WO2022071114A1 - 機械学習装置、摺動面診断装置、推論装置、機械学習方法、機械学習プログラム、摺動面診断方法、摺動面診断プログラム、推論方法、及び、推論プログラム - Google Patents
機械学習装置、摺動面診断装置、推論装置、機械学習方法、機械学習プログラム、摺動面診断方法、摺動面診断プログラム、推論方法、及び、推論プログラム Download PDFInfo
- Publication number
- WO2022071114A1 WO2022071114A1 PCT/JP2021/035082 JP2021035082W WO2022071114A1 WO 2022071114 A1 WO2022071114 A1 WO 2022071114A1 JP 2021035082 W JP2021035082 W JP 2021035082W WO 2022071114 A1 WO2022071114 A1 WO 2022071114A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- data
- learning
- sliding member
- side sliding
- machine learning
- Prior art date
Links
- 238000010801 machine learning Methods 0.000 title claims abstract description 144
- 238000003745 diagnosis Methods 0.000 title claims abstract description 37
- 238000000034 method Methods 0.000 title claims description 69
- 230000001133 acceleration Effects 0.000 claims abstract description 44
- 238000003860 storage Methods 0.000 claims abstract description 32
- 238000013500 data storage Methods 0.000 claims abstract description 23
- 230000008569 process Effects 0.000 claims description 31
- 238000012545 processing Methods 0.000 claims description 23
- 239000000314 lubricant Substances 0.000 claims description 13
- 238000002405 diagnostic procedure Methods 0.000 claims description 9
- 238000012549 training Methods 0.000 abstract description 13
- 230000005856 abnormality Effects 0.000 description 29
- 238000005259 measurement Methods 0.000 description 28
- 210000002569 neuron Anatomy 0.000 description 25
- 238000010586 diagram Methods 0.000 description 18
- 230000002159 abnormal effect Effects 0.000 description 17
- 210000000225 synapse Anatomy 0.000 description 15
- 238000004891 communication Methods 0.000 description 11
- 238000012360 testing method Methods 0.000 description 10
- 238000005070 sampling Methods 0.000 description 8
- 230000002093 peripheral effect Effects 0.000 description 7
- 238000013528 artificial neural network Methods 0.000 description 6
- 230000006870 function Effects 0.000 description 6
- 238000003062 neural network model Methods 0.000 description 6
- 230000008859 change Effects 0.000 description 5
- 238000012805 post-processing Methods 0.000 description 5
- 238000007726 management method Methods 0.000 description 4
- 238000012544 monitoring process Methods 0.000 description 4
- 238000012423 maintenance Methods 0.000 description 3
- 238000007781 pre-processing Methods 0.000 description 3
- 238000011109 contamination Methods 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 238000005461 lubrication Methods 0.000 description 2
- 238000002360 preparation method Methods 0.000 description 2
- 239000013598 vector Substances 0.000 description 2
- 238000005299 abrasion Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 239000007864 aqueous solution Substances 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 239000003795 chemical substances by application Substances 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000556 factor analysis Methods 0.000 description 1
- 238000009472 formulation Methods 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000007477 logistic regression Methods 0.000 description 1
- 230000001050 lubricating effect Effects 0.000 description 1
- 239000010687 lubricating oil Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000000491 multivariate analysis Methods 0.000 description 1
- 239000003921 oil Substances 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 239000000843 powder Substances 0.000 description 1
- 238000003825 pressing Methods 0.000 description 1
- 238000000513 principal component analysis Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Images
Classifications
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F16—ENGINEERING ELEMENTS AND UNITS; GENERAL MEASURES FOR PRODUCING AND MAINTAINING EFFECTIVE FUNCTIONING OF MACHINES OR INSTALLATIONS; THERMAL INSULATION IN GENERAL
- F16C—SHAFTS; FLEXIBLE SHAFTS; ELEMENTS OR CRANKSHAFT MECHANISMS; ROTARY BODIES OTHER THAN GEARING ELEMENTS; BEARINGS
- F16C41/00—Other accessories, e.g. devices integrated in the bearing not relating to the bearing function as such
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F16—ENGINEERING ELEMENTS AND UNITS; GENERAL MEASURES FOR PRODUCING AND MAINTAINING EFFECTIVE FUNCTIONING OF MACHINES OR INSTALLATIONS; THERMAL INSULATION IN GENERAL
- F16C—SHAFTS; FLEXIBLE SHAFTS; ELEMENTS OR CRANKSHAFT MECHANISMS; ROTARY BODIES OTHER THAN GEARING ELEMENTS; BEARINGS
- F16C17/00—Sliding-contact bearings for exclusively rotary movement
- F16C17/12—Sliding-contact bearings for exclusively rotary movement characterised by features not related to the direction of the load
- F16C17/24—Sliding-contact bearings for exclusively rotary movement characterised by features not related to the direction of the load with devices affected by abnormal or undesired positions, e.g. for preventing overheating, for safety
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F16—ENGINEERING ELEMENTS AND UNITS; GENERAL MEASURES FOR PRODUCING AND MAINTAINING EFFECTIVE FUNCTIONING OF MACHINES OR INSTALLATIONS; THERMAL INSULATION IN GENERAL
- F16N—LUBRICATING
- F16N29/00—Special means in lubricating arrangements or systems providing for the indication or detection of undesired conditions; Use of devices responsive to conditions in lubricating arrangements or systems
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F16—ENGINEERING ELEMENTS AND UNITS; GENERAL MEASURES FOR PRODUCING AND MAINTAINING EFFECTIVE FUNCTIONING OF MACHINES OR INSTALLATIONS; THERMAL INSULATION IN GENERAL
- F16N—LUBRICATING
- F16N29/00—Special means in lubricating arrangements or systems providing for the indication or detection of undesired conditions; Use of devices responsive to conditions in lubricating arrangements or systems
- F16N29/04—Special means in lubricating arrangements or systems providing for the indication or detection of undesired conditions; Use of devices responsive to conditions in lubricating arrangements or systems enabling a warning to be given; enabling moving parts to be stopped
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01H—MEASUREMENT OF MECHANICAL VIBRATIONS OR ULTRASONIC, SONIC OR INFRASONIC WAVES
- G01H1/00—Measuring characteristics of vibrations in solids by using direct conduction to the detector
- G01H1/003—Measuring characteristics of vibrations in solids by using direct conduction to the detector of rotating machines
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01H—MEASUREMENT OF MECHANICAL VIBRATIONS OR ULTRASONIC, SONIC OR INFRASONIC WAVES
- G01H17/00—Measuring mechanical vibrations or ultrasonic, sonic or infrasonic waves, not provided for in the preceding groups
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M13/00—Testing of machine parts
- G01M13/04—Bearings
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M99/00—Subject matter not provided for in other groups of this subclass
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/20—Ensemble learning
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F16—ENGINEERING ELEMENTS AND UNITS; GENERAL MEASURES FOR PRODUCING AND MAINTAINING EFFECTIVE FUNCTIONING OF MACHINES OR INSTALLATIONS; THERMAL INSULATION IN GENERAL
- F16C—SHAFTS; FLEXIBLE SHAFTS; ELEMENTS OR CRANKSHAFT MECHANISMS; ROTARY BODIES OTHER THAN GEARING ELEMENTS; BEARINGS
- F16C2233/00—Monitoring condition, e.g. temperature, load, vibration
Definitions
- the present invention relates to a machine learning device, a sliding surface diagnostic device, an inference device, a machine learning method, a machine learning program, a sliding surface diagnostic method, a sliding surface diagnostic program, an inference method, and an inference program.
- a sliding machine element having a sliding surface has been conventionally used as one of many machine elements constituting the machine equipment.
- the sliding machine element includes a fixed-side sliding member and a rotating-side sliding member, and the sliding surfaces of both members slide to realize relative motion of both members. Therefore, the condition of the sliding surface greatly affects the function of the sliding machine element, and it is extremely important to diagnose the condition of the sliding surface and manage it to an appropriate condition from the viewpoint of maintenance of machinery and equipment. Is.
- Patent Document 1 and Patent Document 2 disclose an apparatus for diagnosing a state of a sliding surface in a sliding machine element.
- Patent Document 1 discloses an apparatus for detecting damage to a sliding portion by monitoring physical quantities such as pressure and temperature of the sliding portion between the rotating shaft and the sliding bearing.
- an elastic wave sensor as a sliding state detection means is attached to a bearing device, and lubricating oil is applied to the sliding surface between the bush and the connecting pin in response to a change in vibration detected by the elastic wave sensor.
- a device for detecting a shortage tendency is disclosed.
- the state of the sliding surface is diagnosed based on individual physical quantities (pressure, temperature, and vibration). Therefore, when the change in the state of the sliding surface is expressed in a physical quantity other than the above or is expressed in a plurality of physical quantities in a complex manner, the devices disclosed in Patent Document 1 and Patent Document 2, respectively. I could't handle it.
- diagnosis associated with complex events such as those based on multiple physical quantities, there are large individual differences in the part that depends on the worker's experience (including tacit knowledge) and the worker, and a device that automates such diagnosis. Realization is strongly requested.
- the present invention is a machine learning device and a sliding surface diagnostic device capable of diagnosing the state of a sliding surface in a sliding machine element with high accuracy without depending on the experience of an operator.
- a reasoning device a machine learning method, a machine learning program, a sliding surface diagnostic method, a sliding surface diagnostic program, an inference method, and an inference program.
- the machine learning device is A machine learning device that generates a learning model used in a sliding surface diagnostic device that diagnoses the state of a sliding surface on which a fixed-side sliding member and a rotating-side sliding member slide.
