JP7588522B2 - Machine learning device, pressure tank state estimation device, inference device, machine learning method, machine learning program, pressure tank state estimation method, pressure tank state estimation program, inference method, and inference program - Google Patents

Description

The present invention relates to a machine learning device, a pressure tank state estimation device, an inference device, a machine learning method, a machine learning program, a pressure tank state estimation method, a pressure tank state estimation program, an inference method, and an inference program.

Water supply devices are widely used as devices for supplying water (tap water) to water supply targets such as office buildings, apartment buildings, etc. Water supply devices generally include a pump for pressurizing water, a motor for driving the pump, a control unit for controlling the operation of the pump, and a pressure tank connected downstream of the pump.

Generally, a pressure tank has a diaphragm inside. The diaphragm divides the inside of the container into an air chamber and a water chamber. Air is sealed inside the air chamber beforehand, and water flowing from the pump into the water chamber compresses the air through the diaphragm. A check valve is placed between the pressure tank and the pump, and this check valve and the pressure tank ensure that the water pressure in the water supply pipe is properly maintained even after the pump has stopped, and water is sent to the destination. When water is sent to the destination, the diaphragm protrudes from the air chamber side to the water chamber side.

The diaphragm of such a water supply device repeatedly expands and contracts as the pump operates or water is supplied to the supply destination for many years. The repeated expansion and contraction of the diaphragm gradually deteriorates the diaphragm. Similarly, the fixed portion between the container and the diaphragm gradually deteriorates due to repeated bending. However, since the deterioration of the diaphragm differs depending on the usage conditions, it is difficult to replace it at the optimal timing.
A diaphragm inspection device has been proposed that can accurately diagnose the deterioration state of a diaphragm in a pressure tank and determine the appropriate time to replace it (see, for example, Patent Documents 1 and 2).

Patent No. 3596909 Japanese Patent Application Publication No. 8-86278

The devices disclosed in Patent Documents 1 and 2 diagnose the state of the pressure tank based on individual physical quantities (pressure, motor current, water supply time, etc.). Therefore, the devices disclosed in Patent Documents 1 and 2 cannot handle cases where a change in the state of the pressure tank is expressed in physical quantities other than those mentioned above, or is expressed in a composite manner in multiple physical quantities. Furthermore, diagnosis of a composite event based on multiple physical quantities depends to a large extent on the experience (including tacit knowledge) of the operator and there is a large individual difference between operators, and therefore there is a strong demand for a device that automates such diagnosis.

Furthermore, the devices disclosed in Patent Document 1 and Patent Document 2 each have a pressure sensor for measuring the pressure in the pressure tank. However, the water supply device does not use the pressure value in the pressure tank in controlling operation. Therefore, the water supply device had to take the trouble of providing an expensive pressure sensor only for use during maintenance.

In view of the above-mentioned problems, the present invention aims to provide a machine learning device, a pressure tank state estimation device, an inference device, a machine learning method, a machine learning program, a pressure tank state estimation method, a pressure tank state estimation program, an inference method, and an inference program that enable the state of a pressure tank in a fluid supply device to be estimated with high accuracy without installing expensive sensors or relying on the experience of an operator.

In order to achieve the above object, a machine learning device according to an aspect of the present invention comprises:
A machine learning device that generates a learning model for estimating a pressure of a pressure tank in a fluid supply device including a pump that transfers a fluid and a pressure tank provided on a discharge side of the pump, the machine learning device comprising:
a learning data storage unit that stores a plurality of sets of learning data in which the input data and the output data correspond to each other, the learning data storage unit including input data of the pump pressure for a predetermined period of time and data of the operating status of the pump for the predetermined period of time and output data of the charged pressure of the pressure tank;
a machine learning unit that causes the learning model to learn a correlation between the input data and the output data by inputting a plurality of sets of the learning data to the learning model;
A trained model storage unit that stores the learning model trained by the machine learning unit;
Equipped with.

According to the machine learning device of the present invention, it is possible to provide a learning model capable of inferring (estimating) information on a pressure tank with high accuracy from data on motor current values, data on contact electrical resistance, and data on vibration. Also, according to the pressure tank state estimation device of the present invention, it is possible to estimate the state of a pressure tank with high accuracy without relying on the experience of an operator.

Other objects, configurations and effects will become apparent from the detailed description of the invention described below.

A diagram showing an example of a water supply device 1 to which a pressure tank state estimation system 2 according to this embodiment is applied. A hardware configuration diagram showing an example of a computer 200 that constitutes the machine learning device 4 and the pressure tank state estimation device 5. FIG. 2 is a block diagram showing an example of a machine learning device 4 according to the present embodiment. FIG. 2 is a data configuration diagram showing an example of data (supervised learning) used in the machine learning device 4 according to the present embodiment. FIG. 2 is a schematic diagram showing an example of a neural network model used in the machine learning device 4 according to the present embodiment. 4 is a flowchart showing an example of a machine learning method performed by the machine learning device 4 according to the present embodiment. FIG. 2 is a block diagram showing an example of a pressure tank state estimating device 5 according to the present embodiment. 4 is a flowchart showing an example of a pressure tank state estimation method by the pressure tank state estimation device 5 according to the present embodiment.

Hereinafter, an embodiment of the present invention will be described with reference to the drawings.
The scope necessary for the explanation to achieve the objective of the present invention is shown diagrammatically, and the scope necessary for the explanation of the relevant parts of the present invention will be mainly explained, and the parts where explanation is omitted will be based on publicly known technology.

FIG. 1 shows a fluid supply device 1 to which a pressure tank state estimation system 2 according to an embodiment of the present invention is applied. This fluid supply device 1 is a device for supplying water to buildings such as apartments, offices, commercial facilities, factories, schools, etc., and facilities (water supply targets) such as houses and parks. The fluid supply device 1 introduces water from a fluid supply source 41 through an inlet via an introduction pipe 42. The fluid supply source 41 may be a water tank type using a water tank. A water supply pipe 43 is connected to the discharge port of the fluid supply device 1. The water supply pipe 43 is connected to a fluid supply destination 44 such as a faucet of each building. The fluid supply device 1 pressurizes water from the fluid supply source 41 and supplies the water to each fluid supply destination 44 such as a building.

