Information processing device, machine learning device, inference device, information processing method, machine learning method, and inference method

Abstract

[Problem] The purpose of the present invention is to provide an information processing device, a machine learning device, an inference device, an information processing method, a machine learning method and an inference method which make it possible to verify the quantity of fluid passing through a valve mechanism, verify the presence or absence of unexpected fluid leakage from between a valve body and a valve seat, verify the quantity of the leakage, and verify a sign of unexpected fluid leakage from the valve mechanism. [Solution] An information processing device 250 of the present disclosure predicts the state of a valve mechanism, and derives a valve mechanism state which indicates the state of the valve mechanism, on the basis of input information that is a prescribed variable between the valve body and the valve seat.

Description

Information processing device, machine learning device, inference device, information processing method, machine learning method, and inference method

[0001] The present disclosure relates to an information processing device, a machine learning device, an inference device, an information processing method, a machine learning method, and an inference method.

[0002] Conventionally, a breather valve has been known that has a positive side valve seat having an outlet for the fluid in the container and a positive side valve that can open and close the outlet depending on the pressure in the container (see, for example, Patent Document 1).

[0003] Japanese Patent Application Laid-Open No. 2018-21652

[0004] In a valve mechanism consisting of a valve disc and a valve seat disposed in a fluid flow path, the importance of managing the fluid flowing therethrough requires that the valve mechanism not only reliably close the flow path but also be able to verify the amount of fluid passing through the valve mechanism, be able to verify the presence or absence of unexpected fluid leakage between the valve disc and the valve seat, and be able to verify the amount of leakage in the unlikely event that fluid leaks from the valve mechanism. Furthermore, to prevent or minimize fluid leakage from the valve mechanism due to unforeseen circumstances, it is also required to be able to verify signs of unexpected fluid leakage from the valve mechanism. In the case of handling volatile and flammable fluids, or fluids harmful to humans and the environment, this requirement has become even greater in recent years due to increased concern for animals, plants, and the environment. However, conventional valve mechanisms have a problem in that they are unable to verify the amount of fluid passing through the valve mechanism, the presence or absence of unexpected fluid leakage between the valve disc and the valve seat, the amount of leakage, or signs of unexpected fluid leakage from the valve mechanism until they close and open the flow path. This problem has not been solved at all even in the valve mechanism used in the breather valve that handles volatile fluids, which is shown as an example of a conventional valve mechanism, and exists as a general problem with valve mechanisms.

[0005] The present disclosure has been made to solve the above-mentioned problems, and aims to provide an information processing device, a machine learning device, an inference device, an information processing method, a machine learning method, and an inference method that are capable of verifying the amount of fluid passing through a valve mechanism, verifying whether or not there is unexpected fluid leakage from between the valve body and the valve seat, verifying the amount of leakage, and verifying signs of unexpected fluid leakage from the valve mechanism.

[0006] The information processing device according to the present disclosure is an information processing device that predicts the state of a valve mechanism, and determines a valve mechanism state that indicates the state of the valve mechanism based on input information that is a predetermined variable between a valve disc and a valve seat.

[0007] The information processing device, machine learning device, inference device, information processing method, machine learning method, and inference method disclosed herein can verify the amount of fluid passing through the valve mechanism, verify whether or not there is unexpected fluid leakage from between the valve body and the valve seat, verify the amount of leakage, and verify signs of unexpected fluid leakage from the valve mechanism.

[0008] 6 is a schematic diagram showing a valve system according to embodiment 1. FIG. 7 is a schematic diagram showing the breather valve of FIG. 1. FIG. 8 is a top view showing the introduction section of FIG. 2. FIG. 9 is an enlarged view showing a portion included in the detection device of FIG. 2. FIG. 10 is a top view showing the valve contact member of FIG. 2. FIG. 11 is a configuration diagram showing the machine learning device of FIG. 1. FIG. 12 is a conceptual diagram showing elements of machine learning performed by the machine learning device of FIG. 6. FIG. 13 is a schematic diagram showing the information processing device of FIG. 1. FIG. 14 is a hardware configuration diagram of a computer. FIG. 15 is a flowchart showing a machine learning method by the machine learning device of FIG. 1. FIG. 16 is a flowchart showing a learning valve mechanism state prediction method by the information processing device of FIG. 1. FIG. 17 is a top view showing an introduction section according to embodiment 2.

[0009] Hereinafter, embodiments for carrying out the present disclosure will be described with reference to the drawings. Note that the scope necessary for the explanation to achieve the object of the present disclosure will be schematically shown below, and the scope necessary for explaining the relevant parts of the present disclosure will be mainly explained, and the parts for which explanation is omitted will be considered to be publicly known technologies.

[0010] Embodiment 1. In the description of the valve system of the present disclosure, in embodiment 1, a valve system applied to a breather valve is used. However, the valve system of the present disclosure is not limited to applications to breather valves, and can be used appropriately for general valve mechanisms. The same applies to the valve systems in embodiments 2 and 3 described below.

[0011] 1 is a schematic diagram showing a valve system 200 according to embodiment 1. The valve system 200 includes a breather valve 5, a machine learning device 240, an information processing device 250, and a network 270 connecting the devices together.

[0012] Figure 2 is a schematic diagram showing the breather valve 5 of Figure 1. The breather valve 5 is connected to the opening of the tank 4 via the upstream flange 4b. In Figure 2, a part of the breather valve 5 is shown in cross section along a plane including the axis of the flow path.

[0013] The tank 4 stores a flammable fluid such as a gas or liquid. The tank 4 also contains a volatilized version of the stored fluid. Examples of the flammable gas or liquid include fossil fuels and volatile gases. The shape of the tank 4 may be spherical, cylindrical, rectangular, cuboid, or the like.

[0014] In the first embodiment, the tank 4 is shaped so that the horizontal cross-sectional area of ​​the tank 4 decreases toward the upper end. An attachment portion 4a is provided at the upper end of the tank 4. A through-hole that communicates with the interior of the tank 4 is formed in the attachment portion 4a. This allows the fluid inside the tank 4 to be efficiently discharged through the attachment portion 4a.

[0015] The base of the breather valve 5 is attached to the attachment portion 4a via the upstream flange 4b. In the first embodiment, the tank 4 is on the upstream side, and the breather valve 5 is installed on the downstream side of the tank 4.

[0016] The breather valve 5 can discharge the fluid inside the tank 4 to the atmosphere according to the internal pressure of the tank 4, and can also allow the atmosphere, which is the suction fluid, to flow into the tank 4. In other words, the breather valve 5 can adjust the pressure inside the tank 4.

[0017] In the first embodiment, the tank 4 and the breather valve 5 are installed outdoors and exposed to the atmosphere.

[0018] The breather valve 5 includes a main section 10, an intake section 20, and an exhaust section 40. The main section 10 is a T-shaped piping member. The ends of the T-shaped piping member of the main section 10 are an inlet-side opening end 10a that opens vertically downward, an outlet-side opening end 10b that opens vertically upward, and an intake-side opening end 10c that opens horizontally.

[0019] The main section 10 is installed with the inlet opening end 10a on the lower side and the outlet opening end 10b on the upper side. The flow path leading from the inlet opening end 10a to the outlet opening end 10b is called the main flow path 110. The main section 10 is installed so that the axis of the main flow path 110 is along the vertical direction. The main flow path 110 extends toward the discharge section 40. The main flow path 110 extending to the discharge section 40 will be described later.

[0020] A flow path that branches off from the main flow path 110 and reaches the intake side opening end 10c is referred to as a branch flow path 111. The axis of the branch flow path 111 is aligned horizontally. The intake section 20 is connected to the intake side opening end 10c of the main section 10.

[0021] The intake section 20 has an intake section main body 21 and an intake valve mechanism 22 provided on the intake section main body 21. The intake section main body 21 is a tubular member having two open ends.

[0022] One end of the intake unit main body 21 is a connection side opening end 21a that opens horizontally. The other end of the intake unit main body 21 is formed with an intake port 21b that opens vertically downward. The flow path connecting the connection side opening end 21a and the intake port 21b is referred to as an intake flow path 120. Here, the other end of the intake unit main body 21 that surrounds the intake port 21b is referred to as an intake valve seat 21x.

[0023] The intake valve mechanism 22 has an intake valve 23 which is an on-off valve, an intake valve shaft 24 fixed to the intake valve 23, and an intake valve guide 25 installed in the intake unit main body 21. When the intake valve 23 comes into contact with an intake valve seat 21x which surrounds the intake port 21b, the intake port 21b is closed. When a gap is created between the intake valve 23 and the intake valve seat 21x, the intake port 21b is opened. In other words, the intake valve 23 can open and close the intake flow path 120.

[0024] The intake valve guide 25 supports the intake valve shaft 24 so that it can move vertically. This allows the intake valve 23 to move between an intake closed position and an intake open position as the intake valve shaft 24 moves vertically. The intake valve 23 is an open valve. The intake valve 23 can move toward the intake closed position by falling along the intake valve guide 25 due to the weight of the intake valve 23 and the intake valve shaft 24.

