CN109798970A - Abnormal detector, method for detecting abnormality, abnormality detection system and storage medium - Google Patents
Abnormal detector, method for detecting abnormality, abnormality detection system and storage medium Download PDFInfo
- Publication number
- CN109798970A CN109798970A CN201811248265.8A CN201811248265A CN109798970A CN 109798970 A CN109798970 A CN 109798970A CN 201811248265 A CN201811248265 A CN 201811248265A CN 109798970 A CN109798970 A CN 109798970A
- Authority
- CN
- China
- Prior art keywords
- period
- machine
- data
- vibration
- power spectrum
- 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
Landscapes
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
- Testing And Monitoring For Control Systems (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
Abnormal detector, method for detecting abnormality, abnormality detection system and storage medium are provided.Abnormal detector includes cutting unit, and generation divides the normal vibration data for the normal vibration for being used to show machine to multiple period data of multiple periods;Converter unit carries out Fast Fourier Transform to each period data of the multiple period data generated by cutting unit and calculates multiple power spectrum according to each period;Characteristic spectra computing unit calculates more than one characteristic spectra according to each period according to the multiple power spectrum calculated by converter unit;Model generation unit generates the normal model for being detected to the exception occurred in machine according to the more than one characteristic spectra calculated by characteristic spectra computing unit;The vibration data of index value computing unit, the vibration according to the normal model and for showing machine calculates scheduled index value;And judging unit, whether exception has occurred according in the index value and preset scheduled threshold decision machine.
Description
Technical field
The present invention relates to abnormal detector, method for detecting abnormality, abnormality detection system and it is stored with abnormality detecting program
Storage medium.
Background technique
For example, a kind of known technology, is carried out by the abnormal vibrations in the machine to generator, motor etc.
Detection can detect (for example, referring to patent document 1,2) exception of the machine of component life, component deterioration etc..This
In technology, for example, by using the spectrum waveform of the vibration occurred in machine, and according to the frequency band for showing abnormal vibrations
Whether frequency spectrum has been more than threshold value, to abnormal vibrations can detect.
[existing technical literature]
[patent document]
[patent document 1] (Japan) special open 2009-128103 bulletin
[patent document 2] (Japan) special open 2011-22160 bulletin
Summary of the invention
[subject to be solved by the invention]
However, in the above prior art, exist cannot accurately to abnormal vibrations there is a situation where detect.
For example, being executed as vertical multi-joint robot in the industry robot of compound action, exist and x-axis direction, y-axis side
To and z-axis direction all related situation of each oscillating component (ingredient).In this case, even if in a certain oscillating component
For showing that the frequency spectrum of the frequency band of abnormal vibrations has been more than threshold value, it can not say and exception centainly has occurred in machine.
In view of the above problems, the purpose of an embodiment of the invention is, according to the vibration of machine come accurately
Detect abnormal generation.
[means for solving the problems]
To achieve the goals above, an embodiment of the invention is a kind of vibration for the vibration that basis is used to show machine
The abnormal detector that dynamic data carry out abnormality detection, comprising:
Cutting unit, generation divide the normal vibration data for the normal vibration for being used to show the machine to multiple periods
Multiple period data, each period have predetermined time width;
Converter unit, to each period data in the multiple period data generated by the cutting unit carry out using
The Fast Fourier Transform of window function, to calculate multiple power spectrum according to each period;
Characteristic spectra computing unit is counted according to the multiple power spectrum calculated by the converter unit according to each period
Calculate more than one characteristic spectra;
Model generation unit, according to the more than one characteristic spectra calculated by the characteristic spectra computing unit, generation is used for
The normal model that the exception occurred in the machine is detected;
Index value computing unit, according to the normal model generated by the model generation unit and for showing the machine
Vibration vibration data, calculate scheduled index value;And
Judging unit, according to the index value and preset scheduled threshold value calculated by the index value computing unit,
Judge whether exception has occurred in the machine.
[invention effect]
Abnormal generation can be accurately detected according to the vibration of machine.
Detailed description of the invention
The figure that [Fig. 1] is indicated an example of the overall structure of the abnormality detection system of first embodiment.
The figure that an example that [Fig. 2] constitutes the hardware of the abnormal detector of first embodiment is indicated.
The figure that an example that [Fig. 3] constitutes the function of the abnormal detector of first embodiment is indicated.
[Fig. 4] indicates the figure of an example of vibration data.
[Fig. 5] indicates that the model of first embodiment generates the flow chart of an example of processing.
[Fig. 6] indicates the figure of an example of average power spectra.
[Fig. 7] indicates the figure of an example of maximum power spectrum.
[Fig. 8] indicates the flow chart of an example of the abnormality detection processing of first embodiment.
[Fig. 9] indicates the figure of an example of the Q maximum value of each period.
The figure of an example of the Q value of each frequency number during [Figure 10] expression one.
The figure of an example of the contribution plot of each frequency number during [Figure 11] expression one.
[Figure 12] indicates the figure of the other examples of the overall structure of the abnormality detection system of first embodiment.
The figure that an example that [Figure 13] constitutes the function of the abnormal detector of second embodiment is indicated.
[Figure 14] indicates the flow chart of an example of the abnormality detection processing of second embodiment.
[Figure 15] indicates the figure of an example of output result.
[Figure 16] indicates the figure of the other examples of output result.
[symbol description]
1 abnormality detection system
10 abnormal detectors
20 perception machine (sensing machine)
30 subject machines
101 data obtaining portion
102 frequency conversion parts
103 characteristic spectra calculation parts
104 model generating units
105 index value calculation parts
106 abnormality determiners
107 output sections
110 vibration data storage units
120 model storage units
Specific embodiment
Hereinafter, the embodiments of the present invention are described in detail referring to attached drawing.
[first embodiment]
<overall structure>
Firstly, being illustrated referring to Fig.1 to the overall structure of the abnormality detection system of present embodiment 1.Fig. 1 is to indicate
The figure of an example of the overall structure of the abnormality detection system 1 of one embodiment.
As shown in Figure 1, the abnormality detection system 1 of present embodiment includes abnormal detector 10 and perception machine 20.It is different
Normal detection device 10 and perception machine 20 are carried out such as via network communicable LAN (Local Area Network)
Connection.
Perception machine 20 be to be detected whether have occurred abnormal object, i.e., the vibration of subject machine 30 surveys
The measurement machine of amount.Perceiving machine 20 is, for example, 3-axis acceleration sensor etc., the x-axis direction for measurement object machine 30
Acceleration, subject machine 30 y-axis direction acceleration and subject machine 30 z-axis direction acceleration, thus generate include
The vibration data of the vibration data of the vibration data of x-axis component, the vibration data of y-axis component and z-axis component.In the following, by x-axis
The vibration data of component is expressed as " x-component vibration data ", and the vibration data of y-axis component is expressed as " y-component vibration data ",
And the vibration data of z-axis component is expressed as " z-component vibration data ".
In addition, the vibration data generated can be sent to abnormal detector 10 by perception machine 20.It should be noted that
Perceiving machine 20 for example can carry out subject machine 30 according to predetermined each scheduled time (that is, each sampling period)
Measurement, and generate vibration data.