In this embodiment, the fluid supply device 1 is shown as an example of a water supply device that supplies water, but the fluid supply device 1 may also supply liquids other than water, such as fuel and chemicals, as well as fluids containing gases such as gas, hydrogen, and oxygen.

The fluid supply device 1 includes a pump 12 and a motor 1 as a drive unit for driving the pump 12.
The inverter 14 is connected to a commercial power source 51 via a ground fault circuit interrupter 52.

The fluid supply device 1 has a check valve 15, a flow switch 16, and a pressure tank 17 on the discharge side (downstream side) of the pump 12. The fluid supply device 1 shown in FIG.
, a motor 13, a check valve 15, and a flow switch 16. The fluid supplying device 1 may have two or more sets of pumps 12, motors 13, check valves 15, and flow switches 16. By providing a plurality of pumps 12, if some of the pumps 12 become inoperable, the fluid supplying device 1 can continue to supply water using the other pumps 12 that are operable, thereby avoiding a water outage.

The fluid supply device 1 also has a backflow prevention device 11 upstream of the pump 12. The backflow prevention device 11 is provided in an introduction pipe 42 connected to the suction port of the fluid supply device 1, and prevents water from flowing back from the fluid supply device 1 to the fluid supply source 41. Note that in cases where the fluid supply source 41 is a water tank, for example, the fluid supply device 1 does not need to include the backflow prevention device 11.

The fluid supply device 1 has a check valve 15 in a discharge pipe connected to the discharge port of the pump 12. The check valve 15 prevents backflow of water to the pump 12 when the pump 12 stops. The fluid supply device 1 also has a flow switch 16 downstream (secondary side) of the check valve 15. The flow switch 16 is a flow rate detector that detects when the flow rate of water flowing through the discharge pipe drops to a predetermined value. The fluid supply device 1 also has a pressure tank 17 downstream of the flow switch 16. The pressure tank 17 is a pressure retainer for retaining the pressure on the discharge side of the fluid supply device 1 while the pump 12 is stopped. Here, in this specification, the pressure in the pressure tank and the charged pressure of the pressure tank both refer to the pressure in the air chamber.

The pressure tank 17 according to this embodiment has a diaphragm 17a that divides the inside of the container into an air chamber and a water chamber. Air is sealed inside the air chamber beforehand, and water flowing from the pump 12 into the water chamber compresses the air through the diaphragm 17a. A check valve 1 is provided between the pressure tank 17 and the pump 12.
The check valve 15 and the pressure tank 17 maintain the water pressure in the water supply pipe appropriately even after the pump 12 is stopped, and the water is sent to the fluid supply destination 44. When the water is sent to the fluid supply destination 44, the diaphragm 17a presses from the air chamber side to the water chamber side.

The fluid supplying device 1 includes a control unit 20 that controls the water supplying operation. The control unit 20 includes a CPU (
A circuit board centered on a central processing unit (Central Processing Unit), a dedicated circuit board, or the like may be used. The control unit 20 is supplied with power from a commercial power source 51 via a power line (not shown). However, the control unit 20 may be supplied with power via another component of the fluid supplying device 1, such as the inverter 14.

The control unit 20 and the inverter 14 are connected to each other via a communication line 21. Various set values, a frequency command value as a control command related to the rotation speed of the motor 13, an operation/stop signal, and other signals are sent from the control unit 20 to the inverter 14. The inverter 14 successively sends to the control unit 20 information on the operation status, such as the actual frequency value and current value.

Note that analog signals and/or digital signals may be used as the control signals transmitted and received between the control unit 20 and the inverter 14. For example, an analog signal may be used for the rotation frequency and the like, and a digital signal may be used for an operation stop command and the like.

The fluid supply device 1 having the above configuration is provided with a pressure tank state estimation system 2 that estimates the state of the pressure tank 17. The pressure tank state estimation system 2 includes a measurement device 3, a machine learning device 4, and a pressure tank state estimation device 5.

The measuring device 3 is installed in each part of the fluid supplying device 1, and measures physical quantities and state quantities of each part of the fluid supplying device 1. In the learning phase, the measuring device 3 is connected to the machine learning device 4 and used to measure data for learning, and in the inference phase, the measuring device 3 is connected to the pressure tank state estimating device 5 and used to measure data for estimation. The measuring device 3 includes a pump primary pressure sensor 31, a pump secondary pressure sensor 32, a pump operation status sensor 33, and a usage flow rate sensor 34.
At least one of the pump primary pressure sensor 31 and the pump secondary pressure sensor 32 may be provided as a pump pressure sensor for measuring the pressure related to the pump.
4 may not be provided.

The pump primary pressure sensor 31 is a pressure measuring device for measuring the suction side pressure of the pump 12. The pump primary pressure sensor 31 of this embodiment is provided between the fluid supply source 41 and the backflow prevention device 11. It is preferable that the pump primary pressure sensor 31 can measure time-series data for a predetermined period of time. The predetermined period may include the start-up time of the pump.

The pump secondary pressure sensor 32 is a pressure measuring device for measuring the pressure on the discharge side of the pump 12. In this embodiment, the pump secondary pressure sensor 32 is provided between the flow switch 16 and the pressure tank 17. It is preferable that the pump secondary pressure sensor 32 can measure time-series data for a predetermined period of time. The predetermined period may include the start-up time of the pump.

The pump operation status sensor 33 acquires the operation status of the pump 12. The operation status of the pump 12 is acquired, for example, from the ON/OFF state of the pump 12, the rotation speed of the pump 12, or the current value of the motor 13. It is preferable that the pump operation status sensor 33 is capable of measuring time-series data for a predetermined period of time. The predetermined period may include the start-up time of the pump.
By acquiring the operating status of the pump 12, the amount of fluid discharged from the pump 12 can be estimated.

The pump operation status sensor 33 is preferably capable of acquiring time-series data of the operation status indicating whether the pump 12, the motor 13, or the inverter 14 is in operation or stopped.
The FF command is acquired. The operating state is determined as follows: when the command from the control unit 20 is ON, the engine is in operation; when the command from the control unit 20 is OFF, the engine is in operation;
In the case of FF, it is recognized that the vehicle is stopped.

Furthermore, the pump operation status sensor 33 may acquire time series data of the rotation speed of the pump 12. The rotation speed of the pump 12 may be acquired by measuring the rotation speed of the pump 12 or the motor 13. Furthermore, since the rotation speed of the motor 13 is determined by the output frequency of the inverter 14, the output frequency of the inverter 14 may be acquired.

