CN104200065A - Mechanical vibration signal feature extraction method based on combination of stochastic resonance and kernel principal component analysis - Google Patents
Mechanical vibration signal feature extraction method based on combination of stochastic resonance and kernel principal component analysis Download PDFInfo
- Publication number
- CN104200065A CN104200065A CN201410392086.7A CN201410392086A CN104200065A CN 104200065 A CN104200065 A CN 104200065A CN 201410392086 A CN201410392086 A CN 201410392086A CN 104200065 A CN104200065 A CN 104200065A
- Authority
- CN
- China
- Prior art keywords
- signal
- feature
- vibration signal
- sigma
- index
- 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
Landscapes
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention relates to a mechanical vibration signal feature extraction method based on combination of stochastic resonance and kernel principal component analysis. The method comprises the steps that firstly, the stochastic resonance method is applied for conducting pretreatment on rotor oscillation original signals measured by a sensor, the signal periodicity is improved, and the oscillation signal to noise ratio is improved; then, a time domain feature set is extracted for the pretreated output signals; then the kernel principal component analysis method is adopted for conducting nonlinear feature transformation for the extracted time domain feature set, and therefore the final needed feature set is obtained. The method is applied to feature extraction and failure diagnosis of simulated failure of an engine rotor, the result shows that the feature set extracted through the method is of linear independence, the number of dimensions is smaller, the separability is higher, the precision and efficiency of the failure diagnosis can be effectively improved, and application in engineering practice is facilitated.
Description
Technical field
The invention belongs to mechanical fault diagnosis field, be specifically related to a kind of mechanical oscillation signal feature extracting method combining with core pivot element analysis based on accidental resonance.
Background technology
When the plant equipment recurring structure fault of aeroengine rotor one class, utilizing vibration signal to diagnose is a kind of effective means.But be subject to the impact of the factors such as environment and duty, rotor often bears comparatively complicated load, also there is comparatively serious coupling phenomenon in the vibration that each rotary part produces, in addition due to the restriction of sensor self condition of work, the vibration signal obtaining often exists certain noise.Under certain conditions, for example fault is early stage, and excessively strong noise tends to extraction and the fault of effect characteristics and screens.Therefore vibration signal is carried out to effective pre-service, improve signal to noise ratio (S/N ratio), the fault signature that strengthens signal is the focus of turbine rotator breakdown diagnostic field research.
Conventional vibration signal denoising method has the methods such as wavelet transformation, Blind Signal Separation, EMD decomposition, but the ultimate principle of said method is all from signal, to remove noise contribution, and due to actual vibration signal noise source more complicated, the frequency distribution of noise is also uncertain, and application said method can cause the loss of useful information to a certain extent.
Accidental resonance is a kind of new method of signal process field, and method different from the past is merely by eliminating and suppressing noise and improve signal to noise ratio (S/N ratio).Accidental resonance is the periodic signal that utilizes noise to strengthen to comprise in signal, reach the object that improves output signal-to-noise ratio, thereby avoided signal and noise frequency approach or the condition that signal is too faint under, can there is the situation of useful signal filtering in application filtering algorithm.
Feature extraction is one of basic problem of pattern-recognition, and extracting effective feature is the key that solves pattern recognition problem.And Fault Diagnosis of Aeroengines process is also the process of a pattern-recognition in essence, the diagnostic characteristic that constructs high-quality is also the key that improves its diagnosis efficiency and accuracy thereof.Therefore before carrying out fault diagnosis, must carry out feature extraction by certain method, obtain from different perspectives responsive, the most useful characteristic information, for example.Can, from the angle of Time-domain Statistics characteristic, calculate and extract its conventional Time-domain Statistics feature.But gained temporal signatures collection existence form is simpler, can not effectively reflects faint, nonlinear failure message, and often have the stronger deficiencies such as linear dependence.
Summary of the invention
The technical matters solving
For fear of the deficiencies in the prior art part, the present invention proposes a kind of mechanical oscillation signal feature extracting method combining with core pivot element analysis based on accidental resonance, the technology combining with core pivot element analysis by accidental resonance, realizes pre-service and the Nonlinear feature extraction of vibration signal.For the fault signature of rotor structure, extract and diagnosis, resulting feature set linear independence, dimension is lower, separability is better, has improved precision and the efficiency of rotor structure fault diagnosis.