[0025] When the intake valve 23 is in the intake closing position, the intake valve 23 closes the intake port 21b. That is, the lower surface of the intake valve 23 is in contact with the intake valve seat 21x, so the intake flow path 120 is closed and communication between the intake flow path 120 and the outside of the breather valve 5 is cut off.

[0026] When the intake valve 23 moves upward against its own weight from the intake closed position to the intake open position, a gap is created between the intake valve 23 and the intake valve seat 21xb. In other words, the intake flow path 120 is opened and communicates with the atmosphere outside the breather valve 5.

[0027] The intake side opening end 10c and the connection side opening end 21a are connected, and the intake section 20 is connected to the intake side opening end 10c of the main section 10. As a result, the intake flow path 120 inside the intake section main body 21 communicates with the branch flow path 111 inside the main section 10. In other words, the intake flow path 120 extends horizontally from the branch flow path 111.

[0028] The discharge section 40 is connected to the outlet opening end 10b of the main section 10. The discharge section 40 has a discharge section main body 41, a vent cover 47, a discharge valve mechanism 50 installed on the vent cover 47, a detection device 60, and a valve contact member 70.

[0029] The exhaust unit main body 41 is composed of a tubular inlet 42 and an atmosphere exhaust unit 43 provided to surround one open end of the inlet 42. The exhaust unit main body 41 is installed in an orientation in which the pipe of the inlet 42 is aligned vertically.

[0030] Fig. 3 is a top view showing the introduction portion 42 in Fig. 2. Fig. 4 is an enlarged view showing a portion including the detection device 60 in Fig. 2. Fig. 3 shows the top surface of the introduction portion 42, so the valve contact member 70 is not shown. The explanation will continue based on Figs. 2 to 4.

[0031] The downward-opening end of the inlet section 42 is referred to as the discharge section inlet opening end 42a. The vertically upward-opening end of the inlet section 42 is referred to as the valve side end 42b. The opening of the valve side end 42b is referred to as the valve side opening 42c, and the end face of the valve side end 42b is referred to as the main body end face 42d. The discharge section inlet opening end 42a is connected to the outlet opening end 10b of the main section 10.

[0032] A single detector installation space 42e is formed in the introduction portion 42. The detector installation space 42e opens to the end surface 42d of the main body, passes through the wall of the introduction portion 42 from the opening in the end surface 42d of the main body, and is a space that opens to the outer periphery of the introduction portion 42. A part of the detection device 60 is installed in the detector installation space 42e.

[0033] The detection device 60 has a detector 61 and wiring 62 that transmits a signal from the detector 61 to an input / output device (not shown). The detector 61 can detect a predetermined variable of a detection target.

[0034] At this time, the detector 61 can detect a predetermined variation of the detection target in the direction in which the detector 61 is pointed, and this direction is defined as the detection direction of the detector 61. In other words, the detection direction of the detector 61 is the direction in which the detector 61 can detect a predetermined variation of the detection target.

[0035] One detector 61 and a portion of the wiring 62 extending from the detector 61 are installed in the detector installation space 42e. The detector 61 is installed so that the detection direction of the detector 61 faces from the detector installation space 42e toward the outside of the main body end surface 42d. In other words, the detector 61 can detect a predetermined variable of the detection target located beyond the end surface on the valve-side end 42b side of the introduction part 42.

[0036] The wiring 62 is installed so as to pass through the detector installation space 42e and extend outward from the outer periphery of the introduction part 42. The wiring 62 is connected to an input / output device (not shown). The input / output device can transmit a predetermined variable detected by the detector 61 onto the network 270.

[0037] The network 270 is an information transmission network constructed by wire or wirelessly. The network 270 may be connected to the Internet to enable information transmission. Devices connected to the network 270 can acquire signals from the detector 61 transmitted over the network 270. Other devices connected to the network 270 will be described later.

[0038] A valve contact member 70 is provided at the valve side end portion 42b. Fig. 5 is a top view showing the valve contact member 70 of Fig. 2. The description will continue with reference to Figs. 2 to 5.

[0039] The valve contact member 70 is an annular member. The cross section of the valve contact member 70 taken along a line along the radial direction of the annular shape is L-shaped. That is, the valve contact member 70 is an annular flat plate member bent perpendicular to the plane along the entire inner periphery. The valve contact member 70 is formed in a shape corresponding to the shape of the main body end face 42d.

[0040] The valve contact member 70 is installed in the introduction portion 42 so as to cover the main body end surface 42d of the introduction portion 42 and a portion of the inner circumferential surface of the introduction portion 42 that continues from the main body end surface 42d. Here, the surface of the valve contact member 70 that faces the discharge valve 53 is referred to as a valve-facing surface 70a. The discharge valve 53 will be described later.

[0041] The opening of the detector installation space 42e formed in the main body end surface 42d is closed by installing the valve contact member 70 in the introduction portion 42. That is, the detector 61 is installed in the detector installation space 42e covered by the valve contact member 70. The detection direction of the detector 61 is toward the discharge valve 53 located beyond the valve-facing surface 70a.

[0042] The valve contact member 70 is installed by screwing it to the introduction portion 42. The outer periphery of the valve contact member 70 is provided with attachment points for screwing. However, other well-known methods may be used to install the valve contact member 70 to the introduction portion 42 in addition to screwing. The valve contact member 70 is preferably made of a material that does not significantly reduce the detection capability of the detector 61. For example, the valve contact member 70 is made of stainless steel.

[0043] It is preferable to design the shape of the valve contact member 70 so as not to significantly reduce the detection capability of the detector 61. For example, a significant reduction in the detection capability of the detector 61 can be prevented by thinning the valve contact member 70. As shown in FIG. 2-5, the inner circumferential surface of the introduction portion 42 where the detector installation space 42e is formed is formed to protrude further inward, and the wall thickness of the introduction portion 42 where the detector installation space 42e is formed is thicker than the wall thickness of the remaining portions. The valve contact member 70 is also formed in a shape corresponding to the shape of the main body end surface 42d. However, this is not limited to this. For example, the wall thickness of the introduction portion 42 may be uniform, i.e., the inner circumferential surface of the introduction portion 42 does not have a protruding portion formed inward, and the inner circumferential surface of the introduction portion 42 viewed from the axial direction may be circular. Even in this case, the valve contact member 70 is formed in a shape corresponding to the shape of the main body end surface 42d.

[0044] Here, the valve side end 42b of the inlet portion 42 is the valve seat main body 3. The valve seat main body 3 may be cylindrical and form part of a flow path that is closed by the valve disc, so that the valve side end 42b of the inlet portion 42 forms part of the discharge flow path 140. The main body end surface 42d of the inlet portion 42 is the end of the flow path formed in the valve seat main body 3 and is the main body end surface of the valve seat main body 3 that faces the valve disc. A detector installation space 42e is formed in the valve seat main body 3.

[0045] The valve seat main body 3, which is the valve-side end 42b of the inlet 42, and the valve contact member 70 are referred to as the valve seat 2. Therefore, the valve seat 2 constitutes part of the flow path. The valve seat 2 and the detection device 60 are referred to as the valve seat mechanism 1. In this case, the valve-facing surface 70a of the valve contact member 70 is referred to as the valve seat end surface 2a.

[0046] That is, the valve seat end face 2a faces the underside of the discharge valve 53, which is a valve body, and is in direct contact with the discharge valve 53 when the discharge valve 53 closes the discharge flow path 140. Furthermore, a gap is formed between the discharge valve 53 and the valve seat end face 2a, and when the discharge flow path 140 is opened, the fluid in the tank 4 is discharged through the gap between the discharge valve 53 and the valve seat end face 2a.

[0047] In addition, the detection direction is toward the discharge valve 53, which is a valve body, located beyond the valve seat end face 2a, when the detector 61 is installed in the detector installation space 42e.

[0048] Returning to Figure 2, the explanation will continue. The atmosphere exhaust portion 43 is formed so as to surround the valve side end portion 42b of the introduction portion 42. An opening is formed at the upper end side of the atmosphere exhaust portion 43. The upper end side of the atmosphere exhaust portion 43 is referred to as the valve mechanism side opening end portion 43b. Below the valve mechanism side opening end portion 43b of the atmosphere exhaust portion 43, i.e., in the body portion of the atmosphere exhaust portion 43, a plurality of discharge ports 43a that open horizontally are formed.

[0049] The vent cover 47 is connected to the valve mechanism side opening end 43b. The vent cover 47 closes the opening of the valve mechanism side opening end 43b. A well-known structure can be used to connect the vent cover 47 to the valve mechanism side opening end 43b. A discharge valve mechanism 50 is installed in the vent cover 47.

[0050] The discharge valve mechanism 50 has a discharge valve 53 which is a valve body capable of closing the valve side opening 42c, a discharge valve shaft 54 ​​fixed to the discharge valve 53, and a discharge valve guide 55 installed on the vent cover 47.

