CN109946085A - A kind of method of solid propellant rocket vibration signal sound and vibration noise reduction - Google Patents
A kind of method of solid propellant rocket vibration signal sound and vibration noise reduction Download PDFInfo
- Publication number
- CN109946085A CN109946085A CN201910224953.9A CN201910224953A CN109946085A CN 109946085 A CN109946085 A CN 109946085A CN 201910224953 A CN201910224953 A CN 201910224953A CN 109946085 A CN109946085 A CN 109946085A
- Authority
- CN
- China
- Prior art keywords
- signal
- vibration
- aic
- source
- noise reduction
- 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
Abstract
A kind of method that the present invention proposes solid propellant rocket vibration signal sound and vibration noise reduction, based on blind source separating thinking, it completes model foundation and FastICA-AIC new algorithm designs, and carrying test is carried out on solid engines, vibrating sensor and noise transducer are installed, the separation of signal of vibrating and noise signal is carried out.The present invention solves source signal and source signal number is unknown, and source signal with separate after signal number not necessarily under unanimous circumstances, the problem of sound and vibration noise reduction, has many advantages, such as intelligence, popularity, adaptability, engine condition diagnosis and monitoring can be more effectively carried out, has biggish help to measuring technology is promoted.
Description
Technical field
The invention belongs to Solid Rocket Engine Test observation and control technology field, relate generally in a kind of solid engines test
The noise reduction process of vibration signal sound and vibration removes noise from numerous signals mixed, by useful engine features Signal separator
Out, so as to carry out Digital Signal Analysis and Processing to it, the source signal of effective characterization engine luggine is obtained.
Background technique
Vibration and noise measurement is the main means of engine state monitor and fault diagnosis, studies and solve various vibrations
Problem is the important topic of current field of engineering technology.In engine test, vibration source is more, and Vibration propagation path is complicated, more
Kind vibration source signal mixes in different ways with noise, so that observation signal ingredient is sufficiently complex, and then influences special
The extraction of reference number brings difficulty to engine performance monitoring and fault diagnosis, and therefore, it is necessary to study solid engines test number
According to processing method, noise reduction process is carried out to measurement data, the ambient noise in separation test guarantees the confidence level of measurement data.
Existing vibration signal noise-reduction method has filtering, wavelet analysis, the methods of feature extraction, although can satisfy solid
Test System for Rocket Engine Test requirement, but traditional low-pass filtering can excessively inhibit noise, so that distorted signals, high-pass filtering meeting
So that ambient noise is reinforced simultaneously, small to the accuracy, reliability, error of data in solid engines test want is not met
It asks, and the vibration signal in strengthen the hair test has mutability feature, existing method is often at the noise reduction of steady segment signal
Reason method does not meet Solid Rocket Engine Test vibration signal characteristic.
Summary of the invention
In order to solve the problems existing in the prior art, the present invention proposes a kind of solid propellant rocket vibration signal sound and vibration noise reduction
Method, specially a kind of FastICA-AIC method adapted under solid engines experimental enviroment carries out signal of vibrating and noise
The separation of signal, and applied on the solid propellant rocket of different model, carry out effective Signal separator, it was demonstrated that
It applies the method in our solid propellant rocket tests, can more effectively carry out engine condition diagnosis and monitoring,
There is biggish help to measuring technology is promoted.
The present invention is based on blind source separating thinking, using fast vibration signal separation algorithm FastICA-AIC, from collecting
Mixing vibration signal in obtain independent vibration source feature, isolate mixed signal, obtain source signal, provided reliably for design side
True valid data.
For collected Vibration signal, vibration source signal number judgement is first carried out using AIC algorithm, according to anticipation
Obtained vibration source signal number k, with improved FastICA algorithm and improved learning function operation, to collected vibration
Signal carries out Signal separator, obtains separating resulting.
The technical solution of the present invention is as follows:
A kind of method of the solid propellant rocket vibration signal sound and vibration noise reduction, it is characterised in that: the following steps are included:
Step 1: utilizing M sensor, obtain Vibration signal X, the X=[x of solid propellant rocket1,...,xM]T；
Step 2: the vibration source signal number k in Vibration signal is judged using AIC algorithm；
Step 3: observation signal X being pre-processed, preprocessing process includes removing mean value, decorrelation and whitening processing；
Step 4: establishing a random initial weight vector WM×k；
Step 5: according to formula
WM×k=W*/||W*||
To weight vector WM×kIt is iterated calculating；Wherein E [] is mean operation, and g () is learning function；
Step 6: as weight vector WM×kAfter convergence, according to formula
Obtain k vibration source signal S=[s in observation signal1,...,sk]T。
Further preferred embodiment, a kind of method of solid propellant rocket vibration signal sound and vibration noise reduction, feature
Be: learning function uses g ()=asin ()+bcos (), and parameter a and b are all satisfied 1≤a, b≤2.