Technical scheme
The mechanical oscillation signal feature extracting method combining with core pivot element analysis based on accidental resonance, is characterized in that step is as follows:
Step 1, employing scale transformation stochastic resonance method are done pre-service to original vibration signal, obtain pretreated vibration signal x (n):
Step 2, for pretreated output signal x (n), extract temporal signatures collection, obtain 14 Time-domain Statistics features, include 8 of dimension indexs: pt
1-pt
8, 6 of dimensionless indexs: pt
9-pt
14; Extraction formula is:
Average:
Root-mean-square value:
Root amplitude:
Absolute average:
Measure of skewness:
Kurtosis:
Variance:
peak-to-peak value: pt
8=max{x (n) }-min{x (n) }; Waveform index:
peak index:
Pulse index:
Nargin index:
Measure of skewness index:
Kurtosis index:
Wherein: x (n) is the pretreated vibration signal of accidental resonance, n is n data point in N data point;
Step 3, extracted temporal signatures collection is carried out to the eigentransformation of core pivot element analysis, thereby obtain final required feature set.
Described scale transformation stochastic resonance method is done pre-service to original vibration signal and is: according to frequency compression, than R, calculate double sampling frequency values fsr, obtain iterative computation step-length h=1/fsr, signal is inputted to bistable system, and the step-length h iteration of take obtains system output valve as the pretreated vibration signal x of accidental resonance (n).
The described eigentransformation that temporal signatures collection is carried out to core pivot element analysis is: using temporal signatures collection as original input space X
k, k=1,2 ..., M, M is number of samples, selected Nonlinear Mapping function phi: R
n→ F, N for input intrinsic dimensionality, by Nonlinear Mapping function phi by original input space X
kbe mapped to high-dimensional feature space F: Φ (X
k), k=1,2 ..., in M, then in this feature space, carry out linear pivot analysis, obtain the linear pivot of high-dimensional feature space, the nonlinear principal component that this linearity pivot is the original input space, is the final nonlinear characteristic of extracting gained.
Beneficial effect
A kind of mechanical oscillation signal feature extracting method combining with core pivot element analysis based on accidental resonance that the present invention proposes, first apply accidental resonance method the rotor oscillation original signal of sensor measurement is done to pre-service, strengthen the periodicity of signal, improve the signal to noise ratio (S/N ratio) of vibration signal; Then for pretreated output signal, extract temporal signatures collection; Again extracted temporal signatures centralized procurement is carried out to nonlinear characteristic conversion by core pivot element analysis method, thereby obtain final required feature set.The method is applied in the feature extraction and fault diagnosis of simulated failure of engine rotor, result shows that feature set linear independence, the dimension of the method extraction is lower, separability is stronger, can effectively promote precision and the efficiency of fault diagnosis, be convenient to apply in engineering practice.
Due to the method that the present invention has adopted accidental resonance to combine with core pivot element analysis in vibration signal characteristics extracts, the present invention has the following significant advantage that is different from classic method:
1) utilize accidental resonance can avoid useful feeble signal filtering, but can utilize noise to strengthen the periodic signal comprising in signal, reach the object that improves output signal-to-noise ratio, strengthens signal characteristic;
2) adopt kernel principal component analysis to carry out feature extraction, the nonlinear characteristic of energy better extract vibration signal, the linear dependence of elimination feature, reduces intrinsic dimensionality, improves the separability of feature;
3) gained feature of the present invention can effectively promote precision and the efficiency of vibrating failure diagnosis.
Accompanying drawing explanation
Fig. 1 is aeroengine rotor fault simulation experiment table
1-clutch shaft bearing, 2-the second bearing, 3-the first motor, 4-the second motor, 5-pedestal;
Fig. 2 is 2 dimension classifying qualities of primitive character
Fig. 3 is 2 dimension classifying qualities of feature after accidental resonance pre-service
Fig. 4 is 2 dimension classifying qualities of KPCA feature
Embodiment
Now in conjunction with the embodiments, the invention will be further described for accompanying drawing:
The step of the present embodiment is:
1) adopt accidental resonance method to do pre-service to the rotor original vibration signal of sensor measurement;
2) for pretreated output signal, extract temporal signatures collection;
3) again extracted temporal signatures centralized procurement is carried out to nonlinear characteristic conversion by core pivot element analysis method, thereby obtain final required feature set.