[0051] When the vent cover 47 is connected to the valve mechanism side opening end 43b, the discharge valve guide 55 extends vertically downward from the vent cover 47. The discharge valve guide 55 supports the discharge valve shaft 54 ​​so that the discharge valve shaft 54 ​​is movable along the vertical direction.

[0052] The discharge valve 53 is movable between a discharge closed position and a discharge open position by moving the discharge valve shaft 54 ​​along the vertical direction. The discharge valve 53 is an open valve.

[0053] The discharge valve 53 can move toward the discharge closed position by dropping along the discharge valve guide 55 due to the weight of the discharge valve 53 and the discharge valve shaft 54 ​​.

[0054] When the discharge valve 53 is in the discharge closed position, the discharge valve 53 closes the valve-side opening 42c. Specifically, when the discharge valve 53 is in the discharge closed position, the surface of the discharge valve 53 contacts the valve-facing surface 70a of the valve contact member 70. At this time, the surface of the discharge valve 53 and the valve-facing surface 70a are in continuous contact around the valve-side opening 42c.

[0055] In the first embodiment, the surface of the discharge valve 53 that comes into contact with the valve-facing surface 70a is the underside of the discharge valve 53. That is, the detection device 60 is installed to face the underside of the discharge valve 53, and can detect the state of the discharge valve 53.

[0056] When the discharge valve 53 moves against its own weight from the discharge closed position to the discharge open position, a gap is created between the discharge valve 53 and the valve-side opening 42c.

[0057] The discharge section 40 has a discharge flow path 140 that runs from the discharge section inlet opening end 42 a through the valve side opening 42 c to the plurality of discharge ports 43 a. When the discharge valve 53 is in the discharge closed position, communication between the main flow path 110 extending from the inside of the main section 10 and the discharge flow path 140 is blocked.

[0058] When the discharge valve 53 moves from the discharge closed position toward the discharge open position and a gap is created between the discharge valve 53 and the valve side opening 42c, the main flow path 110 extending from the inside of the main section 10 and the discharge flow path 140 become connected.

[0059] Next, a description will be given of the operation of the breather valve 5. The breather valve 5 operates based on the internal pressure of the tank 4.

[0060] First, when the internal pressure of the tank 4 is normal, the intake valve 23 and the exhaust valve 53 are both closed. In this state, neither the intake valve 23 nor the exhaust valve 53 is operated, and the internal pressure of the tank 4 is maintained.

[0061] Next, we will explain what happens when the internal pressure of the tank 4 becomes lower than the normal pressure. In the intake section 20, the internal pressure of the intake section 20 is low, so the intake valve 23 is pushed by the atmosphere. Therefore, the intake valve 23 moves from the intake closed position toward the intake open position.

[0062] Specifically, the intake valve 23 is pushed up by the atmosphere and rises against its own weight. The total weight of the intake valve 23 and the intake valve shaft 24 is set to a weight that will rise up when pushed up by the atmosphere when the internal pressure of the intake section 20 falls below normal pressure.

[0063] As the intake valve 23 rises against its own weight, the intake passage 120 is opened and the atmosphere enters the inside of the intake section 20 through the intake port 21b.

[0064] On the other hand, because the internal pressure of the discharge part 40 is low, the discharge valve 53 moves downward due to its own weight, and the discharge valve 53 remains in the discharge closed position. That is, the valve-side opening 42c remains closed by the discharge valve 53.

[0065] As a result, air flows in through the intake port 21b, and the internal pressure of the tank 4 increases, causing the intake valve 23 to move to the intake closing position due to its own weight, thereby closing the intake port 21b. In this way, the internal pressure of the tank 4 becomes the pressure set by the weight of the intake valve 23.

[0066] Next, we will explain what happens when the internal pressure of the tank 4 is higher than the normal pressure. At this time, the internal pressure of the intake section 20 is high, so the intake valve 23 remains in the intake closed position due to the internal pressure of the intake section 20 and its own weight. Therefore, the intake port 21b is closed.

[0067] In the discharge section 40, the internal pressure of the tank 4, the main section 10, and the intake section 20 is high, so the discharge valve 53 moves from the discharge closed position to the discharge open position. Specifically, the discharge valve 53 is pushed by the fluid inside the tank 4 and rises against its own weight. The total weight of the discharge valve 53 and the discharge valve shaft 54 ​​is set to a weight that will rise against its own weight when the internal pressure of the main section 10 exceeds the normal pressure.

[0068] Therefore, the fluid inside the tank 4 is discharged from inside the tank 4 through the main flow path 110, the valve-side opening 42c, and the discharge flow path 140. That is, when the internal pressure of the discharge part 40 becomes higher than the normal pressure, the discharge valve mechanism 50 connects the main flow path 110 with the discharge flow path 140. This allows the fluid to be discharged from inside the tank 4 to the downstream side of the discharge valve mechanism 50 through the discharge flow path 140.

[0069] When the fluid in the tank 4 is discharged from the discharge portion 40, the internal pressure of the tank 4 causes the discharge valve 53 to move to the discharge closed position by its own weight, and closes the valve-side opening 42c.

[0070] As a result, the internal pressure of the tank 4 becomes a pressure based on the weight of the discharge valve 53. In this way, the breather valve 5 can maintain the pressure inside the tank 4 at a constant pressure.

[0071] As described above, the fluid inside the tank 4 includes the fluid inside the tank 4 and the fluid that has volatilized and gasified from the fluid. Therefore, the discharged fluid also includes the fluid inside the tank 4 and the fluid that has volatilized and gasified from the fluid.

[0072] Next, the detection device 60 will be described. The detector 61 can constantly detect the states of the discharge valve 53 and the valve seat 2. A predetermined variable of the detection target detected by the detector 61 can be recognized as the state of the valve mechanism, which is the state of the discharge valve 53 or the state between the discharge valve 53 and the valve seat 2.

[0073] For example, a distance sensor capable of measuring the distance to the underside of the discharge valve 53 can be used as the detector 61. In this case, the state of the discharge valve 53 that can be detected from the detection result of the detector 61 is the current distance to the discharge valve 53, i.e., the current position of the discharge valve 53.

[0074] The detector 61 does not have to be a distance sensor. For example, a vibration sensor facing the underside of the discharge valve 53 can be used as the detector 61. The vibration sensor can measure the vibration of the valve-side end 42b facing the discharge valve 53. Even if a vibration sensor is used as the detector 61, it can detect the state of the valve mechanism in the same way as a distance sensor.

[0075] Furthermore, for example, an acoustic sensor facing the underside of the discharge valve 53 may be used as the detector 61. The acoustic sensor may measure the sound between the discharge valve 53. The acoustic sensor can detect the operating sound of the discharge valve 53 and the sound of the fluid between the valve seat and the discharge valve 53. Even when an acoustic sensor is used as the detector 61, it is possible to detect the state of the valve mechanism in the same way as a distance sensor. Although three types of sensors have been exemplified above as sensors that can be used in the detector 61, the types of sensors that can be used in the detector 61 are not limited to the above three types, and various types of sensors can be used in the detector 61.

[0076] The machine learning device 240 is a device that operates as the main subject of the learning phase of machine learning. Fig. 6 is a configuration diagram showing the machine learning device 240 of Fig. 1. Fig. 7 is a conceptual diagram showing the elements of machine learning performed by the machine learning device 240 of Fig. 6. The machine learning device 240 has a machine learning control unit 241, a machine learning communication unit 242, a learning data storage unit 243, and a machine learning model storage unit 244.

[0077] The machine learning control unit 241 generates a learning model 245 that provides output information corresponding to input information. The machine learning control unit 241 uses one or more pieces of learning data 246 to generate the learning model 245. The learning data 246 is made up of learning input information 201a and a learning valve mechanism state 201b corresponding to the learning input information 201a.

[0078] Here, the input information is a predetermined variation of the detection target detected by the detector 61, and indicates a predetermined variation between the discharge valve 53, which is a valve element, and the valve seat 2, which is detected from the valve seat 2. The valve mechanism state indicates the state of the valve mechanism, that is, the state of the discharge valve 53, which is a valve element, and the valve seat 2, and the state between the discharge valve 53 and the valve seat 2. The valve mechanism is made up of the discharge valve 53, which is a valve element, and the valve seat 2, or the valve seat mechanism 1.

[0079] The learning data 246 is data used as training data, verification data, and test data, which are teacher data in supervised learning. The learning valve mechanism state 201b is data used as a correct answer label in supervised learning.

[0080] In addition, the learning input information 201a and the learning valve mechanism state 201b, which is the valve mechanism state corresponding to the learning input information 201a, are the state in which the input information 200a and the valve mechanism state 200b, which is the valve mechanism state corresponding to the input information 200a, are stored in the learning data storage unit 243 as learning data 246.

[0081] The input information 200a and the valve mechanism status 200b are obtained through various product tests, such as examinations and tests, on the breather valve 5. The learning data storage unit 243 in which the input information 200a, the valve mechanism status 200b, the learning input information 201a, the learning valve mechanism status 201b, and the learning data 246 are stored will be described later.