Further preferred embodiment, a kind of method of solid propellant rocket vibration signal sound and vibration noise reduction, feature
It is: takes a, b=1 in learning function.
Further preferred embodiment, a kind of method of solid propellant rocket vibration signal sound and vibration noise reduction, feature
Be: AIC algorithm judges the process of the vibration source signal number k in Vibration signal are as follows:
Step 2.1: EMD decomposition being carried out to Vibration signal X, obtains the intrinsic mode functions of each signal；
Step 2.2: by the intrinsic mode functions of each signal one group of new intrinsic mode functions of compound composition, and to new eigen mode
The Correlation Matrix of function carries out Eigenvalues Decomposition, obtains feature value vector λi；
Step 2.3: k being got into M from 1, substitutes into formula respectively
AIC value when k takes respective value is obtained, is believed by the vibration source in Vibration signal of the corresponding k value of minimum AIC value
Number number, wherein N is sample number.
Beneficial effect
The present invention, which has the special feature that, compared with prior art is: (1) designing the new calculation of ICA method and the identification of source signal number
Method: fast vibration signal separation algorithm FastICA-AIC；(2) " more to divide less " are realized, i.e., to the collected road M source signal point
Separate out k independent signal (wherein M < k)；(3) FastICA-AIC algorithm is applied in solid engines test, in conjunction with test
Data analyze the distribution character of signal of vibrating.
Additional aspect and advantage of the invention will be set forth in part in the description, and will partially become from the following description
Obviously, or practice through the invention is recognized.
Detailed description of the invention
Above-mentioned and/or additional aspect of the invention and advantage will become from the description of the embodiment in conjunction with the following figures
Obviously and it is readily appreciated that, in which:
Fig. 1: the blind source separating flow chart that the present invention chooses；
Fig. 2: what the present invention chose carries the source signal that hammer under test platform taps standard vibration sensor；
Fig. 3: the collected source signal of carrying test platform that the present invention chooses；
Fig. 4: what the present invention chose believes according to the random mixed four tunnels mixing of the carrying collected source signal of test platform
Number；
Fig. 5: the separation signal according to improvement FastICA algorithm solution after mixed that the present invention chooses；
Fig. 6: vacant vibration sensor signal in inventive engine test；
Fig. 7: the time domain vibration signal in inventive engine test when vibrating sensor package fire resisting clod, southern big glue；
Fig. 8: the A-frame of vibration, noise transducer is installed in inventive engine test；
Fig. 9: vibrating sensor mounting blocks in inventive engine test；
Figure 10: the 2nd and the 4th measuring point in inventive engine test；
Figure 11: the 3rd measuring point and the 5th measuring point in inventive engine test；
Figure 12-16: the present invention chooses 4 sensor signals of the first measuring point of first part, is calculated using FastICA-AIC
Method is separated, and the time-domain signal for separating front and back is as shown in figure 12, and the frequency-region signal for separating front and back is as shown in Figure 13 and Figure 14, point
It is as shown in Figure 15 and Figure 16 from the power spectrum of front and back.
Figure 17-21: the present invention chooses totally 8 vibration sensor signals that second part is the first and second measuring points, preceding 6 tunnel
For vibration signal, rear 2 tunnel is noise signal, is separated using FastICA-AIC algorithm, separates the time-domain signal of front and back as schemed
Shown in 17, the frequency-region signal for separating front and back is as shown in Figures 18 and 19, separates the power spectrum of front and back as shown in figs 20 and 21.
Figure 22-24: the vibration engine data for randomly choosing the test of certain model that the present invention chooses, the vibration to acquisition
Sensor signal has carried out blind source separating, and isolated blind source signal time-domain signal is illustrated in fig. 22 shown below, and separates the frequency domain letter of front and back
Number as shown in figs. 23 and 24.
Specific embodiment
The embodiment of the present invention is described below in detail, the embodiment is exemplary, it is intended to it is used to explain the present invention, and
It is not considered as limiting the invention.