Adopt accidental resonance method to do pre-service to original vibration signal, improve Signal-to-Noise, strengthen feeble signal feature.Accidental resonance is a kind of new method of signal process field, and method different from the past is merely by eliminating and suppressing noise and improve signal to noise ratio (S/N ratio).Accidental resonance is the periodic signal that utilizes noise to strengthen to comprise in signal, reach the object that improves output signal-to-noise ratio, thereby avoided signal and noise frequency approach or the condition that signal is too faint under, can there is the situation of useful signal filtering in application filtering algorithm.
Be subject to the restriction of adiabatic approximation theory, classical Stochastic Resonance Theory is only applicable to the situation of signal frequency f < < 1, thereby has greatly restricted the range of application of accidental resonance.In order to solve the restriction of the large parameter signal condition of this class of rotor oscillation, need to adopt scale transformation stochastic resonance method, its basic thought is: define a frequency compression and compare R, according to R, calculate double sampling frequency values fsr, obtain thus iterative computation step-length h=1/fsr, original vibration signal is inputted to bistable system, with step-length h iterative computation, obtain system output valve, this output valve is original vibration signal through the pretreated output signal of accidental resonance.
According to the ultimate principle of scale transformation stochastic resonance, application genetic algorithm is carried out optimizing to frequency compression than R and systematic parameter b, guarantees to obtain optimum output effect.Accidental resonance preprocessing process based on genetic algorithm is as follows:
(1) plant group coding and initialization.Adopt binary coding method, accuracy value is as required determined code length.Span [the b of given b, R value
min, b
max] and [R
min, R
max].Set population quantity S, form at random initial population { b
iand { R
j(i, j=1,2 ..., S).
(2) individual decoding and fitness evaluation.Decoding obtains b, the R value of parent individuality, and carries out fitness evaluation.Here the fitness function of choosing is the signal to noise ratio snr of system output signal, and its computing formula is
S (f in formula
0) be that system is at frequency input signal f
0output power spectrum Y (the f at place
0) amplitude, ground unrest spectrum N (f
0) be Y (f
0) in frequency f
0mean value in the one section of frequency range in left and right, place.
(3) to colony select, crossover and mutation operation.
(4) by { b of colony after upgrading
iand { R
jrepeat the operation of (2)~(3) step, until meet the iterated conditional of setting or reach maximum iteration time, output obtains optimum solution, optimum solution is inputted to bistable system and carry out iterative computation and obtain system output, and this output valve is through the pretreated vibration signal of accidental resonance.
Said as follows for pretreated vibration signal extraction temporal signatures diversity method:
If be x (n) through the pretreated vibration signal of accidental resonance, total n data point, extracts 14 conventional Time-domain Statistics features.Time-domain Statistics feature includes two kinds of dimension and dimensionless.Wherein there are 8 of dimension indexs:
Pt
1-pt
8, 6 of dimensionless indexs: pt
9-pt
14.
(1) average
(2) root-mean-square value
Root-mean-square value refers to the root-mean-square value of signal in one-period.
(3) root amplitude
(4) absolute average
Absolute average refers to the absolute average amplitude of signal in one-period.
(5) measure of skewness
(6) kurtosis
(7) variance
Variance has been reacted the situation that signal fluctuates on average position.
(8) peak-to-peak value
Peak-to-peak value is the poor of maximal value and minimum value.
pt
8=max{x(n)}-min{x(n)} (8)
(9) waveform index
(10) peak index
(11) pulse index
(12) nargin index
(13) measure of skewness index
(14) kurtosis index
Choose above 14 characteristic quantities as temporal signatures collection.Gained temporal signatures collection is carried out the nonlinear characteristic conversion of core pivot element analysis again, thereby obtain final required feature set, method is as follows:
The temporal signatures collection that first two steps are obtained is as original input space X
k(k=1,2 ..., M, M is number of samples), selected Nonlinear Mapping function phi: R
n→ F (N for input intrinsic dimensionality, in the present invention, N is 14), by Nonlinear Mapping function phi by original input space X
kbe mapped to high-dimensional feature space F: Φ (X
k), k=1,2 ..., in M, then in this feature space, carry out linear pivot analysis, obtain the linear pivot of high-dimensional feature space, this linearity pivot is the nonlinear principal component of the original input space, also finally extracts the nonlinear characteristic of gained.
