CN110398331A - Vibratory response frequency domain prediction method and device based on offset minimum binary - Google Patents
Vibratory response frequency domain prediction method and device based on offset minimum binary Download PDFInfo
- Publication number
- CN110398331A CN110398331A CN201910661629.3A CN201910661629A CN110398331A CN 110398331 A CN110398331 A CN 110398331A CN 201910661629 A CN201910661629 A CN 201910661629A CN 110398331 A CN110398331 A CN 110398331A
- Authority
- CN
- China
- Prior art keywords
- vibratory response
- vibration
- vibrating sensor
- data
- frequency domain
- 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.)
- Pending
Links
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
- G01M7/00—Vibration-testing of structures; Shock-testing of structures
- G01M7/02—Vibration-testing by means of a shake table
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M7/00—Vibration-testing of structures; Shock-testing of structures
- G01M7/02—Vibration-testing by means of a shake table
- G01M7/025—Measuring arrangements
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The present invention provides the prediction techniques and device of the multiple spot frequency domain vibratory response under a kind of multi-source load unknown condition based on offset minimum binary.This method is using the frequency domain vibratory response of known measuring point as input, and the frequency domain vibratory response of multiple unknown measuring points is as output, first with the historical data training partial least square model.Under the unknown working condition of multi-source load, using the frequency domain vibratory response of known measuring point as the input of trained partial least square model, output realizes the prediction of the frequency domain vibratory response to multiple unknown measuring points.The device includes multiple vibration sources and multiple vibratory response sensors;The multiple vibration source generates uncorrelated stationary random excitation;The vibrating sensor is fixedly arranged in structure, the vibration of interrecord structure.The present invention has the advantages that realizing under the unknown working condition of multi-source load, the frequency domain vibratory response of multiple measuring points is predicted.
Description
Technical field
Field is predicted in the frequency domain vibratory response that the present invention relates to uncorrelated multi-source load under unknown, is referred in particular to a kind of uncorrelated
Vibratory response frequency domain prediction method and device under multi-source load is unknown based on offset minimum binary.
Background technique
With industry and the development and progress of control technology, the fields such as aerospace, ship, big machinery and bridge
Engineering structure develops to obtain becoming increasingly complex, enlargement, intelligence.It vibrates and must not be in Machine Design, navigation aerospace engineering
Irrespective design factor, mechanical damage, bridge collapse, navigation caused by especially vibratory response is excessive in design and use
Space flight accident is even more commonplace.However the vibratory response of certain nodes under some duty constructions cannot be directly measured,
This makes the difficulty of the control and Vibration Absorption Designing vibrated to node as Machine Design.
For the vibratory response prediction of node, traditionally with the following method: first using experimental method or finite element simulation
Method establishes the kinetics equation of structure, finds out the transmission function of structure, then calculates or predicts using the load working condition of structure
The vibratory response of structure.But traditional method presence has the disadvantage that
One is for complicated engineering structure, the modeling of system, transmission function are sought being not easy to;The second is very
In more situations, the load working condition of structure is also cannot be measured directly, such as guided missile flies in the sky, ocean platform heavy construction
Object is difficult that the external applied load for acting on structure is directly measured or calculated, very when by stormy waves and traffic excitation effect
To because load position does not reach, surveying this dynamic load can not sometimes.
Therefore, how vibratory response frequency domain to be predicted in the case where incoherent multi-source load is unknown, becomes one
A urgent problem to be solved.
Summary of the invention
One of the technical problem to be solved in the present invention is to provide a kind of vibratory response frequency domain based on offset minimum binary pre-
Device is surveyed, realization predicts vibratory response frequency domain in the case where uncorrelated multi-source load is unknown.
The present invention is realized in one of technical problem: a kind of vibratory response frequency domain prediction dress based on offset minimum binary
It sets, including multiple vibration sources and multiple vibratory response sensors;The multiple vibration source is connect with shake table, by described more
A vibration source generates uncorrelated stationary random vibration;The multiple vibrating sensor is fixedly arranged in structure, is passed by the vibration
Sensor records the vibration that multiple vibration sources generate.
Further, the structure includes a uniform disc and a supporting element;The center of the uniform disc and support
Part is affixed.