It should be noted that vibration data is not limited to include the case where the acceleration of three axis.Vibration data is for example
It may include displacement (for example, the displacement of x-axis direction, the displacement in y-axis direction and the displacement in z-axis direction), may also include speed (example
Such as, the speed of the speed of x-axis direction, the speed in y-axis direction and z-axis direction).
Subject machine 30 is the device or equipment for being set to factory, workshop etc..It, can as the specific example of subject machine 30
Enumerate work mechanism (for example, cutting processing machine, bending machine etc.), industrial machine (for example, conveyer, roller press etc.),
Semiconductor manufacturing apparatus, electric calorifie installation, industry robot (for example, vertical multi-joint robot, horizontal articulated robot etc.)
Deng.In addition, as subject machine 30, such as can also be the vehicle of check device, rail vehicle for utilizing vibration to be checked etc.
The device of form.
Abnormal detector 10 is sent out in test object machine 30 according to the vibration data received from perception machine 20
Raw abnormal computer.It should be noted that as abnormal detector 10, it is possible to use such as PLC (Programmable
Logic Controller) etc. control device.
The movement of this system includes " model generation " stage, generates the exception for occurring in test object machine 30
Normal model;And " evaluation " stage, it is detected according to the normal model and vibration data to abnormal, which is included in
The acceleration measured during the work of subject machine 30 etc..Generally speaking, " model generation " stage be subject machine 30 not
Performed offline (offline) processing when work, " evaluation " stage are performed during the work of subject machine 30
(online) is handled online.But it's not limited to that, " model generation " stage and the two stages in " evaluation " stage can also be all
For processed offline, " model generation " stage and the two stages in " evaluation " stage can also all be online processing.
Abnormal detector 10 can generate normal model according to model generation vibration data in " model generation " stage.Mould
Type generation is referred to vibration data, for example, the subject machine 30 worked normally by 20 Duis of perception machine is measured and generated
Vibration data.It should be noted that for model generation vibration data, as indicating the normal dynamic of subject machine 30
The data of work, can also be for by the vibration data of the generations such as user.
In addition, abnormal detector 10 can be according to evaluation vibration data and normal model test object in " evaluation " stage
The exception occurred in machine 30.Evaluation is referred to vibration data, for example, the working online in subject machine 30 by perception machine 20
During the subject machine 30 is measured and the vibration data that generates.
It should be noted that the abnormality detection system 1 of present embodiment may also comprise the object of multiple (plural) types
Machine 30.In this case, as long as the abnormal detector 10 of present embodiment generates just according to each type of subject machine 30
Norm type, and the exception occurred in the subject machine 30 is detected according to each type of subject machine 30.
In addition, the movement of multiple types also can be performed in subject machine 30.For example, just by pair of multiple steps manufacture product
For machine 30, the movement C of the movement A of step A, the movement B and step C of step B can be performed.In this case, this embodiment party
As long as the abnormal detector 10 of formula generates normal model according to each movement, and according to each movement in the subject machine 30
The exception of generation is detected.
In addition, can also be measured by multiple perception machines 20 to the vibration of a subject machine 30.In this case, abnormal
As long as detection device 10 summarizes multiple multiple vibration datas for generating respectively of perception machines 20 for a data, and according to summarizing
The data afterwards carry out the generation of normal model and/or the detection of exception.A data after summarizing refer to, for example,
In the case where being measured by two perception machines 20 to the vibration of a subject machine 30, comprising being surveyed by the 1st perception machine 20
X-component vibration data, y-component vibration data and the z-component vibration data of amount and by the x that measures points of the 2nd perception machine 20
Measure the data of vibration data, y-component vibration data and z-component vibration data.
<hardware composition>
Then, it is illustrated referring to hardware composition of the Fig. 2 to the abnormal detector 10 of present embodiment.Fig. 2 is to indicate
The figure of an example that the hardware of the abnormal detector 10 of first embodiment is constituted.
As shown in Fig. 2, the abnormal detector 10 of present embodiment have input unit 11, display device 12, exterior I/
F13、RAM(Random Access Memory)14、ROM(Read Only Memory)15、CPU(Central
Processing Unit) 16, communication I/F17 and auxilary unit 18.These hardware are communicably carried out by bus 19
Connection.
Input unit 11 is, for example, keyboard, mouse, touch screen etc., for allowing user to input various operations.Display device 12
For example, LCD (Liquid Crystal Display) etc., for showing the processing result of abnormal detector 10.It needs to illustrate
, abnormal detector 10 can also not have at least one of input unit 11 and display device 12.
Exterior I/F13 is the interface interacted with external device (ED).External device (ED) has storage (record) medium 13a etc..
Abnormal detector 10 can carry out the read-write of storage medium 13a via exterior I/F13.Storage medium 13a may be, for example, floppy disk, CD
(Compact Disc), DVD (Digital Versatile Disc), SD storage card, USB storage etc..It should be noted that
The program and/or use of the various functions of the abnormal detector 10 for realizing present embodiment can be stored in storage medium 13a
In the program for the method for detecting abnormality for realizing present embodiment.
RAM14 is the volatile semiconductor memory temporarily saved to program and/data.ROM15 is even if power supply
The non-volatile semiconductor memory that can also be saved to program and/or data is cut off.Exception is for example stored in ROM15
Detection device 10 BIOS (Basic Input/Output System), OS (Operating System) performed when starting
Setting, network settings etc..
CPU16 be by program and/or data from reading out on RAM14 and executing in ROM15, auxilary unit 18 etc.
Reason, to realize whole control and/or the computing device of function of abnormal detector 10.
Communication I/F17 is the interface for communicating abnormal detector 10 and other machines etc..Abnormality detection dress
Vibration data can be received from perception machine 20 via communication I/F17 by setting 10.
Auxilary unit 18 is the non-volatile memory for being stored with program and/or data, may be, for example, HDD (Hard
Disk Drive), SSD (solid state drive) etc..The program and/or data stored in auxilary unit 18 include
Program for realizing the various functions of the abnormal detector 10 of present embodiment, the inspection of the exception for realizing present embodiment
It is the program of survey method, the basic software i.e. OS that abnormal detector 10 is integrally controlled, various for being provided on OS
The application software etc. of function.It should be noted that auxilary unit 18 can pass through scheduled file system, DB (database) etc.
The program and/or data that are stored are managed.
The abnormal detector 10 of present embodiment constitutes the various places, it can be achieved that as described later by with above-mentioned hardware
Reason.
<function composition>
Next, being illustrated referring to function composition of the Fig. 3 to the abnormal detector 10 of present embodiment.Fig. 3 is table
Show the figure of an example that the function of the abnormal detector 10 of first embodiment is constituted.
As shown in figure 3, the abnormal detector 10 of present embodiment have data obtaining portion 101, frequency conversion part 102,
Characteristic spectra calculation part 103, model generating unit 104, index value calculation part 105, abnormality determiner 106 and output section 107.With regard to this
For a little function parts, each function part can be more than one on abnormal detector 10 by being mounted on CPU16 execution
The processing of program and realize.
In addition, the abnormal detector 10 of present embodiment also has vibration data storage unit 110 and model storage unit
120.These storage units for example can all be realized by using auxilary unit 18.It should be noted that in these storage units
At least one can also be realized by using storage device being connect via network with abnormal detector 10 etc..
