormalities in which the occurrence of abnormalities was found at the time when abnormality diagnosis was made, but also abnormal signs in which occurrence of abnormalities was predicted in the future, although within the allowable range that is judged to be normal at the time when abnormal diagnosis was performed.

For example, as shown in FIG. 5, the diagnostic information as an aspect can be comprised with the information which shows either that the fluid pressure actuation valve 10 is normal (no abnormality) and abnormal (abnormality exists). In this case, the diagnostic information is classified as binary, for example, a value indicating that the fluid pressure actuating valve 10 is in a normal state is “0”, and it is determined that the fluid pressure actuating valve 10 is in an abnormal state. What is necessary is just to set the indicated value to "1", and to input the corresponding value in the form associated with the input data by the operator using the work computer PC1. In addition, in this case, the information regarding the specific above content is not necessarily required.

In addition, as shown by the broken line in FIG. 5, the information indicating abnormality among the said diagnostic information may also contain the specific content of the above. As for the above specific content, for example, operation failure of the main valve 11, a failure of the air A circuit, an abnormality of the supply pressure of the air A from the air supply source 14, the solenoid part 3 The supply voltage abnormality, deterioration/damage of the solenoid coil 31, short circuit of the electric circuit, the life of the solenoid part 3, abnormal heat generation of the fluid pressure actuation valve 10, malfunction of the iron core, etc. are included. In this case, the diagnostic information is classified as Multi Valued (3 or more), for example, a value indicating that the fluid pressure actuation valve 10 is in a normal state is "0", and the fluid pressure actuation valve 10 A value indicating an abnormality corresponding to an operation failure of the main valve 11 is “1”, and a value indicating that the fluid pressure actuation valve 10 is an abnormality corresponding to a failure of the air (A) circuit is “2”, hereinafter the same After a value is uniquely determined in advance according to the contents of each above item, the corresponding value is input in a form related to the input data by the operator using the work computer (PC1). By setting such diagnostic information, not only the presence or absence of occurrence of an abnormality, but also information on the specific content of the abnormality when the abnormality occurs (corresponding to abnormality 1/more than 2/more than n shown in Fig. 6) is included in the learning data set can prepare The diagnostic information as an aspect described above is used when learning with a teacher (refer to FIG. 5 ) in machine learning described below.

In addition, as diagnostic information of the fluid pressure actuation valve 10, it is employable other than what was mentioned above. For example, as shown in FIG. 6, the diagnostic information as another aspect can be made into information which shows only that the fluid pressure actuation valve 10 is not abnormal but normal. In this case, since the diagnostic information includes only information indicating that the fluid pressure actuated valve 10 is normal, inevitably, the learning data set including this diagnostic information as output data is the same as when the fluid pressure actuated valve 10 is normal. It is only a data set consisting of input data and output data. Accordingly, since the output data of the training data set in this case is always the same, it will be understood by those skilled in the art that the training data set does not necessarily have the output data as data. The diagnostic information as another aspect is used when learning without a teacher (refer to FIG. 6 ) in machine learning described below.

Optionally, the input data in the training data set includes time series data of the air A pressure (excluding the solenoid valve output side pressure of the air A) time series data, time series data of the control parameters of the solenoid unit 3, Time series data of the temperature of the fluid pressure actuation valve 10 , the total operating time of the fluid pressure actuation valve 10 , the operating time after the last power supply to the fluid pressure actuation valve 10 , the main valve 11 The number of operations, the number of operations of the driving device 12 , the number of operations of the solenoid unit 3 , and the opening/closing time of the main valve 11 may be selectively included.