Further, more vibration sources include one first vibration source and one second vibration source;First vibration source
It connect with shake table, is vibrated by the first vibration source driving structure;Structure band is hammered by second vibration source
Dynamic structural vibration.
The second technical problem to be solved by the present invention is to provide a kind of vibratory response frequency domain based on offset minimum binary pre-
Survey method, realization predict vibratory response frequency domain in the case where uncorrelated multi-source load is unknown.
The present invention is realized in the twos' of technical problem: a kind of vibratory response frequency domain prediction side based on offset minimum binary
Method, the method need to use the prediction meanss as described in claims 1 to 3 is any, and described method includes following steps:
Step S10, uncorrelated stationary random vibration is generated by multiple vibration sources, n vibrating sensor records shake table
Vibration data be historical data, vibration environment is recorded as history operating condition;
Step S20, by n1The vibration data of a vibrating sensor is as known vibratory response data, by n2A vibrating sensing
The vibration data of device is as unknown vibratory response data;Wherein n1For positive integer, n2For positive integer, and n=n1+n2;
Step S30, offset minimum binary training pattern is established, using vibratory response data known in historical data as partially minimum
Two multiply the input data of training pattern, using vibratory response data unknown in historical data as the defeated of offset minimum binary training pattern
Data out obtain the regression relation coefficient square in historical data between known vibratory response data and unknown vibratory response data
Battle array;
Step S40, using regression relation coefficient matrix and known vibratory response data to unknown vibratory response data into
Row prediction.
Further, the step S10 specifically:
The m uncorrelated steady random vibrations that magnitude is gradually increased are generated by the first vibration source and the second vibration source joint
Dynamic load;
The vibratory response size of n vibrating sensor interrecord structure isAnd it calculates auto-power spectrum and isAnd vibratory response size and auto-power spectrum are denoted as historical data, vibration environment is recorded as history
Operating condition;
Wherein m is positive integer, and q indicates that the first vibration source and the second vibration source combine vibration number, n1≤ q, ω are indicated
Vibration frequency,Indicate the vibratory response size of the 1st vibrating sensor,Indicate the vibratory response of n-th of vibrating sensor
Size,Indicate the auto-power spectrum of the 1st vibrating sensor,Indicate n-th of vibrating sensor from power
Spectrum.
Further, the step S20 specifically:
By n1The vibratory response size of a vibrating sensor interrecord structure is denoted asAnd it calculates auto-power spectrum and isAnd as known vibratory response data;
By n2The vibratory response size of a vibrating sensor interrecord structure is denoted asAnd calculate auto-power spectrum
ForAnd as unknown vibratory response data;
Indicate n-th1The vibratory response size of a vibrating sensor,Indicate n-th1The vibration of+1 vibrating sensor
Response magnitude,Indicate the vibratory response size of n-th of vibrating sensor,Indicate n-th1A vibrating sensor
Auto-power spectrum,Indicate n-th1The auto-power spectrum of+1 vibrating sensor,Indicate n-th of vibration
The auto-power spectrum of dynamic sensor;Wherein n1For positive integer, n2For positive integer, and n=n1+n2。
Further, in the step S30, described to establish offset minimum binary training pattern specific as follows:
If initial characteristic data X0It is N × n1Tie up matrix, Y0It is N × n2Matrix is tieed up, that is, shares N number of sample pair, X0Middle sample
Feature is n1Dimension, Y0Middle sample characteristics are n2Dimension.And X and Y are initial data by standardization (subtracting mean value, divided by standard deviation etc.)