[0082] The machine learning control unit 241 can cause the learning model 245 to learn the correlation between the learning input information 201a and the learning valve mechanism state 201b in one or more pieces of learning data 246. In this way, the machine learning control unit 241 can generate a trained learning model 245.

[0083] The learning model 245 employs a neural network structure. The learning model 245 has an input layer 245a, an intermediate layer 245b, and an output layer 245c. The input layer 245a has neurons in a number corresponding to the number of parameters of the learning input information 201a. The output layer 245c has neurons in a number corresponding to the number of conditions of the learning valve mechanism state 201b.

[0084] Between each layer, synapses (not shown) are stretched to connect each neuron, and weights can be assigned to each synapse.

[0085] The machine learning control unit 241 adjusts a group of weight parameters consisting of the weights of each synapse through machine learning. The group of weight parameters is reflected in the learning model 245.

[0086] In the machine learning control unit 241, each parameter of the learning input information 201a is input to each neuron of the input layer 245a, and a learning result valve mechanism state 201c indicating the state of the valve mechanism corresponding to the learning input information 201a is output through the learning model 245. The machine learning control unit 241 examines the learning result valve mechanism state 201c and adjusts the weight of each synapse based on the examination result.

[0087] When the learning model 245 is configured as a regression model, the learning valve mechanism state 201b is output as a numerical value normalized to a predetermined range (for example, 0 to 1). When the learning model 245 is configured as a classification model, the learning valve mechanism state 201b is output as a score (accuracy) for each class as a numerical value normalized to a predetermined range (for example, 0 to 1).

[0088] The learning data storage unit 243 can store a plurality of pieces of learning data 246 as a database. The specific configuration of the database that constitutes the learning data storage unit 243 can be designed as appropriate.

[0089] Input information 200a obtained in advance through product testing or the like and a valve mechanism state 200b corresponding to the input information 200a are input to the learning data storage unit 243. The input information 200a and the valve mechanism state 200b input to the learning data storage unit 243 are stored as learning input information 201a and learning valve mechanism state 201b, respectively. Therefore, the learning valve mechanism state 201b corresponds to the learning input information 201a. In this way, one piece of learning data 246 is stored.

[0090] The learning input information 201a and the learning valve mechanism state 201b are based on the input information 200a obtained by the detector 61 during product testing of the breather valve 5, etc., and the valve mechanism state 200b actually observed and measured during product testing, etc.

[0091] The above-mentioned product tests are carried out during the design, prototyping, and pre-shipment product inspection of the breather valve 5. Predetermined variables of the detection target detected by the detector 61 through various product tests are acquired as input information 200a.

[0092] In addition, the amount of fluid passing between the discharge valve 53 and the valve seat 2, the presence or absence of unexpected fluid leakage from between the valve disc and the valve seat, the amount of leakage, and signs of unexpected fluid leakage from between the valve disc and the valve seat, which are observed or measured simultaneously with the detection of the variables during the product test, are acquired as the valve mechanism state 200b. In the product test, the breather valve 5 and the machine learning device 240 are connected to be able to exchange information, and variables detected by the detector 61 during the product test are sequentially input as input information 200a to the learning data storage unit 243. Then, the valve mechanism state 200b corresponding to the input information 200a determined by the judgment of the operator or the like is input to the learning data storage unit 243. Note that the variables detected in the product test may be stored in a storage device (not shown), and after determining the corresponding valve state, the operator may input the information to the learning data storage unit 243 as the input information 200a and the valve mechanism state 200b corresponding to the input information 200a. In this case, the breather valve 5 and the machine learning device 240 do not necessarily need to be connected to be able to exchange information.

[0093] In the product test, the detector 61 detects the predetermined variables of the detection target even when the opening degree of the discharge valve 53 is zero, that is, even when the flow path is closed by the discharge valve 53. In other words, in the product test, the detector 61 constantly detects the variables of the detection target, and therefore it is possible to obtain time-series data of those variables.

[0094] In the product test, the opening degree of the discharge valve 53 and the flow rate of the fluid discharged from the discharge valve 53 are actually measured using a measuring instrument, and the amount of fluid passing between the discharge valve 53 and the valve seat 2 obtained based on the measurement results can be obtained as the valve mechanism state 200b.

[0095] In addition, an operator who can determine the state of the valve seat 2 and the discharge valve 53 can obtain the results of observation of the discharge valve 53 and the valve seat 2, such as whether the state of the valve seat 2 and the discharge valve 53 is normal or abnormal, and if it is abnormal, what kind of abnormal state it is, as the valve mechanism state 200b.

[0096] Furthermore, if the state between the discharge valve 53 and the valve seat 2 is abnormal, the presence or absence of fluid passing between the discharge valve 53 and the valve seat 2 can be regarded as the presence or absence of unexpected fluid leakage, and the presence or absence of unexpected fluid leakage can be obtained as the valve mechanism state 200b.

[0097] Furthermore, if the state between the discharge valve 53 and the valve seat 2 is abnormal, the amount of fluid passing between the discharge valve 53 and the valve seat 2 can be regarded as the amount of unexpected fluid leakage, and the amount of unexpected fluid leakage can be obtained as the valve mechanism state 200b.

[0098] Furthermore, if the state between the discharge valve 53 and the valve seat 2 is abnormal and the amount of fluid passing between the discharge valve 53 and the valve seat 2 is not zero, a sign of unexpected fluid leakage can be acquired as the valve mechanism state 200b. That is, if the state between the discharge valve 53 and the valve seat 2 is abnormal and the amount of fluid passing between the discharge valve 53 and the valve seat 2 is not zero, the valve mechanism state 200b is acquired as a sign of unexpected fluid leakage for time-series input information over a certain period of time when the amount of fluid passing between the discharge valve 53 and the valve seat 2 was zero prior to that.

[0099] Product tests are also conducted to reproduce the occurrence of defects. For example, there are product tests that simulate the case where some kind of foreign matter is trapped between the valve seat 2 and the discharge valve 53, or product tests that are conducted in a state where at least one of the valve seat 2 and the discharge valve 53 is corroded. Furthermore, there are also product tests that operate the valve mechanism for a long period of time, simulating actual operating conditions, to check the state of deterioration over time and durability.

[0100] In such product tests, it is determined that the condition of the discharge valve 53 and the valve seat 2 is abnormal due to a malfunction or deterioration over time, and the presence or absence of unexpected fluid leakage, the amount of unexpected fluid leakage, and signs of unexpected fluid leakage can be obtained as the valve mechanism status 200b.

[0101] Furthermore, input information 200a obtained from not only product tests but also currently operating breather valves 5 and information that can be obtained during maintenance inspections, etc., can be obtained as valve mechanism status 200b. Alternatively, data created by simulation may be used.

[0102] In this way, the correlation between the input information 200a obtained from product testing or the like and the valve mechanism state 200b is valid because it is obtained as a result of using an actual machine. Therefore, it is desirable to use such data as learning data 246 for machine learning.

[0103] The variables that are the input information 200a are specifically determined by the type of sensor used in the detector 61. Since the detector 61 can be a distance sensor, a vibration sensor, or an acoustic sensor, a predetermined variable corresponding to the type of sensor is obtained as the input information 200a.

[0104] The valve mechanism status 200b may be indicated using a numerical value, an error code, or the like. For example, if the valve mechanism status 200b is a quantitative value, such as a flow rate or a leakage amount, the valve mechanism status 200b may be indicated by the specific amount. For example, the valve mechanism status 200b may be indicated as 1 if the valve mechanism is in a normal state, and as 0 if the valve mechanism is in an abnormal state.

[0105] For example, when indicating a sign of an unexpected fluid leak, the absence of any sign of the leak may be indicated as 0, and the detection of the sign of the leak may be indicated as 1. Alternatively, the absence of any sign of an unexpected fluid leak may be indicated as 0, and the frequent detection of the sign of the leak with a high probability as 100, and any value between these may be used.

[0106] The machine learning control unit 241 can extract any one or more pieces of learning data 246 from the plurality of pieces of learning data 246 stored in the learning data storage unit 243 and use them for machine learning.

[0107] The machine learning model storage unit 244 is a database that stores the trained learning model 245 generated by the machine learning control unit 241, that is, the adjusted weight parameter group.

[0108] The machine learning communication unit 242 is a communication interface unit. The machine learning communication unit 242 can transmit and receive various data by connecting to an external device via the network 270. The trained learning model 245 stored in the machine learning model storage unit 244 is provided to the information processing device 250 via the network 270, a storage medium, etc.

[0109] In addition, in FIG. 6, the learning data storage unit 243 and the machine learning model storage unit 244 are shown as separate storage units, but these may also be configured as a single storage unit.

[0110] 8 is a schematic diagram showing the information processing device 250 of FIG. 1. The information processing device 250 is a device that operates as a main body of the inference phase of machine learning. The information processing device 250 uses a learning model 245 generated by the machine learning device 240 to predict a new valve mechanism state 202b corresponding to newly input prediction input information 202a. The prediction input information 202a is a variable acquired by the detector 61 in the currently operating breather valve 5. In other words, the information processing device 250 determines a new valve mechanism state 202b based on the prediction input information 202a in the currently operating breather valve 5.

