CN111537205A - Method and device for diagnosing machine - Google Patents
Method and device for diagnosing machine Download PDFInfo
- Publication number
- CN111537205A CN111537205A CN201911340744.7A CN201911340744A CN111537205A CN 111537205 A CN111537205 A CN 111537205A CN 201911340744 A CN201911340744 A CN 201911340744A CN 111537205 A CN111537205 A CN 111537205A
- Authority
- CN
- China
- Prior art keywords
- measurement data
- diagnosing
- abnormality
- evaluation value
- pump
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 29
- 238000005259 measurement Methods 0.000 claims abstract description 105
- 230000005856 abnormality Effects 0.000 claims abstract description 73
- 238000003745 diagnosis Methods 0.000 claims abstract description 42
- 238000005070 sampling Methods 0.000 claims abstract description 10
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 80
- 238000011156 evaluation Methods 0.000 claims description 52
- 230000001186 cumulative effect Effects 0.000 claims description 47
- 238000007689 inspection Methods 0.000 claims description 38
- 238000012545 processing Methods 0.000 claims description 35
- 230000002159 abnormal effect Effects 0.000 claims description 28
- 238000010606 normalization Methods 0.000 claims description 24
- 238000010801 machine learning Methods 0.000 claims description 16
- 230000008569 process Effects 0.000 claims description 6
- 230000009467 reduction Effects 0.000 claims description 2
- 238000004891 communication Methods 0.000 description 20
- 239000010865 sewage Substances 0.000 description 19
- 238000012806 monitoring device Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 6
- 238000012544 monitoring process Methods 0.000 description 6
- 238000012706 support-vector machine Methods 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 238000002405 diagnostic procedure Methods 0.000 description 3
- 238000005086 pumping Methods 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 238000009434 installation Methods 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- 230000001932 seasonal effect Effects 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000007599 discharging Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000007667 floating Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000002955 isolation Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
- 238000011144 upstream manufacturing Methods 0.000 description 1
- 238000004804 winding Methods 0.000 description 1
Images
Classifications
-
- 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
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F04—POSITIVE - DISPLACEMENT MACHINES FOR LIQUIDS; PUMPS FOR LIQUIDS OR ELASTIC FLUIDS
- F04B—POSITIVE-DISPLACEMENT MACHINES FOR LIQUIDS; PUMPS
- F04B49/00—Control, e.g. of pump delivery, or pump pressure of, or safety measures for, machines, pumps, or pumping installations, not otherwise provided for, or of interest apart from, groups F04B1/00 - F04B47/00
- F04B49/06—Control using electricity
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F04—POSITIVE - DISPLACEMENT MACHINES FOR LIQUIDS; PUMPS FOR LIQUIDS OR ELASTIC FLUIDS
- F04B—POSITIVE-DISPLACEMENT MACHINES FOR LIQUIDS; PUMPS
- F04B51/00—Testing machines, pumps, or pumping installations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Landscapes
- Engineering & Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Mechanical Engineering (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- Medical Informatics (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Testing And Monitoring For Control Systems (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
- Control Of Non-Positive-Displacement Pumps (AREA)
- Control Of Positive-Displacement Pumps (AREA)
Abstract
The invention provides a method and a device for diagnosing mechanical equipment, which can accurately diagnose abnormality based on various reasons even with a small physical quantity. A method for diagnosing a machine, comprising: a sampling step of sampling n types (n ≧ 2 integer) of measurement data groups representing characteristics of the mechanical equipment in time series; a standardization step of standardizing the sampled measurement data set; and a diagnosis step of drawing the normalized measurement data sets in an n-dimensional coordinate system, and performing a primary diagnosis of normality or abnormality of the machine based on which side of a boundary threshold value set in advance in the n-dimensional coordinate system the drawn measurement data set exists.
Description
Technical Field
The present invention relates to a method and an apparatus for diagnosing a machine, and more particularly to a method and an apparatus for diagnosing a machine suitable for diagnosing a rotating machine such as an underwater pump.
Background
A manhole pump system, which is an example of a sewage pump system, includes: a water storage unit that stores sewage flowing in from the inflow pipe; a plurality of pumps for discharging the sewage stored in the water storage part to the outflow pipe; a water level gauge for measuring a water level of the sewage stored in the water storage portion; and a control panel having a control device for performing a sewage transport control in which any one of the pumps is started to discharge sewage to the outflow pipe when the water level measured by the water level gauge reaches a pump start water level, and the pump is stopped when the water level reaches a pump stop water level.
In such a manhole pump system, two pumps are generally provided, and a control device is configured to alternately operate these pump devices each time sewage is delivered.
Documents of the prior art
Patent document
Patent document 1: japanese patent laid-open No. 2001-34338
Patent document 2: japanese laid-open patent publication No. 2010-236191
The control device provided in the conventional inspection well pump apparatus described above determines whether or not each pump is abnormal based on whether or not a physical quantity such as a drive current value of the pump, an operation time of the pump from a start water level to a stop water level, and a temperature of the pump exceeds a predetermined threshold. Further, in order to measure a physical quantity suitable for the type of the assumed abnormality, measurement processing using various sensors is required, which is very complicated.