The pressure of the air A (except the pressure on the solenoid valve output side of the air A) is preferably the pressure of the air A flowing through each part of the fluid pressure actuation valve 10, specifically, air At least one of the pressure on the solenoid valve input side of the air (A) supplied from the supply source (14) to the solenoid valve (1), and the differential pressure between the pressure on the solenoid valve input side of the air (A) and the pressure on the solenoid valve output side of the air (A) It is preferable to include

The control parameters of the solenoid unit 3 are various parameter information for controlling the solenoid unit 3, specifically, the supply voltage supplied to the solenoid coil 31, the current value when the solenoid coil 31 is energized, the solenoid It is preferable to include at least one of the resistance value when the coil 31 is de-energized, the operating time of the solenoid unit 3 , and the magnetic strength generated in the solenoid unit 3 . Among these, the supply voltage supplied to the solenoid coil 31 can be acquired by the voltage sensor 43 described above, and the current value when the solenoid coil 31 is energized, and the resistance value when the solenoid coil 31 is not energized. The magnetism generated in the solenoid part 3 can be acquired by the above-mentioned current/resistance sensor 44, and the operation time of the solenoid part 3 can be acquired by the operation time meter 47 mentioned above. The intensity of can be acquired by the magnetic sensor 46 described above.

The temperature of the fluid pressure actuation valve 10 indicates a value of the internal temperature in particular of the fluid pressure actuation valve 10 , and can be acquired by the temperature sensor 45 described above.

The total operating time of the fluid pressure actuation valve 10 and the operating time after the last power supply to the fluid pressure actuation valve 10 is turned on can be obtained by the above-described operating time meter 47, and the main valve ( The number of operations 11) and the number of operations of the driving device 12 and the number of operations of the solenoid part 3 can be obtained by the above-described operation counter 48, and the opening and closing times of the main valve 11 are not shown. It can be obtained by using a timer that is not

As described above, increasing the type of input data generally contributes to improving the estimation precision of the learned model obtained after machine learning, but adopting input data with a low degree of correlation with diagnostic information is rather a learned model. may impede the improvement of the estimation precision of Accordingly, the number and type of data employed as input data should be appropriately selected in consideration of the state of the fluid pressure actuating valve 10 to which the learned model is applied, and the like.

The training data set storage unit 202 associates a plurality of data for constituting the training data set acquired by the training data set acquisition unit 201 as one training data set by associating the related input data and output data, and stores It is a database for About the specific structure of the database which comprises this data set storage unit for learning, it can adjust suitably. For example, in Fig. 4, for convenience of explanation, this learning data set storage unit 202 and a learning completed model storage unit 204 described later are shown as separate storage means, but these are a single storage medium ( database) can also be configured.

The learning unit 203 learns the correlation between the input data and the output data included in the plurality of learning data sets by executing machine learning using the plurality of learning data sets stored in the learning data set storage unit 202. To create a learning completion model. In the present embodiment, as a specific method of machine learning, teacher learning using a neural network is employed, as will be described and shown in detail later. However, the specific method of machine learning is not limited thereto, and other learning methods can be employed as long as the correlation between input and output can be learned from the training data set. For example, ensemble learning (random forest, boosting, etc.) may be used.

The learned model storage unit 204 is a database for storing the learned model generated in the learning unit 203 . The learned model stored in this learned model storage unit 24 is applied to the real system through a communication line including the Internet or a storage medium, as required. The specific application aspect of the learning completion model with respect to a real system (data processing system 300) is described in detail later.

Next, the learning method in the learning unit 203 using the several data sets for learning obtained as mentioned above is demonstrated centering around learning with a teacher. Fig. 7 is a diagram showing an example of a neural network model for learning with a teacher implemented in the machine learning apparatus according to an embodiment of the present invention. The neural network in the neural network model shown in Fig. 7 includes l neurons (x1 to xl) in the input layer, m neurons (y11 to y1m) in the first intermediate layer, and n neurons (y21) in the second intermediate layer. y2n), and o neurons (z1-zo) in the output layer. The first intermediate layer and the second intermediate layer are also called hidden layers, and the neural network may have a plurality of hidden layers in addition to the first intermediate layer and the second intermediate layer, or only the first intermediate layer is a hidden layer. okay In addition, although FIG. 7 exemplifies a neural network model in which a plurality (o) output layers are set, for example, when the diagnostic information described above is specified from a value of 1, that is, the number of teacher data described later In the case of one (t1 only), the number of neurons in the output layer can also be one (z1 only).

Also, between the input layer and the first intermediate layer, between the first intermediate layer and the second intermediate layer, and between the second intermediate layer and the output layer, there are nodes connecting the neurons between the layers, and each node has a weight wi (i is a natural number) is correspondingly built.