The present invention mainly have passed through the following six stage in the course of the research:
Step 1: FastICA algorithm realizes process；
Step 2: carrying test platform, igniter shock signal, vibration are simulated by small vibrating platform and PULSE software
Steady segment signal, random signal, sinusoidal signal, square-wave signal etc., acquire valid data, mix unlike signal, mixed moment at random
Battle array is random to be generated, and mixing source is formed, and is used as judgment criteria by signal mean square error (SMSE) and time delay related coefficient (DDCC),
Signal blind source separating is carried out with FastICA algorithm；
Step 3: differentiate the collected valid data of carrying platform with AIC source signal number criterion,
It identifies source signal number, verifies AIC method feasibility；
Step 4: establishing the mathematical model of fast vibration signal separating method, design ICA method and source signal number are identified
New algorithm FastICA-AIC, and work out corresponding software, realize computing function；
Step 5: carrying engine test, the mixed signal of vibration and noise signal in certain model engine test is acquired
(observation signal) carries out a point discrete data with FastICA-AIC method and analyzes, the reliability of verifying this method in the application；
Step 6: the expansion of FastICA-AIC method is applied in other model engine tests, interpretation of result is carried out.
1) during for the first step, following methods are taken:
(1) blind source separating (BSS) --- FastICA implementation process
When engine test, vibration source wide variety, the source signals S such as engine ignition impact, ambient noise is by different
Propagation ducts are superimposed, and by installing sensor acquisition on the engine, obtain mixing observation signal X, see to mixing
It surveys signal and carries out whitening processing, separated using blind source separation method, the vibration source signal estimated after being separated.
Blind source separating process is as shown in Figure 1.
(2) FastICA algorithm is realized
The realization of FastICA algorithm needs to carry out BSS pretreatment, including removes mean value, decorrelation, whitening processing.
Blind source separating (BSS) preprocessing process: the application of blind source separation method has certain supposed premise, is based on each road
It is mutually indepedent between source signal, therefore decorrelation processing need to be carried out；In order to simplify algorithm complexity, it is assumed that mixed signal and source
Signal mean value is 0, so that the mean value of gaussian variable is 0, that is, carries out mean value and pre-processes；Equally, in order to improve the convergence of algorithm
Speed need to carry out whitening processing so that each component is uncorrelated.
BSS problem includes two layers of meaning: identifying source and source separation, the determination of source number is the key that blind source separating problem
Point is the problem that important step and the present invention before blind source separating is implemented focus on solving, in conjunction with engine test
The case where specific environment, the number of source signal can not artificially determine, establishes model in conjunction with ICA algorithm based on AIC criterion,
Design FastICA-AIC new algorithm.
2) during for second step, following method is taken:
Carry test platform:
Since in Solid Rocket Motor Ground Test, the noise of generation has the feature that range is wide, crest frequency is high,
In order to accurately obtain the signal data of noise and ambient noise, it is desirable that noise measuring system should have wider passband and
Biggish dynamic range.
It for the feasibility of verification algorithm, needs to carry test platform, acquires valid data.Currently, our test is flat
Platform is made of computer, PULSE noise and vibration multi-analyser system, power amplifier, vibration mechine and vibrating sensor.
Signal mean square error (SMSE) formula is as follows:
Time delay related coefficient (DDCC) formula is as follows:
Wherein: skFor source signal,To estimate that signal, K are source signal number accordingly, τ is time delay value.
3) during for third step, following methods is taken to carry out the judgement of signal source number:
The determination of source number is the key point of blind source separating problem, is the important step before blind source separating is implemented, and this
Invent the problem focused on solving.Because in conjunction with the specific environment of engine test, the type in source is more and of source signal
Number can not determine that the number of source signal is related to noise source number, if the number of source signal is k, be adjusted by model selection
Parameter preset completes the identification of source signal number with AIC criterion, with best illustration data but can include at least freely to join
Several models is sufficiently approved by academia.In order to avoid excessive de-noising, the selection of parameter preset k is critically important, if source signal
Number k preset big, although fitting result may show more excellent, be likely to occur excessive de-noising phenomenon.Source signal
Number k preset it is small, noise is retained it is excessive, it is as a result certainly bad.Complexity and this models fitting number of the AIC standard in model
According to Optimality tradeoff on fully consider, still using AIC standard selection source signal number k.AIC criterion is base
In the identifying source method that characteristic value is determined according to information theory criterion.Its basic thought is the cost letter for establishing dimensionality of signal space
Number, by obtaining the estimation of source number to cost function optimizing.
Akaike information theory criterion (AIC) is by minimizing following cost function preference pattern
Wherein, λiFor the characteristic value of observation signal, N is sample number, and when k is from 1 to m value, the smallest AIC (k) institute is right
What is answered is signal subspace dimension n, the i.e. identifying source of signal.