The data generated later.If first principal component axial vector of X and Y is respectivelyWith WithIt is
Unit vector, then byWithFirst pair of principal component of X and Y can be representedWithWherein
According to above it is assumed that the solution thought of CCA is to makeWithBetween correlation maximization, i.e.,The solution thought of PCA is to make respectivelyWithRespective variance is maximum, i.e., The solution thought of offset minimum binary is
It is expressed mathematically as:
It can be found out by the method for introducing Lagrange multiplierIt can finally solveIt is symmetrical matrix XTYYTX
The corresponding feature vector of maximum eigenvalue,It is YTXXTThe corresponding feature vector of the maximum eigenvalue of Y.It is finding out
It afterwards, can be in the hope of first pair of relevant principal component of X and YIt is as follows:
It can be X and Y respectively to their principal component according to principal component regression thoughtCarry out regression modeling:
HereIt is different fromBut there is certain relationship between them, and E, G are residual matrix.It utilizesBetween have correlation Y is changed to the principal component to XCarry out regression modeling:
For regression equation (3) (4) (5), can be calculated with least square methodIt is as follows:
It can be derived from the result (6) (7) (8) found outBetween relationship are as follows:
WhereinIt is that X is projected outDirection vector, andIn the case where returning thought (so that residual error E is as small as possible) according to most
What small square law was found out, be generally not identical relationship between the two.Later by principal component in XInexplicable residual error portion E
As new X, principal component in YInexplicable residual error portion F is returned as new Y according to the method for front, is recycled past
Again, until residual error F reaches required precision or principal component score has reached the upper limit (order of initial X), algorithm terminates.If last
K principal component is shared, then a series of vectors are represented by
Wherein subscript is differentBe it is mutually orthogonal,It is also orthogonal, butBe generally not it is orthogonal (this be also with
Different place in the expression formula of PCA).Original X, Y can finally be indicated are as follows:
It utilizesRelationship formula (10) (11) is write as matrix form:
X=TPT+E (12)
Y=TRT+ F=XWRT+ F=XA+F (13)
That is the regression equation of X → Y, wherein A=WRT.The W being calculated in algorithmic procedure, the value of R is gathered, just
It can use offset minimum binary to be predicted, i.e., for the data x newly inputted, calculate each principal component first with W,
I.e.Then it substitutes intoThe prediction of vector y can be found out
Value, or it is directly substituted into yT=xTA。
Further, the step S40 specifically:
Unknown vibratory response data are predicted using regression relation coefficient matrix and known vibratory response data,
In the corresponding environmental working condition to be measured of unknown vibratory response data it is identical as the loading position of history operating condition, and be uncorrelated steady
Random vibration.
The present invention has the advantages that
1, it is predicted by frequency domain of the partial least square model to unknown vibratory response, does not need the transmitting of identification system
Function, magnitude of load or load position greatly reduce the difficulty of prediction, and precision of prediction is high.
2, it is predicted, realized while predicted multiple unknown by frequency domain of the partial least square model to unknown vibratory response
The frequency domain vibratory response situation in oscillation point.
3, it is based on partial least square model, convenient for more to two groups of variable numbers and observe data there are multiple correlation
The less situation of sample size predicted.
Detailed description of the invention
The present invention is further illustrated in conjunction with the embodiments with reference to the accompanying drawings.
Fig. 1 is a kind of structural schematic diagram of the vibratory response frequency domain prediction device based on offset minimum binary of the present invention.
Fig. 2 is a kind of flow chart of the vibratory response frequency domain prediction method based on offset minimum binary of the present invention.
Fig. 3 is the response prediction result of first passage in the embodiment of the present invention two and the comparison figure of legitimate reading.
Fig. 4 is the response prediction result of second channel in the embodiment of the present invention two and the comparison figure of legitimate reading.
Fig. 5 is the overproof comparison of decibel of the response prediction result and legitimate reading of first passage in the embodiment of the present invention two
Figure.
Fig. 6 is the overproof comparison of decibel of the response prediction result and legitimate reading of second channel in the embodiment of the present invention two
Figure.
Description of symbols:
A kind of vibratory response frequency domain prediction device based on offset minimum binary of 100-, 1- shake table, 2- vibration source, 3- vibration
Sensor, 11- uniform disc, 12- supporting element, the first vibration source of 21-, the second vibration source of 22-.