Furthermore, the pump operation status sensor 33 may acquire time-series data of a motor current value supplied to the motor 13 which is the drive source of the pump 12. By acquiring the motor current value, it is possible to acquire the workload of the pump 12. It is also possible to recognize whether the pump 12 is running dry, in a blocked state, or in an overflow state.

The usage flow rate sensor 34 measures the flow rate of the fluid used at the fluid supply destination 44 to which the fluid is supplied from the pressure tank 17 during a predetermined period.
The usage flow rate sensor 34 may measure a total usage flow rate obtained by adding up the flow rates of the multiple fluid supply destinations 44. It is preferable that the usage flow rate be measured using time series data. For example, the usage flow rate sensor 34 may be provided upstream of the fluid supply to the multiple fluid supply destinations 44, as shown in FIG.
The usage flow rate sensor 34 may be provided in each room of a housing complex such as an apartment or condominium, with each room serving as a fluid supply destination 44. The usage flow rate sensor 34 may use a water meter corresponding to each room. The predetermined period may include the start-up time when the pump is started.

The usage flow rate sensor 34 may estimate the usage flow rate from other measurement data instead of directly measuring the usage flow rate. For example,
The flow rate of fluid may be estimated from time series data by measuring the entry and exit records of people at multiple fluid supply destinations 44. Also, sensors may be attached to the toilet, bathroom shower, kitchen faucet, etc. in each room, and the flow rate of fluid may be estimated from the amount of usage. Furthermore, the flow rate of fluid may be estimated from the amount of electricity or gas used in each room.

The measuring device 3 is configured to measure physical quantities and state quantities of each part at a predetermined measurement period, for example, and output the measurement value at the time of measurement to the machine learning device 4 and the pressure tank state estimation device 5 each time the measurement period elapses. Note that the measurement periods of the pump primary pressure sensor 31, the pump secondary pressure sensor 32, the pump operation status sensor 33, and the usage flow rate sensor 34 constituting the measuring device 3 may be the same or different. Furthermore, instead of outputting discrete measurement values each time the measurement period elapses as described above, the measuring device 3 may output continuous measurement values such as analog signals.

The machine learning device 4 operates as a subject of the learning phase, and generates, by machine learning, the learning model 6 used when estimating the state of the pressure tank 17. The machine learning device 4 can employ "supervised learning" as a machine learning method.

The pressure tank state estimation device 5 operates as the subject of the inference phase, and estimates the state of the pressure tank 17 from the measurement values measured by the measuring device 3, using the trained learning model 6 generated by the machine learning device 4. The trained learning model 6 is provided to the pressure tank state estimation device 5 from the machine learning device 4 via any communication network, recording medium, etc.

FIG. 2 is a diagram showing a computer 200 constituting the machine learning device 4 and the pressure tank state estimation device 5.
FIG. 1 is a hardware configuration diagram showing an example of the

Each of the machine learning device 4 and the pressure tank state estimation device 5 is configured by a general-purpose or dedicated computer 200. As shown in FIG. 2, the computer 200 includes, as its main components, a bus 210, a processor 212, a memory 214, an input device 216, a display device 218, a storage device 220, a communication I/F (interface) unit 222,
The computer 200 includes an external device I/F unit 224, an I/O (input/output) device I/F unit 226, and a media input/output unit 228. Note that the above components may be omitted as appropriate depending on the application of the computer 200.

The processor 212 is one or more arithmetic processing devices (CPU, MPU, GPU, DSP, etc.).
The memory 21 is made up of a plurality of memory modules, each of which functions as a control unit for controlling the entire computer 200.
The internal memory 4 stores various data and programs 230 and is composed of, for example, a volatile memory (DRAM, SRAM, etc.) that functions as a main memory, and a non-volatile memory (ROM, flash memory, etc.).

The input device 216 is, for example, a keyboard, a mouse, a numeric keypad, an electronic pen, etc. The display device 218 is, for example, a liquid crystal display, an organic EL display, electronic paper, a projector, etc. The input device 216 and the display device 218 may be integrated together, such as a touch panel display. The storage device 220 is, for example, a HDD, an SSD, etc., and stores various data required for the execution of the operating system and the program 230.

The communication I/F unit 222 is connected to a network 240 such as the Internet or an intranet by wire or wirelessly, and transmits and receives data to and from other computers in accordance with a predetermined communication standard. The external device I/F unit 224 is connected to an external device 250 such as a printer or scanner by wire or wirelessly, and transmits and receives data to and from the external device 250 in accordance with a predetermined communication standard. The wireless communication means typically uses an international standard communication means. International standard communication means include IEEE802.15.4, IEEE802.15.1, IEEE802.15.2, IEEE802.15.3, IEEE802.15.4, IEEE802.15.5, IEEE802.15.6, IEEE802.15.7, IEEE802.15.8, IEEE802.15.9, IEEE802.15.10, IEEE802.15.11, IEEE802.15.12, IEEE802.15.13, IEEE802.15.14, IEEE802.15.15, IEEE802.15.16, IEEE802.15.17, IEEE802.15.18, IEEE802.15.19, IEEE802.15.26, IEEE802.15.27, IEEE802.15.28, IEEE802.15.29, IEEE802.16.29, IEEE802.16.30, IEEE802.16.40, IEEE802.16.51, IEEE802.16.52, IEEE802.16.63, IEEE802.16.74, IEEE802.16.85, IEEE802.16.96
02.15.11a, 11b, 11g, 11n, 11ac, 11ad, ISO/IEC1
There are also Bluetooth standards such as IEEE 802.15.4g and IEEE 4513-3-10.
Bluetooth (registered trademark), Bluetooth Low Energy, Wi-Fi, ZigB
ee (registered trademark), Sub-GHz, EnOcean (registered trademark), etc. can also be used. The I/O device I/F unit 226 is connected to an I/O device 260 such as various sensors and actuators, and transmits and receives various signals and data, such as detection signals from sensors and control signals to actuators, between the I/O device 260. The media input/output unit 228 is composed of a drive device such as a DVD drive, a CD drive, etc., and transmits and receives data such as DVDs,
The optical disk drive 200 reads and writes data from and to a medium 270 such as a CD.