[0111] The information processing device 250 includes an information processing control unit 251, an information processing storage unit 255, and an information processing communication unit 256. The information processing control unit 251 includes an information acquisition unit 252, an information prediction unit 253, and an output processing unit 254.

[0112] The information acquisition unit 252 acquires data acquired by the detector 61 and output from the detection device 60 as prediction input information 202a. The prediction input information 202a is input information for determining a new valve mechanism state. Here, the valve mechanism state corresponding to the prediction input information 202a is referred to as a new valve mechanism state 202b.

[0113] The information prediction unit 253 predicts a new valve mechanism state 202b by inputting the prediction input information 202a acquired by the information acquisition unit 252 into the learning model 245. The output processing unit 254 outputs the new valve mechanism state 202b predicted by the information prediction unit 253 to the information processing communication unit 256.

[0114] The information prediction unit 253 can select and use one learning model 245 from among a plurality of learning models 245 stored in the information processing storage unit 255 .

[0115] The information processing storage unit 255 is a database that stores the trained learning model 245 used by the information prediction unit 253. The information processing storage unit 255 can store multiple trained learning models 245 input from the machine learning device 240.

[0116] Each of the multiple learning models 245 is a multiple trained model that differs, for example, in machine learning method, type of data included in learning input information 201a, type of data included in learning valve mechanism state 201b, etc.

[0117] The information processing storage unit 255 may be substituted by a storage unit of an external computer such as a server-type computer or a cloud-type computer. In this case, the information prediction unit 253 can access the storage unit of the external computer to acquire the learning model 245.

[0118] The information processing communication unit 256 is connected to be able to communicate with devices outside the valve system 200 via a network 270. The information processing communication unit 256 is a communication interface unit that transmits and receives various types of data. The information processing communication unit 256 can output the new valve mechanism status 202b output by the output processing unit 254 to the information processing communication unit 256.

[0119] 9 is a diagram showing the hardware configuration of the computer 900. The machine learning device 240 and the information processing device 250 of the valve system 200 are configured by a general-purpose or dedicated computer 900.

[0120] The computer 900 includes a bus 910, a processor 912, a memory 914, an input device 916, an output device 917, a display device 918, a storage device 920, a communication interface unit 922, an external device interface unit 924, an input / output device interface unit 926, and a media input / output unit 928. Note that the above components may be omitted as appropriate depending on the application of the computer 900.

[0121] The processor 912 is composed of one or more arithmetic processing devices (CPU (Central Processing Unit), MPU (Micro-processing unit), DSP (Digital Signal Processor), GPU (Graphics Processing Unit), etc.) and operates as a control unit that oversees the entire computer 900.

[0122] The memory 914 stores various data and programs 930, and is composed of, for example, a volatile memory (DRAM, SRAM, etc.) that functions as a main memory, a non-volatile memory (ROM), a flash memory, etc.

[0123] The input device 916 is formed of, for example, a keyboard, a mouse, a numeric keypad, an electronic pen, etc., and functions as an input unit. The output device 917 is formed of, for example, a sound output device including audio, a vibration device, etc., and functions as an output unit. The display device 918 is formed of, for example, a liquid crystal display, an organic EL display, electronic paper, a projector, etc., and functions as an output unit.

[0124] The input device 916 and the display device 918 may be configured as an integrated unit, such as a touch panel display. The storage device 920 is configured, for example, with an HDD, an SSD (Solid State Drive), or the like, and functions as a storage unit. The storage device 920 stores various data necessary for the execution of the operating system and the program 930.

[0125] The communication interface unit 922 is connected to a network 940 such as the Internet or an intranet by wire or wirelessly, and functions as a communication unit that transmits and receives data to and from other computers in accordance with a predetermined communication standard. The network 940 may be the same as the network 270.

[0126] The external device interface unit 924 is connected to an external device 950 such as a camera, printer, scanner, or reader / writer via a wired or wireless connection, and functions as a communication unit that sends and receives data to and from the external device 950 in accordance with a predetermined communication standard.

[0127] The input / output device interface unit 926 is connected to input / output devices 960 such as various sensors and actuators, and functions as a communication unit that transmits and receives various signals and data, such as detection signals from sensors and control signals to actuators, between the input / output devices 960.

[0128] The media input / output unit 928 is configured by a drive device such as a DVD drive or a CD drive, and reads and writes data from and to a medium 970, which is a storage medium such as a DVD or a CD.

[0129] In the computer 900 having the above configuration, the processor 912 loads a program 930 stored in the storage device 920 into the memory 914, executes the program, and controls each unit of the computer 900 via the bus 910. Note that the program 930 may be stored in the memory 914 instead of the storage device 920.

[0130] The program 930 may be recorded on a medium 970 in an installable file format or an executable file format and provided to the computer 900 via the media input / output unit 928. The program 930 may be provided to the computer 900 by downloading it via the communication interface unit 922 over the network 940.

[0131] Furthermore, the computer 900 may implement the various functions that are realized by the processor 912 executing the program 930 using hardware such as an FPGA or an ASIC.

[0132] The computer 900 may be, for example, a desktop computer or a portable computer, and may be any type of electronic device. The computer 900 may be a client computer, a server computer, or a cloud computer. The computer 900 may also be applied to devices other than the machine learning device 240 and the information processing device 250 of the valve system 200.

[0133] Next, the machine learning method will be described. Fig. 10 is a flowchart showing the machine learning method performed by the machine learning device 240 of Fig. 1.

[0134] First, the operator inputs one or more sets of learning data 246 in advance and stores them in the learning data storage unit 243. The number of sets of learning data 246 to be stored is set taking into consideration the inference accuracy required for the ultimately obtained learning model 245. Preferably, multiple sets of learning data 246 are stored.

[0135] In the machine learning method, a learning model preparation step is performed in step S100. The machine learning control unit 241 prepares a pre-learning learning model 245. In the prepared learning model 245, the weights of each synapse are set to initial values.

[0136] Next, a machine learning process is performed in step S110. In the machine learning process, a learning data acquisition process is first performed in step S111. The machine learning control unit 241 randomly acquires one piece of learning data 246 from the multiple pieces of learning data 246 stored in the learning data storage unit 243.

[0137] Next, an inference result output process is performed as step S112. The machine learning control unit 241 inputs the learning input information 201a included in the acquired piece of learning data 246 to the input layer 245a of the prepared learning model 245. As a result, a learning result valve mechanism state 201c is output as the inference result from the output layer 245c of the learning model 245.

[0138] The learning result valve mechanism state 201c output as the inference result is generated by the learning model 245 before or during learning. Therefore, the learning result valve mechanism state 201c is different from the learning valve mechanism state 201b, which is the correct label included in the learning data 246.

[0139] Next, a weight adjustment step is performed in step S113. The machine learning control unit 241 compares the learning valve mechanism state 201b in the learning data 246 acquired in step S111, which is the correct label, with the learning result valve mechanism state 201c output as the inference result in step S112. Based on this comparison, the machine learning control unit 241 performs backpropagation, which is a process of adjusting the weights of each synapse, and performs machine learning.

[0140] As a result, the machine learning control unit 241 causes the learning model 245 to learn the correlation between the learning input information 201a and the learning valve mechanism state 201b.

[0141] Next, in step S114, a machine learning termination determination step is performed. The machine learning control unit 241 determines whether a predetermined learning termination condition has been met. This determination is made based on, for example, an evaluation value of an error function based on the learning valve mechanism state 201b, which is the correct label, and the learning result valve mechanism state 201c, the remaining number of unlearned learning data 246 stored in the learning data storage unit 243, etc.

[0142] In step S114, if the machine learning control unit 241 determines that the learning termination condition is not satisfied and that machine learning is to be continued, that is, if step S114 returns No, the process returns to step S111. In this way, the processes of steps S111 to S114 are performed multiple times on the unlearned learning data 246 for the learning model 245 under learning.

[0143] On the other hand, in step S114, if the machine learning control unit 241 determines that the learning termination condition is met and that machine learning is to be terminated, that is, if step S114 returns Yes, the processing proceeds to step S120.

[0144] Then, in step S120, a trained model storage step is performed. The machine learning control unit 241 stores the trained learning model 245 in which the weights of each synapse have been adjusted, i.e., the learning model 245 reflecting the adjusted weight parameter group, in the machine learning model storage unit 244. This completes the machine learning method.

[0145] Next, a method for predicting the new valve mechanism state 202b of the valve system 200 by the information processing device 250 will be described. Fig. 11 is a flowchart showing a learning valve mechanism state prediction method by the information processing device 250 of Fig. 1.

[0146] The learning valve mechanism state prediction method, which is an information processing method, is a method for determining the valve mechanism state indicating the state of the valve mechanism, and can obtain the valve mechanism state based on input information, which is a predetermined variable between the discharge valve 53, which is the valve body, and the valve seat 2.