Specific embodiment
It please refers to shown in Fig. 1 to Fig. 6, a kind of vibratory response frequency domain prediction device 100 based on offset minimum binary of the present invention
One of preferred embodiment, including a shake table 1, multiple vibration sources 2 and n vibrating sensor 3, wherein n is positive integer;Institute
It states vibration source 2 to connect with shake table 1, drives shake table 1 to vibrate by the vibration source 2, and then generate uncorrelated steady random
Vibration;Stationary signal is divided into that tight steady and width is steady, and tight smoothly condition is too stringent in the signal processing, impracticable, of the invention
Finger beam is steady, need to steadily meet three conditions: one is mean value is the constant being unrelated with the time, the second is square bounded, thirdly
It is that auto-correlation function is unrelated with the starting point of signal time, it is only related with the time difference;The variance of wide stationary signal and side be also
It is unrelated with the time;The vibrating sensor 3 is fixedly arranged on shake table 1, is recorded vibration source 2 by the vibrating sensor 3 and is existed
The vibration generated on shake table 1;The angle of the vibrating sensor immobilizes, and can reflect the main vibration of the shake table
It is dynamic;The position and direction that the vibration source group is vibrated every time immobilize.
The shake table 1 includes a uniform disc 11 and a supporting element 12;The center of the uniform disc 11 and support
Part 12 is affixed;The supporting element 12 is simply supported beam, has the advantages that damping ratio is small, can be considered linear structure.
The vibration source 2 includes one first vibration source 21 and one second vibration source 22;First vibration source 21 and vibration
The supporting element 12 of dynamic platform 1 connects, and first vibration source 21 drives the uniform disc 11 of shake table 1 to vibrate by supporting element 12;
Hammering uniform disc 11 by second vibration source 22 drives uniform disc 11 to vibrate;First vibration source 21 is preferably
For shake table excitation;Second vibration source 22 is preferably PCB power hammer, taps the homogeneous circle by PCB power hammer of manually taking
Disk 11 generates vibration.
A kind of one of preferred embodiment of the vibratory response frequency domain prediction method based on offset minimum binary of the present invention, including such as
Lower step:
Step S10, uncorrelated stationary random vibration is generated by multiple vibration sources, n vibrating sensor records shake table
Vibration data be historical data, vibration environment is recorded as history operating condition;
Step S20, by n1The vibration data of a vibrating sensor is as known vibratory response data, by n2A vibrating sensing
The vibration data of device is as unknown vibratory response data;Wherein n1For positive integer, n2For positive integer, and n=n1+n2;
Step S30, offset minimum binary training pattern is established, using vibratory response data known in historical data as partially minimum
Two multiply the input data of training pattern, using vibratory response data unknown in historical data as the defeated of offset minimum binary training pattern
Data out obtain the regression relation coefficient square in historical data between known vibratory response data and unknown vibratory response data
Battle array;
Step S40, using regression relation coefficient matrix and known vibratory response data to unknown vibratory response data into
Row prediction.
The step S10 specifically:
The m uncorrelated steady random vibrations that magnitude is gradually increased are generated by the first vibration source and the second vibration source joint
Dynamic load;
The vibratory response size of n vibrating sensor record shake table isAnd it calculates auto-power spectrum and isAnd vibratory response size and auto-power spectrum are denoted as historical data, vibration environment is recorded as history
Operating condition;
Wherein m is positive integer, and q indicates that the first vibration source and the second vibration source combine vibration number, n1≤ q, ω are indicated
Vibration frequency,Indicate the vibratory response size of the 1st vibrating sensor,Indicate the vibratory response of n-th of vibrating sensor
Size,Indicate the auto-power spectrum of the 1st vibrating sensor,Indicate n-th of vibrating sensor from power
Spectrum.
The step S20 specifically:
By n1The vibratory response size of a vibrating sensor record shake table is denoted asAnd it calculates auto-power spectrum and isAnd as known vibratory response data;
By n2The vibratory response size of a vibrating sensor record shake table is denoted asAnd it calculates from power
Spectrum isAnd as unknown vibratory response data;Power spectral density Sxx
(f): reflection correlation function expresses random signal itself and other signals in the inner link of different moments in time domain;When random
When signal mean value is zero, auto-correlation function and Power spectral density Fourier transform pair each other.