4) for taking following methods in the 4th step and six-step process:
The mathematical model of fast vibration signal separating method, design ICA method and source signal number are established according to the present invention
The new algorithm FastICA-AIC of identification, and work out corresponding software realization computing function；By designing targetedly test side
Case carries out carrying test on solid engines, installs vibrating sensor and noise transducer, acquires and vibrates in engine test
With the mixed signal (observation signal) of noise signal, a point discrete data is carried out with FastICA-AIC method and is analyzed, verifying we
The reliability of method in the application；It is applied on other different model engines, has carried out effective Signal separator, it was demonstrated that
It applies the method in our solid propellant rocket tests, can more effectively carry out engine condition diagnosis and monitoring.
FastICA separation algorithm described above is N number of signal to be isolated from N number of source signal, but answer engine is practical
In, collected observation signal all the way is mixed signal, includes a variety of source signals, so we need to propose a kind of few point
More FastICA-AIC separation algorithm, so that its actual conditions for meeting strengthen the hair test data, more has practicability.
By improving ICA algorithm, matrix dimension is changed to M*k by M*M dimension, to realize to the collected road M source signal
Isolate k independent signal (wherein M < k).
Independent component analysis (Independent Component Analysis；ICA), it is the independence in basis signal source
Property isolates each source signal from mixed signal, and ICA problem can be described simply are as follows: it is assumed that M sensor measures M observation
Signal X=[x1,...,xM]T, each observation signal is k Independent sources signal S=[s1,...,sk]TLinear hybrid, i.e.,
X=AS
Wherein A is the unknown hybrid matrix of M × k, and it is all unknown that ICA problem, which is exactly in source signal vector S and hybrid matrix A,
In the case where, it is desirable to a separation matrix W can be found, mutually independent source signal can be isolated from mixed signal, i.e.,
WTX=S
ICA solves the problems, such as that this one step of key is to establish the criterion that can measure separating resulting independence and corresponding separation
Algorithm, according to different independence criterions, ICA has different separation algorithms, and what application was wide at present is based on the fast of negentropy
Fast ICA algorithm, i.e. FastICA.
Because FastICA algorithm first discusses negentropy principle using negentropy maximum as a search direction.
From information theory theory: in the stochastic variable of the variances such as all, the entropy of gaussian variable is maximum, thus we can use entropy
Non-Gaussian system is measured, the amendment form of entropy, i.e. negentropy are commonly used.According to central-limit theorem, if a stochastic variable X is by many phases
Mutual independent stochastic variable Si(i=1,2,3 ... N) the sum of composition, as long as SiWith limited mean value and variance, though then its
For which kind of distribution, stochastic variable X is compared with SiCloser to Gaussian Profile.In other words, SiIt is stronger compared with the non-Gaussian system of X.Therefore, it is separating
In the process, the mutual independence can indicating separating resulting by the measurement of the non-Gaussian system between separating resulting, works as non-Gaussian system
When measurement reaches maximum, then show that the separation to each isolated component is completed.
The definition of negentropy:
Ng(Y)=H (YGauss)-H(Y)
In formula, YGaussBe one with Y have mutually homoscedastic Gaussian random variable, H () for stochastic variable differential entropy
Ng(Y)={ E [g (Y)]-E [g (YGauss)]}2
According to information theory, in having mutually homoscedastic stochastic variable, the stochastic variable of Gaussian Profile has maximum
Differential entropy.When Y has Gaussian Profile, Ng(Y)=0；The non-Gaussian system of Y is stronger, and differential entropy is smaller, Ng(Y) value is bigger, institute
With Ng(Y) it can be used as estimating for stochastic variable Y non-Gaussian system.
Wherein, E [] is mean operation；G () is Nonlinear Learning function, can use g1(y)=tanh (a1Or g y)2
(y)=yexp (- y2/ 2) or g3(y)=y3Equal nonlinear functions, according to the difference of signal waveform, functional form that g () is taken
It can affect to separating effect.Here, 1≤a1≤ 2, usual we take a1=1.Quick ICA learning rules are to look for one
A direction is so as to WTX (Y=WTX) there is maximum non-Gaussian system.It can be obtained by the iterative formula of FastICA algorithm after simplification:
W*=E { Xg (WTX)}-E{g'(WTX)}W
W=W*/||W*||
Wherein, the number of source signal is determined before carrying out blind source separating according to AIC criterion function, if in AIC standard
The number k of source signal is selected, then to this number k, calculates maximum likelihood function L ,-ln (L) is entropy at this time, according to-ln (L)
={ E [g (Y)]-E [g (YGauss)]}2Carry out next step calculating.