In the computer 200 having the above configuration, the processor 212 executes the program 23.
2.0 into a work memory area of memory 214, executes it, and controls each part of computer 200 via bus 210. Note that program 230 may be stored in storage device 220 instead of memory 214. Program 230 may be recorded on a non-transitory recording medium such as a CD or DVD in an installable file format or an executable file format, and provided to computer 200 via media input/output unit 228. Program 230 may be provided to computer 200 by downloading it via network 240 through communication I/F unit 222.

The computer 200 is, for example, a desktop computer or a portable computer, and is an electronic device of any type. The computer 200 may be a client computer, a server computer, or a cloud computer.
0 may be applied to devices other than the machine learning device 4 and the pressure tank state estimation device 5.

(Machine learning device 4)
FIG. 3 is a block diagram showing an example of the machine learning device 4 according to the first embodiment.

The machine learning device 4 includes a learning data acquisition unit 400, a learning data storage unit 401, a machine learning unit 402, and a trained model storage unit 403. The machine learning device 4 includes, for example,
The learning data acquisition unit 400 is configured as the computer 200 shown in FIG.
The machine learning unit 402 is composed of the processor 212, and includes a learning data storage unit 401 and a trained model storage unit 402.
03 is composed of a storage device 220.

The learning data acquisition unit 400 is an interface unit that is connected to various external devices via the communication network 300 and acquires learning data including at least input data. The external devices are, for example, a measuring device 3 provided in the fluid supply device 1, a measuring device 3 provided in a test device 7 that simulates the fluid supply device 1, and an operator terminal 8 used by an operator. When the pressure tank state estimation device 5 is connected to the communication network 300, the learning data acquisition unit 400 may acquire learning data from the measuring device 3 provided in the fluid supply device 1 that is the estimation target of the pressure tank state estimation device 5.

Each part of the test device 7 is provided with a pump primary pressure sensor 31, a pump secondary pressure sensor 32, a pump operation status sensor 33, and a usage flow rate sensor 34 as the measurement device 3. The test device 7 also has a pressure sensor 35 inside the pressure tank that measures the pressure inside the pressure tank 17. Since it would be expensive to provide the pressure sensor 35 inside the pressure tank in all fluid supply devices 1, the pressure sensor 35 inside the pressure tank is used in the test device 7, and the data is applied to fluid supply devices 1 that do not have the pressure sensor 35 inside the pressure tank to estimate the pressure inside the pressure tank.

The learning data storage unit 401 stores the learning data acquired by the learning data acquisition unit 400 in a
Alternatively, it is a database that stores a plurality of sets. Note that the specific configuration of the database that constitutes the learning data storage unit 401 may be designed as appropriate.

The machine learning unit 402 performs machine learning using the learning data stored in the learning data storage unit 401. That is, the machine learning unit 402 inputs a plurality of sets of learning data to the learning model 6, and causes the learning model 6 to learn the correlation between the input data included in the learning data and the estimated information of the fluid supplying device 1, thereby generating a trained learning model 6. In this embodiment, a case where a neural network is adopted as a specific method of supervised learning by the machine learning unit 402 will be described.

The trained model storage unit 403 is a database that stores the trained learning model 6 generated by the machine learning unit 402. The trained learning model 6 stored in the trained model storage unit 403 is provided to an actual system (e.g., the pressure tank state estimation device 5) via any communication network, recording medium, etc. Note that, although the training data storage unit 401 and the trained model storage unit 403 are shown as separate storage units in FIG. 3, they may be configured as a single storage unit.

FIG. 4 is a data configuration diagram showing an example of data (supervised learning) used by the machine learning device 4 according to this embodiment.

The learning data includes, as input data, at least the suction side primary pressure data of the pump 12 for a predetermined period, the discharge side secondary pressure data of the pump 12 for a predetermined period, data on the operating status of the pump 12 for a predetermined period, and data on the flow rate of the fluid used at the fluid supply destination 44. The predetermined periods for the suction side primary pressure data of the pump 12, the discharge side secondary pressure data of the pump 12, data on the operating status of the pump 12, and data on the flow rate of the fluid used at the fluid supply destination 44 as input data are basically set to the same period, but if a relationship that generates a time difference between each piece of data is recognized, they may be set to a period corresponding to the time difference. Note that the input data may further include other data in addition to the suction side primary pressure data of the pump 12, the discharge side secondary pressure data of the pump 12, data on the operating status of the pump 12, and data on the flow rate of the fluid used at the fluid supply destination 44, and may include, for example, a sound generated when the pump 12 is operating, the temperature of at least one of the pump 12 or the motor 13, the temperature of the pump 12 ...
The data may further include at least one of the data of lubricant pressure, lubricant flow rate, lubricant temperature, and lubricant property values (color, particle size distribution) of 2.

The suction side primary pressure data of the pump 12 for a predetermined period of time is composed of a plurality of measured values (pump suction side primary pressure values) measured by the pump primary pressure sensor 31 at a plurality of measurement time points within the predetermined period of time. The suction side primary pressure data of the pump 12 is composed of, for example, data arranged in chronological order by measurement time, linked to a measurement time indicating the measurement time point. In addition, in a case where the pump primary pressure sensor 31 outputs continuous measured values, the learning data acquisition unit 400 converts the measured values into data at a predetermined sampling period. As a result, the suction side primary pressure data of the pump 12 for a predetermined period of time is composed of a plurality of measured values (pump suction side primary pressure values) at a plurality of sampling time points within the predetermined period of time.

The discharge side secondary pressure data of the pump 12 for a predetermined period of time is composed of a plurality of measured values (pump discharge side secondary pressure values) measured by the pump secondary pressure sensor 32 at a plurality of measurement time points within the predetermined period of time. The discharge side secondary pressure data of the pump 12 is composed of, for example, data arranged in chronological order by measurement time, linked to a measurement time indicating the measurement time point. In addition, in a case where the pump secondary pressure sensor 32 outputs continuous measured values, the learning data acquisition unit 400 converts the measured values into data at a predetermined sampling period. As a result, the discharge side secondary pressure data of the pump 12 for a predetermined period of time is composed of a plurality of measured values (pump discharge side secondary pressure values) at a plurality of sampling time points within the predetermined period of time.