Indicate n-th1The vibratory response size of a vibrating sensor,Indicate n-th1The vibration of+1 vibrating sensor
Response magnitude,Indicate the vibratory response size of n-th of vibrating sensor,Indicate n-th1A vibrating sensor
Auto-power spectrum,Indicate n-th1The auto-power spectrum of+1 vibrating sensor,It indicates n-th
The auto-power spectrum of vibrating sensor;Wherein n1For positive integer, n2For positive integer, and n=n1+n2。
In the step S30, described to establish offset minimum binary training pattern specific as follows:
If initial characteristic data X0It is N × n1Tie up matrix, Y0It is N × n2Matrix is tieed up, that is, shares N number of sample pair, X0Middle sample
Feature is n1Dimension, Y0Middle sample characteristics are n2Dimension.And X and Y are initial data by standardization (subtracting mean value, divided by standard deviation etc.)
The data generated later.If first principal component axial vector of X and Y is respectivelyWith WithIt is
Unit vector, then byWithFirst pair of principal component of X and Y can be representedWithWherein
According to above it is assumed that the solution thought of CCA is to makeWithBetween correlation maximization, i.e.,The solution thought of PCA is to make respectivelyWithRespective variance is maximum, i.e., The solution thought of offset minimum binary isNumber
It is indicated on are as follows:
It can be found out by the method for introducing Lagrange multiplierIt can finally solveIt is symmetrical matrix XTYYTX
The corresponding feature vector of maximum eigenvalue,It is YTXXTThe corresponding feature vector of the maximum eigenvalue of Y.It is finding out
It afterwards, can be in the hope of first pair of relevant principal component of X and YIt is as follows:
It can be X and Y respectively to their principal component according to principal component regression thoughtCarry out regression modeling:
HereIt is different fromBut there is certain relationship between them, and E, G are residual matrix.It utilizesBetween have correlation Y is changed to the principal component to XCarry out regression modeling:
For regression equation (3) (4) (5), can be calculated with least square methodIt is as follows:
It can be derived from the result (6) (7) (8) found outBetween relationship are as follows:
WhereinIt is that X is projected outDirection vector, andIn the case where returning thought (so that residual error E is as small as possible) according to minimum
What square law was found out, be generally not identical relationship between the two.Later by principal component in XInexplicable residual error portion E makees
For new X, principal component in YInexplicable residual error portion F is returned as new Y according to the method for front, is recycled past
Again, until residual error F reaches required precision or principal component score has reached the upper limit (order of initial X), algorithm terminates.If last total
There is k principal component, then a series of vectors are represented by
Wherein subscript is differentBe it is mutually orthogonal,It is also orthogonal, butBe generally not it is orthogonal (this be also with
Different place in the expression formula of PCA).Original X, Y can finally be indicated are as follows:
It utilizesRelationship formula (10) (11) is write as matrix form:
X=TPT+E (12)
Y=TRT+ F=XWRT+ F=XA+F (13)
That is the regression equation of X → Y, wherein A=WRT.The W being calculated in algorithmic procedure, the value of R is gathered, just
It can use offset minimum binary to be predicted, i.e., for the data x newly inputted, calculate each principal component first with W,
I.e.Then it substitutes intoThe prediction of vector y can be found out
Value, or it is directly substituted into yT=xTA。
The step S40 specifically:
Unknown vibratory response data are predicted using regression relation coefficient matrix and known vibratory response data,
In the corresponding environmental working condition to be measured of unknown vibratory response data it is identical as the loading position of history operating condition, and be uncorrelated steady
Random vibration.
The applicable elements that partial least square model accordingly predicts multiple spot vibration are as follows:
1, system must be linearly invariant;
2, each point of load position under environmental working condition to be measured is constant, the load that each point of load applies be stationary random excitation and
It is irrelevant;
3, the position and direction of the point of load applied under history operating condition are identical as under environmental working condition to be measured, each load
The load that point applies is stationary random excitation and irrelevant;
4, the number of known measuring point is more than or equal to the number of the point of load, i.e. n1≥m;
5, p group independent experiment seeks known measuring point (n1) arrive unknown measuring point (n2) regression relation coefficient matrix D, and p >=
n1;
6, the vibratory response of multiple known measuring points under uncorrelated multi-source load excitation can be measured.