[0147] First, in step S200, an input information for prediction acquisition process is carried out. Variable data acquired by detection device 60 is input to information processing device 250 as input information for prediction 202a, and information acquisition unit 252 acquires input information for prediction 202a.

[0148] Next, a prediction step is carried out in step S210. The information prediction unit 253 inputs the prediction input information 202a acquired in step S200 into the learning model 245. As a result, the information prediction unit 253 predicts a new valve mechanism state 202b corresponding to the prediction input information 202a.

[0149] Next, an output processing step is carried out in step S220. As the output processing, the output processing unit 254 displays the new valve mechanism state 202b generated in step S210 on a display screen (not shown) or the like. This completes the prediction of the new valve mechanism state 202b and its output. The new valve mechanism state 202b may be sent to a specified email address by being written in an email or the like.

[0150] Furthermore, since the detection device 60 constantly outputs a predetermined variable, the prediction input information acquisition step is carried out in accordance with this.

[0151] The present disclosure can also be provided in the form of a machine learning program that is a program that causes a computer 900 to function as each part of a machine learning device 240, or a machine learning program that is a program that causes a computer 900 to execute each step of a machine learning method.

[0152] In addition, the present disclosure can also be provided in the form of a valve mechanism status output program, which is a program that causes the computer 900 to function as each part of the information processing device 250, or a valve mechanism status output program, which is a program that causes the computer 900 to execute each step of the information processing method according to embodiment 1.

[0153] Furthermore, the present disclosure can be provided not only in the form of the information processing device 250, the information processing method, or the information processing program according to embodiment 1, but also in the form of an inference device, an inference method, or an inference program used to infer the valve mechanism state. In this case, the inference device, the inference method, or the inference program may include a memory 914 and a processor 912, and the processor 912 may execute a series of processes.

[0154] The series of processes includes an information acquisition process (information acquisition step) for acquiring input information, an inference process (inference step) for inferring the valve mechanism state using the learning model 245 stored in the memory 914, and an output process (output step) for outputting the inferred valve mechanism state to the external device 950. Note that in the output process, the valve mechanism state may be output onto the network 940 via the communication interface unit 922.

[0155] The information processing device 250 according to the first embodiment predicts the state of the valve mechanism and can obtain the valve mechanism state indicating the state of the valve mechanism based on input information that is a predetermined variable between the discharge valve 53, which is the valve disc, and the valve seat 2. This makes it possible to verify the amount of fluid passing through the valve mechanism, the presence or absence of unexpected fluid leakage from between the valve disc and the valve seat, the amount of leakage, and signs of unexpected fluid leakage from the valve mechanism.

[0156] The information processing device 250 according to the first embodiment includes one or more detectors 61 capable of detecting a predetermined change in a detection target in a detection direction. The valve seat 2 constitutes a part of the flow path, and the valve seat 2 has a valve seat end surface 2a with which the flow path is closed when the valve disc, or discharge valve 53, comes into contact. The valve seat 2 also has one or more detector installation spaces 42e, each of which has a detector 61 installed therein. The detection direction, when the detector 61 is installed in the detector installation space 42e, is toward the valve disc, or discharge valve 53, located beyond the valve seat end surface 2a. This allows the detector 61 to detect a predetermined change between the valve seat 2 and the valve disc, or discharge valve 53. This allows for more accurate verification of the amount of fluid passing through the valve mechanism, the presence or absence of unexpected fluid leakage from between the valve disc and the valve seat, the amount of leakage, and signs of unexpected fluid leakage from the valve mechanism.

[0157] According to the information processing device 250 of the first embodiment, at least one of the one or more detectors 61 uses any one of a distance sensor, a vibration sensor, and an acoustic sensor. Although various types of sensors can be used for the detector 61, using any one of a distance sensor, a vibration sensor, and an acoustic sensor allows a sensor that is commercially available and easily available to be used as the detector 61. Therefore, the design, manufacturing, and maintenance of the valve seat mechanism 1 can be easily handled.

[0158] According to the information processing device 250 of the first embodiment, the valve mechanism state includes at least one of the flow rate of the fluid flowing between the discharge valve 53 serving as a valve disc and the valve seat 2, the presence or absence of unexpected fluid leakage from between the discharge valve 53 serving as a valve disc and the valve seat 2, the amount of unexpected fluid leakage from between the discharge valve 53 serving as a valve disc and the valve seat 2, and signs of unexpected fluid leakage from between the discharge valve 53 serving as a valve disc and the valve seat 2. This makes it possible to easily and quickly verify the state of the valve mechanism based on the output of the information processing device 250.

[0159] According to the information processing device 250 of the first embodiment, when input information is input, the valve mechanism state is determined using a learning model in which the correlation between the input information and the valve mechanism state is learned by machine learning, and the determined valve mechanism state is output. This allows a machine learning method based on accumulated information to be used to verify the state of the valve mechanism. Therefore, it is possible to more accurately verify the amount of fluid passing through the valve mechanism, the presence or absence of unexpected fluid leakage from between the valve disc and the valve seat, the amount of leakage, and signs of unexpected fluid leakage from the valve mechanism.

[0160] The machine learning device 240 according to the first embodiment generates a learning model 245 for inferring the valve mechanism state indicating the state of the valve mechanism. The device also includes a learning data storage unit 243 that stores multiple sets of learning data 246, each of which is composed of input information, which is a predetermined variable between the discharge valve 53 (the valve disc) and the valve seat 2, and the valve mechanism state corresponding to the input information. The device also includes a machine learning model storage unit that inputs multiple sets of learning data 246, thereby causing the learning model to learn the correlation between the input information and the valve mechanism state, and stores the learned learning model 245. This allows the learning model 245, which has learned the correlation between the input information and the valve mechanism information, to be generated and stored. Therefore, the trained learning model 245 can be used to more accurately verify the amount of fluid passing through the valve mechanism, the presence or absence of unexpected fluid leakage from between the valve disc and the valve seat, the amount of leakage, and signs of unexpected fluid leakage from the valve mechanism.

[0161] The inference device according to the first embodiment includes a memory 914 and at least one processor 912, and infers a valve mechanism state indicating the state of the valve mechanism. The at least one processor 912 also executes an information acquisition process for acquiring input information, which is a predetermined variable between the discharge valve 53, which is the valve disc, and the valve seat 2, an inference process for inferring the valve mechanism state using a learning model 245 based on machine learning stored in the memory 914 upon acquiring the input information, and an output process for outputting the inferred valve mechanism state. This makes it possible to verify the amount of fluid passing through the valve mechanism, the presence or absence of unexpected fluid leakage from between the valve disc and the valve seat, the amount of leakage, and signs of unexpected fluid leakage from the valve mechanism.

[0162] According to the inference method of embodiment 1, the method is executed by an inference device including a memory 914 and at least one processor 912, and infers a valve mechanism state indicating the state of the valve mechanism. The at least one processor 912 also executes an information acquisition step of acquiring input information, which is a predetermined variable between the discharge valve 53, which is the valve disc, and the valve seat 2. Upon acquiring the input information, the method executes an inference step of inferring the valve mechanism state using a learning model 245 based on machine learning stored in the memory 914. The method also executes an output step of outputting the inferred valve mechanism state. This makes it possible to verify the amount of fluid passing through the valve mechanism, the presence or absence of unexpected fluid leakage from between the valve disc and the valve seat, the amount of leakage, and signs of unexpected fluid leakage from the valve mechanism.

[0163] According to the information processing method of the first embodiment, a valve mechanism state indicating the state of the valve mechanism is determined. The valve mechanism state is also obtained based on input information that is a predetermined variable between the discharge valve 53, which is the valve disc, and the valve seat 2. This makes it possible to verify the amount of fluid passing through the valve mechanism, the presence or absence of unexpected fluid leakage from between the valve disc and the valve seat, the amount of leakage, and signs of unexpected fluid leakage from the valve mechanism.

[0164] According to the machine learning method of the first embodiment, a learning model 245 for inferring a valve mechanism state indicating the state of the valve mechanism is generated. Furthermore, in a state in which one or more pieces of learning data 246 are stored, each of which is composed of input information, which is a predetermined variable between the discharge valve 53 serving as a valve disc, and the valve seat 2, and a valve mechanism state corresponding to the input information, a machine learning process is executed in which multiple sets of learning data 246 are input, thereby causing the learning model to learn the correlation between the input information and the valve mechanism state. Furthermore, a learned model storage process is executed in which the learning model 245 learned in the machine learning process is stored in the machine learning model storage unit 244. This makes it possible to verify the amount of fluid passing through the valve mechanism, verify the presence or absence of unexpected fluid leakage from between the valve disc and the valve seat, verify the amount of leakage, and verify signs of unexpected fluid leakage from the valve mechanism.

[0165] In the first embodiment, a neural network is used as the learning model 245. However, this is not limiting. Other machine learning models may be used as the learning model 245. Examples of other machine learning models include tree-type models such as decision trees and regression trees, ensemble learning such as bagging and boosting, neural network-type models (including deep learning) such as recurrent neural networks, convolutional neural networks, and LSTM (Long Short Term Memory), clustering-type models such as hierarchical clustering, non-hierarchical clustering, k-nearest neighbors, and k-means, multivariate analysis such as principal component analysis, factor analysis, and logistic regression, and support vector machines.