In addition, since the threshold value for determining an abnormality is also controlled by the environment of the inspection well in which the pump is installed, it is difficult to set the threshold values uniformly. For example, in a region where the amount of water taken in per unit time is large and a region where the amount of water taken in per unit time is small, a difference in the frequency of starting the pump and the operation time is caused, and therefore, there is a problem that it is difficult to perform accurate determination if the determination is performed with a constant threshold value.
In particular, in a remote monitoring system including a communication device and a server for managing pump operation data transmitted from each communication device in a control panel of each inspection well pump facility, the remote monitoring system being capable of monitoring an operation state by accessing the server from a terminal owned by a manager, there is a problem in that: in order to independently determine various abnormalities of the pumps provided in the individual inspection well pump devices, it is necessary to transmit a very large amount of physical quantity, and therefore, the number of sensors and the capacity of transmission data increase, and it is difficult to set an appropriate threshold value.
The problem is not limited to the inspection well pump device, but is a problem common to various mechanical devices.
Disclosure of Invention
In view of the above-described problems, an object of the present invention is to provide a method and an apparatus for diagnosing a machine, which can accurately diagnose an abnormality due to various causes even with a small physical quantity.
In order to achieve the above object, a first characteristic configuration of a diagnostic method for a machine tool according to the present invention is characterized by comprising: a sampling step of sampling n kinds of measurement data groups representing characteristics of mechanical equipment in time series, wherein n is an integer greater than or equal to 2; a standardization step of standardizing the sampled measurement data set; and a diagnosis step of drawing the normalized measurement data sets in an n-dimensional coordinate system, and performing a primary diagnosis of normality or abnormality of the machine based on which side of a boundary threshold value set in advance in the n-dimensional coordinate system the drawn measurement data set exists.
By normalizing the n types of measurement data sets sampled in time series in the sampling step in the normalization step, it is possible to extract, as feature quantities, measurement data sets from which influences due to, for example, variations inherent in the machine and influences of the installation environment are excluded, draw points representing these feature quantities, that is, the n types of measurement data sets, as a plurality of points in time series in the n-dimensional coordinate system in the diagnosis step, and appropriately diagnose whether the machine is normal or abnormal once using a boundary threshold set in advance as an index.
The second characteristic feature is that, in addition to the first characteristic feature, statistical data required for the normalization process used in the normalization step is calculated based on a measurement data set of a latest predetermined period when the normalization process is executed.
By performing the normalization processing of the measurement data set based on the statistical data obtained based on the measurement data set of the latest predetermined period, it is possible to eliminate the influence due to the passage of time such as seasonal variation, for example, and to perform a highly reliable diagnosis.
The third characteristic feature is the first or second characteristic feature, wherein, in the diagnosing step, each time each measurement data group is diagnosed as abnormal once, a new cumulative evaluation value is obtained by multiplying a predetermined abnormal reference value by a weighting coefficient based on a distance between each measurement data group and the boundary threshold value and adding the result to a current cumulative evaluation value, and each time each measurement data group is diagnosed as normal in the diagnosing step, a new cumulative evaluation value is obtained by subtracting a predetermined normal return evaluation value from the current cumulative evaluation value, and normal or abnormal of the machine is finally diagnosed based on the finally obtained cumulative evaluation value, an initial value of which is 0.
If a mild abnormality such as an abnormality is diagnosed in one diagnosis and then normal is diagnosed, there is also an abnormality that is continuously diagnosed as an abnormality and finally causes a serious failure. In this case, the degree of abnormality such as an abnormality requiring maintenance in the near future can be diagnosed by adding a value obtained by multiplying a predetermined abnormality reference value by a weighting coefficient based on the distance between the measurement data group and the boundary threshold value every time an abnormality is diagnosed in one diagnosis, subtracting a predetermined normal return evaluation value every time a normal state is diagnosed, calculating a cumulative evaluation value obtained by this calculation, and finally diagnosing the normality or abnormality of the machine based on this cumulative evaluation value.
The fourth characteristic feature is characterized in that, in addition to any one of the first to third characteristic features, a diagnostic map is provided in which the outside of the boundary threshold set in the n-dimensional coordinate system is divided into a plurality of regions, each of the regions is associated with one of the causes of an abnormality, and in the diagnosing step, the cause of an abnormality is diagnosed based on the region in which each measurement data group is drawn.
The feature point of the normalized n types of measurement data sets is that the cause of an abnormality can be estimated from which region of the diagnostic map is divided outside the boundary threshold value indicating the normal region.
The fifth characteristic feature is characterized in that, in addition to any one of the first to fourth characteristic features, in the diagnosing step, the boundary threshold value is automatically generated by a machine learning device based on the measurement data group input as learning data.
For example, when a measurement data set in a state where the machine is operating properly is input to the machine learning device, a boundary threshold value for identifying normality and abnormality is automatically generated.
The sixth characteristic feature is, in addition to any one of the first to fifth characteristic features, that each of the measurement data constituting the measurement data group is measurement data that converges to a predetermined range in a normal state of the machine.
As long as the measurement data group is measurement data that converges to a predetermined range during normal operation of the machine, a boundary threshold value with high reliability can be set.
The seventh characteristic feature is, in addition to any one of the first to sixth characteristic features, that the mechanical device is a pump device that does not involve adjustment of a rotation speed.
The present invention can be suitably used for abnormality diagnosis of a pump device in which the start and stop of the pump device are switched depending on the presence or absence of power supply.