Feature extraction based on core pivot element analysis, concrete steps are as follows:
1) select suitable kernel function, calculate nuclear matrix K, and then acquisition is gone after average
2) right
computation of characteristic values and proper vector;
3) according to eigenwert accumulation contribution rate, be greater than 85%, extract maximum several eigenwert characteristic of correspondence vectors.
4) be calculated as follows the projection of mapping (enum) data in extracted proper vector direction, this projection inputs the nonlinear principal component of sample matrix X, is the nonlinear characteristic of final gained vibration signal.
In formula: V
kfor proper vector, the input vector that x is luv space, the mapping vector that Φ (x) is high-dimensional feature space,
for related coefficient, M is number of samples.
Example:
The present embodiment mainly verifies that the vibration performance extracting method combining with core pivot element analysis based on accidental resonance can improve feature discriminability, eliminates the linear dependence between feature, reduces intrinsic dimensionality, thereby can improve fault diagnosis precision and efficiency.As shown in Figure 1, its basic composition comprises aeroengine rotor fault simulation experiment table: pedestal 5, the first motor 3, the second motor 4, clutch shaft bearing 1, the second bearing 2, shaft coupling, wheel disc etc.Fig. 1 is the structural representation of testing table, wherein, and B
i(i=1 ..., 7) and be bearing seat, D
i(i=1 ..., 4) and be rotor disk, P
1, P
2, P
3be 3 vibration transducers, P
4for speed probe, J
1, J
2, J
3for shaft coupling.
The speed setting of the first motor 3 is 1500rpm, and the speed setting of the second motor 4 is 2400rpm, for the cooperative state of simulated engine HP&LP Rotor.P
4sensor collection be the tach signal of interior axle, P
1, P
2, P
3vibration transducer is electric vortex type displacement sensor, and sample frequency is 2000Hz.Model rotor is normal, uneven, rubbing and the loosening 200 groups of sample datas of totally 4 kinds of states of bearing seat, 50 groups of samples of every kind of state, and wherein 30 groups for training, and all the other 20 groups as test.Below P is chosen in checking
1the vibration signal of sensor measurement.
For ease of comparative descriptions, define following 3 kinds of different feature sets:
(1) feature set I: measure original vibration signal and directly calculate 14 temporal signatures composition characteristic collection I of gained.
(2) feature set II: measure original vibration signal and calculate again 14 temporal signatures composition characteristic collection II of gained after accidental resonance pre-service.
(3) feature set III: adopt KPCA method to convert and extract gained feature composition characteristic collection III feature set II, be greater than 85% by eigenwert accumulation contribution rate, can choose 2 feature composition characteristic collection III, in Table 1.
The parameter of table 1 KPCA feature extraction
Eigenwert sequence number | Eigenvalue λ i | Contribution rate (%) | Contribution rate of accumulative total (%) |
1 | 1.362 | 58.30 | 58.30 |
2 | 0.837 | 35.83 | 94.13 |
3 | 0.106 | 4.54 | 98.67 |
For above 3 category feature collection, the validity that this method is described from separability index and classification diagnosis performance two aspects of sample set.
First the separability of above 3 stack features sample sets is analyzed.Define separability parameter between average class:
Separability parameter is used for measuring the separability size of given feature samples collection.In formula, s is pattern class number; d
ijit is the inter-object distance between i quasi-mode and j quasi-mode; r
i, r
jbe respectively sample in i, j quasi-mode and, apart from the ultimate range of mode top, be the radius of the smallest sphere of holding this class sample.
Table 2 is the separability parameter of three kinds of feature samples collection.As can be seen from Table 2, accidental resonance pre-service and KPCA feature extraction all can effectively improve the separability of sample set.
The separability desired value of table 2 feature set
Feature set | Feature set I | Feature set II | Feature set III |
Separability index ρ | 0.536 | 0.857 | 0.931 |
In order to see intuitively the separability size of different characteristic sample set, feature samples is normalized, then project to 2 dimensional planes and show.Fig. 2, Fig. 3 and Fig. 4 are respectively primitive character sample set I, the drop shadow effect based on the pretreated feature samples collection of accidental resonance II, feature samples collection III based on accidental resonance pre-service and KPCA conversion at 2 dimensional planes.