- a learning data storage unit that stores one or a plurality of sets of learning data including at least data and data of vibration generated in at least one of the fixed-side sliding member and the rotating-side sliding member in the predetermined period as input data.
- a machine learning unit that causes the learning model to learn the correlation between the input data and the diagnostic information of the sliding surface by inputting one or a plurality of sets of the learning data into the learning model. It includes a learned model storage unit that stores the learning model trained by the machine learning unit.
- the sliding surface diagnostic apparatus is It is a sliding surface diagnostic device that diagnoses the state of the sliding surface on which the fixed side sliding member and the rotating side sliding member slide by using the learning model generated by the machine learning device.
- An input data acquisition unit that acquires data and input data including data of vibration generated in at least one of the fixed-side sliding member and the rotating-side sliding member in the predetermined period.
- the input data acquired by the input data acquisition unit is input to the learning model, and an inference unit for inferring diagnostic information on the sliding surface is provided.
- a learning model capable of inferring (estimating) diagnostic information on a sliding surface with high accuracy from motor current value data, contact electrical resistance data, and vibration data.
- the state of the sliding surface can be diagnosed with high accuracy without depending on the experience of the operator.
- FIG. 1 is a schematic configuration diagram showing an example of a first sliding machine element 1A to which the sliding surface diagnostic system 2 according to the first embodiment is applied.
- FIG. 2 is a schematic configuration diagram showing an example of a second sliding machine element 1B to which the sliding surface diagnostic system 2 according to the first embodiment is applied.
- the sliding machine elements 1A and 1B are a first sliding machine element 1A (see FIG. 1) in which the sliding surfaces 100 and 110 are in line contact with each other and a second sliding machine in which the sliding surfaces 100 and 110 are in surface contact. It is roughly classified into element 1B (see FIG. 2).
- Each of the first and second sliding machine elements 1A and 1B includes a fixed-side sliding member 10 having a fixed-side sliding surface 100 and a rotating-side sliding member 11 having a rotating-side sliding surface 110. ..
- sliding surface is used as a term including not only the case of indicating both the fixed side sliding surface 100 and the rotating side sliding surface 110 but also the case of indicating either one of them. ..
- the fixed side sliding member 10 is fixed to a support member (not shown).
- the rotating side sliding member 11 is fixed to the rotating shaft 12, and the rotating shaft 12 is connected to the motor 13 as a drive source.
- the rotating side sliding member 11 is rotated integrally with the rotating shaft 12 by supplying a current from the motor control device 14 to the motor 13 and driving the motor 13 to rotate.
- the fixed side sliding surface 100 and the rotating side sliding surface 110 slide via the lubricant.
- the lubricant may be any substance, such as oil, water, aqueous solution and the like.
- the first sliding machine element 1A has sliding surfaces 100 and 110 in line contact, and constitutes, for example, a sliding bearing that receives a radial load.
- the fixed-side sliding member 10 and the rotating-side sliding member 11 are formed in an annular shape and are arranged concentrically around the rotating shaft 12.
- the rotating side sliding member 11 is fixed to the rotating shaft 12 so as to cover the outer peripheral surface of the rotating shaft 12.
- the outer peripheral surface of the rotating side sliding member 11 corresponds to the rotating side sliding surface 110.
- the rotating shaft 12 itself may function as the rotating side sliding member 11 without separating the rotating side sliding member 11 and the rotating shaft 12, and in that case, the outer peripheral surface of the rotating shaft 12 may be used.
- the fixed side sliding member 10 is arranged so as to cover the rotating side sliding surface 110.
- the inner peripheral surface of the fixed side sliding member 10 corresponds to the fixed side sliding surface 100.
- the second sliding mechanical element 1B has surface-contact sliding surfaces 100 and 110, and constitutes, for example, a sliding bearing or a mechanical seal that receives a thrust load.
- the fixed-side sliding member 10 and the rotating-side sliding member 11 are formed in an annular shape and are arranged coaxially with the rotating shaft 12.
- the donut-shaped side surface of the fixed-side sliding member 10 located on the rotating-side sliding member 11 side corresponds to the fixed-side sliding surface 100.
- the rotating side sliding member 11 is fixed to the outer peripheral surface of the rotating shaft 12.
- the donut-shaped side surface located on the fixed side sliding member 10 side of the rotating side sliding member 11 corresponds to the rotating side sliding surface 110.
- the fixed side sliding member 10 and the rotating side sliding member 11 are supported by the urging member 15 so that the fixed side sliding surface 100 and the rotating side sliding surface 110 are pressed in the axial direction of the rotating shaft 12.
- the urging member 15 may be provided on the fixed side sliding member 10 side instead of or in addition to being provided on the rotating side sliding member 11. Further, as shown in FIG. 2, the urging member 15 may be omitted without being provided on the rotating side sliding member 11.
- a sliding surface diagnosis system 2 for diagnosing the states of the sliding surfaces 100 and 110 is provided for the sliding machine elements 1A and 1B having the above configuration.
- the sliding surface diagnostic system 2 includes a measuring device 3, a machine learning device 4, and a sliding surface diagnostic device 5.
- the measuring device 3 is installed in each part of the sliding machine elements 1A and 1B, and measures the physical quantity and the state quantity of each part of the sliding machine elements 1A and 1B.
- the measuring device 3 is connected to the machine learning device 4 and used for learning data measurement, and in the inference phase, it is connected to the sliding surface diagnostic device 5 and used for diagnostic data measurement.
- the measuring device 3 includes a motor current sensor 30, a contact electrical resistance measuring circuit 31, an acoustic emission (hereinafter referred to as “AE”) measuring circuit 32, and an acceleration sensor 33.
- AE acoustic emission
- the motor current sensor 30 is attached to, for example, a power line connecting the motor 13 and the motor control device 14, and measures the motor current value supplied from the motor control device 14 to the motor 13.
- the motor current sensor 30 may be provided inside the motor 13 or inside the motor control device 14.
- the contact electrical resistance measuring circuit 31 is electrically connected to the fixed side sliding member 10 and the rotating side sliding member 11, respectively, and applies a predetermined voltage to the fixed side sliding member 10 and the rotating side sliding member 11.
- the electrical resistance of contact with the member 11 is measured.
- the measured value of the contact electrical resistance for example, when there is not enough lubricant between the fixed side sliding member 10 and the rotating side sliding member 11, the resistance value (0 ⁇ ) indicating the contact state is measured. Will be done. Further, when a sufficient lubricant is present between the fixed side sliding member 10 and the rotating side sliding member 11, a resistance value (for example, on the order of 100 k ⁇ ) indicating an insulating state is measured.
- the AE measurement circuit 32 includes an AE sensor 320 fixed to the fixed side sliding member 10 and a signal processing circuit 321 electrically connected to the AE sensor 320.
- the AE sensor 320 detects an AE wave generated in the fixed side sliding member 10 and outputs it as an AE signal.
- the signal processing circuit 321 has, for example, a preamplifier and a discriminator, and measures an AE wave by amplifying the AE signal output by the AE sensor 320 through a predetermined filter.
- the AE sensor 320 may be fixed to the rotating side sliding member 11 in place of or in addition to the fixed side sliding member 10. Further, although the AE sensor 320 is attached to the outer peripheral cylindrical surface of the fixed side sliding member 10 in the examples of FIGS.
- the attachment position and attachment form of the AE sensor 320 may be appropriately changed.
- the AE sensor 320 may be attached to, for example, the axial end surface of the fixed-side sliding member 10 or may be attached to the fixed-side sliding member 10 via a member such as an adapter or a casing.
- the acceleration sensor 33 is fixed to the fixed-side sliding member 10 and measures the acceleration generated in the fixed-side sliding member 10.
- the acceleration sensor 33 is, for example, a sensor that measures the acceleration of one axis, and is fixed to the fixed side sliding member 10 so as to measure the acceleration in the direction orthogonal to the sliding surfaces 100 and 110 (arrows D1 and D2). Will be done.
- the direction indicated by the arrow D1 is the radial direction of the fixed-side sliding member 10 and the load direction of the rotating shaft 12.
- the direction indicated by the arrow D2 is the axial direction of the fixed side sliding member 10, and the rotating side sliding member 11 and the fixed side sliding member 10 are pressed against each other.
- the acceleration sensor 33 may be fixed to the rotating side sliding member 11 in place of or in addition to the fixed side sliding member 10. Further, although the acceleration sensor 33 is attached to the outer peripheral cylindrical surface of the fixed side sliding member 10 in the examples of FIGS. 1 and 2, the attachment position and attachment form of the acceleration sensor 33 may be appropriately changed.
- the acceleration sensor 33 may be attached to, for example, the axial end surface of the fixed-side sliding member 10 or may be attached to the fixed-side sliding member 10 via a member such as an adapter or a casing.
- the measuring device 3 measures, for example, the physical quantity and the state quantity of each part in a predetermined measurement cycle, and each time the measurement cycle elapses, the measured value at the measurement time is measured by the machine learning device 4 and the sliding surface diagnostic device 5. It is configured to be able to output to.
- the measurement cycles of the motor current sensor 30, the contact electrical resistance measuring circuit 31, the AE measuring circuit 32, and the acceleration sensor 33 constituting the measuring device 3 may be the same or different. Further, the measuring device 3 may output continuous measured values such as an analog signal instead of outputting discrete measured values each time the measurement cycle elapses as described above.
- the machine learning device 4 operates as the main body of the learning phase, and generates a learning model 6 used when diagnosing the states of the sliding surfaces 100 and 110 by machine learning.