AIC=2k-2ln (L)
The improvement and application of learning function:
Generated vibration is random vibration in engine working process, by continuously distributed frequencies all in considered frequency band
Sinusoidal wave component in rate has nonlinear and nonstationary feature.In practical implementation, for the ease of analyzing and applying, often see
Statistical property as the vibration signal steadily traversed, i.e. engine does not change with time, and can use single sample
Time history indicates the statistical property of random vibration.But for the vibration signal with " noise " signal, due to " noise "
Influence, be likely to result in phase difference, include cosine signal, thus learning function selection SIN function combined with cosine function
Form, separating effect is best, from hereafter we analysis comparison in can be very good verifying this point.Select g (x)=asin
(x) when+bcos (x), here, 1≤a, b≤2, usual we take a, b=1 that can greatly shorten runing time, improve and calculate effect
Rate, runing time improves between 200 times to 100 times, so that single analysis time < 40s.
The different learning function runing times (s) of table 1
| Number of run/functional form | Tangent function | asin(x)+bcos(x) | Exponential function |
| 1 | 1263.467088 | 18.706969 | 31.538490 |
| 2 | 432.838639 | 10.863066 | 11.932021 |
| 3 | 145.981759 | 3.260864 | 9.390147 |
| 4 | 413.521170 | 2.537447 | 4.457204 |
| 5 | 167.566621 | 1.556501 | 3.371089 |
| 6 | 1416.71772 | 11.162860 | 9.538435 |
Based on the above principles, step of the present invention is obtained are as follows:
Step 1: utilizing M sensor, obtain Vibration signal X, the X=[x of solid propellant rocket1,...,xM]T。
Step 2: the vibration source signal number k in Vibration signal is judged using AIC algorithm；
Step 2.1: EMD decomposition being carried out to Vibration signal X, obtains the intrinsic mode functions of each signal；
Step 2.2: by the intrinsic mode functions of each signal one group of new intrinsic mode functions of compound composition, and to new eigen mode
The Correlation Matrix of function carries out Eigenvalues Decomposition, obtains feature value vector λi；
Step 2.3: k being got into M from 1, substitutes into formula respectively
AIC value when k takes respective value is obtained, is believed by the vibration source in Vibration signal of the corresponding k value of minimum AIC value
Number number, wherein N is sample number.
Step 3: observation signal X being pre-processed, preprocessing process includes removing mean value, decorrelation and whitening processing；
Step 4: establishing a random initial weight vector WM×k；
Step 5: according to formula
WM×k=W*/||W*‖
To weight vector WM×kIt is iterated calculating；Wherein E [] is mean operation, and g () is learning function；Wherein learn
Function uses g ()=asin ()+bcos (), and parameter a and b are all satisfied 1≤a, and b≤2 take a, b=1 here；
Step 6: as weight vector WM×kAfter convergence, according to formula
Obtain k vibration source signal S=[s in observation signal1,...,sk]T。
Fig. 3 is 3 source signals, and the 4 tunnel mixed signals generated by this group of source signal mixing by improving as shown in figure 4, calculate
What method was realized divides more results as shown in Figure 5 less.It can be apparent from by intuitive time domain signal waveform, the innovatory algorithm
Separating effect is fine, can clearly identify source signal shown in Fig. 3.
FastICA-AIC algorithm proposed by the present invention is applied in strengthen the hair test by we, is verified it and is applied in test
Feasibility.
Using the data of the shaking platform carried in laboratory, method validation is carried out.
Standard vibration sensor is installed on the vibration excitor of laboratory, during vibration excitor work, standard vibration is passed
Sensor additional external is excited by impact, i.e., taps standard vibration sensor with hammer, collected in this way by standard vibration sensor
Vibration data only contain and determine known two excitation source signals, as shown in Figure 2.With AIC algorithm to collected vibration
The AIC that dynamic data are handled exports result are as follows: AIC=[4.2947,0.0087,0.0098,0.0105,0.0115,
0.0122,0.0126], the results showed that there is minimum value in the 2nd data point [0.0087], according to the decision criteria of AIC algorithm,
So that the smallest number of AIC value is source signal number, i.e., the signal includes that there are two Independent sources signals, with actual conditions one
It causes.
Carrying vibrating sensor is given below and noise transducer carries out certain model Solid Rocket Motor Ground Test, wraps
Include three test preparation, test process and test result analysis parts.
A. test prepares
Test prepares to specifically include that the preparation of sensor, the selection of point position, pick up calibration and tooling preparation etc..
Two kinds of sensor is selected in this test, and one is vibrating sensor, another kind is noise transducer；Vibration
Sensor acquires the vibration signal in engine working process, the various voice signals during noise transducer acquisition test.