The evaluation index that multiple spot vibration is accordingly predicted based on partial least square model:
In order to verify the correctness and accuracy of prediction, need for prediction data to be compared with truthful data, due to this
Experimental data is the data of frequency domain, and the standard for industrially generalling use relative error 3dB carries out judging whether to comply with standard.Assuming that
R* is truthful data, and r is prediction data, then 3dB standard is as follows:
If inequality is set up, illustrate the relative error of prediction within 3dB, prediction is accurate;If inequality is not
It sets up, illustrates that the relative error of prediction is more than 3dB, prediction is inaccurate.
Other than industrial common 3dB standard, there are also the common error analysis evaluations such as MARE, SD and RMSE to refer to
Mark, calculation are as follows:
Wherein rkFor the value of k-th of component of true value r,For the estimated value of k-th of component of true value r.ekFor kth
The true value of a component and the relative error of predicted value,For the relative error mean value of true value and estimated value.Three above refers to
It is mathematically of equal value although mark calculation has difference.
The two of a kind of preferred embodiment of the vibratory response frequency domain prediction device and method based on offset minimum binary of the present invention,
Including an independent ball-type noise excitation source, an independent suspended type vibration platform, a shake table and 9 vibrating sensors;Institute
Vibrating sensor is stated to be fixedly arranged on shake table;The shake table is shaken by ball-type noise excitation source and suspended type vibration platform
It is dynamic;
There are 3 kinds of magnitude excitations in ball-type noise excitation source, and magnitude is gradually increased;The suspended type vibration platform has 5 kinds
Magnitude excitation, and magnitude is gradually increased;When noise excitation and vibrational excitation combination loading, the amount of noise excitation and vibrational excitation
Grade combination of two, forms 15 kinds of different magnitudes, to realize the complicated vibroacoustic environment of simulation, tests for response prediction
Research;Measure the exciting force of vibrational excitation, the exciting acceleration of vibrational excitation and swashing for acoustically-driven respectively by vibrating sensor
Vibration acoustic pressure, and response is measured by acceleration transducer, and record corresponding test result data.At this in 15 under operating condition,
P=14 group operating condition is chosen as historical data, i.e. the number p=14 of independent experiment, 1 group is surveyed for being used as under work condition environment to be measured
Examination uses.N=9 response measuring point, is grouped by the specific data that 15 groups of n=9 channels are collected by experiment first,
Choose n1Response data of the response data of=7 measuring points (vibrating sensor) as known measuring point, n2The response of=2 measuring points
Response data of the data as unknown measuring point.Data are frequency domain data, and each group of each channel sampled data of data is 1601
A, frequency values are 0Hz to 6.4KHz from low to high.
The relationship between known response and unknown response is trained with p=14 group, i.e., to two channels in 9 channels
Response is predicted that visible response prediction result substantially meets 3dB requirement compared with legitimate reading into attached drawing 6 from attached drawing 3.
In conclusion the present invention has the advantages that
1, it is predicted by frequency domain of the partial least square model to unknown vibratory response, does not need the transmitting of identification system
Function, magnitude of load or load position greatly reduce the difficulty of prediction, and precision of prediction is high.
2, it is predicted, realized while predicted multiple unknown by frequency domain of the partial least square model to unknown vibratory response
The frequency domain vibratory response situation in oscillation point.
3, it is based on partial least square model, convenient for more to two groups of variable numbers and observe data there are multiple correlation
The less situation of sample size predicted.
Although specific embodiments of the present invention have been described above, those familiar with the art should be managed
Solution, we are merely exemplary described specific embodiment, rather than for the restriction to the scope of the present invention, it is familiar with this
The technical staff in field should be covered of the invention according to modification and variation equivalent made by spirit of the invention
In scope of the claimed protection.
Claims (7)
1. a kind of vibratory response frequency domain prediction device based on offset minimum binary, it is characterised in that: including multiple vibration sources and
Multiple vibratory response sensors;The multiple vibration source is connect with shake table, is generated by the multiple vibration source uncorrelated flat
Steady random vibration;The multiple vibrating sensor is fixedly arranged in structure, is recorded multiple vibration sources by the vibrating sensor and is produced
Raw vibration.
2. the vibratory response frequency domain prediction device based on offset minimum binary as described in claim 1, it is characterised in that: the knot
Structure includes a uniform disc and a supporting element;The center of the uniform disc and supporting element are affixed.