The neural network in the neural network model according to the present embodiment uses the training data set, the time series data of the valve opening degree of the main valve 11, the time series data of the pressure on the solenoid valve output side of the air A, and the fluid The correlation between the diagnostic information of the pneumatic actuation valve 10 is learned. Specifically, each of the time series data of the valve opening degree of the main valve 11 as the state variable and the time series data of the pressure on the output side of the solenoid valve of air A is mapped to the neurons of the input layer, and the values of the neurons in the output layer is a method of calculating the output value of a general neural network, that is, the value of the output side neuron is the value of the input side neuron connected to the neuron, and the weight wi associated with the node connecting the output side neuron and the input side neuron Calculation as the sum of the sequence of multiplication values is calculated by using a method that is performed for all neurons other than the neurons in the input layer. In addition, when inputting the state variable to the neurons of the input layer, the format in which the information acquired as the state variable is input can be appropriately set in consideration of the precision of the completed learning model to be generated, and the like. Specifically, in order to adjust the number of neurons corresponding to each input data, or to adjust the number of neurons to a value that can correspond to the neurons, preprocessing may be performed on specific input data.

Then, the calculated values of the o neurons z1 to zo in the output layer, that is, in the present embodiment, one or more diagnostic information and teacher data t1 to t1 to one or more diagnostic information constituting a part of the learning data set. , respectively, to obtain an error, and to adjust the weight wi corresponding to each node (backpropagation) so that the obtained error is small.

Then, when a predetermined condition such as repeating the above-described series of steps a predetermined number of times or that the error becomes smaller than an allowable value is satisfied, learning is finished and the neural network model ( All corresponding weights wi) are stored in the learned model storage unit 204 as the learned model.

(machine learning method)

In connection with the above, the present invention provides a machine learning method. The machine learning method which concerns on this invention is demonstrated below with reference to FIG.5 (learning phase), FIG.6 (learning phase), and FIG.7, FIG.8. 8 is a flowchart showing an example of a machine learning method according to an embodiment of the present invention. In the machine learning method shown below, although description is performed based on the machine learning apparatus 200 mentioned above, about the structure used as a premise, it is not limited to the said machine learning apparatus 200. In addition, this machine learning method is realized by using a computer, and various things are applicable as a computer, for example, the computer device which comprises the external device 15, work computer PC1, or the microcontroller 70. or a server device disposed on a network. In addition, about the specific configuration of this computer, for example, an arithmetic unit composed of at least a CPU or GPU, a storage unit composed of volatile or nonvolatile memory, etc., a communication unit for communicating with a network or other equipment; One including a bus for connecting each of these devices can be employed.

Teacher learning as a machine learning method according to the present embodiment is a preliminary preparation for starting machine learning, first, a desired number of learning data sets (refer to FIG. 5 ) are prepared, and the prepared plurality of learning data sets are used for learning. It stores in the data set storage unit 202 (step S11). The number of training data sets prepared here may be set in consideration of the inference precision required for the finally obtained learning completed model.

Several methods can be employed to prepare the training dataset used for this teacher-education training. For example, when the opening/closing operation of the main valve 11 is performed over a predetermined period by a stroke test or the like, when an abnormality occurs in the specific fluid pressure driving valve 10, or when an operator recognizes an abnormality In the present invention, various types of information of the fluid pressure actuating valve 10 in a predetermined period at that time are acquired using a plurality of sensors 4 or the like, and the operator transmits diagnostic information in a form correlating with these information to a work computer. By specifying and inputting using (PC1) or the like, the input data and output data constituting the learning data set (for example, the value of the output data in this case is “1”) are prepared. And by repeating such an operation, a method of preparing a desired number of training data sets can be employed. In addition, as a method of preparing a data set for learning, other than this method, various methods such as acquiring a data set for learning by actively creating an abnormal state in the fluid pressure actuation valve 10 can be adopted. there is. However, since various types of information of the fluid pressure actuation valve 10 have a tendency peculiar to each of the fluid pressure actuation valves 10 in many cases, the target for acquiring the data constituting the data set for learning is a machine to be described later. It is preferable to collect only from the fluid pressure actuation valve 10 of 1 to which the learned model obtained through learning is to be applied. In addition, as a data set for learning, it is comprised of input/output data when an abnormality has occurred, and input data and output data ( For example, the value of the output data in this case includes a predetermined number of training data sets composed of "0").