Influence for physics abatement engine operation noise to vibration signal, we prevent the partial vibration sensor used
Shield, means of defence include the big glue of package fire resisting clod and package south.
It can be seen that from the vacant vibration sensor signal of Fig. 6 i.e. by vibrating sensor far from engine, be mounted on such as Fig. 8
On bracket, vibration brought by igniting shock and ambient noise is only experienced, it is found that amplitude in ± 10g, shows that ambient noise is true
Vibration signal can be interfered in fact.
Fig. 7 is time domain vibration signal when vibrating sensor wraps up fire resisting clod, southern big glue, is shown by physical protection,
After part interference is fallen in isolation, amplitude decrease to some degree demonstrates again that ambient noise can interfere vibration signal.
But either installation noise transducer is analyzed, or carries out artificial physics protection, can all expend a large amount of people
Power, material resources and financial resources.Using FastICA-AIC algorithm of the present invention, noise reduction process can be carried out in later data processing, saved
Human and material resources improve test efficiency.
This test has selected five point positions, the first measuring point and the second measuring point in front skirt on the engine altogether, this two
A point position is 90 ° symmetrical；In the middle part of cylinder section, the two point positions are also 90 ° symmetrical for third measuring point and the 4th measuring point；Peace
When filling the vibrating sensor of this four measuring points, it is close to engine surface with vibrating sensor mounting blocks shown in Fig. 4；5th measuring point
Noise transducer all the way is only installed and installation direction is towards vibration source --- engine makes it far from from testing jig and engine, uses
Tripod shown in Fig. 8 is fixed.It draws and illustrates sensor fixing structure
B. process is tested
Test process is installed comprising sensor, line and protection and data acquisition etc..
The each measuring point of first four measuring point installs four sensors, and all in axial direction, four sensors are successively are as follows: vibrating sensing
Vibrating sensor, the noise transducer that device, the big glue of the vibrating sensor of fire resisting clod protection and south protect；2nd and the 4th measuring point passes
Sensor is as shown in Figure 10 after installing；3rd and the 5th measuring point sensor is as shown in figure 11 after installing；
All the sensors test process is working properly, and acquisition signal is normal.
C. test result analysis
Noise transducer acquires signal measurement unit and can be indicated with pa (Pa) or decibel (dB), vibrating sensor in test
Acquiring signal measurement unit is mostly acceleration (g) to indicate.In vibration analysis, it is also often used the unit decibel for indicating differential
(dB), for the ease of comparing vibration signal and noise signal in a magnitude, vibration and noise is believed by following equation (5)
Number it is separately converted to decibel (dB).
Wherein, VsdB is the decibel value after the conversion of vibrating sensor voltage value, and VndB is the conversion of noise transducer voltage value
Decibel value afterwards, V are the physical quantity voltage value of vibration and noise signal transducer acquisition, and Vr is that common voltage value 0.775V, ρ are
The conversion coefficient of corresponding noise transducer.In order to compare vibration and noise signal under same magnitude, we will according to formula (5)
The unit of vibration and noise signal is completely converted into decibel (dB).
Test data, which is divided into two parts, to be analyzed.
First part is 4 sensor signals of the first measuring point, is separated using FastICA-AIC algorithm, before separation
Time-domain signal afterwards is as shown in figure 12, and the frequency-region signal for separating front and back is as shown in Figure 13 and Figure 14, separates the power spectrum of front and back such as
Shown in Figure 15 and Figure 16.First three road signal be respectively vibrating sensor time-domain signal, fire resisting clod protection vibrating sensor when
The vibrating sensor time-domain signal of the big glue protection of domain signal and south, is finally noise transducer time-domain signal all the way.
By comparing amplitude under same frequency ingredient situation of change and whether there is or not the generations of new frequency content, it can be determined that work
Condition situation.
The noise signal frequency as can be seen that in Figure 13 (4) is compared from acquisition signal spectrum figure and the spectrogram for separating signal
Compose with spectrum signal in Figure 14 (2) from major frequency components to maximum value corresponding to frequency it is consistent, and each frequency distribution is consistent,
400Hz and 660Hz are concentrated on, spectrum signal basic one in the vibration signals spectrograph in Figure 13 (2) and Figure 13 (3) and Figure 14 (3)
It causes.
Power spectral density is to describe a kind of method of statistical property of the random vibration on frequency domain, it describes each of signal
Distribution situation of the power that frequency component is included on frequency domain, reflects the energy variation of signal.From acquisition power spectrum signal
With the comparison of splitting signal power spectrum as can be seen that the noise power spectrum in Figure 15 (4) and power spectrum signal base in Figure 16 (1)
This is consistent, and the vibration signal power spectrum in Figure 15 (2) and Figure 15 (3) and power spectrum signal in Figure 16 (2) are almost the same.Show to lead to
FastICA-AIC algorithm is crossed, can effectively be separated ambient noise signal.