3. the vibratory response frequency domain prediction device based on offset minimum binary as described in claim 1, it is characterised in that: described more
A vibration source includes one first vibration source and one second vibration source;First vibration source is connect with structure, passes through described
The vibration of one vibration source driving structure;The vibration of structure driving structure is hammered by second vibration source.
4. a kind of vibratory response frequency domain prediction method based on offset minimum binary, it is characterised in that: the method need to be used as weighed
Benefit requires 1 to 3 any prediction meanss, and described method includes following steps:
Step S10, uncorrelated stationary random vibration, the vibration data of n vibrating sensor record are generated by multiple vibration sources
For historical data, vibration environment is recorded as history operating condition;
Step S20, by n1The vibration data of a vibrating sensor is as known vibratory response data, by n2A vibrating sensor
Vibration data is as unknown vibratory response data;Wherein n1For positive integer, n2For positive integer, and n=n1+n2;
Step S30, offset minimum binary training pattern is established, using vibratory response data known in historical data as offset minimum binary
The input data of training pattern, using vibratory response data unknown in historical data as the output number of offset minimum binary training pattern
According to obtaining the regression relation coefficient matrix in historical data between known vibratory response data and unknown vibratory response data;
Step S40, unknown vibratory response data are carried out using regression relation coefficient matrix and known vibratory response data pre-
It surveys.
5. the vibratory response frequency domain prediction method based on offset minimum binary as claimed in claim 4, it is characterised in that: the step
Rapid S10 specifically:
The m uncorrelated stationary random vibrations that magnitude is gradually increased are generated by the first vibration source and the second vibration source joint
Load;
The vibratory response size of n vibrating sensor interrecord structure isAnd it calculates auto-power spectrum and is
And vibratory response size and auto-power spectrum are denoted as historical data, vibration environment is recorded as history operating condition;
Wherein m is positive integer, and q indicates that the first vibration source and the second vibration source combine vibration number, n1≤ q, ω indicate vibration frequency
Rate,Indicate the vibratory response size of the 1st vibrating sensor,Indicate the vibratory response size of n-th of vibrating sensor,Indicate the auto-power spectrum of the 1st vibrating sensor,Indicate the auto-power spectrum of n-th of vibrating sensor.
6. the vibratory response frequency domain prediction method based on offset minimum binary as claimed in claim 5, it is characterised in that: the step
Rapid S20 specifically:
By n1The vibratory response size of a vibrating sensor interrecord structure is denoted asAnd it calculates auto-power spectrum and isAnd as known vibratory response data;
By n2The vibratory response size of a vibrating sensor interrecord structure is denoted asAnd it calculates auto-power spectrum and isAnd as unknown vibratory response data;
Indicate n-th1The vibratory response size of a vibrating sensor,Indicate n-th1The vibratory response of+1 vibrating sensor
Size,Indicate the vibratory response size of n-th of vibrating sensor,Indicate n-th1A vibrating sensor from
Power spectrum,Indicate n-th1The auto-power spectrum of+1 vibrating sensor,Indicate n-th of vibration
The auto-power spectrum of sensor;Wherein n1For positive integer, n2For positive integer, and n=n1+n2。
7. the vibratory response frequency domain prediction method based on offset minimum binary as claimed in claim 4, it is characterised in that: the step
Rapid S40 specifically:
Unknown vibratory response data are predicted using regression relation coefficient matrix and known vibratory response data, wherein not
Know that the corresponding environmental working condition to be measured of vibratory response data is identical as the loading position of history operating condition, and is uncorrelated steady random
Vibration.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910661629.3A CN110398331A (en) | 2019-07-22 | 2019-07-22 | Vibratory response frequency domain prediction method and device based on offset minimum binary |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910661629.3A CN110398331A (en) | 2019-07-22 | 2019-07-22 | Vibratory response frequency domain prediction method and device based on offset minimum binary |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110398331A true CN110398331A (en) | 2019-11-01 |