As can be seen from Figure 2, four kinds of state overlap ratios are more serious, and separability is poor, and corresponding separability parameter value is just lower; And in Fig. 4, it is overlapping that three state there is no, separability is better, and therefore corresponding separability parameter value is just higher; In Fig. 3, the overlapping situation of three state and separability are slightly poor compared with Fig. 4, better than the result of Fig. 2.
Table 3 be the diagnostic accuracy of three kinds of feature sets and velocity ratio.Adopt 120 groups of data as training sample, other 80 groups of data are as test sample book.Sorter adopts one-to-many support vector machine classifier, and wherein, kernel function is selected radial basis function, σ=0.25, C=100.
Diagnostic accuracy and the speed of three kinds of feature sets of table 3
From the contrast of table 3, can find out, for same support vector machine diagnostic model, no matter feature set III is that precision or the speed of classification diagnosis is all better than feature set II and I, and the result of feature set II is better than feature set I.
These results suggest that it is all effective that accidental resonance pre-service and the conversion of KPCA nonlinear characteristic are extracted for vibration signal characteristics.Accidental resonance, by improving the signal to noise ratio (S/N ratio) of output signal, reaches the object that strengthens periodic signal feature, thereby is expected to obtain better feature extraction effect; The nonlinear characteristic information of KPCA eigentransformation energy better extract vibration signal, and can eliminate the correlativity between feature.Therefore, the Feature Extraction Technology that two kinds of methods are combined and obtained, just can guarantee gained feature set linear independence, dimension is lower, separability is better, finally can improve precision and the efficiency of engine rotor vibrating failure diagnosis.
Claims (3)
1. the mechanical oscillation signal feature extracting method combining with core pivot element analysis based on accidental resonance, is characterized in that step is as follows:
Step 1, employing scale transformation stochastic resonance method are done pre-service to original vibration signal, obtain pretreated vibration signal x (n):
Step 2, for pretreated output signal x (n), extract temporal signatures collection, obtain 14 Time-domain Statistics features, include 8 of dimension indexs: pt
1-pt
8, 6 of dimensionless indexs: pt
9-pt
14; Extraction formula is:
Average:
Root-mean-square value:
Root amplitude:
Absolute average:
Measure of skewness:
Kurtosis:
Variance:
peak-to-peak value: pt
8=max{x (n) }-min{x (n) }; Waveform index:
peak index:
Pulse index:
Nargin index:
Measure of skewness index:
Kurtosis index:
Wherein: x (n) is the pretreated vibration signal of accidental resonance, n is n data point in N data point;
Step 3, extracted temporal signatures collection is carried out to the eigentransformation of core pivot element analysis, thereby obtain final required feature set.
2. the mechanical oscillation signal feature extracting method combining with core pivot element analysis based on accidental resonance according to claim 1, it is characterized in that: described scale transformation stochastic resonance method is done pre-service to original vibration signal and is: according to frequency compression, than R, calculate double sampling frequency values fsr, obtain iterative computation step-length h=1/fsr, signal is inputted to bistable system, and the step-length h iteration of take obtains system output valve as the pretreated vibration signal x of accidental resonance (n).