The data on the operating status of the pump 12 during a predetermined period is preferably at least one of data on the operating state of the pump 12 indicating whether the pump 12 is operating or stopped, data on the rotation speed of the pump 12 during a predetermined period, and data on the motor current value supplied to the motor 13, which is the drive source of the pump 12, during a predetermined period. In this embodiment, the data on the operating status of the pump 12 is:
Data on the operating state of the pump 12, indicating whether the pump 12 is operating or stopped, data on the number of revolutions of the pump 12 during a predetermined period, and data on the rotation speed of the motor 11, which is the driving source of the pump 12, during the predetermined period.
In the following description, the data is assumed to consist of motor current value data supplied to terminal 3.

The operational status data indicating whether the pump 12 is operating or stopped during a predetermined period is composed of a plurality of measurement values (operating/stopped, etc.) measured by the pump operation status sensor 33 at a plurality of measurement time points within the predetermined period. The operational status data is, for example, linked to a measurement time indicating the measurement time point and composed of an array of data arranged in chronological order by measurement time. Note that in a case where the pump operation status sensor 33 outputs continuous measurement values, the learning data acquisition unit 400 converts the measurement values into data at a predetermined sampling period. As a result, the operational status data during a predetermined period is composed of a plurality of measurement values (operating/stopped, etc.) at a plurality of sampling time points within the predetermined period.

The data on the rotation speed of the pump 12 during a predetermined period is composed of a plurality of measurement values (rotation speeds) measured by the pump operation status sensor 33 at a plurality of measurement time points within the predetermined period. The data on the rotation speed is, for example, linked to a measurement time indicating the measurement time point, and composed as an array of data arranged in chronological order by the measurement time. Note that, in a case where the pump operation status sensor 33 outputs continuous measurement values, the learning data acquisition unit 400 converts the measurement values into data at a predetermined sampling period. As a result, the data on the rotation speed during a predetermined period is composed of a plurality of measurement values (rotation speeds) at a plurality of sampling time points within the predetermined period.

The motor current value data supplied to the motor 13, which is the drive source of the pump 12, during a specified period of time is composed of multiple measurement values (current values) measured by the pump operation status sensor 33 at multiple measurement points within the specified period. The current value data is, for example, linked to the measurement time indicating the measurement point, and composed as an array of data arranged in chronological order by measurement time. Note that in cases where the pump operation status sensor 33 outputs continuous measurement values, the learning data acquisition unit 400 digitizes the measurement values at a specified sampling period. As a result,
The data on the current value for a given period of time is made up of a plurality of measured values (current values) at a plurality of sampling points within the given period of time.

The data of the flow rate of the fluid used at the fluid supply destination 44 to which the fluid is supplied from the pressure tank 17 during a predetermined period is composed of a measured value of the total usage flow rate obtained by adding up the flow rates of the multiple fluid supply destinations 44. The data of the flow rate of the fluid used at the fluid supply destination 44 to which the fluid is supplied from the pressure tank 17 is composed of, for example, data arranged in chronological order by measurement time, linked to a measurement time indicating the time point of measurement. Note that in cases where the usage flow rate sensor 34 outputs continuous measurement values, the learning data acquisition unit 400 converts the measurement values into data at a predetermined sampling period. As a result, the data of the flow rate of the fluid used at the fluid supply destination 44 to which the fluid is supplied from the pressure tank 17 during a predetermined period is composed of multiple measurement values (flow rates of the fluid used at the fluid supply destination 44) at multiple sampling points within the predetermined period.

When "supervised learning" is adopted as the machine learning, the learning data further includes, as output data associated with the input data, data on the enclosed pressure of the pressure tank 17 during a predetermined period measured by the pressure sensor 35 in the pressure tank of the test device 7. In supervised learning, the output data is referred to as, for example, teacher data or correct answer label.

The data of the enclosed pressure of the pressure tank 17 is point data of the enclosed pressure at a point included in a specified period or at a point later than the specified period, or time series data of the enclosed pressure during a period including a specified period or at a point later than the specified period.

Therefore, as shown in FIG. 4, the learning data according to this embodiment includes primary pressure data on the suction side of the pump 12 for a predetermined period, secondary pressure data on the discharge side of the pump 12 for a predetermined period,
The data is configured by associating input data including data on the operating status of the pump 12 over a specified period of time and data on the flow rate of the fluid used at the fluid supply destination 44 with output data including data on the sealed pressure of the pressure tank 17 over a specified period of time.

Here, the input data included in the learning data (the suction side 1 of the pump 12 in a predetermined period)
The correlation between the data on the filled pressure of the pressure tank 28 and the data on the secondary pressure of the pump 12 for a specified period, data on the operating status of the pump 12 for a specified period, and data on the flow rate of the fluid used at the fluid supply destination 44) is described below.

The secondary pressure on the discharge side of the pump 12 during a predetermined period varies depending on the primary pressure on the suction side of the pump 12 during the predetermined period, the operating conditions of the pump 12 during the predetermined period, and the charged pressure of the pressure tank 17. However, measuring the charged pressure of the pressure tank 17 requires an expensive sensor, and has traditionally been largely dependent on the experience of the operator. Therefore, by measuring the primary pressure on the suction side of the pump 12 during a predetermined period, the operating conditions of the pump 12 during the predetermined period, and the secondary pressure on the discharge side of the pump 12 during the predetermined period using learning data as input data, it becomes possible to estimate the charged pressure of the pressure tank 17. In addition, by measuring the flow rate of the fluid used at the fluid supply destination 44 during a predetermined period, it becomes possible to estimate the charged pressure of the pressure tank 17 with higher accuracy.

When acquiring the above-mentioned learning data, the learning data acquiring unit 400 acquires as input data from the measuring device 3, the suction side primary pressure data of the pump 12 for a predetermined period, the operating status data of the pump 12 for a predetermined period, the discharge side secondary pressure data of the pump 12 for a predetermined period, and the flow rate data of the fluid used at the fluid supply destination 44 for a predetermined period, all of which are measured by the measuring device 3 provided in the fluid supply device 1 or the test device 7. When an operator estimates the charged pressure of the pressure tank 17 for a predetermined period when the input data was measured by the measuring device 3 and inputs the estimated value to the operator terminal 8, the learning data acquiring unit 400 acquires the estimated value inputted at the operator terminal 8 as output data (teacher data) from the operator terminal 8. The learning data acquiring unit 400 then configures one learning data by associating the input data with the output data, and stores the learning data in the learning data storage unit 401.