Second part is totally 8 vibration sensor signals of the first and second measuring points, and preceding 6 tunnel is vibration signal, and rear 2 tunnel is
Noise signal, the time-domain signal for separating front and back is as shown in figure 17, and the frequency-region signal for separating front and back is as shown in Figures 18 and 19, before separation
Power spectrum afterwards is as shown in figs 20 and 21.
The noise signal frequency as can be seen that in Figure 18 (1) is compared from acquisition signal spectrum figure and the spectrogram for separating signal
Compose, Figure 18 (2) almost the same with spectrum signal in Figure 19 (2)) in vibration signals spectrograph and spectrum signal in Figure 19 (3) it is basic
Unanimously.Wherein, the major frequency components in Figure 19 (1) are the integral multiple of 50Hz, consistent with Hz noise.
It can be seen that the noise power in Figure 20 (1) from acquisition power spectrum signal and splitting signal power spectrum comparison
Compose, in vibration signal power spectrum and Figure 21 (3) in Figure 20 (2) spectrum signal base almost the same with spectrum signal in Figure 21 (2)
This is consistent.
It is tested by our carrying, demonstrates the feasibility of FastIAC-AIC algorithm, and then be generalized to different model
Engine Block Test occasion, it is not necessary to transfer manpower every time and carry out physical protection, be contribution of the invention.We randomly choose certain type
The vibration engine data of number test, have carried out blind source separating to the vibration sensor signal of acquisition, when isolated blind source signal
Domain signal is as shown in figure 22, separates the frequency-region signal of front and back as shown in figs. 23 and 24.
Although the embodiments of the present invention has been shown and described above, it is to be understood that above-described embodiment is example
Property, it is not considered as limiting the invention, those skilled in the art are not departing from the principle of the present invention and objective
In the case where can make changes, modifications, alterations, and variations to the above described embodiments within the scope of the invention.
Claims (4)
1. a kind of method of solid propellant rocket vibration signal sound and vibration noise reduction, it is characterised in that: the following steps are included:
Step 1: utilizing M sensor, obtain Vibration signal X, the X=[x of solid propellant rocket1,...,xM]T；
Step 2: the vibration source signal number k in Vibration signal is judged using AIC algorithm；
Step 3: observation signal X being pre-processed, preprocessing process includes removing mean value, decorrelation and whitening processing；
Step 4: establishing a random initial weight vector WM×k；
Step 5: according to formula
WM×k=W*/||W*||
To weight vector WM×kIt is iterated calculating；Wherein E [] is mean operation, and g () is learning function；
Step 6: as weight vector WM×kAfter convergence, according to formula
Obtain k vibration source signal S=[s in observation signal1,...,sk]T。
2. a kind of method of solid propellant rocket vibration signal sound and vibration noise reduction according to claim 1, it is characterised in that: learn
It practises function and uses g ()=asin ()+bcos (), and parameter a and b are all satisfied 1≤a, b≤2.
3. a kind of method of solid propellant rocket vibration signal sound and vibration noise reduction according to claim 2, it is characterised in that: learn
It practises in function and takes a, b=1.