- the machine learning device 4 can adopt both "supervised learning” and “unsupervised learning” as a machine learning method. In this embodiment, “supervised learning” is adopted, and in the second embodiment described later, “unsupervised learning” is adopted.
- the sliding surface diagnostic device 5 operates as the main body of the inference phase, and the sliding surfaces 100 and 110 are measured from the measured values measured by the measuring device 3 using the learned learning model 6 generated by the machine learning device 4. Diagnose the condition of.
- the learned learning model 6 is provided from the machine learning device 4 to the sliding surface diagnostic device 5 via an arbitrary communication network, recording medium, or the like.
- FIG. 3 is a hardware configuration diagram showing an example of a computer 200 constituting the machine learning device 4 and the sliding surface diagnostic device 5.
- Each of the machine learning device 4 and the sliding surface diagnostic device 5 is composed of a general-purpose or dedicated computer 200.
- the computer 200 has, 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, and an external device. It includes a device I / F unit 224, an I / O (input / output) device I / F unit 226, and a media input / output unit 228.
- the above components may be omitted as appropriate depending on the intended use of the computer 200.
- the processor 212 is composed of one or a plurality of arithmetic processing units (CPU, MPU, GPU, DSP, etc.) and operates as a control unit that controls the entire computer 200.
- the memory 214 stores various data and the program 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 composed of, for example, a keyboard, a mouse, a numeric keypad, an electronic pen, and the like.
- the display device 218 is composed of, for example, a liquid crystal display, an organic EL display, electronic paper, a projector, or the like.
- the input device 216 and the display device 218 may be integrally configured as in the touch panel display.
- the storage device 220 is composed of, for example, an HDD, an SSD, or the like, and stores various data necessary for executing 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 / receives data to / from another computer according to a predetermined communication standard.
- the external device I / F unit 224 is connected to an external device 250 such as a printer or a scanner by wire or wirelessly, and transmits / receives data to / from the external device 250 in accordance with a predetermined communication standard.
- the I / O device I / F unit 226 is connected to the I / O device 260 such as various sensors and actuators, and is connected to the I / O device 260, for example, a detection signal by a sensor or a control signal to the actuator. Sends and receives various signals and data.
- the media input / output unit 228 is composed of a drive device such as a DVD drive or a CD drive, and reads / writes data to / from the media 270 such as a DVD or a CD.
- the processor 212 calls and executes the program 230 in the work memory area of the memory 214, and controls each part of the computer 200 via the bus 210.
- the program 230 may be stored in the storage device 220 instead of the memory 214.
- the program 230 may be recorded on a non-temporary recording medium such as a CD or DVD in an installable file format or an executable file format, and may be provided to the computer 200 via the media input / output unit 228.
- the program 230 may be provided to the computer 200 by downloading via the network 240 via the communication I / F unit 222.
- the computer 200 is composed of, for example, a stationary computer or a portable computer, and is an electronic device of any form.
- the computer 200 may be a client type computer, a server type computer, or a cloud type computer.
- the computer 200 may be applied to devices other than the machine learning device 4 and the sliding surface diagnostic device 5.
- FIG. 4 is a block diagram showing an example of the machine learning device 4 according to the first embodiment.
- the first sliding machine element 1A is the target of the learning model 6, but the second sliding machine element 1B may be the target of the learning model 6.
- the machine learning device 4 includes a learning data acquisition unit 40, a learning data storage unit 41, a machine learning unit 42, and a learned model storage unit 43.
- the machine learning device 4 is composed of, for example, the computer 200 shown in FIG.
- the learning data acquisition unit 40 is composed of the communication I / F unit 222 or the I / O device I / F unit 226, and the machine learning unit 42 is composed of the processor 212, and is composed of the learning data storage unit 41 and the learning data storage unit 41.
- the trained model storage unit 43 is composed of a storage device 220.
- the learning data acquisition unit 40 is an interface unit that is connected to various external devices via a communication network 20 and acquires learning data including at least input data.
- the external device is used by, for example, a measuring device 3 provided on the first sliding machine element 1A, a measuring device 3 provided on a test device 7 simulating the first sliding machine element 1A, and an operator. It is a worker terminal 8 or the like.
- the sliding surface diagnostic device 5 is connected to the communication network 20
- the learning data acquisition unit 40 is connected to the first sliding machine element 1A to be diagnosed by the sliding surface diagnostic device 5. Learning data may be acquired from the provided measuring device 3.
- test device 7 is provided with a motor current sensor 30, a contact electric resistance measuring circuit 31, an AE measuring circuit 32, and an acceleration sensor 33 as the measuring device 3.
- the test device 7 is a device simulating the first sliding machine element 1A, it is configured as, for example, a device for performing a block-on-ring type one-way slip friction test. ..
- the learning data storage unit 41 is a database that stores one or a plurality of sets of learning data acquired by the learning data acquisition unit 40.
- the specific configuration of the database constituting the learning data storage unit 41 may be appropriately designed.
- the machine learning unit 42 carries out machine learning using the learning data stored in the learning data storage unit 41. That is, by inputting a plurality of sets of learning data into the learning model 6, the machine learning unit 42 inputs the correlation between the input data included in the learning data and the diagnostic information of the sliding surfaces 100 and 110 into the learning model 6. By training, a trained learning model 6 is generated.
- a neural network is adopted as a specific method of supervised learning by the machine learning unit 42 will be described.
- the trained model storage unit 43 is a database that stores the trained learning model 6 generated by the machine learning unit 42.
- the trained learning model 6 stored in the trained model storage unit 43 is provided to an actual system (for example, a sliding surface diagnostic device 5) via an arbitrary communication network, a recording medium, or the like.
- an actual system for example, a sliding surface diagnostic device 5
- the learning data storage unit 41 and the trained model storage unit 43 are shown as separate storage units in FIG. 4, they may be configured by a single storage unit.
- FIG. 5 is a data configuration diagram showing an example of data (supervised learning) used in the machine learning device 4 according to the first embodiment.
- the training data is, as input data, data of the motor current value supplied to the motor 13 in a predetermined period, data of contact electrical resistance between the fixed side sliding member 10 and the rotating side sliding member 11 in a predetermined period, and data. And, at least the data of the vibration generated in at least one of the fixed side sliding member 10 and the rotating side sliding member 11 in a predetermined period is included.
- the predetermined period for each of the motor current value data, contact electrical resistance data, and vibration data as input data is basically set to the same period, but there is a relationship that a time difference occurs between each data. If it is permitted, the period corresponding to the time difference may be set.
- the input data may further include other data in addition to the motor current value data, the contact electrical resistance data, and the vibration data.
- the fixed side sliding member 10 and the rotating side sliding member 11 may be included.
- the motor current value data in a predetermined period is composed of a plurality of measured values (motor current values) measured by the motor current sensor 30 at a plurality of measurement points within the predetermined period.
- the data of the motor current value is, for example, associated with the measurement time indicating the measurement time point, and is configured as the data of an array arranged in chronological order at the measurement time.
- the learning data acquisition unit 40 converts the measured value into data at a predetermined sampling cycle.
- the data of the motor current value in the predetermined period is composed of a plurality of measured values (motor current values) at a plurality of sampling points in the predetermined period.
- the contact electrical resistance data in a predetermined period is composed of a plurality of measured values (contact electrical resistance) measured by the contact electrical resistance measuring circuit 31 at a plurality of measurement time points within the predetermined period.
- the contact electrical resistance data is, for example, associated with the measurement time indicating the measurement time point, and is configured as data in an array arranged in chronological order at the measurement time.
- the learning data acquisition unit 40 converts the measured value into data at a predetermined sampling cycle.
- the contact electrical resistance data in the predetermined period is composed of a plurality of measured values (contact electrical resistance) at a plurality of sampling points in the predetermined period.
- the vibration data in the predetermined period is at least one of the AE wave data in the predetermined period and the acceleration data generated in the direction orthogonal to the sliding surfaces 100 and 110 in the predetermined period.
- the vibration data will be described as being composed of AE wave data and acceleration data.
- the vibration data may use data other than the AE wave data and the acceleration data as long as it represents the vibration generated in at least one of the fixed side sliding member 10 and the rotating side sliding member 11.
- the AE wave data in a predetermined period is composed of a plurality of measured values (AE waves) measured by the AE measurement circuit 32 at a plurality of measurement time points within the predetermined period.
- the AE wave data is, for example, associated with a measurement time indicating a measurement time point, and is configured as data in an array arranged in chronological order at the measurement time.
- the learning data acquisition unit 40 converts the measured values into data at a predetermined sampling cycle.
- the AE wave data in the predetermined period is composed of a plurality of measured values (AE waves) at a plurality of sampling points in the predetermined period.
- Acceleration data in a predetermined period is composed of a plurality of measured values (acceleration) measured by an acceleration sensor 33 at a plurality of measurement points within a predetermined period.
- the acceleration data is, for example, associated with the measurement time indicating the measurement time point, and is configured as data in an array arranged in chronological order at the measurement time.
- the learning data acquisition unit 40 converts the measured values into data at a predetermined sampling cycle.
- the acceleration data in the predetermined period is composed of a plurality of measured values (acceleration) at a plurality of sampling points in the predetermined period.
- the learning data is output data associated with the input data, and the states of the sliding surfaces 100 and 110 in a predetermined period are any of a plurality of states. Further includes diagnostic information indicating that there is.
- the output data is referred to as, for example, teacher data or correct label in supervised learning.