FIG. 5 is a schematic diagram showing an example of a neural network model used in the machine learning device 4 according to this embodiment.

The learning model 6 is configured as a neural network model shown in Fig. 5. The neural network model includes l neurons (x1 to xl) in the input layer, m neurons (y11 to y1m) in the first hidden layer, and n neurons (y2
The input layer is made up of o neurons (z1 to zo) in the output layer.

Each neuron in the input layer is associated with a respective input data included in the learning data. Each neuron in the output layer is associated with a respective output data included in the learning data. Note that a predetermined pre-processing may be performed on the input data before it is input to the input layer, and a predetermined post-processing may be performed on the output data after it is output from the output layer.

The first and second hidden layers are also called hidden layers. In addition to the first and second hidden layers, the neural network may further include a plurality of hidden layers.
Only the first hidden layer may be a hidden layer. Synapses that connect neurons in each layer are laid between the input layer and the first hidden layer, between the first hidden layer and the second hidden layer, and between the second hidden layer and the output layer, and a weight wi (i is a natural number) is assigned to each synapse.

A neural network model uses training data to learn the correlation between the input data and output data by inputting input data contained in the training data into an input layer, and comparing the output data output from the output layer as the inference result with the output data (teacher data) contained in the training data.

Specifically, input data included in the learning data is input to each neuron in the input layer, and the value of each neuron in the output layer is calculated by performing a process for all neurons other than the input layer, in which the value of the neuron on the input side connected to the neuron in question is calculated as the sum of a sequence of multiplication values of the value of the neuron on the input side connected to the neuron in question and the weight wi associated with the synapse connecting the neuron on the output side and the neuron on the input side.

Then, the values (z1 to zo) output to each neuron in the output layer as the inference results are compared with the values (t1 to to) of the teacher data corresponding to each of the output data included in the learning data to determine the error, and a process (backpropagation) is performed to adjust the weights wi associated with each synapse so that the error is reduced.

When a predetermined learning end condition is met, such as the above series of steps being repeated a predetermined number of times or the above error being smaller than a tolerance, the machine learning is terminated and the trained neural network model (all weights w i associated with each synapse)
is generated as:

(Machine learning methods)
6 is a flowchart showing an example of a machine learning method performed by the machine learning device 4 according to this embodiment. The machine learning method corresponds to the learning phase in FIG.

First, in step S11, the learning data acquisition unit 400 prepares a desired number of pieces of learning data as advance preparation for starting machine learning, and stores the prepared learning data in the learning data storage unit 401. The number of pieces of learning data to be prepared here may be set in consideration of the inference accuracy required for the learning model 6 to be finally obtained.

There are several methods for preparing training data. For example,
When estimating the charged pressure of the pressure tank 17 in a specific fluid supply device 1 or test device 7, various measured values in a period before and after the estimation are obtained using the measurement device 3, and the operator uses the operator terminal 8 to input the estimated results in correspondence with these measured values, thereby preparing input data and output data that constitute learning data. By repeating this operation, it is possible to prepare multiple sets of learning data. As another method, for example, it is also possible to obtain learning data by intentionally setting the charged pressure of the pressure tank of the fluid supply device 1 or test device 7 to a predetermined value.

Next, in step S12, the machine learning unit 402 prepares a learning model 6 before learning in order to start machine learning. The learning model 6 before learning prepared here is configured with the neural network model exemplified in FIG. 5, and the weights of each synapse are set to an initial value. Each neuron in the input layer is associated with the following input data included in the learning data: suction side primary pressure data of the pump 12 during a predetermined period, discharge side secondary pressure data of the pump 12 during a predetermined period, data on the operating status of the pump 12 during a predetermined period, and data on the flow rate of the fluid used at the fluid supply destination 44 during a predetermined period. Each neuron in the output layer is associated with the respective data on the enclosed pressure of the pressure tank 17 as output data included in the learning data.

Next, in step S<b>13 , the machine learning unit 402 acquires, for example, one piece of learning data at random from the multiple sets of learning data stored in the learning data storage unit 401 .

Next, in step S14, the machine learning unit 402 inputs the input data included in one piece of learning data to the input layer of the prepared learning model 6 before learning (or during learning). As a result, output data is output as an inference result from the output layer of the learning model 6, but the output data is generated by the learning model 6 before learning (or during learning). Therefore, in the state before learning (or during learning), the output data output as an inference result indicates information different from the output data (teacher data) included in the learning data.

Next, in step S15, the machine learning unit 402 performs machine learning by comparing the output data (teacher data) included in one of the learning data acquired in step S12 with the output data output from the output layer as an inference result in step S130 and adjusting the weight of each synapse. In this way, the machine learning unit 402 causes the learning model 6 to learn the correlation between the input data and the output data.

Next, in step S16, the machine learning unit 402 determines whether or not it is necessary to continue machine learning based on, for example, the error between the output data and the teacher data and the remaining number of unlearned learning data stored in the learning data storage unit 401.

In step S16, if the machine learning unit 402 determines to continue the machine learning (No in step S16), the process returns to step S13, and steps S13 to S15 are performed multiple times on the learning model 6 being trained using untrained learning data. On the other hand, in step S16, if the machine learning unit 402 determines to end the machine learning (No in step S1
6 (Yes), and proceed to step S17.

Then, in step S17, the machine learning unit 402 stores the trained learning model 6 generated by adjusting the weights associated with each synapse in the trained model storage unit 43.
6 is completed. In the machine learning method, step S11 is a learning data storage step, steps S12 to S16 are machine learning steps, and step S1
Step 7 corresponds to the learned model storage step.

As described above, according to the machine learning device 4 and the machine learning method of this embodiment, the suction side primary pressure data of the pump 12 for a predetermined period as input data included in the learning data,
A learning model 6 can be provided that can infer (estimate) the sealed pressure of the pressure tank 17 with high accuracy from data on the secondary pressure of the discharge side of the pump 12 for a specified period of time, data on the operating status of the pump 12 for a specified period of time, and data on the flow rate of the fluid used at the fluid supply destination 44 for a specified period of time.

(Pressure Tank State Estimation Device 5)
FIG. 7 is a block diagram showing an example of a pressure tank state estimating device 5 according to this embodiment.