The eighth characteristic feature is the inspection well pump device according to the seventh characteristic feature, wherein the pump device is a well pump device that repeats starting and stopping of the pump according to a water level.
The method can be suitably used for abnormality diagnosis of inspection well pump equipment which is repeatedly started and stopped according to the water level.
The ninth characteristic feature is the eighth characteristic feature in that the measurement data set includes a pump current value and a reduction rate of the water level or a pump operation time when the pump is operated once.
By using the pump current value and the pump operation time when the pump is operated once as the measurement data group, it is possible to appropriately diagnose the abnormality of the inspection well pump equipment using a small number of measurement data groups.
A first characteristic configuration of a diagnostic device for a machine tool according to the present invention is characterized by comprising: a normalization processing unit that normalizes n types of measurement data sets representing characteristics of a machine device, which are sampled in time series, where n is an integer of 2 or more; and a diagnosis processing unit that draws the normalized measurement data sets in an n-dimensional coordinate system, and performs a primary diagnosis of normality or abnormality of the machine based on which side of a boundary threshold value set in advance in the n-dimensional coordinate system the drawn measurement data set exists.
The second characteristic configuration is characterized in that, in addition to the first characteristic configuration, the diagnostic processing unit obtains a new cumulative evaluation value by multiplying a predetermined abnormality reference value by a weighting coefficient based on a distance between each measurement data group and the boundary threshold value and adding the new cumulative evaluation value to the current cumulative evaluation value each time each measurement data group is diagnosed as abnormal once, obtains a new cumulative evaluation value by subtracting a predetermined abnormality reference value from the current cumulative evaluation value each time each measurement data group is diagnosed as normal in the diagnosing step, and finally diagnoses whether the machine is normal or abnormal based on the finally obtained cumulative evaluation value, an initial value of the cumulative evaluation value being 0.
The third characteristic feature is characterized in that, in addition to the first or second characteristic feature, the diagnostic apparatus includes a diagnostic map in which an area outside the boundary threshold set in the n-dimensional coordinate system is divided into a plurality of areas, each area being associated with one of the causes of an abnormality, and the diagnostic processing unit diagnoses the cause of an abnormality based on the area in which each measurement data set is drawn.
The fourth characteristic feature is characterized in that, in addition to any one of the first to third characteristic features, the diagnosis processing unit includes a machine learning device that automatically generates the boundary threshold value based on the measurement data group input as learning data.
As described above, according to the present invention, there are provided a method and a device for diagnosing a machine, which can accurately diagnose an abnormality based on various causes even with a small physical quantity.
Drawings
Fig. 1 is an explanatory view of a well pump device.
Fig. 2 is an explanatory diagram of an abnormality diagnosis device for a well pump device.
Fig. 3 is an explanatory diagram showing the steps of the inspection well pump device abnormality diagnosis method.
Fig. 4(a) is an explanatory view of the measurement data set, and fig. 4(b) is an enlarged explanatory view of a main part of the measurement data set.
Fig. 5 is an explanatory diagram of the normalization processing.
Fig. 6 is an explanatory diagram of one diagnosis.
Fig. 7 is an explanatory diagram of the cumulative evaluation value and the final diagnosis.
Fig. 8 is an explanatory diagram of the diagnostic chart.
Description of the symbols
10 inspection well pump equipment
PA and PB pump
18. 19 water level meter
21 control part
22 storage section
24 communication unit
30 mobile terminal
40 monitoring device
41 communication unit
42 data processing part
44 diagnostic part
46 standardization processing part
48 diagnosis processing part
Detailed Description
Hereinafter, a method for diagnosing a machine and a device for diagnosing a machine according to the present invention will be described by taking a well pump device as an example.
Fig. 1 shows an inspection well pump arrangement 10. The inspection well pump device 10 includes: a catch basin 12 as a water reservoir for storing sewage flowing from the sewage inflow pipe 11 on the upstream side; two pumps PA, PB for pumping the sewage stored in the inspection well 12 to the sewage outflow pipe 13 on the downstream side; and water level gauges 18, 19 that measure the water level of the sewage stored in the inspection well 12.
A first suction pipe 15b, a first elbow pipe 15c, and a first horizontal pipe 15d are flange-connected to the discharge elbow pipe 15a of the first pump PA, and the first horizontal pipe 15d is flange-connected to the sewage outflow pipe 13 via the header pipe 13 a. A check valve 15e is provided between the first pumping pipe 15b and the first elbow pipe 15 c.
A second suction pipe 17b and a second elbow pipe 17c are flange-connected to the discharge elbow pipe 17a of the second pump PB, and the second elbow pipe 17c is flange-connected to the sewage outflow pipe 13 via the header pipe 13 a. A check valve 17e is provided between the second pumping pipe 17b and the second elbow pipe 17 c.
A level gauge 18 of a plunge pressure type or a bubble type is provided at the bottom of the inspection well 12. The water level of the sewage stored in the inspection well 12 is continuously detected by the water level gauge 18. The floating type water gauge 19 is provided as a backup water gauge for detecting the abnormally high water level HHWL.
A control panel device 200 for housing a control panel 20 including a control unit 21 for performing sewage conveyance control for controlling the pumps PA and PB to pressure-feed the sewage accumulated in the inspection well 12 to the sewage outflow pipe 13 is provided in the vicinity of the inspection well 12.