- the diagnostic information is at least one of diagnostic information regarding wear of the sliding surfaces 100 and 110, diagnostic information regarding seizure of the sliding surfaces 100 and 110, and diagnostic information regarding contamination of the lubricant.
- the diagnostic information is configured as information indicating that the state of the sliding surfaces 100 and 110 is either normal or abnormal.
- the diagnostic information is classified into two values, for example, the value indicating that the sliding surfaces 100 and 110 are in a normal state is set to "0", and the sliding surfaces 100 and 110 are in an abnormal state.
- the value indicating is defined as "1".
- the diagnostic information will be described as representing either normal or abnormal.
- the learning data according to the present embodiment includes normal and abnormal data including motor current value data, contact electrical resistance data, AE wave data, and acceleration data. It is configured in association with output data including diagnostic information indicating that it is one of the above.
- the abnormalities here are not only ex post facto abnormalities that were found to occur at the time of diagnosis, but also within the permissible range that is judged to be normal at the time of diagnosis, but it seems that future abnormalities were predicted. It may also include signs of abnormality.
- the information indicating the abnormality includes, for example, a plurality of abnormalities according to the specific content and degree of the abnormality so as to correspond to the abnormality 1 / abnormality 2 / ... / abnormality n shown in FIG. You may go out.
- the abnormality include, for example, an abnormality related to wear, an abnormality related to seizure, or an abnormality related to stain of the lubricant.
- the diagnostic information is classified into multiple values (integers of 3 or more), and for example, the values indicating that the sliding surfaces 100 and 110 are in a normal state are set to "0", and it is an abnormality related to wear.
- the indicated value is set to "1”
- the value indicating that the abnormality is related to seizure is set to "2”
- the values are similarly defined as values according to the content of each abnormality.
- the specific degree of abnormality includes, for example, in the case of an abnormality related to wear, the degree of wear is classified into multiple levels.
- the diagnostic information is classified into multiple values (integers of 3 or more), for example, the value indicating that the sliding surfaces 100 and 110 are in a normal state is set to "0", and the degree of wear is low.
- a value indicating that the abnormality is abnormal is set to "1”
- a value indicating that the degree of wear is an abnormality at a medium level is set to "2”
- the values are similarly defined as values according to the degree of each abnormality.
- the input data (motor current value data, contact electrical resistance data, AE wave data, and acceleration data in a predetermined period) included in the learning data and the diagnostic information of the sliding surfaces 100 and 110 The correlation between them will be described.
- the motor current value is supplied to the motor 13 from the motor control circuit 14 when the motor 13 is rotationally driven at a predetermined rotation speed
- the motor current value changes according to the states of the sliding surfaces 100 and 110.
- the data of the motor current value indirectly represents the frictional resistance of the sliding surfaces 100 and 110, and is mainly used for monitoring the frictional resistance of the sliding surfaces 100 and 110.
- the contact electrical resistance continuously changes depending on whether the fixed side sliding member 10 and the rotating side sliding member 11 are in an insulated state or a contact state. Therefore, the contact electrical resistance data is mainly used to monitor the lubrication state of the lubricant on the sliding surfaces 100 and 110.
- the sliding surfaces 100 and 110 are suddenly soiled. Vibration is generated and measured as a change in acceleration or AE wave. Further, when the sliding surfaces 100 and 110 are damaged due to seizure, it is measured as an AE wave. Therefore, the acceleration data and the AE wave data are mainly used for monitoring the contamination state of the lubricant on the sliding surfaces 100 and 110 and the seizure state of the sliding surfaces 100 and 110.
- the main monitoring applications for individual data regarding motor current value data, contact electrical resistance data, AE wave data, and acceleration data are as described above, but the states of the sliding surfaces 100 and 110 are sliding.
- the moving surfaces 100 and 110 change in a complicated manner due to various influences such as the usage environment, usage conditions, and maintenance status.
- the learning data is configured to include the motor current value data, the contact electrical resistance data, the AE wave data, and the acceleration data as input data for the sliding surfaces 100 and 110. Even when the state changes in a complicated manner due to various influences as described above, it is possible to diagnose the state of the sliding surfaces 100 and 110 with high accuracy.
- the learning data acquisition unit 40 acquires the above learning data, the data of the motor current value in a predetermined period measured by the first sliding machine element 1A or the measuring device 3 provided in the test device 7. , Contact electrical resistance data, AE wave data, and acceleration data are acquired from the measuring device 3 as input data. Further, the worker diagnoses the state of the sliding surfaces 100 and 110 during a predetermined period in which the input data is measured by the measuring device 3, and the diagnosed diagnosis result (for example, normal or abnormal) is used as the worker terminal 8. Then, the learning data acquisition unit 40 acquires the diagnostic information based on the diagnostic result input by the worker terminal 8 as output data (teacher data) from the worker terminal 8. Then, the learning data acquisition unit 40 constitutes one learning data by associating these input data with the output data, and stores them in the learning data storage unit 41.
- FIG. 6 is a schematic diagram showing an example of a neural network model used in the machine learning device 4 according to the first embodiment.
- the learning model 6 is configured as a neural network model shown in FIG.
- the neural network model consists of l neurons (x1 to xl) in the input layer, m neurons (y11 to y1m) in the first intermediate layer, and n neurons (y21 to y2n) in the second intermediate layer. , And o neurons (z1 to zo) in the output layer.
- Each neuron in the input layer is associated with each of the input data included in the learning data.
- Each neuron in the output layer is associated with each of the output data included in the learning data. It should be noted that the input data before being input to the input layer may be subjected to a predetermined pre-processing, or the output data after being output from the output layer may be subjected to a predetermined post-processing.
- the first intermediate layer and the second intermediate layer are also called hidden layers, and the neural network may have a plurality of hidden layers in addition to the first intermediate layer and the second intermediate layer. 1 Only the intermediate layer may be used as a hidden layer.
- synapses connecting the neurons of each layer are stretched between the input layer and the first intermediate layer, between the first intermediate layer and the second intermediate layer, and between the second intermediate layer and the output layer.
- a weight wi (i is a natural number) is associated with each synapse.
- the neural network model uses the training data to input the input data included in the training data to the input layer, and the output data output from the output layer as the inference result and the output included in the training data. By comparing with the data (teacher data), the correlation between the input data and the output data is learned.
- each neuron in the input layer is input with each of the input data included in the learning data.
- the value of each neuron in the output layer is the product of the value of the neuron on the input side connected to the neuron and the weight wi associated with the synapse connecting the neuron on the output side and the neuron on the input side. It is calculated by performing the process of calculating as the sum of several sequences of all neurons except the input layer.
- the values (z1 to zo) output to each neuron in the output layer as the inference result are compared with the values of the teacher data (t1 to to) corresponding to each of the output data included in the learning data.
- a process (backpro vacation) is performed in which an error is obtained and the weight wi associated with each synapse is adjusted so that the error becomes small.
- FIG. 7 is a flowchart showing an example of a machine learning method by the machine learning device 4 according to the first embodiment.
- the machine learning method corresponds to the learning phase of FIG.
- step S100 the learning data acquisition unit 40 prepares a desired number of learning data as a preliminary preparation for starting machine learning, and the prepared learning data is stored in the learning data storage unit 41.
- the number of learning data prepared here may be set in consideration of the inference accuracy required for the finally obtained learning model 6.
- the method of preparing learning data For example, when an abnormality occurs in the sliding surfaces 100 and 110 of the specific first sliding machine element 1A and the test device 7, or when the operator recognizes a sign of the abnormality, various types in the period before and after that.
- the input data constituting the learning data Prepare the output data (for example, the value of the output data in this case is "1"). Then, by repeating such work, it is possible to prepare a plurality of sets of learning data.
- the learning data not only when an abnormality has occurred, but also when no abnormality has occurred, that is, the first sliding machine element 1A and the sliding surfaces 100 and 110 of the test device 7 are in a normal state.
- a plurality of sets of learning data composed of the input data and the output data (for example, the value of the output data in this case is “0”) are prepared.
- the machine learning unit 42 prepares the learning model 6 before learning in order to start machine learning.
- the pre-learning learning model 6 prepared here is composed of the neural network model illustrated in FIG. 6, and the weight of each synapse is set to the initial value.
- Each neuron in the input layer is associated with motor current value data, contact electrical resistance data, AE wave data, and acceleration data as input data included in the training data.
- Each neuron in the output layer is associated with each of the diagnostic information as output data included in the learning data.
- step S120 the machine learning unit 42 randomly acquires, for example, one learning data from a plurality of sets of learning data stored in the learning data storage unit 41.
- step S130 the machine learning unit 42 inputs the input data included in one learning data to the input layer of the prepared pre-learning (or during learning) learning model 6.
- output data is output as an inference result from the output layer of the learning model 6, and the output data is generated by the learning model 6 before (or during learning) learning. Therefore, in the state before (or during learning) the learning, the output data output as the inference result shows information different from the output data (teacher data) included in the learning data.
- step S140 the machine learning unit 42 receives the output data (teacher data) included in one learning data acquired in step S120, and the output data output as the inference result from the output layer in step S130.
- Machine learning is performed by comparing the data and adjusting the weight of each synapse.
- the machine learning unit 42 causes the learning model 6 to learn the correlation between the input data and the output data (diagnosis information of the sliding surfaces 100 and 110).
- the diagnostic information constituting the teacher data is defined as a binary classification in which the normal state is indicated by "0" and the abnormal state is indicated by "1", one for learning selected in step S120.
- the value of the output data included in the data set is "1”, but the value of the output data output from the output layer is a predetermined value of 0 to 1, specifically, a value such as "0.63". It is assumed that it has been output.