The pressure tank state estimation device 5 includes an input data acquisition unit 500, an inference unit 501, a trained model storage unit 502, and an output processing unit 503. The pressure tank state estimation device 5 is configured, for example, by the computer 200 shown in FIG. 2. In this case, the input data acquisition unit 500
is configured by the communication I/F unit 222 or the I/O device I/F unit 226, the inference unit 501 and the output processing unit 503 are configured by the processor 212, and the trained model storage unit 502 is configured by
The pressure tank state estimation device 5 is configured as a storage device 220. The pressure tank state estimation device 5 may be incorporated in the fluid supply device 1, or may be incorporated in a higher-level management device of the fluid supply device 1 (for example, an equipment controller, an equipment management system that manages multiple pieces of equipment, etc.).

The input data acquisition unit 500 is connected to the measurement device 3 (pump primary pressure sensor 31, pump secondary pressure sensor 32, pump operation status sensor 33, and usage flow rate sensor 34) provided in the fluid supply device 1, and acquires input data (intake side primary pressure data of the pump 12, discharge side secondary pressure data of the pump 12, and the like for a predetermined period of time) based on the measured values measured by the measurement device 3.
2 and data on the flow rate of the fluid used at the fluid supply destination 44).

The inference unit 501 converts the input data acquired by the input data acquisition unit 500 into a learning model 6
and performs an inference process to infer information about the pressure tank 17. For the inference process, a machine learning device 4 and a trained learning model 6 that has been trained through supervised learning using a machine learning method are used.

The inference unit 501 not only has a function of performing inference processing using the learning model 6, but also has a pre-processing function of adjusting the input data acquired by the input data acquisition unit 500 into a desired format, etc., and inputting the adjusted data to the learning model 6 as pre-processing of the inference processing, and a post-processing function of applying a predetermined logical formula or calculation formula to the value of the output data output from the learning model 6 as post-processing of the inference processing, thereby finally estimating the state of the pressure tank 17. The inference result of the inference unit 501 is
It is preferable to store the results in the learned model memory unit 502 or in another memory device (not shown), and past inference results can be used as learning data for online learning or re-learning, for example, to further improve the inference accuracy of the learning model 6.

The trained model storage unit 502 is a database that stores trained learning models 6 used in the inference process of the inference unit 501. The number of trained models 6 stored in the trained model storage unit 502 is not limited to one. For example, the number of trained models 6 may be different depending on the number of input data, or
A plurality of learning models 6 using different machine learning techniques may be stored and selectively made available for use.

The output processing unit 503 performs processing to output the inference result of the inference unit 501, i.e., the state of the pressure tank 17. As a specific output means, various means can be adopted. For example, the output processing unit 503 may notify the operator of the state of the pressure tank 17 by display or sound, or may transmit the state of the pressure tank 17 as history to, for example, the fluid supplying device 1 or a higher-level management device (not shown) of the fluid supplying device 1, store it in a storage unit of the management device of the fluid supplying device 1, or use it for driving control of the motor 13.

(Pressure tank status estimation method)
FIG. 8 is a flowchart showing an example of a pressure tank state estimation method performed by the pressure tank state estimation device 5 according to this embodiment.

When a current is supplied to the motor 13, the motor 13 is driven to rotate, and the pump 12 is rotated. At this time, the series of pressure tank state estimation methods shown in Fig. 8 may be executed at a predetermined estimation timing by the pressure tank state device 5. The estimation timing may be, for example, at a predetermined time interval, or when a predetermined event occurs (when an operator issues an operation instruction, when a maintenance operation is performed, etc.).

First, in step S21, the input data acquisition unit 500 acquires input data (the pump 12 pressure sensor 31, the pump secondary pressure sensor 32, the pump operation status sensor 33, and the usage flow rate sensor 34) based on the measured values measured by the measurement device 3.
The data of the suction side primary pressure of the pump 12, the discharge side secondary pressure of the pump 12, the data of the operating status of the pump 12, and the data of the flow rate of the fluid used at the fluid supply destination 44 are acquired.
The input data is required to include measurement values measured by the measuring device 3 for the same period as the predetermined period or for a period longer than the predetermined period.

Next, in step S22, the inference unit 501 performs preprocessing on the input data and inputs the preprocessed data to the input layer of the learning model 6, and obtains the output data output from the output layer of the learning model 6.

Next, in step S23, the inference unit 501 estimates the pressure inside the pressure tank 17.

As described above, the pressure tank state estimation device 5 and pressure tank state estimation method of this embodiment can estimate the state of the pressure tank 17 with high accuracy without relying on the experience of the operator.

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

In the above embodiment, a case has been described in which a neural network is adopted as a specific method of machine learning by the machine learning unit 402, but the machine learning unit 402 may adopt any other machine learning method. Examples of other machine learning methods include tree types such as decision trees and regression trees, ensemble learning such as bagging and boosting, neural network types (including deep learning) such as recurrent neural networks and convolutional neural networks, clustering types 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,
Examples include support vector machines.

The present invention can also be provided in the form of a program (machine learning program) 230 for causing a computer 200 shown in Fig. 2 to execute each step of the machine learning method according to the embodiment. The present invention can also be provided in the form of a program (pressure tank state estimation program) 230 for causing a computer 200 shown in Fig. 2 to execute each step of the pressure tank state estimation method according to the embodiment.

(Inference device, inference method, and inference program)
The present invention can be provided not only in the form of the pressure tank state estimation device 5 (pressure tank state estimation method or pressure tank state estimation program) according to the above embodiment, but also in the form of an inference device (inference method or inference program) used to estimate the state of the pressure tank 17. In this case, the inference device (inference method or inference program) can include a memory and a processor, and the processor executes a series of processes. The series of processes includes data on the suction side primary pressure of the pump 12, data on the discharge side secondary pressure of the pump 12, data on the operating status of the pump 12, and data on the fluid supply destination 44 during a predetermined period.
An input data acquisition process (input data acquisition step) for acquiring input data including data on the flow rate of the fluid used in the pressure tank 1.
and an inference process (inference step) for inferring state estimation information of the seventh embodiment.

By providing it in the form of an inference device (inference method or inference program), it can be easily applied to various devices compared to the case of implementing the pressure tank state estimation device 5. It can be naturally understood by those skilled in the art that when the inference device (inference method or inference program) infers the state of the pressure tank 17, the inference method implemented by the inference unit 501 of the pressure tank state estimation device 5 may be applied using the machine learning device 4 and the trained learning model 6 generated by the machine learning method according to the above embodiment.