Vibration data storage unit 110 stores model generation vibration data and evaluation with vibration data.These moulds
Type generation vibration data and evaluation with vibration data be include as perception machine 20 according to measured by each sampling period plus
The time series data of speed (acceleration, the acceleration in y-axis direction and the acceleration of axis direction of x-axis direction).In other words,
For vibration data, for example, can be the time in horizontal axis, the longitudinal axis is the acceleration (acceleration of the acceleration, y-axis direction of x-axis direction
Degree or z-axis direction acceleration) time domain in be indicated.
It should be noted that being stored for model generation vibration data and evaluation vibration data in vibration data
It does not need to be divided into different data and stored in portion 110.For example, in a vibration data, it can be by some time
Between width data (e.g., including moment t=t1To moment t=t2In a period of measured acceleration vibration data) make
For model generation vibration data, and by the data of another time width (e.g., including moment t=t3To moment t=t4's
During measured acceleration vibration data) be used as evaluation vibration data.
Here, (model is generated with vibration data or evaluation vibration the vibration data stored in vibration data storage unit 110
Data) one be illustrated in Fig. 4.Fig. 4 is the figure for indicating an example of vibration data.
Fig. 4 (a) is an example of x-component vibration data contained by vibration data.Fig. 4 (b) is y-component contained by vibration data
An example of vibration data.Fig. 4 (c) is an example of z-component vibration data contained by vibration data.Such as Fig. 4 (a)~Fig. 4 (b) institute
Show, the vibration data of each component is the time series data for the acceleration that horizontal axis is the time, the longitudinal axis is each component.It needs to illustrate
It is, by the time width of a vibration data (that is, measurement start time to the measurement of acceleration contained by the vibration data terminates
The time width at moment) it is expressed as sampling period.
Data obtaining portion 101 can obtain model from vibration data storage unit 110 in " model generation " stage and generate with vibration
Data.In addition, data obtaining portion 101 can obtain evaluation vibration data from vibration data storage unit 110 in " evaluation " stage.
Frequency conversion part 102 will be obtained in " model generation " stage and the generation of " evaluation " stage by data obtaining portion 101
The sampling period of vibration data (model generates vibration data or evaluates vibration data) is divided into scheduled period unit
Period data.In addition, frequency conversion part 102 used for each period data quick Fu of window (window) function
Vertical leaf transformation (FFT:Fast Fourier Transform), so that each its is converted into frequency domain according to each window.
It accordingly, can be spectral intensity according to each window acquisition longitudinal axis in data during one, the power that horizontal axis is frequency
Spectrum.For example, can get L power spectrum in the case where including L window in data during one.
It should be noted that period data can be generated according to the vibration data of each component.For example, that will be obtained by data
Portion 101 obtain vibration data sampling period be divided into N number of period data in the case where, can be respectively by x-component vibration number
According to sampling period, the sampling period of the sampling period of y-component vibration data and z-component vibration data be divided into N number of period,
Data during thus generating.So in this case, produce by x-component vibration data carried out N segmentation N number of period data,
Y-component vibration data has been subjected to N number of period data of N segmentation and has been carried out z-component vibration data N number of period of N segmentation
Data.
Characteristic spectra calculation part 103 is in " model generation " stage and " evaluation " stage according to each period data and according to by frequency
The spectra calculation that rate transformation component 102 obtains is used to indicate the characteristic power spectrum of scheduled characteristic.As characteristic power spectrum, can arrange
Enumerating indicates the average power spectra according to the average value of each window power spectrum obtained and indicates to be obtained according to each window
The maximum power spectrum of the maximum value of the power spectrum obtained.Accordingly, it can be obtained according to each period divided by frequency conversion part 102 special
Property power spectrum.
According to the characteristic spectra obtained by characteristic spectra calculation part 103, (period is single in " model generation " stage for model generating unit 104
Multiple characteristic spectras of position) generate normal model.At this point, (Japan) special open 2016-164772 for example can be used in model generating unit 104
Method disclosed in number bulletin simultaneously generates normal model according to multiple characteristic spectras.
Later, model generating unit 104 stores normal model generated to model storage unit 120.It should be noted that
Normal model is also referred to as " Profile ".
Index value calculation part 105 is in " evaluation " stage according to the normal model stored in model storage unit 120 and by characteristic
It composes the characteristic power spectrum that calculation part 103 obtains and calculates scheduled index value according to each period.It, can as scheduled index value
Enumerate the Q statistical magnitude (Q value) of each frequency band in this period and/or the maximum value of the Q value in each this period (Q maximum
Value).
Abnormality determiner 106 can determine whether the index value calculated by index value calculation part 105 has been more than scheduled threshold value.
In the case where being determined as that index value has been more than threshold value, can detect as exception has occurred in subject machine 30.
Output section 107 is such as the exportable chart drawn to the index value calculated as index value calculation part 105.
As output object, such as display device 12 can be enumerated etc..
<model generation processing>
Then, the model generation processing for generating normal model is illustrated referring to Fig. 5.Fig. 5 is to indicate that first is real
The model for applying mode generates the flow chart of an example handled.
Firstly, data obtaining portion 101 obtains model generation vibration data (step from vibration data storage unit 110
S101).It should be noted that including as described above x-component vibration data, y-component vibration number in model generation vibration data
According to and z-component vibration data.
Then, frequency conversion part 102 generates vibration data (the model generation vibration that will be obtained by data obtaining portion 101
Data) sampling period be divided into data (step S102) during scheduled period unit.Here, as scheduled period,
As long as such as being the time width comprising 65536 data values (acceleration value).In the following, will include 65536 data values
Time width as during one.
For example, producing and shaking to x-component in the case where sampling period is divided into N number of period of period 1 to period N
Dynamic data have carried out N number of period data of N segmentation, have carried out N number of period data of N segmentation to y-component vibration data and to z
Component vibration data has carried out N number of period data of N segmentation.In present embodiment, period 1 to the phase is divided into sampling period
Between N N number of period situation be column be illustrated.
In the following, each period data divided based on x-component vibration data are expressed as " data during x-component ", it will
Each period data divided based on y-component vibration data are expressed as " data during y-component ", and will be vibrated based on z-component
Data and each period data divided are expressed as " data during z-component ".
Next, frequency conversion part 102 used the fast Flourier of window function to become for each period data
It changes, calculates the power spectrum (step S103) for having arrived frequency domain according to each window transform whereby.
More specifically, frequency conversion part 102 for example using include 2048 data values time width as window width,
And overlapping (overlap) rate is set as 50%, Fast Fourier Transform is carried out thus in accordance with each window to calculate power spectrum.
Accordingly, such as a period data, L=65536/ (2048/2)=64 power spectrum can be calculated.That is, being directed to N number of x
Each of data can calculate L power spectrum during component.It equally, can for each of data during N number of y-component
Calculate L power spectrum.Equally, L power spectrum can be calculated for each of data during N number of z-component.
Then, characteristic spectra calculation part 103 is according to each period data and according to the power spectrum calculated by frequency conversion part 102
To calculate the characteristic power spectrum (step S104) for indicating scheduled characteristic.In the following, calculating average function as characteristic power spectrum
Rate spectrum and maximum power spectrum.
Average power spectra can obtain by way of according to average value of each period data to calculate L power spectrum.
More specifically, for being based on data during one and L power spectrum calculating, by according to each frequency calculating spectral intensity
Average value, can get average power spectra.
Maximum power spectrum can obtain by way of according to maximum value of each period data to calculate L power spectrum.
More specifically, for being based on data during one and L power spectrum calculating, by according to each frequency calculating spectral intensity
Maximum value, can get maximum power spectrum.
Accordingly, average power spectra can be calculated for each of data during N number of x-component and maximum power is composed.Together
Sample can calculate average power spectra for each of data during N number of y-component and maximum power is composed.Equally, for N number of z
Each of data can calculate average power spectra and maximum power spectrum during component.
It should be noted that characteristic spectra calculation part 103 can also calculate average power spectra and maximum power spectrum in it is any one
It is a.The reason is that for example, the abnormal vibrations occurred in having grasped subject machine 30 in advance be steady-state vibration in the case where,
The exception of the subject machine 30 can be detected by using the normal model generated based on average power spectra.Equally, example
Such as, in the case that the abnormal vibrations occurred in having grasped subject machine 30 in advance are vibration bursts, by using based on maximum
Power spectrum and the normal model generated can detect the exception of the subject machine 30.
Here, the one of average power spectra is illustrated in Fig. 6.Fig. 6 is the figure for indicating an example of average power spectra.
Fig. 6 (a) is according to data average power spectra calculated during an x-component.Fig. 6 (b) is according to a y-component
Period data average power spectra calculated.Fig. 6 (c) is according to data average power spectra calculated during a z-component.Such as
Shown in Fig. 6 (a)~Fig. 6 (c), the average power spectra of each component is that horizontal axis is frequency number, the data that the longitudinal axis is spectral intensity.It needs
Illustrate, frequency number is the number for indicating scheduled frequency band.
In the following, " x-component mean power will be expressed as according to data average power spectra calculated during an x-component
Spectrum ", will be expressed as " y-component average power spectra " according to data average power spectra calculated during a y-component, and by basis
Data average power spectra calculated is expressed as " z-component average power spectra " during one z-component.
In addition, the one of maximum power spectrum is illustrated in Fig. 7.Fig. 7 is the figure for indicating an example of maximum power spectrum.
Fig. 7 (a) is according to the maximum power spectrum calculated of data during an x-component.Fig. 7 (b) is according to a y-component
Period data maximum power spectrum calculated.Fig. 7 (c) is according to the maximum power spectrum calculated of data during a z-component.Such as
Shown in Fig. 7 (a)~Fig. 7 (c), it is frequency number, the data that the longitudinal axis is spectral intensity that the maximum power spectrum of each component, which is horizontal axis,.
In the following, " x-component maximum power will be shown as according to data maximum power stave calculated during an x-component
Spectrum ", will be shown as " y-component maximum power spectrum " according to data maximum power stave calculated during a y-component, and by basis
Data maximum power stave calculated is shown as " z-component maximum power spectrum " during one z-component.
As shown in Figure 6 and Figure 7, for maximum power spectrum compared with average power spectra, the spectral intensity at each frequency number is higher.
It should be noted that other than average power spectra and maximum power are composed, making for characteristic spectra calculation part 103
For characteristic spectra, such as standard deviation power spectrum can also be calculated, maximum change specific power spectrum etc..Standard deviation power spectrum refers to, needle
The standard deviation that spectral intensity is calculated the L power spectrum calculated based on data during one according to each frequency is obtained
Power spectrum.Maximum changes specific power spectrum and refers to, according to each frequency between adjacent window (window that a part has been overlapped)
Maximum value power spectrum obtained to calculate the difference of spectral intensity.
Then, model generating unit 104 is according to the multiple average power spectras and multiple maximums calculated by characteristic spectra calculation part 103
Power spectrum generates normal model (step S105).At this point, (Japan) special open 2016- for example can be used in model generating unit 104
Method disclosed in No. 164772 bulletins simultaneously generates normal model according to multiple average power spectras and multiple maximum powers spectrum.
For example, with regard to N number of x-component average power spectra, N number of y-component average power spectra, N number of z-component average power spectra, N number of x
For component maximum power spectrum, N number of y-component maximum power spectrum and N number of z-component maximum power spectrum, as long as N is indicated
The lot data of six variables of a batch (batch), and use model disclosed in (Japan) special open 2016-164772 bulletin
Generation method.
More specifically, when n=1, N when, just according to data x-component obtained during the x-component of period n
Average power spectra and x-component maximum power spectrum, according to data y-component average power spectra obtained during the y-component of period n and
Y-component maximum power spectrum and maximum according to data z-component average power spectra obtained during the z-component of period n and z-component
For power spectrum, if the lot data of six variables as a batch in period n, and use (Japan) special open
Model generating method disclosed in 2016-164772 bulletin.
Accordingly, normal model is produced by model generating unit 104.Normal model generated is stored in model storage unit
120。
It should be noted that in present embodiment, as characteristic power spectrum, to using average power spectra and maximum power to compose
The case where the two power spectrum, is illustrated, still, such as in any being used only in average power spectra and maximum power spectrum
In the case where a power spectrum, as long as the lot data for the ternary for indicating N number of batch, and use (Japan) special open
Model generating method disclosed in 2016-164772 bulletin.It should be noted that of variable contained by a batch
The number of variable contained by the number and vibration data of (quantity) by characteristic power spectrum is counted to determine.For example, in characteristic power spectrum
Number be S and vibration data contained by variable number be T in the case where, the number of variable contained by a batch be S ×
T。
<abnormality detection processing>
Next, referring to Fig. 8 to using normal model to come at the abnormality detection of abnormal generation of test object machine 30
Reason is illustrated.Fig. 8 is the flow chart for indicating an example of abnormality detection processing of first embodiment.
Firstly, data obtaining portion 101 obtains evaluation vibration data (step S201) from vibration data storage unit 110.It needs
It is noted that as described above including x-component vibration data, y-component vibration data and z-component vibration in evaluation vibration data
Dynamic data.
Then, frequency conversion part 102 generates the vibration data (evaluation vibration data) that will be obtained by data obtaining portion 101
Sampling period be divided into data (step S202) during scheduled period unit.As scheduled period, with " model life
At " stage is same, for example, comprising the time width of 65536 data values (acceleration value).
In the following, data during M period for being divided into period 1 to period M as sampling period, with " model generation "
Stage is same, and data are expressed as " data during x-component " during being divided to x-component vibration data, will be to y-component
Data are expressed as " data during y-component " during vibration data is divided, and will have been carried out to z-component vibration data point
Data are expressed as " data during z-component " during cutting.
Then, frequency conversion part 102 is implemented to have used the Fast Fourier Transform of window function for each period data,
The power spectrum (step S203) of frequency domain has been transformed to thus in accordance with each window calculation.At this point, frequency conversion part 102 using with "
Model generation " stage same window width and Duplication carry out Fast Fourier Transform.Accordingly, same with " model generation " stage
Sample, for example, L power spectrum can be calculated for data during one.
It should be noted that the quick of window function is utilized to for the implementation of each period data in present embodiment
The case where Fourier transform, is illustrated, and but not limited to this.For example, small echo change can also be carried out for each period data
Change (wavelet transform) etc..
Later, characteristic spectra calculation part 103 according to each period data according to the power spectrum calculated by frequency conversion part 102 come
Calculate the characteristic power spectrum (step S204) for indicating scheduled characteristic.At this point, characteristic spectra calculation part 103 can calculate and " model life
At " stage same characteristic power spectrum.In the following, calculating average power spectra and maximum power spectrum as characteristic power spectrum.
Accordingly, average power spectra can be calculated for each of data during M x-component and maximum power is composed.Equally,
Average power spectra and maximum power spectrum can be calculated for each of data during M y-component.Equally, for M z-component
Each of period data can calculate average power spectra and maximum power spectrum.
Then, index value calculation part 105 is according to the normal model stored in model storage unit 120 and by characteristic spectra calculation part
The 103 characteristic power spectrum (average power spectra and maximum power spectrum) obtained simultaneously calculate scheduled index value according to each period
(step S205).In the following, the Q value as scheduled index value, to each frequency number for calculating the frequency band in expression this period
With indicate this period each period number Q maximum value the case where be illustrated.However, being not limited to Q as index value
The maximum value of statistic and/or Q statistical magnitude, such as T can also be used2Statistic, T2The maximum value etc. of statistic.
For the Q value of each frequency number during certain, each frequency number of each characteristic power spectrum can be passed through
Total being indicated of contribution plot (contribution plot) (contribution plot of Q statistical magnitude).For example, within this period, if
The contribution plot of certain frequency number f in x-component average power spectra is " Q11(f) ", frequency number in y-component average power spectra
The contribution plot of f is " Q12(f) ", the contribution plot of the frequency number f in z-component average power spectra is " Q13(f) ", x-component is maximum
The contribution plot of frequency number f in power spectrum is " Q21(f) ", the contribution plot of the frequency number f in y-component maximum power spectrum
For " Q22(f) " contribution plot of the frequency number f and in z-component maximum power spectrum is " Q23(f) ", the then Q of frequency number f
Value can be by Q11(f)+Q12(f)+Q13(f)+Q21(f)+Q22(f)+Q23(f) it indicates.
In addition, the Q maximum value of each period number be this period number n represented during each Q (f) in n maximum
Value.
Then, abnormality determiner 106 judges whether the index value calculated by index value calculation part 105 has been more than scheduled threshold
It is worth (step S206).It should be noted that threshold value can be set according to each index value.That is, using Q value as index value
In the case where Q maximum value, the threshold value for the Q value of each frequency number can be set and the Q for each period number is maximum
The threshold value of value.
Next, output section 107 is for example exportable to have carried out drafting to the index value calculated by index value calculation part 105
(step S207) such as charts.
Here, as output section 107 output result an example, indicate the chart quilt of the Q maximum value of each period number
It is shown in Fig. 9.Number is horizontal axis, the chart that Q maximum value is the longitudinal axis during chart shown in Fig. 9 is.In the example shown in Fig. 9,
Threshold value as the Q maximum value for each period number sets " 10000 ".In this case, can by abnormality determiner 106
It detects and exception has occurred at number " 164 " during being more than the threshold value.
Accordingly, it is had occurred in the knowable subject machine 30 of the user of the abnormal detector 10 of present embodiment abnormal
Period.
In addition, the other examples of the output result as output section 107, the Q of each frequency number during certain is indicated
The chart of value is shown in Figure 10.Chart shown in Fig. 10 is that frequency number is horizontal axis, the chart that Q value is the longitudinal axis.Shown in Figure 10
Example in, set " 200 " as the threshold value of Q value for each frequency number.In this case, by abnormality determiner
106 can be detected be more than the threshold value frequency number " 150 " at exception has occurred.
Accordingly, such as in the case where having carried out associated situation with the frequency band when exception occurs to abnormal classification in advance, this
The abnormal type occurred in the knowable subject machine 30 of the user of the abnormal detector 10 of embodiment.That is, right
In the case where being determined as frequency band when certain exception having occurred in machine 30, the abnormal detector 10 of present embodiment
User also would know that the abnormal type occurred in the subject machine 30.
Here, as described above, the Q value of each frequency number can pass through the contribution plot Q at the frequency number11~Q13And Q21
~Q23Total be indicated.The one of each contribution plot is illustrated in Figure 11.Figure 11 (a) be certain during each frequency number tribute
Offer figure Q11.Figure 11 (b) is the contribution plot Q of each frequency number in this period12.Figure 11 (c) is each frequency in this period
The contribution plot Q of number13.Figure 11 (d) is the contribution plot Q of each frequency number in this period21.Figure 11 (e) is in this period
The contribution plot Q of each frequency number22.Figure 11 (f) is the contribution plot Q of each frequency number in this period23.By according to every
A frequency number calculates Q11~Q13And Q21~Q23It is total, the Q value at the frequency number can be calculated.In abnormal generation quilt
In the case where detecting, the user of the abnormal detector 10 of present embodiment by referring to each frequency number contribution
Figure, it is to be understood that the contribution degree of which variable of which characteristic power spectrum is higher.
<other examples of abnormality detection system 1>
Here, the other examples of the overall structure of the abnormality detection system 1 of 2 pairs of present embodiments are illustrated referring to Fig.1.
Figure 12 is the figure for indicating the other examples of the overall structure of abnormality detection system 1 of first embodiment.
As shown in figure 12, the abnormality detection system 1 of present embodiment have abnormal detector 10, perception machine 20 and
Display device 40 may be, for example, the composition that can be communicatedly connected via the network N of WWW (Internet) etc..Change speech
It, the model based on abnormal detector 10 generates processing and abnormality detection processing can be used as cloud service (cloud serVice)
It is provided to the user of display device 40.
In the abnormality detection system 1 shown in Figure 12, finger that abnormal detector 10 will be calculated by index value calculation part 105
The judgement result of scale value and abnormality determiner 106 is sent to display device 40.Accordingly, such as basis can be shown in display device 40
Output result represented by the chart of Fig. 9~as shown in Figure 11.It should be noted that as display device 40, such as can make
With PC (personal computer), smart phone, tablet terminal etc..
<summary of first embodiment>
As described above, the abnormality detection system 1 of present embodiment for example can be offline according to for indicating subject machine 30
Regular event vibration data generate normal model.In addition, the abnormality detection system 1 of present embodiment can also be according to by perceiving
The vibration data and the normal model that the movement for the subject machine 30 that machine 20 is worked online by measurement obtains are to the object
The exception of machine 30 is detected.Accordingly, in the abnormality detection system of present embodiment 1, according to the vibration of subject machine 30,
Abnormal generation can accurately be detected.
[second embodiment]
Next, being illustrated to second embodiment.In second embodiment, in the case where detecting exception
It has determined and abnormal variable (for example, the variable for indicating x-component, y-component and z-component of acceleration etc.) has had occurred and in this base
The case where power spectrum and normal model for the variable being determined are shown on plinth is illustrated.Accordingly, such as abnormal detector
10 user can confirm the power spectrum and normal model that abnormal variable has occurred, and can be used as studying carefully bright generation
The reference of abnormal reason, determining abnormal position etc..
It should be noted that being mainly illustrated to the item being different from the first embodiment in second embodiment.Separately
Outside, identical symbol is imparted to constituent element same as the first embodiment, and it is illustrated to be omitted.
<function composition>
Firstly, the function composition of the abnormal detector 10 of 3 pairs of present embodiments is illustrated referring to Fig.1.Figure 13 is table
Show the figure of an example that the function of the abnormal detector 10 of second embodiment is constituted.
As shown in figure 13, the abnormal detector 10 of present embodiment also has determining section 108.The function part can by by
CPU16 executes the processing for the more than one program being mounted on abnormal detector 10 and realizes.
Determining section 108 can be in the case where abnormality determiner 106 detects abnormal generation to abnormal change has occurred
(a possibility that more precisely, being abnormal generation higher variable) is measured to be determined.
In addition, the power spectrum of the exportable variable determined by determining section 108 in the output section 107 of present embodiment and the variable
Normal model, using as output result.In addition, as output object, such as display device 12 can be enumerated etc..Accordingly, may be used
Power spectrum and change of abnormal variable a possibility that (more precisely, being abnormal generation higher variable) has occurred in display
The normal model of amount.
<abnormality detection processing>
Then, referring to Fig.1 4, it is to the abnormal generation for using normal model test object machine 30 and different detecting
The abnormality detection processing of the power spectrum of abnormal variable and the normal model of the variable has occurred in display in the case where normal generation
It is illustrated.Figure 14 is the flow chart for indicating an example of abnormality detection processing of second embodiment.It should be noted that Figure 14
Step S201~step S206 it is identical as Fig. 8, so illustrating to be omitted to it.
Abnormal situation is detected (namely it is decided that for the index of ID during certain by abnormality determiner 106 in step S206
The case where value has been more than threshold value) under, abnormal frequency number has occurred during determination shown in this period ID in determining section 108
A possibility that (more precisely, being abnormal generation higher frequency number) (step S301).Here, such as in index value it is
In the case where Q maximum value, as long as the Q value in each frequency number of this period ID is that highest frequency number is true by determining section 108
It is set to and abnormal frequency number has occurred.Equally, such as in the case where index value is Q value, as long as determining section 108 should
Q value in each frequency number of period ID is that highest frequency number is determined as that abnormal frequency number has occurred.
Then, it is determined that determining during portion 108 is shown in the period ID for identified frequency in above-mentioned steps S301
The highest variable (step S302) of the contribution degree of the Q value of number.If the frequency number is f, the Q value of frequency number f is such as
It for example can be by Q described in upper11(f)+Q12(f)+Q13(f)+Q21(f)+Q22(f)+Q23(f) it indicates.So the contribution degree of variable x can
By Q11(f)+Q21(f) it indicates, the contribution degree of variable y can be by Q12(f)+Q22(f) it indicates, the contribution degree of variable z can be by Q13(f)+
Q23(f) it indicates.Determining section 108 is determined the variable of the highest contribution degree in these contribution degrees.In the following, also will be by determining
The variable that portion 108 determines is expressed as " abnormal that variable occurs ".
Next, the abnormal power spectrum that variable occurs and the exception that the output of output section 107 is determined by determining section 108 occur
The normal model of variable, using as output result (step S303).Here, export result one is illustrated in Figure 15.Such as Figure 15 institute
Show, as output as a result, overlapping shows that the normal model of variable occurs for the abnormal power spectrum that variable occurs and the exception.According to
This, the normal model that the abnormal power spectrum that variable occurs and the exception variable can occur for user on one side is compared, on one side
As the reference for studying carefully bright abnormal the reason of occurring, the position for determining exception etc..In other words, user can be to different in frequency domain
Often when movement and it is normal when movement be compared, actual unusual condition can accurately be confirmed whereby.
At this point, user for example can also be by the specified range for wanting confirmation such as mouse, to specified range progress
Amplification.It accordingly, can be more detailed to the abnormal power spectrum that variable occurs and the progress such as difference of normal model of generation variable extremely
Thin confirmation.
In addition, user for example can also be by executing display handover operation etc., to shown in Figure 16 (a)~Figure 16 (c) at this time
Output result shown.Output shown in Figure 16 (a) is the result is that the abnormal power spectrum that variable occurs and the exception become
The difference absolute value of each frequency of the normal model of amount.Output shown in Figure 16 (b) is the result is that the abnormal power spectrum that variable occurs
The ratio of each frequency of the normal model of variable occurs extremely with this.Output shown in Figure 16 (c) is the result is that abnormal occur variable
Power spectrum and this extremely occur variable normal model each frequency difference.By referring to these outputs as a result, user
It can be as the reference for studying carefully bright abnormal the reason of occurring, the position for determining exception etc..
<summary of second embodiment>
As described above, the abnormality detection system 1 of present embodiment can be shown in the case where detecting exception as defeated
The normal model of variable occurs for the abnormal power spectrum that variable occurs of result and the exception out.Accordingly, user can be in frequency band
Exception when movement and it is normal when movement be compared, and can be used as study carefully bright abnormal the reason of occurring, determine it is abnormal
Position etc. reference.
It should be noted that during exception has occurred, will be determined in above-mentioned steps S301 in present embodiment
The highest variable of contribution degree of Q value of frequency number variable occurs as abnormal, but not limited to this.For example, can also press
According to the contribution degree for the Q value from the sequence of high to low (or from down to height), will be located at front (or below) (that is, contribution degree compared with
It is high) S variable variable occurs as abnormal.Accordingly, for example, by according to for the Q value contribution degree from it is high to low (or from
Down to height) sequence show each abnormal power spectrum and normal model that variable occurs, user can be as bright different for studying carefully
The reference of the reason of often occurring, the position for determining exception etc..
Based on above-mentioned, a kind of abnormal detector is provided, is carried out according to the vibration data for showing the vibration of machine different
Often detection, the abnormal detector include cutting unit, generate the normal vibration for the normal vibration that will be used to show the machine
Dynamic data are divided to multiple period data of multiple periods, and each period has predetermined time width;Transformation is single
Member carries out having used the quick of window function to each period data in the multiple period data generated by the cutting unit
Fourier transform, to calculate multiple power spectrum according to each period;Characteristic spectra computing unit, according to single by the transformation
Multiple power spectrum that member calculates calculate more than one characteristic spectra according to each period;Model generation unit, according to by institute
The more than one characteristic spectra of characteristic spectra computing unit calculating is stated, is generated for being detected to the exception occurred in the machine
Normal model;Index value computing unit, it is described according to the normal model generated by the model generation unit and for showing
The vibration data of the vibration of machine calculates scheduled index value;And judging unit, it is calculated according to by the index value computing unit
Index value and preset scheduled threshold value, judge whether exception has occurred in the machine.
The characteristic spectra computing unit is calculated according to the multiple power spectrum and according to each period for indicating
The maximum power spectrum of the average power spectra of the average value of multiple power spectrum and the maximum value for indicating the multiple power spectrum is stated,
The index value computing unit according to calculated by the characteristic spectra computing unit average power spectra and maximum power spectrum, Yi Jisuo
It states normal model and calculates scheduled index value.
The index value computing unit calculate the Q statistical magnitude as the index value, it is described during Q statistical magnitude most
Big value, T2Statistic and it is described during T2At least one of maximum value of statistic.
The abnormal detector also has output unit, the Q statistics of each frequency for exporting the vibration
At least one of the maximum value of amount and the Q statistical magnitude of each period.
The index value is the maximum value of Q statistical magnitude or Q statistical magnitude, and the abnormal detector also has determination unit,
In the case where by the judging unit being judged to that exception has occurred in the machine, in a period of the exception has occurred, really
Determining Q statistical magnitude is maximum frequency, and true from high to low sequence according to the contribution degree of the Q statistical magnitude for identified frequency
Determine the variable of predetermined number, in multiple power spectrum that the output unit output is calculated by the converter unit with by it is described really
The corresponding power spectrum of variable and normal model corresponding with the variable determined by the determination unit that order member determines.
Output unit output and the corresponding power spectrum of the variable and normal model corresponding with the variable it is every
At least one of difference absolute value, ratio and difference of a frequency.
A kind of method for detecting abnormality is also provided, wherein according to the vibration data for showing the vibration of machine to exception into
The abnormal detector of row detection executes following steps, it may be assumed that generates the normal vibration for the normal vibration that will be used to show the machine
Dynamic data are divided to the segmentation step of multiple period data of multiple periods, and each period has the predetermined time wide
Degree;Each period data in the multiple period data generated by the segmentation step are carried out having used the quick of window function
Fourier transform, to calculate the shift step of multiple power spectrum according to each phase;It is calculated according to by the shift step
Multiple power spectrum and according to each period calculate more than one characteristic spectra characteristic spectra calculate step;According to by the characteristic
It is normal for being detected to the exception occurred in the machine that spectrum calculates the more than one characteristic spectra generation that step calculates
The model generation step of model;According to the normal model generated by the model generation step and the vibration for showing the machine
The index value that dynamic vibration data calculates scheduled index value calculates step;And calculate what step calculated according to by the index value
Abnormal judgment step whether has occurred in machine described in index value and preset scheduled threshold decision.
A kind of abnormality detection system is also provided, including machine and the measurement machine measured to the vibration of the machine, tool
Have: cutting unit, generate the normal vibration data for the normal vibration for being used to show the machine are divided it is more to multiple periods
A period data, each period have predetermined time width;Converter unit, to what is generated by the cutting unit
Each period data in multiple period data used the Fast Fourier Transform of window function, thus according to each institute
Multiple power spectrum are calculated during stating;Characteristic spectra computing unit, according to the multiple power spectrum calculated by the converter unit, according to every
A period calculates more than one characteristic spectra;Model generation unit, according to one calculated by the characteristic spectra computing unit
A above characteristic spectra, generates the normal model for being detected to the exception occurred in the machine;Index value calculates single
Member, according to the vibration data of the normal model and the vibration for showing the machine that are generated by the model generation unit, meter
Calculate scheduled index value;And judging unit, according to the index value that is calculated by the index value computing unit and preset pre-
Fixed threshold value judges whether exception has occurred in the machine.
In addition, also providing a kind of storage medium for being stored with abnormality detecting program, which is that computer can
Reader can make computer execute above-mentioned method for detecting abnormality.
Embodiment illustrated of the invention is illustrated above, but the present invention is not limited to specifically disclosed realities
Mode is applied, without departing from the range that claims are recorded, can also carry out various deformations and/or change to it.
Claims (9)
1. a kind of abnormal detector carries out abnormality detection, the exception according to the vibration data for showing the vibration of machine
Detection device includes
Cutting unit, generate the normal vibration data for the normal vibration for being used to show the machine are divided it is more to multiple periods
A period data, each period have predetermined time width;
Converter unit carries out having used window to each period data in the multiple period data generated by the cutting unit
The Fast Fourier Transform of function, to calculate multiple power spectrum according to each period;
Characteristic spectra computing unit calculates one according to each period according to the multiple power spectrum calculated by the converter unit
A above characteristic spectra;
Model generation unit is generated according to the more than one characteristic spectra calculated by the characteristic spectra computing unit for institute
State the normal model that the exception occurred in machine is detected;
Index value computing unit, according to the normal model generated by the model generation unit and the vibration for showing the machine
Dynamic vibration data calculates scheduled index value;And
Judging unit, according to the index value and preset scheduled threshold value calculated by the index value computing unit, judgement
Whether exception has occurred in the machine.
2. abnormal detector as described in claim 1, wherein
The characteristic spectra computing unit is calculated according to each period for indicating the multiple according to the multiple power spectrum
The maximum power spectrum of the average power spectra of the average value of power spectrum and the maximum value for indicating the multiple power spectrum,
The index value computing unit according to calculated by the characteristic spectra computing unit average power spectra and maximum power spectrum and
The normal model calculates scheduled index value.
3. abnormal detector as claimed in claim 1 or 2, wherein
The index value computing unit calculate Q statistical magnitude, it is described during the maximum value of Q statistical magnitude, T2Statistic and described
During T2At least one of maximum value of statistic, as the index value.
4. abnormal detector as claimed in claim 3, also includes
Output unit exports the Q statistical magnitude and the Q statistical magnitude of each period of each frequency of the vibration
At least one of maximum value.
5. abnormal detector as claimed in claim 4, wherein
The index value is the maximum value of Q statistical magnitude or Q statistical magnitude,
The abnormal detector also has determination unit, is being judged to that exception has occurred in the machine by the judging unit
In the case where, in a period of the exception has occurred, determine that Q statistical magnitude is maximum frequency, and according to for having determined
The contribution degree of the Q statistical magnitude of frequency determines the variable of predetermined number from high to low sequence,
In multiple power spectrum that output unit output is calculated by the converter unit with determined by the determination unit
The corresponding power spectrum of variable and normal model corresponding with the variable determined by the determination unit.
6. abnormal detector as claimed in claim 5, wherein
Each frequency of output unit output and the variable corresponding power spectrum and normal model corresponding with the variable
At least one of difference absolute value, ratio and difference of rate.
7. a kind of method for detecting abnormality, wherein carried out abnormality detection according to the vibration data for showing the vibration of machine different
Normal detection device executes following step, it may be assumed that
Generation divides the normal vibration data for the normal vibration for being used to show the machine to multiple period numbers of multiple periods
According to segmentation step, each period have predetermined time width;
Each period data in the multiple period data generated by the segmentation step are carried out having used the fast of window function
Fast Fourier transform, to calculate the shift step of multiple power spectrum according to each period;
According to the multiple power spectrum calculated by the shift step, more than one characteristic spectra is calculated according to each period
Characteristic spectra calculates step;
The more than one characteristic spectra that step calculates is calculated according to by the characteristic spectra, is generated for occurring in the machine
The model generation step for the normal model that exception is detected;
Based on the vibration data of the normal model and the vibration by showing the machine that are generated by the model generation step
The index value for calculating scheduled index value calculates step;And
The index value and preset scheduled threshold value that step calculates are calculated according to by the index value, is judged in the machine
Whether abnormal judgment step is had occurred.
8. a kind of storage medium for being stored with computer-readable program, wherein the computer-readable program executes computer
Following step, it may be assumed that
Generation divides the normal vibration data for the normal vibration for being used to show machine to multiple period data of multiple periods
Segmentation step, each period have predetermined time width;
Each period data in the multiple period data generated by the segmentation step are carried out having used the fast of window function
Fast Fourier transform, to calculate the shift step of multiple power spectrum according to each period;
According to the multiple power spectrum calculated by the shift step, more than one characteristic spectra is calculated according to each period
Characteristic spectra calculates step;
The more than one characteristic spectra that step calculates is calculated according to by the characteristic spectra, is generated for occurring in the machine
The model generation step for the normal model that exception is detected;
Based on the vibration data of the normal model and the vibration by showing the machine that are generated by the model generation step
The index value for calculating scheduled index value calculates step;And
The index value and preset scheduled threshold value that step calculates are calculated according to by the index value, is judged in the machine
Whether abnormal judgment step is had occurred.
9. a kind of abnormality detection system, including machine and the measurement machine measured to the vibration of the machine, the abnormal inspection
Examining system includes
Cutting unit, generate the normal vibration data for the normal vibration for being used to show the machine are divided it is more to multiple periods
A period data, each period have predetermined time width;
Converter unit carries out having used window to each period data in the multiple period data generated by the cutting unit
The Fast Fourier Transform of function, to calculate multiple power spectrum according to each period;
Characteristic spectra computing unit calculates one according to each period according to the multiple power spectrum calculated by the converter unit
A above characteristic spectra;
Model generation unit is generated according to the more than one characteristic spectra calculated by the characteristic spectra computing unit for institute
State the normal model that the exception occurred in machine is detected;
Index value computing unit, according to the normal model generated by the model generation unit and the vibration for showing the machine
Dynamic vibration data calculates scheduled index value;And
Judging unit, according to the index value and preset scheduled threshold value calculated by the index value computing unit, judgement
Whether exception has occurred in the machine.
Applications Claiming Priority (4)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2017-221642 | 2017-11-17 | ||
JP2017221642 | 2017-11-17 | ||
JP2018-128340 | 2018-07-05 | ||
JP2018128340A JP7358725B2 (en) | 2017-11-17 | 2018-07-05 | Abnormality detection device, abnormality detection method, abnormality detection program and abnormality detection system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109798970A true CN109798970A (en) | 2019-05-24 |
CN109798970B CN109798970B (en) | 2021-04-02 |
Family
ID=66556234
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811248265.8A Active CN109798970B (en) | 2017-11-17 | 2018-10-24 | Abnormality detection device, abnormality detection method, abnormality detection system, and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109798970B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112665706A (en) * | 2020-11-30 | 2021-04-16 | 武汉第二船舶设计研究所(中国船舶重工集团公司第七一九研究所) | Marine platform vibration monitoring and analyzing method and system |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPS5862528A (en) * | 1981-10-09 | 1983-04-14 | Sumitomo Metal Ind Ltd | Monitoring method for periodical motion body |
JP2004020424A (en) * | 2002-06-18 | 2004-01-22 | Mitsubishi Chemicals Corp | Processing method for vibration signal |
CN102607845A (en) * | 2012-03-05 | 2012-07-25 | 北京工业大学 | Bearing fault characteristic extracting method for redundantly lifting wavelet transform based on self-adaptive fitting |
CN102606891A (en) * | 2012-04-11 | 2012-07-25 | 广州东芝白云自动化***有限公司 | Water leakage detector, water leakage detecting system and water leakage detecting method |
CN104595112A (en) * | 2013-10-30 | 2015-05-06 | 通用电气公司 | Wind turbine and method for evaluating health status of blades thereon |
-
2018
- 2018-10-24 CN CN201811248265.8A patent/CN109798970B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPS5862528A (en) * | 1981-10-09 | 1983-04-14 | Sumitomo Metal Ind Ltd | Monitoring method for periodical motion body |
JP2004020424A (en) * | 2002-06-18 | 2004-01-22 | Mitsubishi Chemicals Corp | Processing method for vibration signal |
CN102607845A (en) * | 2012-03-05 | 2012-07-25 | 北京工业大学 | Bearing fault characteristic extracting method for redundantly lifting wavelet transform based on self-adaptive fitting |
CN102606891A (en) * | 2012-04-11 | 2012-07-25 | 广州东芝白云自动化***有限公司 | Water leakage detector, water leakage detecting system and water leakage detecting method |
CN104595112A (en) * | 2013-10-30 | 2015-05-06 | 通用电气公司 | Wind turbine and method for evaluating health status of blades thereon |
Non-Patent Citations (1)
Title |
---|
陈运胜: "发电机传动轴承的异常振动谱特征提取算法", 《国外电子测量技术》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112665706A (en) * | 2020-11-30 | 2021-04-16 | 武汉第二船舶设计研究所(中国船舶重工集团公司第七一九研究所) | Marine platform vibration monitoring and analyzing method and system |
CN112665706B (en) * | 2020-11-30 | 2023-04-11 | 武汉第二船舶设计研究所(中国船舶重工集团公司第七一九研究所) | Vibration monitoring and analyzing method and system for maritime work platform |
Also Published As
Publication number | Publication date |
---|---|
CN109798970B (en) | 2021-04-02 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US11624731B2 (en) | Method for predicting remaining life of numerical control machine tool | |
CN107449958B (en) | Abnormality determination device and abnormality determination method | |
US10525563B2 (en) | Abnormality-detecting device and method for tool of machine tool | |
JP7358725B2 (en) | Abnormality detection device, abnormality detection method, abnormality detection program and abnormality detection system | |
CN106441896A (en) | Characteristic vector extraction method for rolling bearing fault mode identification and state monitoring | |
US8942939B2 (en) | Real-time detection system and the method thereof | |
EP1705468A2 (en) | Abnormality detection device and method having a judgment model | |
JPWO2015177870A1 (en) | Manufacturing equipment diagnosis support device | |
EP3825800A1 (en) | Anomaly detection device, method of detecting anomaly, and program | |
Yao et al. | Real-time chatter detection and automatic suppression for intelligent spindles based on wavelet packet energy entropy and local outlier factor algorithm | |
CN104657526A (en) | Method for predicting machining quality of machine tool | |
CN110398384A (en) | The machine health monitoring of rotating machinery | |
CN116296364A (en) | Gear fault diagnosis method, system, motor controller and computer medium | |
JP2016200451A (en) | Signal processing method, signal processing device, and cutting work abnormality detection device | |
CN109798970A (en) | Abnormal detector, method for detecting abnormality, abnormality detection system and storage medium | |
CN112630579A (en) | Aging test method and system for servo driver | |
Aparisi et al. | Interpreting the out-of-control signals of multivariate control charts employing neural networks | |
CN112987682A (en) | Control method, control device and mechanical equipment | |
CN116561646A (en) | Rolling bearing multidimensional parameter automatic detection line and rolling bearing grade evaluation method | |
Dorado-Vicente et al. | Comparing planar pocketing tool paths via acceleration measurement | |
Kuzio et al. | Identification of fault frequency variation in the envelope spectrum in the vibration-based local damage detection in possible changing load/speed conditions | |
JP2009300192A (en) | Crack detecting device and crack detecting method | |
CN111947929B (en) | Method for analyzing working condition sensitivity of vibration characteristic evaluation index of rotary machine | |
JP7230371B2 (en) | Abnormality detection device, abnormality detection method, abnormality detection program and abnormality detection system | |
JP2021076597A (en) | Detecting rotor anomalies by determining vibration trends during transient speed operation |
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 |