The control panel 20 is provided with a control unit 21, a storage unit 22, and a communication unit 24. The storage unit 22 stores control information from the control unit 21, water level information from the water level gauges 18 and 19, and the like. The communication unit 24 includes: a transmitting unit that transmits various information stored in the storage unit 22 to the remote monitoring apparatus 40, and a receiving unit that receives a control command from the monitoring apparatus 40.
It is preferable to use a wireless communication medium such as a mobile phone network as a communication medium for connecting the communication unit 24 and the monitoring device 40, and the monitoring device 40 and the communication unit 24 are connected to each other via the communication medium, and further, the portable communication terminal 30 provided by the administrator of the inspection well pump device 10 and the monitoring device 40 can be connected to each other via the wireless communication medium.
The control panel 20 is connected to the pumps PA and PB by ac power supply lines L1 and L2, and the control panel 20 is connected to the level gauges 18 and 19 by a signal line S.
When it is detected that the water level measured by the water level gauge 18 reaches the predetermined pump start water level HWL, the control unit 21 controls the power supply from the power supply line L1 so as to start one pump PA of the pumps PA and PB, and when it is detected that the water level reaches the pump stop water level LWL lower than the pump start water level HWL, the control unit 21 stops the power supply and stops the one pump PA.
When it is detected again that the water level reaches the pump start water level HWL after that, the control unit 21 controls the supply of power from the power supply line L2 to start the other pump PB, and when it is detected that the water level reaches the pump stop water level LWL, the supply of power is stopped to stop the pump PB. That is, the control unit 21 alternately controls the operation of the pumps PA and PB.
When the pump start water level HWL cannot be detected due to a failure of the water level gauge 18 or the like, or when a large amount of rainwater exceeding the drainage capacity of 1 pump flows into the inspection well 12 due to a concentrated heavy rain, and the water level gauge 19 measures that the water level has reached the abnormally high water level HHWL, the control unit 21 operates both the pumps PA and PB at the same time.
The control unit 21 samples the water level information detected by the water level meters 18 and 19 in time series at intervals of, for example, 1 minute, and stores the sampled information in the storage unit 22, and also stores time series operation information such as the start timing and stop timing of the pumps PA and PB and the operation time from start to stop in the storage unit 22.
As shown in fig. 2, the communication unit 24 included in the control panel 20 of each inspection well pump device 10 is configured to transmit the water level information and the operation information stored in the storage unit 22 to the monitoring device 40 at predetermined intervals.
The monitoring device 40 functions as a diagnostic device for the inspection well pump device 10, and includes: a communication unit 41 that communicates with the communication unit 24 of each inspection well pump device 10 and the portable communication terminal 30 of the administrator; a database DB storing water level information and operation information transmitted from the communication unit 24 of each inspection well pump device 10; a data processing unit 42 that exchanges data with the database DB; and a diagnosis unit 44 for diagnosing whether each well pump device 10 is operating normally based on the water level information and the operation information stored in the database DB.
The diagnosis unit 44 includes a normalization processing unit 46 and a diagnosis processing unit 48. The normalization processing unit 46 is configured to normalize n types (n ≧ 2 integer) of measurement data sets representing characteristics of the inspection well pump device 10, which are sampled in time series.
The diagnosis processing unit 48 is configured to include a machine learning device that automatically generates a boundary threshold value for diagnosing whether the inspection well pump device is normal or abnormal based on a measurement data set input as learning data, and the machine learning device is configured to draw the normalized measurement data sets in the n-dimensional coordinate system, respectively, and to perform a primary diagnosis of the normal or abnormal condition of the inspection well pump device 10 based on which side of the boundary threshold value preset in the n-dimensional coordinate system the drawn measurement data set exists. The Machine learning device is composed of a computer executing a Class-One Support Vector Machine (One-Class Support Vector Machine) algorithm.
The diagnosis processing unit 48 is configured to multiply a predetermined abnormality reference value by a weighting coefficient based on a distance between each measurement data group and a boundary threshold value and add the result to the current cumulative evaluation value to obtain a new cumulative evaluation value each time a diagnosis is performed, and to finally diagnose whether the inspection well pump device 10 is normal or abnormal based on the cumulative evaluation value obtained by subtracting the predetermined abnormality reference value each time each measurement data group is diagnosed as being normal in the diagnosis.
The diagnostic processing unit 48 is configured to diagnose the cause of the abnormality based on the region in which each measurement data group is drawn.
Fig. 3 shows a flow of a series of diagnostic processes executed by the monitoring device 40. When the measurement data for 1 day is received from each well pump device 10 and finally stored in the database DB (SA1), a feature quantity, which is a measurement data set used for abnormality determination, is extracted for each pump of each well pump device 10 (SA2), and a mean value and a variance value, which are statistics required for normalization processing, are calculated from the measurement data set stored in the database DB in the latest predetermined period, and the feature quantity for 1 day, which is a determination target, is normalized (SA 3).
The normalized feature amounts are sequentially input to the machine learning device, and a determination is made once based on a set boundary threshold value (SA5), and if it is determined to be abnormal, the cumulative evaluation value is added in consideration of the degree of abnormality (SA6), and if it is determined to be normal, the cumulative evaluation value is subtracted (SA 7).