Family
ID=68324873
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910661629.3A Pending CN110398331A (en) | 2019-07-22 | 2019-07-22 | Vibratory response frequency domain prediction method and device based on offset minimum binary |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110398331A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111220312A (en) * | 2020-02-28 | 2020-06-02 | 中国铁道科学研究院集团有限公司 | Bolt state diagnosis method and system |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104021395A (en) * | 2014-06-20 | 2014-09-03 | 华侨大学 | Target tracing algorithm based on high-order partial least square method |
CN107085633A (en) * | 2017-04-12 | 2017-08-22 | 华侨大学 | The device and method of multiple spot vibratory response frequency domain prediction based on SVMs |
CN109827777A (en) * | 2019-04-01 | 2019-05-31 | 哈尔滨理工大学 | Rolling bearing fault prediction technique based on Partial Least Squares extreme learning machine |
-
2019
- 2019-07-22 CN CN201910661629.3A patent/CN110398331A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104021395A (en) * | 2014-06-20 | 2014-09-03 | 华侨大学 | Target tracing algorithm based on high-order partial least square method |
CN107085633A (en) * | 2017-04-12 | 2017-08-22 | 华侨大学 | The device and method of multiple spot vibratory response frequency domain prediction based on SVMs |
CN109827777A (en) * | 2019-04-01 | 2019-05-31 | 哈尔滨理工大学 | Rolling bearing fault prediction technique based on Partial Least Squares extreme learning machine |
Non-Patent Citations (3)
Title |
---|
张洪等: "基于PLS的滚动轴承装配质量预测", 《控制工程》 * |
徐江敏等: "基于偏最小二乘回归的冲压件回弹预测研究", 《热加工工艺》 * |
李军等: "基于小波核偏最小二乘回归方法的混沌系统建模研究", 《物理学报》 * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111220312A (en) * | 2020-02-28 | 2020-06-02 | 中国铁道科学研究院集团有限公司 | Bolt state diagnosis method and system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107256204A (en) | The experimental provision and method of multiple spot vibratory response frequency domain prediction based on transmission function | |
US8393206B1 (en) | Dry wind tunnel system | |
US11237079B2 (en) | Device for analyzing sensitivity of object using frequency response and analyzing method using the same | |
Døssing | Structural testing | |
CN105159865B (en) | Experimental rig and its load recognition method that a kind of sound and vibration load joint applies | |
CN107092738B (en) | Experimental device and method for vibration response frequency domain prediction based on multiple linear regression | |
CN107085633A (en) | The device and method of multiple spot vibratory response frequency domain prediction based on SVMs | |
Lage et al. | Force magnitude reconstruction using the force transmissibility concept | |
CN110398331A (en) | Vibratory response frequency domain prediction method and device based on offset minimum binary | |
US6845671B2 (en) | Inverse method to estimate the properties of a flexural beam and the corresponding boundary parameters | |
Prato et al. | A reliable sampling method to reduce large sets of measurements: a case study on the calibration of digital 3-axis MEMS accelerometers | |
US7590495B2 (en) | Inverse method to calculate material properties using a non-resonant technique | |
Ligeikis et al. | Modeling and experimental validation of a pylon subassembly mockup with multiple nonlinearities | |
RU2658125C1 (en) | Method for determining parameters of natural tones of structure vibrations in resonant tests | |
Faircloth et al. | Experimental modal analysis of appropriate boundary conditions for the evaluation of cross-laminated timber panels for an in-line approach | |
Sadati et al. | Identification of a nonlinear joint in an elastic structure using optimum equivalent linear frequency response function | |
RU2715369C1 (en) | Method for determining partial frequencies of a controlled surface of an aircraft and device for its implementation | |
Peeters et al. | The performance of time domain system identification methods applied to operational data | |
Karle et al. | An analysis of transducer mass loading effect inshaker testing | |
JP2003161670A (en) | Evaluation method for response and feature for auxiliary vibration table | |
JP3240757B2 (en) | Structure vibration test apparatus and method | |
Ben-Haim | Identification of certain polynomial nonlinear structures by adaptive selectively-sensitive excitation | |
Lage et al. | A force identification approach for multiple-degree-of-freedom systems | |
JP4092878B2 (en) | Shaking table, control device therefor, and control method | |
Hung et al. | Identification of dynamic systems from data composed by combination of their response compoments |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20191101 |