- step S140 if the same input data is input to the input layer of the learning model 6 being trained, the learning model being trained so that the value output from the output layer approaches "1".
- the weight associated with each synapse of 6 is adjusted.
- step S150 the machine learning unit 42 determines whether or not it is necessary to continue machine learning, for example, an error between the output data and the teacher data, or not stored in the learning data storage unit 41. Judgment is made based on the remaining number of learning data for learning.
- step S150 determines in step S150 that machine learning is to be continued (No in step S150)
- the process returns to step S120, and the steps S120 to S140 are unlearned for the learning model 6 being learned. Perform multiple times using the data.
- step S150 determines in step S150 that the machine learning unit 42 ends machine learning (Yes in step S150)
- the process proceeds to step S160.
- step S160 the machine learning unit 42 stores the learned learning model 6 generated by adjusting the weights associated with each synapse in the learned model storage unit 43, and is shown in FIG. 7. Finish a series of machine learning methods.
- step S100 corresponds to a learning data storage process
- steps S110 to S150 correspond to a machine learning process
- step S160 corresponds to a learned model storage process.
- the motor current value data, the contact electrical resistance data, and the vibration (AE wave and acceleration in the present embodiment) data is provided. Therefore, it is possible to provide a learning model 6 capable of inferring (estimating) the diagnostic information of the sliding surfaces 100 and 110 with high accuracy.
- FIG. 8 is a block diagram showing an example of the sliding surface diagnostic device 5 according to the first embodiment.
- the first sliding machine element 1A is the target of the learning model 6, but the second sliding machine element 1B may be the target of the learning model 6.
- the sliding surface diagnostic device 5 includes an input data acquisition unit 50, an inference unit 51, a learned model storage unit 52, and an output processing unit 53.
- the sliding surface diagnostic device 5 is composed of, for example, the computer 200 shown in FIG.
- the input data acquisition unit 50 is composed of the communication I / F unit 222 or the I / O device I / F unit 226, and the inference unit 51 and the output processing unit 53 are composed of the processor 212, and the trained model storage is performed.
- the unit 52 is composed of a storage device 220.
- the sliding surface diagnostic device 5 may be incorporated in the motor control device 14, or is a management device higher than the motor control device 14 (for example, a controller for mechanical equipment, an equipment management system for managing a plurality of mechanical equipment). Etc.) may be incorporated.
- the input data acquisition unit 50 is connected to a measuring device 3 (motor current sensor 30, contact electric resistance measuring circuit 31, AE measuring circuit 32, and acceleration sensor 33) provided in the first sliding machine element 1A. It is an interface unit that acquires input data (motor current value data, contact electrical resistance data, AE wave data, and acceleration data in a predetermined period) based on the measured values measured by the measuring device 3.
- a measuring device 3 motor current sensor 30, contact electric resistance measuring circuit 31, AE measuring circuit 32, and acceleration sensor 33
- the inference unit 51 inputs the input data acquired by the input data acquisition unit 50 into the learning model 6 and performs an inference process for inferring the diagnostic information of the sliding surfaces 100 and 110.
- a machine learning device 4 and a learned learning model 6 in which supervised learning is performed by a machine learning method are used.
- the inference unit 51 not only has a function of performing inference processing using the learning model 6, but also adjusts the input data acquired by the input data acquisition unit 50 into a desired format or the like as a preprocessing of the inference processing, and the learning model 6 By applying a predetermined logical formula or calculation formula to the value of the output data output from the learning model 6 as the pre-processing function to be input to and the post-processing of the inference processing, the states of the sliding surfaces 100 and 110 are finally finalized. It also includes a post-processing function to make a judgment.
- the inference result of the inference unit 51 is preferably stored in the learned model storage unit 52 or another storage device (not shown), and the past inference result is, for example, further improved in the inference accuracy of the learning model 6. Therefore, it can be used as learning data used for online learning and re-learning.
- the trained model storage unit 52 is a database that stores the trained learning model 6 used in the inference processing of the inference unit 51.
- the number of learning models 6 stored in the trained model storage unit 52 is not limited to one. For example, a plurality of learning models 6 having different numbers of input data or different machine learning methods may be stored and selectively available.
- the output processing unit 53 performs output processing for outputting the inference result of the inference unit 51, that is, the diagnostic information of the sliding surfaces 100 and 110.
- various means can be adopted.
- the output processing unit 53 for example, notifies the operator of diagnostic information by display or sound, or as a diagnostic history of the first sliding machine element 1A, for example, is higher than the motor control device 14 or the motor control device 14. It may be transmitted to a management device (not shown), stored in a storage unit of the motor control device 14 or the management device of the motor control device 14, or used for drive control of the motor 13.
- FIG. 9 is a flowchart showing an example of a sliding surface diagnosis method by the sliding surface diagnosis device 5 according to the first embodiment.
- the diagnostic information of the sliding surfaces 100 and 110 will be described as being defined as a binary classification representing either normal “0” or abnormal “1”.
- the sliding surface diagnosis method corresponds to the inference phase of FIG.
- a current is supplied from the motor control device 14 to the motor 13 to drive the motor 13 to rotate, so that the rotating side sliding member 11 of the first sliding machine element 1A is in a rotated state.
- the diagnosis timing may be, for example, at predetermined time intervals or when a predetermined event occurs (at the time of an operator's operation instruction, maintenance operation, etc.).
- the input data acquisition unit 50 receives input data based on the measured values measured by the measuring device 3 (motor current sensor 30, contact electrical resistance measuring circuit 31, AE measuring circuit 32, and acceleration sensor 33). (Data of motor current value, data of contact electric resistance, data of AE wave, and data of acceleration in a predetermined period) are acquired. At this time, the input data is required to include the measured value measured by the measuring device 3 for the same period as the predetermined period or for a longer period than the predetermined period.
- step S210 the inference unit 51 preprocesses the input data, inputs it to the input layer of the learning model 6, and acquires the output data output from the output layer of the learning model 6.
- step S220 the inference unit 51 compares the value of the output data (the number between 0 〜 1) with a predetermined threshold value as an example of post-processing of supervised learning, and outputs, for example. If the value of the data is less than a predetermined threshold value, it is determined that the diagnostic information is "normal”, and if it is equal to or higher than the predetermined threshold value, it is determined that the diagnostic information is "abnormal". Is output as an inference result.
- a predetermined threshold value as an example of post-processing of supervised learning
- step S230 the output processing unit 53 determines whether the diagnostic information of the sliding surfaces 100 and 110, which is the inference result of the inference unit 51, represents "normal” or "abnormal", and is "normal”. If it is determined, the process proceeds to step S240, and if it is determined to be "abnormal", the process proceeds to step S250.
- step S240 the output processing unit 53 outputs information indicating "normal". Note that step S240 may be omitted.
- step S250 the output processing unit 53 outputs information indicating "abnormality".
- steps S200 correspond to an input data acquisition step
- steps S210 to S220 correspond to an inference step
- steps S230 to S250 correspond to an output processing step.
- the states of the sliding surfaces 100 and 110 are diagnosed with high accuracy without depending on the experience of the operator. be able to.
- the machine learning device 4 Similar to the first embodiment (see FIG. 4), the machine learning device 4 includes a learning data acquisition unit 40, a learning data storage unit 41, a machine learning unit 42, and a learned model storage unit 43. Be prepared.
- FIG. 10 is a data configuration diagram showing an example of data (unsupervised learning) used in the machine learning device 4 according to the second embodiment.
- the learning data includes only input data when the diagnostic information indicates that the states of the sliding surfaces 100 and 110 in a predetermined period are in a predetermined state. That is, the learning data may be configured not to include the output data, and the presence / absence of the output data and the format of the output data can be appropriately selected according to the machine learning method adopted by the machine learning device 4.
- the diagnostic information is configured as information indicating that the state of the sliding surfaces 100 and 110 is either normal or abnormal.
- the learning data includes data on the motor current value and data on the contact electrical resistance in a predetermined period when the states of the sliding surfaces 100 and 110 are normal. It consists only of AE wave data and input data including acceleration data.
- the learning data acquisition unit 40 acquires the above-mentioned learning data, the data of the motor current value in a predetermined period measured by the sliding machine elements 1A and 1B or the measuring device 3 provided in the test device 7.
- the contact electrical resistance data, the AE wave data, and the acceleration data are acquired from the measuring device 3 as input data.
- the worker diagnoses the state of the sliding surfaces 100 and 110 in the predetermined period in which the input data is measured by the measuring device 3, and inputs to the worker terminal 8 that the diagnosed diagnosis result is normal.
- the learning data acquisition unit 40 constitutes one learning data only with the input data, and stores it in the learning data storage unit 41.
- the machine learning unit 42 By inputting one or more sets of learning data into the learning model 6, the machine learning unit 42 inputs the input data included in the learning data and diagnostic information indicating that the states of the sliding surfaces 100 and 110 are normal. By letting the learning model 6 learn the correlation of the above, the learned learning model 6 is generated.
- an autoencoder is adopted as a specific method of unsupervised learning by the machine learning unit 42 will be described.
- FIG. 11 is a schematic diagram showing an example of an autoencoder model used in the machine learning device 4 according to the second embodiment.
- the learning model 6 is configured as an autoencoder model shown in FIG.
- the autoencoder model consists of l neurons (x1 to xl) in the input layer, m neurons (y1 to ym) in the middle layer, and o neurons (z1 to zo) in the output layer. Will be done.
- the number of neurons in the input layer and the output layer is the same, which is larger than the number of neurons in the middle layer.