A final abnormality determination is made as to whether or not the cumulative evaluation value calculated in this way exceeds a preset cumulative abnormality threshold (SA8), and if the cumulative abnormality threshold is exceeded, the cause of the abnormality is specified based on a diagnostic map showing the correlation with the cause of the abnormality in advance (SA9), and an alarm, which is the content of the abnormal state, is transmitted together with the cause of the abnormality to a mobile terminal or the like owned by the administrator (SA 10). The alarm notification is transmitted as an email via the mailer provided in the communication unit 41.
The diagnostic unit 44 will be described in detail below.
Fig. 4(a) shows operation data of the inspection well pump device for 24 hours from 0 am to 0 pm on the next day. The fluctuation of the water level, the operation timing and operation time of the pumps PA and PB, the current value of the pump PA, and the current value of the pump PB are shown in this order from the top.
Fig. 4(b) is an enlarged display for facilitating understanding of the relationship between the fluctuation of the water level and the operation timing and operation time of the pumps PA and PB shown in fig. 4 (a). And if the water storage level of the inspection well reaches HWL, starting the pump PA, and if the water storage level of the inspection well is reduced to LWL, stopping the pump PA. Next, the pump PB is started when the stored water level reaches HWL, and stopped when the water level falls to LWL. During the period when the water level is reduced from HWL to LWL, either pump is activated.
When the amount of the sewage to be fed to the inspection well is large, the operation time of the pump is long, and the speed of lowering the water level during the operation of the pump is low. Hereinafter, for simplification, the speed of lowering the water level during the pump operation will be simply referred to as "the slope of the water level".
The water level information and the operation information stored in the storage unit 22 of each well pump device 10 are transmitted to the monitoring device 40 via the communication unit 24, and are stored in the database DB via the data processing unit 42.
From such data, the data processing unit 42 extracts and transmits characteristic quantities, which are n kinds (n ≧ 2, but n ≧ 2 in the present embodiment) of measurement data sets representing characteristics of the inspection well pump device 10 sampled in time series, as "slopes of water levels" ((HWL-LWL)/operating times) and "current values" of the pumps PA and PB at the time of activation, respectively, to the normalization processing unit 46 in units of 1 day. The slope of the water level of the inspection well and the current value of the pump as the measurement data set are measurement data which have a correlation with each other and converge in a predetermined range when the pump is normal.
As shown in fig. 5, the normalization processing unit 46 uses the average value μ and the variance σ for normalization of the data as a total of the feature values stored in the database DB during the last three months2Calculation is performed to normalize the "slope of water level" and the "current value" by the expression (x- μ)/σ with respect to the characteristic amount x.
Although not limited to the last past three months, the statistical data (average value, variance value) required for the normalization processing is preferably calculated based on the measurement data set of the last predetermined period when the normalization processing is performed, and the influence due to the passage of time such as seasonal variation is eliminated, thereby enabling highly reliable diagnosis.
When the feature values "slope of water level" and "current value" normalized by the normalization processing unit 46 are input to the diagnosis processing unit 48, the normalized feature values are plotted in a two-dimensional coordinate system indicating the slopes of current and water level, respectively, and a normal or abnormal condition is diagnosed once based on which side of a boundary threshold value set in advance in the two-dimensional coordinate system the plotted measurement data set exists.
By normalizing the n types of measurement data sets sampled in time series in the sampling step in the normalization step, it is possible to extract, as feature quantities, measurement data sets from which influences due to, for example, variations inherent to mechanical equipment, influences of installation environments, and the like are excluded, draw points representing these feature quantities, that is, two types of measurement data sets, as a plurality of points in time series in a two-dimensional coordinate system in the diagnosis step, and appropriately diagnose whether the inspection well pump device is normal or abnormal at one time using a boundary threshold value set in advance as an index.
Fig. 6 shows a case where a circle (indicated by a thick line) having a predetermined radius with the origin (the average value of the current and the average value of the slope of the water level) of a two-dimensional coordinate system in which the vertical axis represents the current and the horizontal axis represents the slope of the water level as the center is used as a boundary threshold, and it is determined that the drawn feature amount is normal when the drawn feature amount is located inside the boundary threshold, and it is determined that the drawn feature amount is abnormal when the drawn feature amount is located outside the boundary threshold.
The diagnosis processing unit 48 includes a machine learning device that automatically generates the boundary threshold value based on the measurement data set input as the learning data. As the Machine learning apparatus, in addition to the above-described One-Class Support Vector Machine (One-Class Support Vector Machine), a computer that executes an algorithm such as an IF (Isolation Forest) method or an RC (Robust Covariance) method can be used.
By performing machine learning, a normal data space in which normal measurement data (training data) is mapped, that is, an internal space of a boundary threshold value is generated in a mapping space (feature space) shown in fig. 6. In the example of fig. 6, the result of learning 1-year feature data of a manhole pump device at 100 locations where two pumps are installed is shown as learning data.
The diagnosis processing unit 48 multiplies and adds a predetermined abnormality reference value Vnb by a weighting coefficient W based on a distance between a position of each measurement data group and a boundary threshold value based on the following mathematical expression every time the measurement data group is diagnosed to be abnormal at one time, calculates a cumulative evaluation value V by subtracting a predetermined normal recovery evaluation value Vpb every time each measurement data group is diagnosed to be normal at one time, and finally diagnoses whether the inspection well pump device is normal or abnormal based on the cumulative evaluation value.