When step S11 is completed, in order to start learning in the learning unit 203 next, the neural network model before learning is prepared (S12). The neural network model before learning prepared here has, for example, the structure shown in FIG. 7 as its structure, and the weight of each node is set as an initial value. Then, from a plurality of training data sets stored in the training data set storage unit 202, for example, a training data set of 1 is randomly selected (step S13), and the input data in the 1 training data set is prepared It is input to the input layer (refer FIG. 7) of the neural network model before learning (step S14).

Here, since the value of the output layer (refer to Fig. 7) generated as a result of the step S14 is generated by the neural network model before learning, in most cases, a value different from the desired result, that is, a value different from the correct diagnosis information. A value representing information. Then, next, machine learning is implemented using the diagnostic information as teacher data in the learning data set of 1 acquired in step S13, and the value of the output layer generated in step S13 (step S15). The machine learning performed here is, for example, a process of comparing the values of the output layer with the diagnostic information constituting the teacher data, and adjusting the weights associated with each node in the neural network model prior to learning so that a desirable output layer is obtained (back program). pargation) may be used. In addition, the number and format of the values output to the output layer of the neural network model before learning are the same number and format as the teacher data in the learning data set as the learning object.

If the machine learning mentioned here is specifically exemplified, if the diagnostic information constituting the teacher data is normal, it is defined as “0”, and when it is abnormal, it is composed of a certain value (binary classification), and When the value of the output data in the learning data set of 1 selected in step S13 is "1", the value of the output layer is a predetermined value of 0 to 1, specifically, for example, a value of "0.63" is output. Therefore, in step S15, if the same input data is input to the input layer, the weight associated with each node of the neural network model being trained is adjusted so that the value obtained by the neural network model being trained is close to "1". do.

When machine learning is performed in step S15, whether or not it is necessary to further machine learning is specified based on, for example, the remaining number of unlearned learning data sets stored in the learning data set storage unit 202 ( step S16). And when machine learning is continued (No in step S16), it returns to step S13, and when terminating machine learning (Yes in step S16), it moves to step S17. In the case of continuing the machine learning, the steps S13 to S15 are performed a plurality of times with respect to the neural network model being trained using the unlearned learning data set. The precision of the finally generated trained model is generally increased in proportion to this number of times.

When machine learning is finished (Yes in step S16), the neural network generated by adjusting the weights associated with each node by a series of processes is stored as a learned model in the learned model storage unit 204 ( Step S17), a series of learning processes are ended. The learning completion model memorized here can be applied to and used in the data processing system 300 to be described later.

In the learning process and machine learning method of the machine learning device described above, in order to generate one learned model, the precision is obtained by repeatedly executing the machine learning process a plurality of times for one (before learning) neural network model. It has been described and shown to obtain a fully trained model that is improved and applied to the data processing system 300 . However, the present invention is not limited to this acquisition method. For example, a plurality of learned models subjected to machine learning a predetermined number of times are stored as one candidate in the learned model storage unit 204, and a data set for judging validity is input to the plurality of learned model groups. You may make it generate|occur|produce (the value of the neuron of the output layer), and compare and examine the precision of the value specified in the output layer, and you may make it select one best learning completion model applied to the data processing system 300 . In addition, the data set for validity judgment should just be comprised from the same data set as the data set for learning used for learning, and what is not used for learning.

As described above, by applying the machine learning apparatus and machine learning method according to the present embodiment, from various data acquired by the plurality of sensors 4 formed in the proper position of the fluid pressure actuation valve 10, It is possible to obtain a learning completion model capable of accurately deriving diagnostic information indicating whether abnormalities and abnormalities (including precursors of abnormalities) occur.