[0166] Embodiment 2. In describing the valve system 200 of the present disclosure, the valve system 200 used in the breather valve 5 is also used in the second embodiment. However, the valve system 200 of the present disclosure is not limited to applications to the breather valve 5, and can be used as appropriate for general valve mechanisms.

[0167] The information processing device 250 of the second embodiment differs from the valve system 200 of the first embodiment in that it outputs a new valve mechanism state 202b without using a learning model obtained by machine learning.

[0168] The information processing device 250 can output a new valve mechanism state 202b corresponding to the acquired prediction input information 202a.

[0169] The information prediction unit 253 determines a new valve mechanism state 202b using a predetermined determination method based on the prediction input information 202a input and acquired by the information acquisition unit 252. These determination methods do not use machine learning methods, and therefore do not use a learning model. The output processing unit 254 can output the determined new valve mechanism state 202b to the information processing communication unit 256.

[0170] In the information processing device 250, the input information for prediction 202 a acquired by the information acquisition unit 252 can also be sequentially stored in the information processing storage unit 255. This allows the stored input information for prediction 202 a to be treated as time-series data. While the information processing storage unit 255 in the first embodiment functions as a database for the learning model 245, in the second embodiment, the information processing storage unit 255 functions as a storage device that can store information.

[0171] The predetermined determination methods implemented by the information prediction unit 253 include a method based on a comparison with stored waveform data, a method based on a threshold value to determine whether something is normal or abnormal, a method based on a calculation using coefficients to calculate physical quantities, and other well-known statistical analysis calculations.

[0172] The information processing storage unit 255 stores in advance information such as waveform data, thresholds, or coefficients required for the determination method implemented by the information prediction unit 253. The stored waveform data includes, for example, time series data of a predetermined variable detected by the detector 61 that may be a sign of an unexpected fluid leak.

[0173] The information prediction unit 253 can use this waveform data to implement a predetermined determination method. For example, the corresponding valve mechanism state can be determined based on whether the stored waveform data or characteristics of the waveform data match the waveform data or characteristics of the waveform data based on the prediction input information 202a. If they match, a new valve mechanism state 202b can be determined, indicating that there is a sign of an unexpected fluid leak.

[0174] In addition, the stored threshold values ​​include one in which a state in which the discharge valve 53 opens the flow path for a certain period of time or more is considered to be an abnormal state, and that time is stored as the threshold value, or one in which a state in which the discharge valve 53 continues to perform micro-movements for a certain period of time or more is considered to be an abnormal state, and that time is stored as the threshold value.

[0175] The information prediction unit 253 can use this threshold value to implement a predetermined determination method. For example, it determines whether the opening of the flow path by the discharge valve 53 or the slight movement of the discharge valve 53 continues for a time period equal to or greater than the threshold value based on the prediction input information 202a. If a time period equal to or greater than the threshold value has passed, it can determine a new valve mechanism state 202b indicating that the valve mechanism is abnormal.

[0176] The stored coefficients include those required for calculating physical quantities, such as a coefficient required for calculating the flow rate of the fluid flowing out from between the discharge valve 53 and the valve seat 2.

[0177] The information prediction unit 253 can use this coefficient to implement a predetermined determination method. For example, the information prediction unit 253 can calculate and determine a flow rate using the stored coefficient. The calculated flow rate can be used as the flow rate of the fluid flowing out of the valve mechanism to determine the new valve mechanism state 202b.

[0178] The information processing control unit 251 can determine the new valve mechanism state 202b based on the prediction input information 202a, using pre-stored waveform data, thresholds, and coefficients. Furthermore, according to the second embodiment, the pre-stored waveform data, thresholds, and coefficients are not limited to those described above, and can be selected as appropriate. Furthermore, in order to determine the new valve mechanism state 202b, well-known arithmetic and statistical methods can be used, not limited to the determination methods described above.

[0179] Since machine learning is not used in the second embodiment, the configurations relating to the learning model 245, the machine learning device 240, and the inference device in the first embodiment are not used. The other configurations of the second embodiment are the same as those of the first embodiment, and therefore description thereof will be omitted.

[0180] According to the information processing device 250 of the second embodiment, when input information is input, the valve mechanism state is determined from the input information based on a predetermined determination method that does not use a machine learning method, and the determined valve mechanism state is output. This makes it possible to verify the state of the currently installed valve mechanism based on past knowledge of the valve mechanism. Therefore, it is possible to more accurately verify the amount of fluid passing through the valve mechanism, the presence or absence of unexpected fluid leakage from between the valve disc and the valve seat, the amount of leakage, and signs of unexpected fluid leakage from the valve mechanism.

[0181] Embodiment 3. In describing the valve system 200 of the present disclosure, the third embodiment also uses the valve system 200 used in the breather valve 5. However, the valve system 200 of the present disclosure is not limited to applications to the breather valve 5, and can be used as appropriate for general valve mechanisms.

[0182] The third embodiment differs from the first or second embodiment in that three detectors 61 are installed on the valve seat 2. Fig. 12 is a top view showing the introduction section 42 according to the third embodiment.

[0183] The detection device 60 has three detectors 61, an input / output device (not shown), and wiring 62 that transmits signals from each detector 61 to the input / output device. Three detector installation spaces 42e are formed in the introduction section 42. One detector 61 and the wiring 62 connected to the detector 61 are installed in each detector installation space 42e.

[0184] The valve contact member 70 is formed in a shape corresponding to the shape of the main body end surface 42d. The openings of the multiple detector installation spaces 42e formed in the main body end surface 42d are closed by installing the valve contact member 70 in the introduction portion 42. That is, the multiple detectors 61 are installed in the detector installation spaces 42e covered by the valve contact member 70. Each of the multiple detectors 61 is installed along the valve-facing surface 70a. As shown in FIG. 12 , the inner circumferential surface of the introduction portion 42 where each detector installation space 42e is formed protrudes further inward, and the wall thickness of the introduction portion 42 where the detector installation space 42e is formed is thicker than the wall thickness of the remaining portions. The valve contact member 70 is also formed in a shape corresponding to the shape of the main body end surface 42d. However, this is not limited to this. For example, the wall thickness of the introduction portion 42 may be uniform, i.e., the inner peripheral surface of the introduction portion 42 may not have any inward protruding portions, and the inner peripheral surface of the introduction portion 42 may be circular when viewed in the axial direction. Even in this case, the valve contact member 70 is formed in a shape corresponding to the shape of the main body end surface 42d.

[0185] Each of the multiple detectors 61 may be the same type of sensor, for example, all three detectors 61 may be distance sensors. The other configurations of the third embodiment are the same as those disclosed in the first embodiment, and therefore description thereof will be omitted.

[0186] In the third embodiment, three detectors 61 are installed in the introduction section 42. However, this is not limited to this. The number of detectors 61 arranged in the introduction section 42 may be two or more, i.e., multiple. Also, one or multiple detectors 61 may be installed in one detector installation space 42e. In the third embodiment, multiple detectors 61 of the valve seat mechanism 1 are installed along the valve seat end surface 2a. This allows for more detailed investigation of slight tilts and vibrations of the discharge valve 53. Therefore, the behavior of the valve disc can be observed in more detail.

[0187] The multiple detectors 61 of the valve seat mechanism 1 in the third embodiment are sensors of the same type. This makes it easy to compare and examine the signals obtained from the detectors 61. Therefore, even slight tilts or deviations of the discharge valve 53 can be easily detected in more detail.

[0188] It should be noted that the three detectors 61 in the third embodiment are all the same type of sensor. However, this is not limited to this. For example, the types of sensors of the three detectors 61 may be one or more distance sensors and one or more vibration sensors. In other words, multiple types of sensors may be used for the multiple detectors 61.

[0189] In the valve seat mechanism 1 according to the third embodiment, at least two detectors 61 are different types of sensors, which allows the behavior of the discharge valve 53 to be examined from multiple angles.

[0190] Furthermore, the valve system 200 and valve seat mechanism 1 in the first to third embodiments are provided in a breather valve 5. However, this is not limited to this. The valve system 200 and valve seat mechanism 1 in the first to third embodiments can be provided in any valve mechanism that closes the flow path by closing the valve seat end surface 2a of the valve seat 2 with a valve member and opens the flow path by creating a gap between the valve member and the valve seat 2. Examples of such valve mechanisms include rotary valves such as ball valves and diaphragm valves.

[0191] Furthermore, in the valve system 200 according to the first to third embodiments, the input information is a predetermined variable detected by the detector 61 installed in the detector installation space 42e formed in the valve seat 2. However, this is not limited to this. The detector 61 may be installed anywhere on the valve seat 2 or the valve mechanism.

[0192] (Other Embodiments) The present disclosure is not limited to the above-described embodiment, and various modifications can be made without departing from the spirit and scope of the present disclosure, all of which are included in the technical concept of the present disclosure.