- Each neuron in the input layer is associated with each of the input data included in the learning data.
- the intermediate layer is also called a hidden layer, and may have a plurality of hidden layers. Further, synapses connecting between the neurons of each layer are stretched between the input layer and the intermediate layer, and between the intermediate layer and the output layer, and each synapse has a weight wi (i is a natural number). Is associated with.
- the auto-encoder model uses the training data to input the input data included in the training data to the input layer, and the output data output from the output layer as the inference result and the input included in the training data. By comparing with the data and performing the process of adjusting the weight wi, the pattern and the tendency of the input data are learned.
- FIG. 12 is a flowchart showing an example of a machine learning method by the machine learning device 4 according to the second embodiment.
- step S300 the learning data acquisition unit 40 prepares a desired number of learning data as a preliminary preparation for starting machine learning, and the prepared learning data is stored in the learning data storage unit 41.
- learning is performed by using the measuring device 3 to acquire various measured values in a predetermined period when the sliding surfaces 100 and 110 in the specific sliding machine elements 1A and 1B and the test device 7 are in a normal state. Prepare the input data that composes the data. Then, by repeating such work, it is possible to prepare a plurality of sets of learning data.
- the machine learning unit 42 prepares the learning model 6 before learning in order to start machine learning.
- the pre-learning learning model 6 prepared here is composed of the autoencoder model illustrated in FIG. 11, and the weight of each synapse is set to the initial value.
- Each neuron in the input layer is associated with motor current value data, contact electrical resistance data, AE wave data, and acceleration data as input data included in the training data.
- step S320 the machine learning unit 42 randomly acquires, for example, one learning data from a plurality of sets of learning data stored in the learning data storage unit 41.
- step S330 the machine learning unit 42 inputs the input data included in one learning data to the input layer of the prepared pre-learning (or during learning) learning model 6.
- output data is output as an inference result from the output layer of the learning model 6, and the output data is generated by the learning model 6 before (or during learning) learning. Therefore, in the state before (or during learning) the learning, the output data output as the inference result shows information different from the input data included in the learning data.
- step S340 the machine learning unit 42 compares the input data included in one learning data acquired in step S320 with the output data output as the inference result from the output layer in step S330.
- Machine learning is performed by adjusting the weight of each synapse.
- step S350 the machine learning unit 42 determines whether or not it is necessary to continue machine learning. As a result, when it is determined to continue (No in step S350), the process returns to step S320, the steps S320 to S340 are performed on the learning model 6 being learned, and it is determined to end machine learning (step S350). Yes), proceed to step S360.
- step S300 corresponds to a learning data storage process
- steps S310 to S350 correspond to a machine learning process
- step S360 corresponds to a learned model storage process.
- the motor current value data, the contact electrical resistance data, and the vibration (AE wave and acceleration in the present embodiment) data is provided. Therefore, it is possible to provide a learning model 6 capable of inferring (estimating) the diagnostic information of the sliding surfaces 100 and 110 with high accuracy.
- the sliding surface diagnostic device 5 Similar to the first embodiment (see FIG. 8), the sliding surface diagnostic device 5 includes an input data acquisition unit 50, an inference unit 51, a learned model storage unit 52, and an output processing unit 53.
- the inference unit 51 inputs the input data acquired by the input data acquisition unit 50 into the learning model 6 and performs an inference process for inferring the diagnostic information of the sliding surfaces 100 and 110.
- the machine learning device 4 and the learned learning model 6 in which unsupervised learning is performed by the machine learning method are used.
- FIG. 13 is a flowchart showing an example of a sliding surface diagnosis method by the sliding surface diagnosis device 5 according to the second embodiment.
- the diagnostic information of the sliding surfaces 100 and 110 will be described as being defined as a binary classification indicating that it is either normal or abnormal.
- step S400 the input data acquisition unit 50 receives input data based on the measured values measured by the measuring device 3 (motor current sensor 30, contact electrical resistance measuring circuit 31, AE measuring circuit 32, and acceleration sensor 33).
- the measuring device 3 motor current sensor 30, contact electrical resistance measuring circuit 31, AE measuring circuit 32, and acceleration sensor 33.
- step S410 the inference unit 51 preprocesses the input data, inputs it to the input layer of the learning model 6, and acquires the output data output from the output layer of the learning model 6.
- step S420 the inference unit 51 obtains the difference between the feature amount based on the input data and the feature amount based on the output data as an example of post-processing of unsupervised learning.
- a feature quantity is, for example, a parameter expressed as a feature vector in a multidimensional space, and a difference is expressed as a distance between feature vectors. Then, the inference unit 51 determines that the diagnostic information is "normal” if the difference is less than the predetermined threshold value, and determines that the diagnostic information is "abnormal” if the difference is greater than or equal to the predetermined threshold value. Then, the judgment result is output as an inference result.
- step S430 the output processing unit 53 determines whether the diagnostic information of the sliding surfaces 100 and 110, which is the inference result of the inference unit 51, represents "normal” or "abnormal". As a result, if it is determined to be "normal”, the process proceeds to step S440, and if it is determined to be "abnormal", the process proceeds to step S450 (optional), and the series of sliding surface diagnosis methods shown in FIG. 13 is completed. do.
- steps S400 correspond to an input data acquisition step
- steps S410 to S420 correspond to an inference step
- steps S430 to S450 correspond to an output processing step.
- the states of the sliding surfaces 100 and 110 are diagnosed with high accuracy without depending on the experience of the operator. be able to.
- machine learning has been described.
- Part 42 may employ any other machine learning technique.
- Other machine learning methods include, for example, tree types such as decision trees and regression trees, ensemble learning such as bagging and boosting, neural network types such as recursive neural networks and convolutional neural networks (including deep learning). Examples include hierarchical clustering, non-hierarchical clustering, clustering type such as k-means method and k-means method, principal component analysis, factor analysis, multivariate analysis such as logistic regression, and support vector machine.
- the present invention can also be provided in the form of a program (machine learning program) 230 for causing the computer 200 shown in FIG. 3 to execute each process provided in the machine learning method according to the above embodiment. Further, the present invention can also be provided in the form of a program (sliding surface diagnosis program) 230 for causing the computer 200 shown in FIG. 3 to execute each step provided in the sliding surface diagnosis method according to the above embodiment. ..
- the present invention is not only based on the aspect of the sliding surface diagnostic device 5 (sliding surface diagnostic method or sliding surface diagnostic program) according to the above embodiment, but is also an inference device used for diagnosing the state of the sliding surface. It can also be provided in the form of (inference method or inference program).
- the inference device includes a memory and a processor, and the processor may execute a series of processes. The series of processes acquires input data including data of motor current value in a predetermined period, data of contact electric resistance in a predetermined period, and data of vibration (at least one of AE wave and acceleration) in a predetermined period. It includes an input data acquisition process (input data acquisition process) and an inference process (inference process) for inferring diagnostic information on a sliding surface when input data is acquired by the input data acquisition process.
- an inference device inference method or inference program
- it can be easily applied to various devices as compared with the case where the sliding surface diagnostic device 5 is mounted.
- the inference device inference method or inference program
- the sliding surface is used by using the machine learning device 4 according to the above embodiment and the learned learning model 6 generated by the machine learning method. It is understandable to those skilled in the art that the inference method carried out by the inference unit 51 of the diagnostic apparatus 5 may be applied.
- the present invention can be used for a machine learning device, a sliding surface diagnostic device, an inference device, a machine learning method, a machine learning program, a sliding surface diagnostic method, a sliding surface diagnostic program, an inference method, and an inference program.