In the learning method and machine learning method of the machine learning apparatus 200, "learning with a teacher" has been described, but as a method of generating a completed learning model, other known methods such as a convolutional neural network (CNN) The method of "learning with a teacher" may be used, and as the diagnostic information constituting the output data, only the diagnostic information related to the other aspects described above, that is, the fluid pressure actuating valve 10 is not abnormal but is normal. You may use "learning without a teacher" using a learning data set (refer to FIG. 6 ) including information indicating . Even when "learning without a teacher" is used, diagnostic information in the output data matched with the input data can be obtained only the information of the steady state of the fluid pressure actuation valve 10, the "learning phase" of FIG. As shown by ", a learned model can be obtained by learning the correlation representing the characteristics of the steady state of the input data and the output data. In this case, at the time of reasoning in the data processing system 300 described later, the input data determined not to match the characteristics of the normal state by a predetermined amount is not in the normal state, in other words, in an abnormal state, so that the inference of diagnostic information is performed. can be realized As a specific method of this "teacherless learning", for example, a well-known method using the auto-encoder etc. shown briefly in FIG. 6 can be used, and detailed description is abbreviate|omitted here.

(data processing system)

Next, with reference to FIG. 9 , an example of application of the completed learning model generated by the machine learning apparatus 200 and the machine learning method described above will be described and shown. 9 is a schematic block diagram showing a data processing system according to an embodiment of the present invention.

As the data processing system 300 which concerns on this embodiment, the aspect mounted in the microcontroller 70 of the fluid pressure actuation valve 10 mentioned above is illustrated. In addition, with respect to this data processing system 300, it is also possible to apply at least a part of it to other apparatuses, for example, another apparatus connected to the external device 15 or the fluid pressure actuation valve 10. As shown in FIG.

The data processing system 300 includes at least an input data acquisition unit 301 , an inference unit 302 , a trained model storage unit 303 , and a notification unit 304 .

The input data acquisition unit 301 is an interface unit for acquiring various types of data that are connected to the plurality of sensors 4 included in the fluid pressure actuation valve 10 and output by each sensor 4 . This input data acquisition unit 301 acquires input data including at least time series data of the valve opening degree of the main valve 11 and time series data of the pressure on the solenoid valve output side of the air A. In addition, in the example shown in FIG. 9, it is connected to all the sensors 4 including the main valve opening degree sensor 42 and the 2nd pressure sensor 41 so that input data usable for inference can be acquired, As to which sensor 4 to connect to the input data acquisition unit 301, it can select suitably according to the learning completion model etc. used in the inference unit 302 mentioned later. In addition, it is preferable to store the reasoning result of the reasoning unit 302 in a storage means (not shown), and the stored past reasoning result is, for example, an additional value of the reasoning precision of the learned model in the learned model storing unit 303 . For improvement, it can be used as a training data set used for online learning.

The inference unit 302 is for inferring whether or not an abnormality has occurred in the fluid pressure actuation valve 10 from the various data of the fluid pressure actuation valve 10 acquired by the input data acquisition unit 301 . For this reasoning, for example, a learned model trained using the machine learning apparatus 200 and machine learning method described above is used, and this learned model is a learning model storage unit composed of an arbitrary storage medium. It is stored in (303). In addition, this reasoning unit 302 learns by adjusting the input data acquired by the input data acquisition unit 301 into a desired format or the like as a pre-processing of the inference processing, as well as a function of performing inference processing using the learned model. As a pre-processing function input to the completed model or post-processing of inference processing, by applying a predetermined threshold value to the output value output by the learned model, for example, the occurrence of abnormality (including post-mortem abnormalities and abnormal precursors) occurs ( It also includes a post-processing function that finally determines whether there is an abnormality (normal) or abnormality (abnormality).

The learned model storage unit 303 is a storage medium for storing the learned model used in the inference unit 302 as described above. The number of learned models stored in this learned model storage unit 303 is not limited to one. For example, the number of input data is different or the learning method is different (for example, learning with a teacher and learning without a teacher performed in the machine learning apparatus 200 described above) a plurality of learning completion models may be stored and selectively made available.

The informing unit 304 is for notifying the inference result of the inference unit 302 to an operator or the like. Various specific notification means can be employed, for example, the reasoning result is transmitted to the external device 15 via the communication unit 8, displayed on the GUI of the external device 15, or driven by fluid pressure. By providing a light emitting member, a speaker, etc. in advance in the valve 10 and operating them, the presence or absence of an abnormal occurrence can be informed to an operator etc.

The data processing process by the data processing system provided with the above structure is demonstrated below with reference to FIG. 5 (inference phase), FIG. 6 (inference phase), and FIG. 10 is a flowchart showing an example of a data processing process by the data processing system 300 according to an embodiment of the present invention.

In the fluid pressure actuation valve 10, when the opening/closing operation of the main valve 11 is performed over a predetermined period by, for example, a stroke test or the like, and the diagnosis of the fluid pressure actuation valve 10 is started with it, , the input data acquisition unit 301 acquires various data indicating the state of each part of the fluid pressure actuation valve 10 acquired by the plurality of sensors 4 (step S21). Input data desired by the input data acquisition unit 301 (time series data of the valve opening degree of the main valve 11 in a predetermined period, and time series data of the solenoid valve output side pressure of air A (see Figs. 5 and 6 ) )) at a point in time when the inference unit 302 based on the input data is performed (step S22).

Specifically, the inference unit 302 pre-processes the input data and inputs it to the learning model complete model, and performs post-processing on the output value from the learning model complete model, thereby It is determined whether abnormality and abnormality are included). In the post-processing in learning with a teacher (refer to "inference phase" in Fig. 5), the inference unit 302 sets the output value of the learning model completion model (a number between 0 and 1 in case of binary classification) and a predetermined threshold value. By comparing, for example, if the output value of the learning model completion model is greater than or equal to a predetermined threshold, it is judged as “abnormal (abnormal)” and less than a predetermined threshold as “no abnormality (normal)”, and the judgment result is inferred. output as a result. In addition, in the post-processing in teacher-free learning (refer to "inference phase" in Fig. 6), the reasoning unit 302 is configured to generate the difference between the output value (feature amount) of the learning model completion model and the feature amount based on the input data. (distance) is calculated, and if the difference (distance) is greater than or equal to a predetermined threshold value, it is judged as “abnormal (abnormal)” and when the difference (distance) is less than a predetermined threshold, “no abnormality (normal)”, and the judgment result is output as an inference result .

Then, in step S22, when inference by the inference unit 302 is performed and the reasoning result indicates "No abnormality (normal)" (No in step S23), the series of speculation is continued. It returns to step S21. On the other hand, when the inference result indicates "abnormality (abnormal)" as shown in FIGS. 5 and 6 (Yes in step S23), the inference result is "abnormal" )", that is, that an abnormality (including a post-mortem abnormality and a precursor of abnormality) has occurred in the fluid pressure actuation valve 10 (step S24). Then, after the occurrence of the abnormality is notified in step S24, the flow returns to step S21 so as to continue a series of inferences. In addition, depending on the usage purpose of the fluid pressure actuation valve 10 and the content of the detected abnormality, you may make it respond|action such as stopping the fluid pressure actuation valve 10 at the stage where the abnormality was detected.

(inference device)

The present invention may provide not only with the aspect of the data processing system 300 described above, but also with the aspect of an inference apparatus for performing inference. In that case, the reasoning device may include a memory and at least one processor, and among them, the processor may execute a series of processing. The series of processing refers to the valve opening degree of the main valve 11 when the opening/closing operation of the main valve 11 is performed over a predetermined period in the fluid pressure drive valve 10 by a stroke test or the like, for example. processing of acquiring input data including time series data of the air A and time series data of the pressure on the output side of the solenoid valve of air A, and inferring diagnostic information in the fluid pressure actuation valve 10 when the input data is input including processing. By providing the present invention in the form of the above-described reasoning device, it is possible to apply the data processing system 300 to various fluid pressure actuated valves 10 simply compared to the case where the data processing system 300 is mounted. At this time, when the reasoning device performs the processing for inferring the diagnostic information, using the learning completed model learned by the machine learning device and the machine learning method of the present invention described so far in this manual, data It may be understood by a person skilled in the art that the reasoning method performed by the reasoning unit 302 of the processing system may be applied.

This invention is not limited to embodiment mentioned above, It is possible to carry out various changes within the range which does not deviate from the main point of this invention. And all of them are included in the technical idea of this invention.

One… Solenoid valve, 3... Solenoid part, 4... sensor, 10… hydraulically actuated valves, 11 . . . main valve, 12… (hydraulic) drive device, 14... air supply, 15... external device, 26… input side flow path, 27... output side flow path, 28... exhaust flow path, 30… Solenoid case, 31… Solenoid coil, 32… Moving iron core, 40... A first pressure sensor, 41... a second pressure sensor, 42... Main valve opening sensor, 43… voltage sensor, 44… Current/resistance sensor, 45… temperature sensor, 46… magnetic sensor, 47... Uptime meter (timer), 48... Operation counter (counter), 70… microcontroller, 100… Plumbing, 200… Machine Learning Apparatus, 201… A training data set acquisition unit, 202 . A data set storage unit for training, 203 ... learning unit, 204... Learning completed model memory unit, 300... data processing system, 301... input data acquisition unit, 302 . Inference unit, 303... Learning completed model memory unit, 304 . . . Notification unit, A… Air (drive fluid), PC1… work computer

Claims (8)

A machine learning apparatus applied to a fluid pressure actuation valve having at least a main valve, a drive device for driving the main valve, and an electromagnetic valve for controlling supply and exhaustion of a driving fluid with respect to the drive device,
input data including time series data of the valve opening degree of the main valve in a predetermined period and time series data of the pressure on the output side of the solenoid valve of the driving fluid supplied and discharged from the solenoid valve to the drive device during the predetermined period; a learning data set storage unit for storing a plurality of sets of learning data sets composed of output data comprising diagnostic information of the fluid pressure actuation valve corresponding to input data;
a learning unit for learning a learning model for inferring a correlation between the input data and the output data by inputting a plurality of sets of the training data sets;
and a learned model storage unit for storing the learning model learned by the learning unit.
The method of claim 1,
The diagnostic information is information indicating which of the fluid pressure actuated valve is normal or abnormal.
The method of claim 1,
The diagnostic information is information indicating that the fluid pressure actuating valve is not abnormal but is normal.
4. The method according to any one of claims 1 to 3,
The said predetermined period is a machine learning apparatus which consists of the execution period of the stroke test in the said fluid pressure actuation valve.
5. The method of claim 4,
The stroke test is any one of a partial stroke test and a full stroke test.
A data processing system used for a fluid pressure actuation valve comprising at least a main valve, a drive device for driving the main valve, and an electromagnetic valve for controlling supply and exhaustion of a drive fluid with respect to the drive device, the data processing system comprising:
Acquire input data including time series data of the valve opening degree of the main valve in a predetermined period and time series data of the pressure on the output side of the solenoid valve of the driving fluid supplied and exhausted from the solenoid valve to the drive device during the predetermined period an input data acquisition unit to
The input data acquired by the input data acquisition unit is input into the learned model generated by the machine learning apparatus according to any one of claims 1 to 5, and diagnostic information of the fluid pressure actuation valve is inputted. A data processing system comprising an inference unit for inference.
An inference device used in a fluid pressure actuation valve comprising at least a main valve, a drive device for driving the main valve, and an electromagnetic valve for controlling supply and exhaustion of a drive fluid with respect to the drive device,
The reasoning device comprises a memory and at least one processor;
the at least one processor
Acquire input data including time series data of the valve opening degree of the main valve in a predetermined period and time series data of the pressure on the output side of the solenoid valve of the driving fluid supplied and exhausted from the solenoid valve to the drive device during the predetermined period processing and
an inference device, configured to execute a process of inferring diagnostic information of the fluid pressure actuation valve when the input data is input.
A machine learning method using a computer, which is applied to a fluid pressure actuation valve having at least a main valve, a drive device for driving the main valve, and an electromagnetic valve for controlling supply and exhaustion of a drive fluid to the drive device, comprising:
input data including time series data of the valve opening degree of the main valve in a predetermined period and time series data of the pressure on the output side of the solenoid valve of the driving fluid supplied and discharged from the solenoid valve to the drive device during the predetermined period; a step of storing a plurality of sets of learning data sets composed of output data comprising diagnostic information of the fluid pressure actuation valve corresponding to the input data;
learning a learning model for inferring a correlation between the input data and the output data by inputting a plurality of sets of the training data sets;
and storing the learning model.

Images