1...fluid supply device,
12 ... pump, 13 ... motor, 14 ... inverter, 17 ... pressure tank,
41...Fluid supply source, 44...Fluid supply destination,
2...Pressure tank state estimation system,
3...measuring device, 31...pump primary pressure sensor, 32...pump secondary pressure sensor, 33...pump operation status sensor, 34...usage flow rate sensor,
4...machine learning device, 5...pressure tank state estimation device, 6...learning model,
400...learning data acquisition unit, 401...learning data storage unit, 402...machine learning unit, 40
3...Model storage unit,
500: input data acquisition unit, 501: inference unit, 502: model storage unit, 503: output processing unit,
200...Computer

Claims (14)

A machine learning device that generates a learning model for estimating a pressure of a pressure tank in a fluid supply device including a pump that transfers a fluid and a pressure tank provided on a discharge side of the pump, the machine learning device comprising:
a learning data storage unit that stores a plurality of sets of learning data in which the input data includes pump pressure data for the pump during a predetermined period, data on the operating status of the pump during the predetermined period, and data on the usage flow rate of the fluid used at a fluid supply destination to which the fluid is supplied from the pressure tank during the predetermined period , and data on the filled pressure of the pressure tank as output data, and in which the input data and the output data are associated with each other;
a machine learning unit that causes the learning model to learn a correlation between the input data and the output data by inputting a plurality of sets of the learning data to the learning model;
A trained model storage unit that stores the learning model trained by the machine learning unit.
Machine learning device. The data on the usage flow rate of the fluid included in the input data is
time series data of a total usage flow rate obtained by summing up the measurement results of the usage flow rate measuring meters provided at the plurality of fluid supply destinations;
Time series data of entry and exit records provided at the plurality of fluid supply destinations;
Time series data of faucet usage amounts provided at the plurality of fluid supply destinations;
Time series data of electricity usage provided at the plurality of fluid supply destinations, or
Time series data of gas usage provided at the plurality of fluid supply destinations;
That is,
The machine learning device according to claim 1 . The pump pressure data included in the input data is
Time series data of the suction side pressure of the pump;
At least one of time series data of the discharge side pressure of the pump,
The machine learning device according to claim 1 or 2 . The data on the operation status of the pump included in the input data is
Time series data of the operation state of the pump, which indicates whether the pump is operating or stopped;
Time series data of the rotation speed of the pump, or
Time series data of a motor current value supplied to a motor which is a driving source of the pump.
The machine learning device according to claim 1 . The data on the charged pressure of the pressure tank included in the output data is
Point data of the charged pressure at a time point included in the predetermined period or at a time point after the predetermined period, or
The time series data of the charged pressure in the predetermined period or a period including a point after the predetermined period.
The machine learning device according to claim 1 . The predetermined period is:
a start time when the pump is started;
The machine learning device according to claim 1 . 7. A pressure tank state estimation device for estimating a pressure of a pressure tank in a fluid supply device including a pump for transporting a fluid and a pressure tank provided on a discharge side of the pump, using a learning model generated by the machine learning device according to any one of claims 1 to 6,
an input data acquisition unit that acquires input data including pump pressure data for the pump during a predetermined period and data on an operating status of the pump during the predetermined period;
an inference unit that inputs the input data acquired by the input data acquisition unit into the learning model and infers the charged pressure of the pressure tank;
Pressure tank condition estimator. 1. An inference device used to estimate a pressure of a pressure tank in a fluid supply device including a pump for transferring a fluid and a pressure tank provided on a discharge side of the pump, comprising:
The inference device comprises a memory and a processor;
The processor,
an input data acquisition process for acquiring input data including pump pressure data for the pump during a predetermined period , data on the operating status of the pump during the predetermined period, and data on the usage flow rate of the fluid used at a fluid supply destination to which the fluid is supplied from the pressure tank during the predetermined period ;
When the input data is acquired in the input data acquisition process, an inference process is executed to infer the charged pressure of the pressure tank.
Inference device. A machine learning method for generating a learning model for estimating a pressure of a pressure tank in a fluid supply device including a pump for transporting a fluid and a pressure tank provided on a discharge side of the pump, the method comprising:
a learning data storage step of storing a plurality of sets of learning data in a learning data storage unit , the learning data including input data including pump pressure data for the pump during a predetermined period, data regarding the operating status of the pump during the predetermined period, and data regarding a usage flow rate of the fluid used at a fluid supply destination to which the fluid is supplied from the pressure tank during the predetermined period, and output data including data regarding the filled pressure of the pressure tank, the input data and the output data corresponding to each other;
a machine learning process in which a correlation between the input data and the output data is learned by the learning model through a machine learning unit by inputting a plurality of sets of the learning data into the learning model;
A trained model storage step of storing the learning model trained by the machine learning unit in a trained model storage unit.
Machine learning methods. A pressure tank state estimation method for estimating a pressure of a pressure tank in a fluid supply device including a pump for transporting a fluid and a pressure tank provided on a discharge side of the pump, using a learning model generated by the machine learning device according to any one of claims 1 to 6, comprising:
an input data acquisition step of acquiring input data including time series data of a pump pressure related to the pump during a predetermined period and time series data of an operating status of the pump during the predetermined period;
and an inference step of inputting the input data acquired by the input data acquisition step into the learning model and inferring the charged pressure of the pressure tank.
Pressure tank condition estimation method. 1. An inference method used to estimate a pressure of a pressure tank in a fluid supply device including a pump for transporting a fluid and a pressure tank provided on a discharge side of the pump, comprising:
an input data acquisition step of acquiring input data including pump pressure data for the pump during a predetermined period , data on the operating status of the pump during the predetermined period, and data on the usage flow rate of the fluid used at a fluid supply destination to which the fluid is supplied from the pressure tank during the predetermined period ;
When the input data is acquired in the input data acquisition step, an inference step of inferring a charged pressure of the pressure tank is executed.
Inference methods. A method for causing a computer to execute each step of the machine learning method according to claim 9 ,
Machine learning programs. A method for causing a computer to execute each step of the pressure tank state estimation method according to claim 10 ,
Pressure tank condition estimation program. A method for causing a computer to execute each step of the inference method according to claim 11 ,
Inference program.

Images