Landscapes
- Engineering & Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Mechanical Engineering (AREA)
- Theoretical Computer Science (AREA)
- Software Systems (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Computing Systems (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
- Sliding-Contact Bearings (AREA)
Abstract
Description
固定側摺動部材と回転側摺動部材とが摺動する摺動面の状態を診断する摺動面診断装置に用いられる学習モデルを生成する機械学習装置であって、
所定期間において前記回転側摺動部材の駆動源であるモータに供給されるモータ電流値のデータ 、前記所定期間における前記固定側摺動部材と前記回転側摺動部材との間の接触電気抵抗のデータ、及び、前記所定期間において前記固定側摺動部材及び前記回転側摺動部材の少なくとも一方に生じる振動のデータを入力データとして少なくとも含む学習用データを1又は複数組記憶する学習用データ記憶部と、
前記学習モデルに前記学習用データを1又は複数組入力することで、前記入力データと前記摺動面の診断情報との相関関係を前記学習モデルに学習させる機械学習部と、
前記機械学習部により学習させた前記学習モデルを記憶する学習済みモデル記憶部と、を備える。
上記機械学習装置により生成された学習モデルを用いて、固定側摺動部材と回転側摺動部材とが摺動する摺動面の状態を診断する摺動面診断装置であって、
所定期間における前記回転側摺動部材の駆動源であるモータに供給されるモータ電流値のデータ、前記所定期間における前記固定側摺動部材と前記回転側摺動部材との間の接触電気抵抗のデータ、及び、前記所定期間における前記固定側摺動部材及び前記回転側摺動部材の少なくとも一方に生じる振動のデータを含む入力データを取得する入力データ取得部と、
前記入力データ取得部により取得された前記入力データを前記学習モデルに入力し、前記摺動面の診断情報を推論する推論部と、を備える。
図1は、第1の実施形態に係る摺動面診断システム2が適用された第1の摺動機械要素1Aの一例を示す概略構成図である。図2は、第1の実施形態に係る摺動面診断システム2が適用された第2の摺動機械要素1Bの一例を示す概略構成図である。
図4は、第1の実施形態に係る機械学習装置4の一例を示すブロック図である。図4に示す機械学習装置4では、第1の摺動機械要素1Aを学習モデル6の対象とするが、第2の摺動機械要素1Bを学習モデル6の対象としてもよい。
図7は、第1の実施形態に係る機械学習装置4による機械学習方法の一例を示すフローチャートである。なお、機械学習方法は、図5の学習フェーズに該当する。
図8は、第1の実施形態に係る摺動面診断装置5の一例を示すブロック図である。図8に示す摺動面診断装置5では、第1の摺動機械要素1Aを学習モデル6の対象とするが、第2の摺動機械要素1Bを学習モデル6の対象としてもよい。
図9は、第1の実施形態に係る摺動面診断装置5による摺動面診断方法の一例を示すフローチャートである。図9では、摺動面100、110の診断情報が、正常「0」及び異常「1」のいずれかを表す2値分類として定義されたものとして説明する。なお、摺動面診断方法は、図5の推論フェーズに該当する。
図9に示す一連の摺動面診断方法を終了する。摺動面診断方法において、ステップS200が入力データ取得工程、ステップS210~S220が推論工程、ステップS230~S250が出力処理工程に相当する。
第1の実施形態では、機械学習の手法として、「教師あり学習」を採用した場合について説明したが、本実施形態では、「教師なし学習」を採用した場合について説明する。なお、第2の実施形態に係る摺動機械要素1A、1B並びに摺動面診断システム2を構成する測定装置3、機械学習装置4及び摺動面診断装置5の基本的な構成や動作は、第1の実施形態と同様であるため、以下では第1の実施形態との相違点を中心に説明する。
機械学習装置4は、第1の実施形態(図4参照)と同様に、学習用データ取得部40と、学習用データ記憶部41と、機械学習部42と、学習済みモデル記憶部43とを備える。
図12は、第2の実施形態に係る機械学習装置4による機械学習方法の一例を示すフローチャートである。
摺動面診断装置5は、第1の実施形態(図8参照)と同様に、入力データ取得部50と、推論部51と、学習済みモデル記憶部52と、出力処理部53とを備える。
図13は、第2の実施形態に係る摺動面診断装置5による摺動面診断方法の一例を示すフローチャートである。図13では、摺動面100、110の診断情報が、正常及び異常のいずれかであることを表す2値分類として定義されたものとして説明する。
本発明は上述した実施形態に制約されるものではなく、本発明の主旨を逸脱しない範囲内で種々変更して実施することが可能である。そして、それらはすべて、本発明の技術思想に含まれるものである。
本発明は、上記実施形態に係る摺動面診断装置5(摺動面診断方法又は摺動面診断プログラム)の態様によるもののみならず、摺動面の状態を診断するために用いられる推論装置(推論方法又は推論プログラム)の態様で提供することもできる。その場合、推論装置(推論方法又は推論プログラム)としては、メモリと、プロセッサとを含み、このうちのプロセッサが、一連の処理を実行するものとすることができる。当該一連の処理とは、所定期間におけるモータ電流値のデータ、所定期間における接触電気抵抗のデータ、及び、所定期間における振動(AE波及び加速度の少なくとも1つ)のデータを含む入力データを取得する入力データ取得処理(入力データ取得工程)と、入力データ取得処理にて入力データを取得すると、摺動面の診断情報を推論する推論処理(推論工程)とを含む。
3…測定装置、4…機械学習装置、5…摺動面診断装置、6…学習モデル、
10…固定側摺動部材、11…回転側摺動部材、12…回転軸、13…モータ、
14…モータ制御回路、15…付勢部材、
30…モータ電流センサ、31…接触電気抵抗測定回路、
32…AE測定回路、33…加速度センサ、
40…学習用データ取得部、41…学習用データ記憶部、
42…機械学習部、43…モデル記憶部、
50…入力データ取得部、51…推論部、52…モデル記憶部、53…出力処理部、
100…固定側摺動面、110…回転側摺動面、200…コンピュータ、
Claims (13)
- 固定側摺動部材と回転側摺動部材とが摺動する摺動面の状態を診断する摺動面診断装置に用いられる学習モデルを生成する機械学習装置であって、
所定期間において前記回転側摺動部材の駆動源であるモータに供給されるモータ電流値のデータ、前記所定期間における前記固定側摺動部材と前記回転側摺動部材との間の接触電気抵抗のデータ、及び、前記所定期間において前記固定側摺動部材及び前記回転側摺動部材の少なくとも一方に生じる振動のデータを入力データとして少なくとも含む学習用データを1又は複数組記憶する学習用データ記憶部と、
前記学習モデルに前記学習用データを1又は複数組入力することで、前記入力データと前記摺動面の診断情報との相関関係を前記学習モデルに学習させる機械学習部と、
前記機械学習部により学習させた前記学習モデルを記憶する学習済みモデル記憶部と、を備える、
機械学習装置。 - 前記学習用データは、
前記入力データに対応付けられ、前記所定期間における前記摺動面の状態が複数の状態のうちのいずれかであることを表す前記診断情報を出力データとしてさらに含み、
前記機械学習部は、
前記入力データと前記出力データとの相関関係を教師あり学習により前記学習モデルに学習させる、
請求項1に記載の機械学習装置。 - 前記学習用データは、
前記診断情報が前記所定期間における前記摺動面の状態が所定の状態であることを表すときの前記入力データのみを含み、
前記機械学習部は、
前記入力データと前記摺動面の状態が前記所定の状態であることを表す前記診断情報との相関関係を教師なし学習により前記学習モデルに学習させる、
請求項1に記載の機械学習装置。 - 前記入力データに含まれる前記振動のデータは、
アコースティックエミッション波のデータ、及び、前記摺動面と直交する方向に生じる加速度のデータ、の少なくとも一つである、
請求項1乃至請求項3のいずれか一項に記載の機械学習装置。 - 前記診断情報は、
前記摺動面の摩耗に関する診断情報、前記摺動面の焼付きに関する診断情報、及び、前記摺動面を潤滑する潤滑剤の汚損に関する診断情報である、の少なくとも一つである、
請求項1乃至請求項4のいずれか一項に記載の機械学習装置。 - 請求項1乃至請求項5のいずれか一項に記載の機械学習装置により生成された学習モデルを用いて、固定側摺動部材と回転側摺動部材とが摺動する摺動面の状態を診断する摺動面診断装置であって、
所定期間における前記回転側摺動部材の駆動源であるモータに供給されるモータ電流値のデータ、前記所定期間における前記固定側摺動部材と前記回転側摺動部材との間の接触電気抵抗のデータ、及び、前記所定期間における前記固定側摺動部材及び前記回転側摺動部材の少なくとも一方に生じる振動のデータを含む入力データを取得する入力データ取得部と、
前記入力データ取得部により取得された前記入力データを前記学習モデルに入力し、前記摺動面の診断情報を推論する推論部と、を備える、
摺動面診断装置。 - 固定側摺動部材と回転側摺動部材とが摺動する摺動面の状態を診断するために用いられる推論装置であって、
前記推論装置は、メモリと、プロセッサとを備え、
前記プロセッサは、
所定期間における前記回転側摺動部材の駆動源であるモータに供給されるモータ電流値のデータ、前記所定期間における前記固定側摺動部材と前記回転側摺動部材との間の接触電気抵抗のデータ、及び、前記所定期間における前記固定側摺動部材及び前記回転側摺動部材の少なくとも一方に生じる振動のデータを含む入力データを取得する入力データ取得処理と、
前記入力データ取得処理にて前記入力データを取得すると、前記摺動面の診断情報を推論する推論処理と、を実行する、
推論装置。 - 固定側摺動部材と回転側摺動部材とが摺動する摺動面の状態を診断する摺動面診断装置に用いられる学習モデルを学習する機械学習方法であって、
所定期間における前記回転側摺動部材の駆動源であるモータに供給されるモータ電流値のデータ、前記所定期間における前記固定側摺動部材と前記回転側摺動部材との間の接触電気抵抗のデータ、及び、前記所定期間における前記固定側摺動部材及び前記回転側摺動部材の少なくとも一方に生じる振動のデータを入力データとして少なくとも含む学習用データを、学習用データ記憶部に1又は複数組記憶する学習用データ記憶工程と、
前記学習モデルに前記学習用データを1又は複数組入力することで、前記入力データと前記摺動面の診断情報との相関関係を前記学習モデルに学習させる機械学習工程と、
前記機械学習工程により学習させた前記学習モデルを学習済みモデル記憶部に記憶する学習済みモデル記憶工程と、を備える、
機械学習方法。 - コンピュータに、請求項8に記載の機械学習方法が備える各工程を実行させるための、
機械学習プログラム。 - 請求項1乃至請求項5のいずれか一項に記載の機械学習装置により生成された学習モデルを用いて、固定側摺動部材と回転側摺動部材とが摺動する摺動面の状態を診断する摺動面診断方法であって、
所定期間における前記回転側摺動部材の駆動源であるモータに供給されるモータ電流値のデータ、前記所定期間における前記固定側摺動部材と前記回転側摺動部材との間の接触電気抵抗のデータ、及び、前記所定期間における前記固定側摺動部材及び前記回転側摺動部材の少なくとも一方に生じる振動のデータを含む入力データを取得する入力データ取得工程と、
前記入力データ取得工程により取得された前記入力データを前記学習モデルに入力し、前記摺動面の診断情報を推論する推論工程と、を備える、
摺動面診断方法。 - コンピュータに、請求項10に記載の摺動面診断方法が備える各工程を実行させるための、
摺動面診断プログラム。 - 固定側摺動部材と回転側摺動部材とが摺動する摺動面の状態を推論する推論方法であって、
所定期間における前記回転側摺動部材の駆動源であるモータに供給されるモータ電流値のデータ、前記所定期間における前記固定側摺動部材と前記回転側摺動部材との間の接触電気抵抗のデータ、及び、前記所定期間における前記固定側摺動部材及び前記回転側摺動部材の少なくとも一方に生じる振動のデータを含む入力データを取得する入力データ取得工程と、
前記入力データ取得工程にて前記入力データを取得すると、前記摺動面の診断情報を推論する推論工程と、を備える、
推論方法。 - コンピュータに、請求項12に記載の推論方法が備える各工程を実行させるための、
推論プログラム。
Priority Applications (4)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202180066372.2A CN116324193A (zh) | 2020-09-30 | 2021-09-24 | 机器学习装置、滑动面诊断装置、推论装置、机器学习方法、机器学习程序、滑动面诊断方法、滑动面诊断程序、推论方法及推论程序 |
KR1020237014234A KR20230075496A (ko) | 2020-09-30 | 2021-09-24 | 기계 학습 장치, 미끄럼 이동면 진단 장치, 추론 장치, 기계 학습 방법, 기계 학습 프로그램, 미끄럼 이동면 진단 방법, 미끄럼 이동면 진단 프로그램, 추론 방법 및 추론 프로그램 |
US18/246,648 US20240011821A1 (en) | 2020-09-30 | 2021-09-24 | Machine learning apparatus, sliding-surface diagnosis apparatus, inference apparatus, machine learning method, machine learning program, sliding-surface diagnosis method, sliding-surface diagnosis program, inference method, and inference program |
EP21875416.6A EP4224025A1 (en) | 2020-09-30 | 2021-09-24 | Machine learning device, slide surface diagnosis device, inference device, machine learning method, machine learning program, slide surface diagnosis method, slide surface diagnosis program, inference method, and inference program |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2020164735A JP7044333B1 (ja) | 2020-09-30 | 2020-09-30 | 機械学習装置、摺動面診断装置、推論装置、機械学習方法、機械学習プログラム、摺動面診断方法、摺動面診断プログラム、推論方法、及び、推論プログラム |
JP2020-164735 | 2020-09-30 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2022071114A1 true WO2022071114A1 (ja) | 2022-04-07 |
Family
ID=80950321
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2021/035082 WO2022071114A1 (ja) | 2020-09-30 | 2021-09-24 | 機械学習装置、摺動面診断装置、推論装置、機械学習方法、機械学習プログラム、摺動面診断方法、摺動面診断プログラム、推論方法、及び、推論プログラム |
Country Status (7)
Country | Link |
---|---|
US (1) | US20240011821A1 (ja) |
EP (1) | EP4224025A1 (ja) |
JP (1) | JP7044333B1 (ja) |
KR (1) | KR20230075496A (ja) |
CN (1) | CN116324193A (ja) |
TW (1) | TW202232005A (ja) |
WO (1) | WO2022071114A1 (ja) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE102022113888A1 (de) | 2022-06-01 | 2023-12-07 | Rheinisch-Westfälische Technische Hochschule (RWTH) Aachen, Körperschaft des öffentlichen Rechts | Verfahren und System zur Überwachung eines Gleitlagers |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPS5741740Y2 (ja) * | 1977-11-04 | 1982-09-13 | ||
JPS5913692B2 (ja) * | 1977-02-14 | 1984-03-31 | 株式会社安川電機 | 軸受潤滑状態の余裕度判定方法 |
JPH10318261A (ja) | 1997-05-20 | 1998-12-02 | Hitachi Constr Mach Co Ltd | 軸受用給脂時期判別制御装置 |
JP2007292731A (ja) * | 2006-03-28 | 2007-11-08 | Victor Co Of Japan Ltd | 軸受検査方法及びモータの製造方法 |
JP2010180982A (ja) | 2009-02-06 | 2010-08-19 | Nippon Soken Inc | 回転軸支持構造、および、機関の制御装置 |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP5741740B1 (ja) | 2014-03-14 | 2015-07-01 | オムロン株式会社 | 封止接点装置およびその製造方法 |
JP5913692B1 (ja) | 2015-06-24 | 2016-04-27 | 海寧金永和家紡紹造有限公司 | ソファーカバー |
-
2020
- 2020-09-30 JP JP2020164735A patent/JP7044333B1/ja active Active
-
2021
- 2021-09-24 WO PCT/JP2021/035082 patent/WO2022071114A1/ja active Application Filing
- 2021-09-24 CN CN202180066372.2A patent/CN116324193A/zh active Pending
- 2021-09-24 US US18/246,648 patent/US20240011821A1/en active Pending
- 2021-09-24 KR KR1020237014234A patent/KR20230075496A/ko unknown
- 2021-09-24 EP EP21875416.6A patent/EP4224025A1/en active Pending
- 2021-09-29 TW TW110136221A patent/TW202232005A/zh unknown
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPS5913692B2 (ja) * | 1977-02-14 | 1984-03-31 | 株式会社安川電機 | 軸受潤滑状態の余裕度判定方法 |
JPS5741740Y2 (ja) * | 1977-11-04 | 1982-09-13 | ||
JPH10318261A (ja) | 1997-05-20 | 1998-12-02 | Hitachi Constr Mach Co Ltd | 軸受用給脂時期判別制御装置 |
JP2007292731A (ja) * | 2006-03-28 | 2007-11-08 | Victor Co Of Japan Ltd | 軸受検査方法及びモータの製造方法 |
JP2010180982A (ja) | 2009-02-06 | 2010-08-19 | Nippon Soken Inc | 回転軸支持構造、および、機関の制御装置 |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE102022113888A1 (de) | 2022-06-01 | 2023-12-07 | Rheinisch-Westfälische Technische Hochschule (RWTH) Aachen, Körperschaft des öffentlichen Rechts | Verfahren und System zur Überwachung eines Gleitlagers |
Also Published As
Publication number | Publication date |
---|---|
JP2022056795A (ja) | 2022-04-11 |
TW202232005A (zh) | 2022-08-16 |
KR20230075496A (ko) | 2023-05-31 |
EP4224025A1 (en) | 2023-08-09 |
JP7044333B1 (ja) | 2022-03-30 |
US20240011821A1 (en) | 2024-01-11 |
CN116324193A (zh) | 2023-06-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Eren et al. | A generic intelligent bearing fault diagnosis system using compact adaptive 1D CNN classifier | |
Zhang et al. | Dynamic condition monitoring for 3D printers by using error fusion of multiple sparse auto-encoders | |
CN110554657B (zh) | 一种数控机床运行状态健康诊断系统及诊断方法 | |
EP0573357B1 (fr) | Procédé de diagnostic d'un processus évolutif | |
CN113128561A (zh) | 一种机床轴承故障诊断方法 | |
TW202133968A (zh) | 異常檢知裝置、異常檢知方法以及記錄電腦可讀取程式之記錄媒體 | |
WO2022071114A1 (ja) | 機械学習装置、摺動面診断装置、推論装置、機械学習方法、機械学習プログラム、摺動面診断方法、摺動面診断プログラム、推論方法、及び、推論プログラム | |
Malawade et al. | Neuroscience-inspired algorithms for the predictive maintenance of manufacturing systems | |
Wu et al. | Layer-wise relevance propagation for interpreting LSTM-RNN decisions in predictive maintenance | |
JP2022546052A (ja) | センサアグノスティックなメカニカル機械故障識別 | |
WO2021090765A1 (ja) | 異常検知装置、異常検知方法及びコンピュータプログラム | |
CN113360304A (zh) | 用于对机器的异常状态进行原因分析的方法和计算单元 | |
van den Hoogen et al. | An improved wide-kernel cnn for classifying multivariate signals in fault diagnosis | |
CN116561514A (zh) | 车辆轮毂轴承单元的故障诊断方法、系统、装置及介质 | |
Zahra et al. | Predictive maintenance with neural network approach for UAV propulsion systems monitoring | |
Jia et al. | Multitask convolutional neural network for rolling element bearing fault identification | |
WO2020071066A1 (ja) | 異常判定装置、信号特徴量予測器、異常判定方法、学習モデルの生成方法及び学習モデル | |
Mahesh et al. | Data-driven intelligent condition adaptation of feature extraction for bearing fault detection using deep responsible active learning | |
Han et al. | A study on fault classification system based on deep learning algorithm considering speed and load condition | |
Demircan | Vibration based condition monitoring of pumping systems in textile dyehouses | |
RU2795745C1 (ru) | Независимая от датчика идентификация неисправности механического станка | |
Arslan et al. | Convolutional auto-encoder based degradation point forecasting for bearing data set | |
Jin | Modeling of Machine Life Using Accelerated Prognostics and Health Management (APHM) and Enhanced Deep Learning Methodology | |
Mahamad | Diagnosis, classification and prognosis of rotating machine using artificial intelligence | |
Gituku | Application of Domain Invariant Features in Condition Monitoring of Rolling Element Bearings Under Changing Operating Conditions |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 21875416 Country of ref document: EP Kind code of ref document: A1 |
|
WWE | Wipo information: entry into national phase |
Ref document number: 18246648 Country of ref document: US |
|
ENP | Entry into the national phase |
Ref document number: 20237014234 Country of ref document: KR Kind code of ref document: A |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
ENP | Entry into the national phase |
Ref document number: 2021875416 Country of ref document: EP Effective date: 20230502 |