V＝Vnb×W－Vpb
In the present embodiment, Vnb is set to 1, and Vnb < Vpb < Wmax × Vnb.
If a mild abnormality such as an abnormality is diagnosed in one diagnosis and then normal is diagnosed, there is also an abnormality that is continuously diagnosed as an abnormality and finally causes a serious failure. In this case, the degree of abnormality such as an abnormality requiring maintenance in the near future can be diagnosed by adding a value obtained by multiplying a predetermined abnormality reference value by a weighting coefficient based on the distance between the measurement data group and the boundary threshold value every time an abnormality is diagnosed in one diagnosis, subtracting a predetermined normal return evaluation value every time a normal state is diagnosed, calculating a cumulative evaluation value obtained by this calculation, and finally diagnosing the normality or abnormality of the machine based on this cumulative evaluation value.
Fig. 7 shows transition of the cumulative evaluation value. The accumulated evaluation value is evaluated once for each operation number, and a value of Vnb × W is added to the initial value 0 when an abnormality is determined, and a value of Vpb is subtracted when a normal determination is performed. When the cumulative evaluation value exceeds a predetermined threshold value, in this example, exceeds "10", the final abnormality determination is performed, and the content is transmitted to a mobile terminal or the like owned by the administrator. After the abnormality determination, the primary determination and the final determination are continued.
Fig. 8 shows an example of a diagnostic map. The outside of the boundary threshold shown in the feature space shown in fig. 6 is divided into eight regions, each of which is associated with one of the causes of abnormality. The diagnosis processing unit 48 diagnoses the cause of the abnormality based on the region in which each feature data is drawn. For example, in the example of fig. 8, when the characteristic data is plotted in the area 3, it is diagnosed that the foreign matter is involved in the abnormal state, when the characteristic data is plotted in the area 5, it is diagnosed that the water pressure leakage is abnormal, and when the characteristic data is plotted in the area 8, it is diagnosed that the air pool is abnormal.
In the above-described example, the example in which the two combinations of the pump drive current value and the slope of the water level of the inspection well are used as the measurement data set has been described, but the measurement data set is not limited to these data, and may be set as appropriate. For example, instead of the slope of the water level, the operation time of the pump at each activation associated with the slope of the water level may be used as the measurement data.
The present invention has been described above by taking a manhole pump device as an example of a machine, but in a machine other than the manhole pump device, the present invention can be used for abnormality determination of another machine as long as each measurement data constituting the measurement data group is a measurement data having a correlation with each other and converging within a predetermined range when the machine is normal.
As described above, the method for diagnosing a machine according to the present invention includes: a sampling step of sampling n types (n ≧ 2 integer) of measurement data groups representing characteristics of the mechanical equipment in time series; a standardization step of standardizing the sampled measurement data set; and a diagnosis step of drawing the normalized measurement data sets in an n-dimensional coordinate system, and performing a primary diagnosis of normality or abnormality of the machine based on which side of a boundary threshold value set in advance in the n-dimensional coordinate system the drawn measurement data set exists.
The measurement data constituting the measurement data set are those which have a correlation with each other and which converge within a predetermined range in a normal state of the machine, and the statistical data required for the normalization process used in the normalization step is preferably calculated based on the measurement data set of a predetermined period of time immediately after the normalization process is performed.
The diagnosing step is configured to multiply a predetermined abnormality reference value by a weighting coefficient based on a distance between each measurement data group and a boundary threshold value and add the result to a current cumulative evaluation value to obtain a new cumulative evaluation value each time each measurement data group is diagnosed as abnormal at one time, to subtract a predetermined normal return evaluation value from the current cumulative evaluation value to obtain a new cumulative evaluation value each time each measurement data group is diagnosed as normal in the diagnosing step, and to finally diagnose whether the machine is normal or abnormal based on the finally obtained cumulative evaluation value, an initial value of the cumulative evaluation value being 0.
The diagnostic method further includes a diagnostic map in which an area outside the boundary threshold set in the n-dimensional coordinate system is divided into a plurality of areas, each area being associated with one of the causes of the abnormality, and the diagnosing step is configured to diagnose the cause of the abnormality based on the area in which each measurement data group is drawn.
In the diagnosing step, it is preferable that the boundary threshold value is automatically generated by the machine learning device based on the measurement data set inputted as the learning data.
In the above-described embodiment, the whole of the diagnostic step is executed by the machine learning device, but only the generation of the boundary threshold value may be executed by the machine learning device, or the diagnostic step may be executed using a predetermined boundary value without using the machine learning device.
The above-described embodiments are merely examples of the present invention, and the technical scope of the present invention is not limited to the description, and it is needless to say that the specific configurations of the respective parts represented by a pump, a water level gauge, and the like, the threshold value set for abnormality determination, and the like can be appropriately changed in design within the scope of achieving the operational effect of the present invention.
Claims (13)
1. A method for diagnosing a machine, comprising:
a sampling step of sampling n kinds of measurement data groups representing characteristics of mechanical equipment in time series, wherein n is an integer greater than or equal to 2;
a standardization step of standardizing the sampled measurement data set; and
and a diagnosis step of drawing the normalized measurement data sets in an n-dimensional coordinate system, and performing a primary diagnosis of normality or abnormality of the machine based on which side of a boundary threshold value set in advance in the n-dimensional coordinate system the drawn measurement data set exists.