3. the mechanical oscillation signal feature extracting method combining with core pivot element analysis based on accidental resonance according to claim 1, is characterized in that: the described eigentransformation that temporal signatures collection is carried out to core pivot element analysis is: using temporal signatures collection as original input space X
k, k=1,2 ..., M, M is number of samples, selected Nonlinear Mapping function phi: R
n→ F, N for input intrinsic dimensionality, by Nonlinear Mapping function phi by original input space X
kbe mapped to high-dimensional feature space F: Φ (X
k), k=1,2 ..., in M, then in this feature space, carry out linear pivot analysis, obtain the linear pivot of high-dimensional feature space, the nonlinear principal component that this linearity pivot is the original input space, is the final nonlinear characteristic of extracting gained.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410392086.7A CN104200065A (en) | 2014-08-11 | 2014-08-11 | Mechanical vibration signal feature extraction method based on combination of stochastic resonance and kernel principal component analysis |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410392086.7A CN104200065A (en) | 2014-08-11 | 2014-08-11 | Mechanical vibration signal feature extraction method based on combination of stochastic resonance and kernel principal component analysis |
Publications (1)
Publication Number | Publication Date |
---|---|
CN104200065A true CN104200065A (en) | 2014-12-10 |
Family
ID=52085358
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201410392086.7A Pending CN104200065A (en) | 2014-08-11 | 2014-08-11 | Mechanical vibration signal feature extraction method based on combination of stochastic resonance and kernel principal component analysis |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104200065A (en) |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105004498A (en) * | 2015-07-09 | 2015-10-28 | 西安理工大学 | Vibration fault diagnosis method of hydroelectric generating set |
CN105258789A (en) * | 2015-10-28 | 2016-01-20 | 徐州医学院 | Method and device for extracting vibration signal characteristic frequency band |
CN105628358A (en) * | 2015-12-28 | 2016-06-01 | 北华大学 | Centrifuge rotor fault diagnosis method based on parameter self-adaptive stochastic resonance |
CN105760839A (en) * | 2016-02-22 | 2016-07-13 | 重庆大学 | Bearing fault diagnosis method based on multi-feature manifold learning and support vector machine |
CN108151869A (en) * | 2017-11-27 | 2018-06-12 | 广州航新航空科技股份有限公司 | A kind of mechanical oscillation characteristic index extracting method, system and device |
CN110816588A (en) * | 2019-12-04 | 2020-02-21 | 中车株洲电力机车有限公司 | Locomotive wheel out-of-roundness detection method, device, equipment and system |
CN113075546A (en) * | 2021-03-24 | 2021-07-06 | 河南中烟工业有限责任公司 | Motor vibration signal feature extraction method and system |
CN113514144A (en) * | 2021-07-28 | 2021-10-19 | 郑州轻工业大学 | Unbalance-rubbing coupling fault detection method based on eddy current displacement sensor |
CN113822257A (en) * | 2021-11-24 | 2021-12-21 | 航天智控(北京)监测技术有限公司 | Abnormal point detection method based on combination of dimensionless features and virtual samples |
CN114184870A (en) * | 2021-12-14 | 2022-03-15 | 河北科技大学 | Non-invasive load identification method and equipment |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20090312979A1 (en) * | 2007-12-05 | 2009-12-17 | Min-Chun Pan | Adaptive Real-Time Analysis Method for Dynamic Signals Generated by Rotary Machines |
CN103714255A (en) * | 2013-12-30 | 2014-04-09 | 北京信息科技大学 | Failure prediction method based on nonlinear failure reconstruction |
-
2014
- 2014-08-11 CN CN201410392086.7A patent/CN104200065A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20090312979A1 (en) * | 2007-12-05 | 2009-12-17 | Min-Chun Pan | Adaptive Real-Time Analysis Method for Dynamic Signals Generated by Rotary Machines |
CN103714255A (en) * | 2013-12-30 | 2014-04-09 | 北京信息科技大学 | Failure prediction method based on nonlinear failure reconstruction |
Non-Patent Citations (3)
Title |
---|
BERNHARD SCHOLKOPF等: "Nonlinear component analysis as a kernel eigenvalue problem", 《NEURAL COMPUTATION》 * |
任立通等: "基于随机共振预处理的振动故障特征提取研究", 《振动与冲击》 * |
胡金海等: "粗糙核主元分析方法及其在故障特征提取中的应用", 《振动与冲击》 * |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105004498A (en) * | 2015-07-09 | 2015-10-28 | 西安理工大学 | Vibration fault diagnosis method of hydroelectric generating set |
CN105258789A (en) * | 2015-10-28 | 2016-01-20 | 徐州医学院 | Method and device for extracting vibration signal characteristic frequency band |