[0193] Various aspects of the present disclosure are summarized below as appendices.

[0194] (Supplementary Note 1) An information processing device that predicts the state of a valve mechanism, the information processing device determining a valve mechanism state indicating the state of the valve mechanism based on input information that is a predetermined variable between a valve disc and a valve seat. (Supplementary Note 2) The information processing device according to Supplementary Note 1, comprising one or more detectors that can detect the predetermined variable of a detection target in a detection direction, the valve seat constituting a part of a flow path, the valve seat having a valve seat end face that closes the flow path when the valve disc comes into contact with the valve seat, one or more detector installation spaces being formed in the valve seat, the detector being installed in each of the detector installation spaces, and the detection direction being toward the valve disc that is located beyond the valve seat end face when the detector is installed in the detector installation space. (Supplementary Note 3) The information processing device according to Supplementary Note 2, wherein at least one of the one or more detectors uses any one type of sensor from the group consisting of a distance sensor, a vibration sensor, and an acoustic sensor. (Supplementary Note 4) The information processing device according to any one of Supplementary Note 1 to Supplementary Note 3, wherein the valve mechanism state includes at least one of a flow rate of fluid flowing between the valve disc and the valve seat, presence or absence of unexpected fluid leakage from between the valve disc and the valve seat, an unexpected fluid leakage amount from between the valve disc and the valve seat, and a sign of unexpected fluid leakage from between the valve disc and the valve seat. (Supplementary Note 5) The information processing device according to any one of Supplementary Note 1 to Supplementary Note 4, wherein when the input information is input, the valve mechanism state is determined using a learning model in which a correlation between the input information and the valve mechanism state is learned by machine learning, and the determined valve mechanism state is output. (Supplementary Note 6) A machine learning device that generates a learning model for inferring a valve mechanism state that indicates the state of a valve mechanism, comprising: a learning data storage unit that stores multiple sets of learning data composed of input information that is a predetermined variable between a valve disc and a valve seat, and the valve mechanism state corresponding to the input information; and a machine learning model storage unit that inputs the multiple sets of learning data, thereby causing the learning model to learn the correlation between the input information and the valve mechanism state, and stores the learned learning model.(Supplementary Note 7) An inference device for inferring a valve mechanism state indicating the state of a valve mechanism, comprising a memory and at least one processor, wherein the at least one processor executes the following: an information acquisition process for acquiring input information which is a predetermined variable between a valve disc and a valve seat, an inference process for inferring the valve mechanism state using a learning model by machine learning stored in the memory upon acquiring the input information, and an output process for outputting the inferred valve mechanism state. (Supplementary Note 8) An information processing method for determining a valve mechanism state indicating the state of a valve mechanism, comprising: determining the valve mechanism state corresponding to input information which is a predetermined variable between the valve disc and the valve seat. (Supplementary Note 9) A machine learning method for generating a learning model for inferring a valve mechanism state indicating the state of a valve mechanism, the machine learning method comprising: a machine learning process in which, in a state in which one or more pieces of learning data are stored, the machine learning data is input and the learning model learns the correlation between the input information and the valve mechanism state; and a learned model storage process in which the learning model learned in the machine learning process is stored in a machine learning model storage unit. (Supplementary Note 10) An inference method for inferring a valve mechanism state indicating the state of a valve mechanism, the inference method being executed by an inference device comprising a memory and at least one processor; an information acquisition process in which the at least one processor acquires input information that is a predetermined variable between the valve disc and the valve seat; an inference process in which, upon acquiring the input information, the valve mechanism state is inferred using a machine learning learning model stored in the memory; and an output process in which the inferred valve mechanism state is output.

[0195] 1 Valve seat mechanism, 2 Valve seat, 2a Valve seat end surface, 3 Valve seat body, 4 Tank, 4a Mounting portion, 4b Upstream flange, 5 Breather valve, 10 Main portion, 10a Inlet side opening end, 10b Outlet side opening end, 10c Intake side opening end, 20 Intake portion, 21 Intake portion body, 21a Connection side opening end, 21b Intake port, 21x Intake valve seat, 22 Intake valve mechanism, 23 Intake valve (valve body), 24 Intake valve shaft, 25 Intake valve guide, 40 Discharge portion, 41 Discharge portion body, 42 Introduction portion, 42a Discharge portion inlet side opening end, 42b Valve side end, 42c Valve side opening, 42d Body end surface, 42e Detector installation space, 43 Atmospheric discharge portion, 43a Discharge port, 43b Valve mechanism side opening end, 47 vent cover, 50 exhaust valve mechanism, 53 exhaust valve (valve body), 54 exhaust valve shaft, 55 exhaust valve guide, 60 detection device, 61 detector, 62 wiring, 70 valve contact member, 110 main flow path, 111 branch flow path, 120 intake flow path, 140 exhaust flow path, 200 valve system, 200a input information, 200b valve mechanism state, 201a learning input information, 201b learning valve mechanism state, 201c learning result valve mechanism state, 202a prediction input information, 202b new valve mechanism state, 240 machine learning device, 241 machine learning control unit, 242 machine learning communication unit, 243 learning data storage unit, 244 machine learning model storage unit, 245 learning model, 245a input layer, 245b intermediate layer, 245c output layer, 246 learning data, 250 Information processing device, 251 information processing control unit, 252 information acquisition unit, 253 information prediction unit, 254 output processing unit, 255 information processing storage unit, 256 information processing communication unit, 270 network, 900 computer, 910 bus, 912 processor, 914 memory, 916 input device, 917 output device, 918 display device, 920 storage device, 922 communication interface unit, 924 external device interface unit, 926 input / output device interface unit, 928 media input / output unit, 930 program, 940 network, 950 external device, 960 input / output device, 970 media (non-temporary storage medium).

Claims

1. An information processing apparatus for predicting the state of a valve mechanism, the information processing apparatus obtaining a valve mechanism state indicating the state of the valve mechanism based on input information which is a predetermined variable between a valve body and a valve seat.

2. The information processing apparatus according to claim 1, further comprising one or more detectors capable of detecting the predetermined variable of a detection target in a detection direction, wherein the valve seat forms part of a flow path, a valve seat end face is formed on the valve seat, the flow path being closed when the valve body contacts the valve seat end face, one or more detector installation spaces are formed on the valve seat, the detectors are installed in the respective detector installation spaces, and the detection direction is directed toward the valve body beyond the valve seat end face when the detectors are installed in the detector installation spaces.

3. The information processing apparatus according to claim 2, wherein at least one of the one or more detectors uses one type of sensor selected from a distance sensor, a vibration sensor, and an acoustic sensor.

4. The information processing apparatus according to claim 1, wherein the valve mechanism state includes at least one of a flow rate of fluid flowing between the valve body and the valve seat, presence or absence of unexpected fluid leakage from between the valve body and the valve seat, an amount of unexpected fluid leakage from between the valve body and the valve seat, and an omen of unexpected fluid leakage from between the valve body and the valve seat.

5. The information processing apparatus according to claim 1, wherein when the input information is input, the valve mechanism state is determined using a learning model in which a correlation between the input information and the valve mechanism state has been learned by machine learning, and the determined valve mechanism state is output.

6. A machine learning apparatus for generating a learning model for inferring a valve mechanism state indicating the state of a valve mechanism, the machine learning apparatus comprising: a learning data storage unit storing a plurality of sets of learning data constituted by input information which is a predetermined variable between a valve body and a valve seat and the valve mechanism state corresponding to the input information; and a machine learning model storage unit storing the learning model in which the correlation between the input information and the valve mechanism state has been learned by inputting a plurality of sets of the learning data.

7. An inference device comprising a memory and at least one processor, for inferring a valve mechanism state indicating the state of a valve mechanism, wherein the at least one processor executes an information acquisition process of acquiring input information which is a predetermined variable between a valve body and a valve seat, an inference process of inferring the valve mechanism state using a learning model by machine learning stored in the memory when the input information is acquired, and an output process of outputting the inferred valve mechanism state.

8. An information processing method for determining a valve mechanism state indicating the state of a valve mechanism, the method comprising obtaining the valve mechanism state corresponding to input information which is a predetermined variable between a valve body and a valve seat.

9. A machine learning method for generating a learning model for inferring a valve mechanism state indicating the state of a valve mechanism, the method comprising, in a state where one or more pieces of learning data composed of input information which is a predetermined variable between a valve body and a valve seat and the valve mechanism state corresponding to the input information are stored, a machine learning step of causing a learning model to learn the correlation between the input information and the valve mechanism state when the learning data is input, and a learned model storage step of storing the learning model learned in the machine learning step in a machine learning model storage unit.

10. An inference method executed by an inference device comprising a memory and at least one processor, for inferring a valve mechanism state indicating the state of a valve mechanism, wherein the at least one processor executes an information acquisition step of acquiring input information which is a predetermined variable between a valve body and a valve seat, an inference step of inferring the valve mechanism state using a learning model by machine learning stored in the memory when the input information is acquired, and an output step of outputting the inferred valve mechanism state.

Images