2. The method of diagnosing a mechanical device according to claim 1,
the statistical data required for the normalization process used in the normalization step is calculated based on the measurement data set of the most recent predetermined period when the normalization process is executed.
3. The method of diagnosing a mechanical device according to claim 1 or 2,
in the diagnosing step, each time each measurement data group is diagnosed as abnormal once, a new cumulative evaluation value is obtained by multiplying a predetermined abnormal reference value by a weighting coefficient based on a distance between each measurement data group and the boundary threshold value and adding the result to the current cumulative evaluation value, and each time each measurement data group is diagnosed as normal in the diagnosing step, a new cumulative evaluation value is obtained by subtracting a predetermined normal return evaluation value from the current cumulative evaluation value, and a final diagnosis is performed on the normality or abnormality of the machine based on the finally obtained cumulative evaluation value, an initial value of which is 0.
4. The method of diagnosing a mechanical apparatus according to any one of claims 1 to 3,
a diagnostic map is provided in which the outside of the boundary threshold set in the n-dimensional coordinate system is divided into a plurality of regions, each region being associated with one of causes of an abnormality,
in the diagnosing step, the cause of the abnormality is diagnosed based on the region in which each measurement data set is depicted.
5. The method of diagnosing a mechanical device according to any one of claims 1 to 4,
in the diagnosing step, the boundary threshold value is automatically generated by a machine learning device based on the measurement data group input as learning data.
6. The method of diagnosing a mechanical apparatus according to any one of claims 1 to 5,
each of the measurement data constituting the measurement data group is measurement data that converges to a predetermined range when the machine is normal.
7. The method of diagnosing a mechanical apparatus according to any one of claims 1 to 6,
the mechanical device is a pump device without accompanying speed adjustment.
8. The method of diagnosing a mechanical device according to claim 7,
the pump equipment is inspection well pump equipment which repeatedly starts and stops the pump according to the water level.
9. The method of diagnosing a mechanical device according to claim 8,
the measurement data set includes a pump current value and a reduction speed of the water level or a pump operation time when the pump is operated once.
10. A diagnostic device for a machine tool, comprising:
a normalization processing unit that normalizes n types of measurement data sets representing characteristics of a machine device, which are sampled in time series, where n is an integer of 2 or more; and
and a diagnosis processing unit that draws the normalized measurement data sets in an n-dimensional coordinate system, and performs a primary diagnosis of normality or abnormality of the machine based on which side of a boundary threshold value set in advance in the n-dimensional coordinate system the drawn measurement data set exists.
11. The diagnostic device for mechanical equipment according to claim 10,
the diagnostic processing unit obtains a new cumulative evaluation value by multiplying a predetermined abnormal reference value by a weighting coefficient based on a distance between each measurement data group and the boundary threshold value and adding the new cumulative evaluation value to the current cumulative evaluation value each time each measurement data group is diagnosed as abnormal at one time, obtains a new cumulative evaluation value by subtracting the predetermined abnormal reference value from the current cumulative evaluation value each time each measurement data group is diagnosed as normal in the diagnosing step, and finally diagnoses whether the machine is normal or abnormal based on the finally obtained cumulative evaluation value, the initial value of the cumulative evaluation value being 0.
12. The diagnostic apparatus for mechanical equipment according to claim 10 or 11,
a diagnostic map is provided in which the outside of the boundary threshold set in the n-dimensional coordinate system is divided into a plurality of regions, each region being associated with one of causes of an abnormality,
the diagnosis processing unit diagnoses a cause of abnormality based on a region in which each measurement data set is described.
13. The diagnostic device for mechanical equipment according to any one of claims 10 to 12,
the diagnostic processing unit includes a machine learning device that automatically generates the boundary threshold value based on the measurement data set input as learning data.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP2018-245240 | 2018-12-27 | ||
| JP2018245240A JP7175752B2 (en) | 2018-12-27 | 2018-12-27 | Diagnostic method for mechanical equipment and diagnostic device for mechanical equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111537205A true CN111537205A (en) | 2020-08-14 |
| CN111537205B CN111537205B (en) | 2024-07-09 |
Family
ID=71449137
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911340744.7A Active CN111537205B (en) | 2018-12-27 | 2019-12-23 | Method for diagnosing machine and diagnostic device for machine |
Country Status (2)
| Country | Link |
|---|---|
| JP (1) | JP7175752B2 (en) |
| CN (1) | CN111537205B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113869373A (en) * | 2021-09-07 | 2021-12-31 | 迪希埃(北京)半导体技术有限公司 | Equipment abnormality detection method and device, computer equipment and storage medium |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021029387A1 (en) | 2019-08-09 | 2021-02-18 | 新明和工業株式会社 | Underwater pump system, information processing device, and computer program |