CN105258789B (en) * | 2015-10-28 | 2018-05-11 | 徐州医学院 | A kind of extracting method and device of vibration signal characteristics frequency band |
CN105628358A (en) * | 2015-12-28 | 2016-06-01 | 北华大学 | Centrifuge rotor fault diagnosis method based on parameter self-adaptive stochastic resonance |
CN105760839A (en) * | 2016-02-22 | 2016-07-13 | 重庆大学 | Bearing fault diagnosis method based on multi-feature manifold learning and support vector machine |
CN108151869B (en) * | 2017-11-27 | 2020-07-03 | 广州航新航空科技股份有限公司 | Mechanical vibration characteristic index extraction method, system and device |
CN108151869A (en) * | 2017-11-27 | 2018-06-12 | 广州航新航空科技股份有限公司 | A kind of mechanical oscillation characteristic index extracting method, system and device |
CN110816588A (en) * | 2019-12-04 | 2020-02-21 | 中车株洲电力机车有限公司 | Locomotive wheel out-of-roundness detection method, device, equipment and system |
CN113075546A (en) * | 2021-03-24 | 2021-07-06 | 河南中烟工业有限责任公司 | Motor vibration signal feature extraction method and system |
CN113514144A (en) * | 2021-07-28 | 2021-10-19 | 郑州轻工业大学 | Unbalance-rubbing coupling fault detection method based on eddy current displacement sensor |
CN113514144B (en) * | 2021-07-28 | 2022-07-26 | 郑州轻工业大学 | Unbalance-rubbing coupling fault detection method based on eddy current displacement sensor |
CN113822257A (en) * | 2021-11-24 | 2021-12-21 | 航天智控(北京)监测技术有限公司 | Abnormal point detection method based on combination of dimensionless features and virtual samples |
CN114184870A (en) * | 2021-12-14 | 2022-03-15 | 河北科技大学 | Non-invasive load identification method and equipment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104200065A (en) | Mechanical vibration signal feature extraction method based on combination of stochastic resonance and kernel principal component analysis | |
Liu et al. | Feature extraction of rotor fault based on EEMD and curve code | |
Han et al. | Intelligent fault diagnosis method for rotating machinery via dictionary learning and sparse representation-based classification | |
Qin et al. | The optimized deep belief networks with improved logistic sigmoid units and their application in fault diagnosis for planetary gearboxes of wind turbines | |
CN110792563B (en) | Wind turbine generator blade fault audio monitoring method based on convolution generation countermeasure network | |
Li et al. | Manifold sensing-based convolution sparse self-learning for defective bearing morphological feature extraction | |
CN103575523B (en) | The rotary machinery fault diagnosis method of kurtosis-envelope spectrum analysis is composed based on FastICA- | |
CN104748961A (en) | Gear fault diagnosis method based on SVD decomposition and noise reduction and correlation EEMD entropy features | |
CN105758644A (en) | Rolling bearing fault diagnosis method based on variation mode decomposition and permutation entropy | |
CN102944418B (en) | Wind turbine generator group blade fault diagnosis method | |
CN103245907B (en) | A kind of analog-circuit fault diagnosis method | |
CN103728551B (en) | A kind of analog-circuit fault diagnosis method based on cascade integrated classifier | |
CN105134619A (en) | Failure diagnosis and health evaluation method based on wavelet power, manifold dimension reduction and dynamic time warping | |
CN105928702B (en) | Variable working condition box bearing method for diagnosing faults based on form PCA | |
CN102156042A (en) | Gear fault diagnosis method based on signal multi-characteristic matching | |
CN105004498A (en) | Vibration fault diagnosis method of hydroelectric generating set | |
CN105678343A (en) | Adaptive-weighted-group-sparse-representation-based diagnosis method for noise abnormity of hydroelectric generating set | |
Pan et al. | Compressed sensing based on dictionary learning for reconstructing blade tip timing signals | |
Malik | Wavelet and Hilbert Huang transform based wind turbine imbalance fault classification model using k-nearest neighbour algorithm | |
CN107463872A (en) | A kind of rotating machinery Crack Fault Diagnosis in Shaft method | |
Zhang et al. | Research on the fault diagnosis method for rolling bearings based on improved VMD and automatic IMF acquisition | |
CN116861320A (en) | Rotor fault diagnosis method based on short-time Fourier synchronous compression transformation | |
CN103728135A (en) | Bearing fault feature extraction and diagnosis method of non-negative matrix factorization | |
Bonavita | On the limitations of data-driven weather forecasting models | |
Yang et al. | Change detection in rotational speed of industrial machinery using Bag-of-Words based feature extraction from vibration signals |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20141210 |