4. a kind of method of solid propellant rocket vibration signal sound and vibration noise reduction according to claim 1, it is characterised in that:
AIC algorithm judges the process of the vibration source signal number k in Vibration signal are as follows:
Step 2.1: EMD decomposition being carried out to Vibration signal X, obtains the intrinsic mode functions of each signal；
Step 2.2: by the intrinsic mode functions of each signal one group of new intrinsic mode functions of compound composition, and to new intrinsic mode functions
Correlation Matrix carry out Eigenvalues Decomposition, obtain feature value vector λi；
Step 2.3: k being got into M from 1, substitutes into formula respectively
AIC value when k takes respective value is obtained, using the corresponding k value of minimum AIC value as the vibration source signal in Vibration signal
Number, wherein N is sample number.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910224953.9A CN109946085A (en) | 2019-03-24 | 2019-03-24 | A kind of method of solid propellant rocket vibration signal sound and vibration noise reduction |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910224953.9A CN109946085A (en) | 2019-03-24 | 2019-03-24 | A kind of method of solid propellant rocket vibration signal sound and vibration noise reduction |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109946085A true CN109946085A (en) | 2019-06-28 |
Family
ID=67011569
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910224953.9A Pending CN109946085A (en) | 2019-03-24 | 2019-03-24 | A kind of method of solid propellant rocket vibration signal sound and vibration noise reduction |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109946085A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112113784A (en) * | 2020-09-22 | 2020-12-22 | 天津大学 | Equipment state monitoring method based on equipment acoustic signals and EMD |
| CN112349292A (en) * | 2020-11-02 | 2021-02-09 | 深圳地平线机器人科技有限公司 | Signal separation method and device, computer readable storage medium, electronic device |
| CN113639945A (en) * | 2021-06-28 | 2021-11-12 | 上海宇航系统工程研究所 | Spacecraft random vibration test condition design method based on empirical mode decomposition |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102735937A (en) * | 2012-06-29 | 2012-10-17 | 中国科学院半导体研究所 | Method for measuring signal phase difference |
-
2019
- 2019-03-24 CN CN201910224953.9A patent/CN109946085A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102735937A (en) * | 2012-06-29 | 2012-10-17 | 中国科学院半导体研究所 | Method for measuring signal phase difference |
Non-Patent Citations (2)
| Title |
|---|
| 于欣欣: "交流电机变频调速系统振动信号的盲源分离", 《中国优秀硕士学位论文全文数据库 工程科技II辑》 * |
| 卫莹 等: "基于盲源分离方法的声振降噪算法研究", 《探索科学》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112113784A (en) * | 2020-09-22 | 2020-12-22 | 天津大学 | Equipment state monitoring method based on equipment acoustic signals and EMD |
| CN112349292A (en) * | 2020-11-02 | 2021-02-09 | 深圳地平线机器人科技有限公司 | Signal separation method and device, computer readable storage medium, electronic device |
| CN112349292B (en) * | 2020-11-02 | 2024-04-19 | 深圳地平线机器人科技有限公司 | Signal separation method and device, computer readable storage medium and electronic equipment |
| CN113639945A (en) * | 2021-06-28 | 2021-11-12 | 上海宇航系统工程研究所 | Spacecraft random vibration test condition design method based on empirical mode decomposition |
| CN113639945B (en) * | 2021-06-28 | 2024-02-09 | 上海宇航系统工程研究所 | Spacecraft random vibration test condition design method based on empirical mode decomposition |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Batel | Operational modal analysis-another way of doing modal testing | |
| CN109946085A (en) | A kind of method of solid propellant rocket vibration signal sound and vibration noise reduction | |
| Li et al. | A study of the noise from diesel engines using the independent component analysis | |
| Hermans et al. | Modal testing and analysis of structures under operational conditions: industrial applications | |
| Bonato et al. | Use of cross-time–frequency estimators for structural identification in non-stationary conditions and under unknown excitation | |
| CN104763532B (en) | The method and apparatus of internal combustion engine, preferably petrol engine pinking for identification | |
| Dayong et al. | Extraction of fault component from abnormal sound in diesel engines using acoustic signals | |
| CN110441018B (en) | Time-frequency analysis method for fire impact response data | |
| CN106771598B (en) | A kind of Adaptive spectra kurtosis signal processing method | |
| CN109684898A (en) | Aero-engine and its vibration signal blind separating method and device | |
| CN104406792A (en) | Solid rocket engine fault diagnosis method | |
| Yao et al. | An adaptive seismic signal denoising method based on variational mode decomposition | |
| CN111007390B (en) | Analog circuit fault diagnosis model based on algebraic method | |
| CN107843406B (en) | Cavity modal wave motion characteristic determination method based on pulse pressure correlation function | |
| CN109973251A (en) | A kind of blind source separation method for solid propellant rocket vibration signal noise reduction | |
| Li et al. | A spectrum detection approach for bearing fault signal based on spectral kurtosis | |
| KR100727488B1 (en) | Waveform synthesis methods for shock response spectrum over a short time interval, a digital recursive filter for a shock response history and its applications | |
| Drouet et al. | Experimental modeling of Wiener filters estimated on an operating diesel engine | |
| CN111323227A (en) | Method for extracting fault features of aeroengine rotor | |
| Siano et al. | A nonlinear black-box modeling method for knock detection in spark-ignition engines | |
| JP4996134B2 (en) | Natural vibration mode extraction method, natural vibration mode extraction apparatus, and natural vibration mode extraction program | |
| CN111931288B (en) | Method for measuring distribution similarity of time-frequency domain of fire impact response | |
| RU2331893C1 (en) | Method of discrete component separation in signal spectre and device for its implementation | |
| Li et al. | Pattern recognition on diesel engine working condition by using a novel methodology—Hilbert spectrum entropy | |
| Chauhan | Parameter estimation and signal processing techniques for operational modal analysis |
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: 20190628 |