| JP7431391B2 (en) * | 2020-07-29 | 2024-02-15 | ＮｉＸ ＪＡＰＡＮ株式会社 | Manhole pump deterioration prediction method |
| KR102627669B1 (en) * | 2021-07-05 | 2024-01-24 | (주)비에이에너지 | A method of building a machine learning model for the prevention and diagnosis of an energy storage device accident, and a computer program capable of preventing and diagnosing an energy storage device accident based on the machine learning model |
| CN115163513B (en) * | 2022-06-02 | 2024-02-23 | 北京和利时系统集成有限公司 | Fault diagnosis method and device for subway submersible sewage pump |
| WO2024095929A1 (en) * | 2022-11-01 | 2024-05-10 | Ｊｆｅスチール株式会社 | Anomaly diagnosis device and anomaly diagnosis method |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6216066B1 (en) * | 1998-07-01 | 2001-04-10 | General Electric Company | System and method for generating alerts through multi-variate data assessment |
| JP2001334999A (en) * | 2000-02-01 | 2001-12-04 | General Electric Co <Ge> | Warning generation system on trend performance analysis |
| JP2015034513A (en) * | 2013-08-09 | 2015-02-19 | 株式会社クボタ | Abnormality detection device, polluted water transfer pump device and monitoring device |
| JP2015226229A (en) * | 2014-05-28 | 2015-12-14 | 富士通株式会社 | Base station device |
| JP2018024055A (en) * | 2016-08-10 | 2018-02-15 | 三菱重工工作機械株式会社 | Abnormality detection device and method for tool of machine tool |
| JP2018055150A (en) * | 2016-09-26 | 2018-04-05 | 株式会社クボタ | Notification device, monitoring system and notification method |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2005345154A (en) | 2004-05-31 | 2005-12-15 | Kyushu Electric Power Co Inc | Method and device for detecting omen of abnormality |
| JP5337853B2 (en) * | 2011-09-12 | 2013-11-06 | 株式会社日立製作所 | Power plant diagnostic device and power plant diagnostic method |
| US8757469B2 (en) | 2011-10-26 | 2014-06-24 | GM Global Technology Operations LLC | Binary classification of items of interest in a repeatable process |
-
2018
- 2018-12-27 JP JP2018245240A patent/JP7175752B2/en active Active
-
2019
- 2019-12-23 CN CN201911340744.7A patent/CN111537205B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6216066B1 (en) * | 1998-07-01 | 2001-04-10 | General Electric Company | System and method for generating alerts through multi-variate data assessment |
| JP2001334999A (en) * | 2000-02-01 | 2001-12-04 | General Electric Co <Ge> | Warning generation system on trend performance analysis |
| JP2015034513A (en) * | 2013-08-09 | 2015-02-19 | 株式会社クボタ | Abnormality detection device, polluted water transfer pump device and monitoring device |
| JP2015226229A (en) * | 2014-05-28 | 2015-12-14 | 富士通株式会社 | Base station device |
| JP2018024055A (en) * | 2016-08-10 | 2018-02-15 | 三菱重工工作機械株式会社 | Abnormality detection device and method for tool of machine tool |
| JP2018055150A (en) * | 2016-09-26 | 2018-04-05 | 株式会社クボタ | Notification device, monitoring system and notification method |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113869373A (en) * | 2021-09-07 | 2021-12-31 | 迪希埃(北京)半导体技术有限公司 | Equipment abnormality detection method and device, computer equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| JP2020107072A (en) | 2020-07-09 |
| JP7175752B2 (en) | 2022-11-21 |
| CN111537205B (en) | 2024-07-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111537205B (en) | Method for diagnosing machine and diagnostic device for machine | |
| US11341836B2 (en) | Persistent monitoring and real time low latency local control of centrifugal hydraulic pump, remote monitoring and control, and collecting data to produce performance profiles | |
| JP7498321B2 (en) | Manhole pump diagnostic method and manhole pump diagnostic device | |
| US11047392B2 (en) | System and method for determining a use condition for an appliance | |
| US6721683B2 (en) | Pump motor diagnosis | |
| JP4542819B2 (en) | Hydraulic machine, system and method for monitoring the health status of a hydraulic machine | |
| CN112119220B (en) | Monitoring module and method for identifying operating situations in a wastewater pumping station | |
| JP6234732B2 (en) | Abnormality detection device, sewage transfer pump device and monitoring device | |
| US20140379300A1 (en) | Pump efficiency determining system and related method for determining pump efficiency | |
| US9127678B2 (en) | Fast-response pump monitoring and in-situ pump data recording system | |
| EP3567173B1 (en) | Alarm management module for a wastewater pumping station | |
| GB2314412A (en) | Method of monitoring pump performance | |
| KR20160026492A (en) | Anomaly detection method using confidence interval estimation based on time series analysis | |
| KR100567434B1 (en) | On-Line Monitoring System for Pump Stations | |
| KR101321349B1 (en) | Pump performance estimation method and system using a water level sensor | |
| JP2024019788A (en) | Manhole pump diagnostic method and manhole pump diagnostic device | |
| JP2022075128A (en) | Checkup method and checkup device for manhole pump | |
| JP2024019789A (en) | Manhole pump diagnostic method and manhole pump diagnostic device | |
| CN115443362A (en) | Method for monitoring and controlling operation of a pump station | |
| KR100875131B1 (en) | System for diagnosing sewer automatically and flux flow state diagnosis method therefor | |
| JP2023090211A (en) | Insulation resistance evaluation method and electric motor diagnostic method | |
| KR102566249B1 (en) | Valve abnormality detection system and method therefor | |
| JP7542516B2 (en) | How to diagnose your pump | |
| US20240264028A1 (en) | Device for monitoring a room | |
| KR102706611B1 (en) | Smart pump control panel with artificial intelligence function |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |