CN117116291A - Sound signal processing method of sand-containing water flow impulse turbine - Google Patents
Sound signal processing method of sand-containing water flow impulse turbine Download PDFInfo
- Publication number
- CN117116291A CN117116291A CN202311055700.6A CN202311055700A CN117116291A CN 117116291 A CN117116291 A CN 117116291A CN 202311055700 A CN202311055700 A CN 202311055700A CN 117116291 A CN117116291 A CN 117116291A
- Authority
- CN
- China
- Prior art keywords
- signal
- imf
- component
- noise
- wavelet
- 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
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 title claims abstract description 57
- 230000005236 sound signal Effects 0.000 title claims abstract description 36
- 239000004576 sand Substances 0.000 title claims abstract description 29
- 238000003672 processing method Methods 0.000 title claims abstract description 7
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 41
- 230000007704 transition Effects 0.000 claims abstract description 20
- 230000002159 abnormal effect Effects 0.000 claims abstract description 5
- 239000013049 sediment Substances 0.000 claims description 28
- 239000002245 particle Substances 0.000 claims description 18
- 238000004364 calculation method Methods 0.000 claims description 16
- 238000000034 method Methods 0.000 claims description 12
- 238000012545 processing Methods 0.000 claims description 12
- 238000005070 sampling Methods 0.000 claims description 3
- 238000012544 monitoring process Methods 0.000 abstract description 2
- 238000004088 simulation Methods 0.000 description 10
- 238000001228 spectrum Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 5
- 238000005299 abrasion Methods 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 4
- 238000009826 distribution Methods 0.000 description 3
- 230000003116 impacting effect Effects 0.000 description 3
- 238000011161 development Methods 0.000 description 2
- 230000003628 erosive effect Effects 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000000576 coating method Methods 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 230000003449 preventive effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 238000010183 spectrum analysis Methods 0.000 description 1
- 230000001131 transforming effect Effects 0.000 description 1
Classifications
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F03—MACHINES OR ENGINES FOR LIQUIDS; WIND, SPRING, OR WEIGHT MOTORS; PRODUCING MECHANICAL POWER OR A REACTIVE PROPULSIVE THRUST, NOT OTHERWISE PROVIDED FOR
- F03B—MACHINES OR ENGINES FOR LIQUIDS
- F03B11/00—Parts or details not provided for in, or of interest apart from, the preceding groups, e.g. wear-protection couplings, between turbine and generator
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L19/00—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis
- G10L19/02—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis using spectral analysis, e.g. transform vocoders or subband vocoders
- G10L19/0212—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis using spectral analysis, e.g. transform vocoders or subband vocoders using orthogonal transformation
- G10L19/0216—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis using spectral analysis, e.g. transform vocoders or subband vocoders using orthogonal transformation using wavelet decomposition
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/03—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters
- G10L25/06—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters the extracted parameters being correlation coefficients
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/03—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters
- G10L25/18—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters the extracted parameters being spectral information of each sub-band
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/27—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the analysis technique
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/48—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use
- G10L25/51—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use for comparison or discrimination
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Human Computer Interaction (AREA)
- Computational Linguistics (AREA)
- Signal Processing (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Acoustics & Sound (AREA)
- Multimedia (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Mechanical Engineering (AREA)
- General Engineering & Computer Science (AREA)
- Chemical & Material Sciences (AREA)
- Combustion & Propulsion (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The invention relates to an acoustic signal processing method of a sand-containing water flow impulse turbine, and belongs to the technical field of turbine state monitoring. Collecting sound signals at the rotating wheel blades of the water turbine set under different working conditions; ICEEMDAN decomposition is carried out on the original sound signal to obtain a series of natural mode functions; calculating the correlation coefficient and variance contribution rate of the IMF, and finding out a minimum value point; dividing the IMF into a high-frequency noise component, a transition signal component and a useful signal component by a minimum value point; carrying out wavelet threshold denoising on the transition signal component, and reconstructing the transition signal component and the useful signal component to obtain a noise reduction signal; performing Fourier transform on the noise reduction signal to obtain a spectrogram; calculating sample entropy characteristic values of noise reduction signals under different working conditions to obtain a reference value; and acquiring sound signals at the rotating wheel blades during running of the water turbine unit, calculating a sample entropy characteristic value after noise reduction of the signals to obtain an actual value, comparing the actual value with a reference value, and judging that abnormal shutdown and sand avoidance occur when the actual value is larger than the reference value.
Description
Technical Field
The invention belongs to the technical field of state monitoring of water turbines, and particularly relates to an acoustic signal processing method of a sand-containing water flow impulse water turbine.
Background
The high-speed development of the economy and society and the gradual expansion of industrial production make the demands of various industries on electric power energy become vigorous. The water and electricity is used as a green clean energy source, and plays an important role in meeting the environmental protection requirement and solving the shortage of power supply in China. In recent years, with the increasing capacity of hydropower installations, stable operation of units has received high attention. Most of hydropower stations in China are built on rivers, and when flood occurs, sediment content in the rivers increases rapidly, and part of sediment with larger particle size inevitably enters the water turbine along with water flow, so that the sediment becomes an important factor affecting safety of the hydroelectric generating set. The cavitation erosion degree is increased when the sand-carrying water flow causes serious abrasion of the water turbine, so that the running efficiency of the unit is greatly reduced, and irrecoverable safety accidents occur in serious cases. Therefore, it is necessary to study the sound signal of the sand-containing water flow impacting the turbine runner blade, analyze the spectral characteristics of the sand with different particle diameters, and take preventive measures in advance to avoid the occurrence of safety accidents.
Regarding the serious sediment erosion problem caused by the sand-containing water flow to the water turbine, the current research is mainly focused on the numerical simulation additional experiment, the relation between the abrasion rate and the water turbine material, the development of novel abrasion-resistant materials and coatings and the like, and the related analysis is less by utilizing the sound signals of the sand-containing water flow impacting the runner blades. When the sediment spectrum analysis is performed by utilizing the sound signal of the sediment water flow flowing through the turbine runner, the collected sound signal contains a large amount of background noise, and noise reduction is required to be performed on the original measured signal in order to obtain a purer useful signal. Because the sound of the sand-containing water flow flowing through the turbine runner is a nonlinear non-stationary signal, the traditional EMD decomposition method can generate serious modal aliasing phenomenon, and an ideal processed signal is difficult to obtain.
Disclosure of Invention
The invention aims to provide an acoustic signal processing method of a water flow impulse turbine containing sand, which solves the problems that ideal processed signals are difficult to obtain and the stop and sand avoidance revenues are difficult to obtain through the signals in the prior art.
In order to solve the technical problems, the invention adopts the following technical scheme: 1. the sound signal processing method of the sand-containing water flow impulse turbine is characterized by comprising the following steps of:
s1: collecting sound signals at the rotating wheel blades of the water turbine set under different working conditions;
s2: ICEEMDAN decomposition is carried out on the original sound signal to obtain a series of natural mode functions;
s3: calculating the correlation coefficient and variance contribution rate of the IMF, and finding out a minimum value point;
s4: dividing the IMF into a high-frequency noise component, a transition signal component and a useful signal component by a minimum value point;
s5: carrying out wavelet threshold denoising on the transition signal component, and reconstructing the transition signal component and the useful signal component to obtain a noise reduction signal;
s6: performing Fourier transform on the noise reduction signal to obtain a spectrogram;
s7: calculating sample entropy characteristic values of noise reduction signals under different working conditions to obtain a reference value;
s8: and acquiring sound signals at the rotating wheel blades when the water turbine unit operates, calculating a sample entropy characteristic value after noise reduction of the signals in an operating state according to the processing of the S2-S6, obtaining an actual value, comparing the actual value with a reference value, and judging that abnormal shutdown and sand avoidance occur when the actual value is larger than the reference value.
The further technical scheme is that in the step S1, the working conditions are respectively three types of sediment free, sediment particle size of 1-3mm and sediment particle size of 2-4mm, and sampling data are randomly acquired under different working conditions.
The further technical scheme is that the ICEEMDAN decomposition in step S2 is specifically:
s201: there is an original signal x to be decomposed, to which i sets of white noise w are added (i) To obtain
x (i) ＝x+β o E 1 (w (i) )
Wherein beta is 0 To add the ratio of the signal-to-noise ratio of noise relative to the original signal to the standard deviation of the added noise; w (w) (i) (i=1, 2,3, … …, i) is a series of gaussian white noise with a mean of 0 and a unit variance of 1;
s202: using EMD for signal x (i) Decomposing, calculating local mean value of the signal, thereby obtaining first order residual error r 1 And a first order modal component IMF 1 ：
r 1 ＝<M(x (i) )>
IMF 1 ＝x-r 1
Wherein M () is the local mean of the decomposed signal; the average value of all iteration times is calculated;
s203: adding white noise w to the first order residual (i) Calculate the second order residual r 2 And a second order modal component IMF 2 ：
r 2 ＝<M(r 1 +β 1 E 2 (w (i) ))>
IMF 2 ＝r 1 -r 2 ＝r 1 -<M(r 1 +β 1 E 2 (w (i) ))>
Wherein beta is 1 To add the ratio of the signal-to-noise ratio of noise relative to the original signal to the standard deviation of the added noise;
s204: and so on, calculate the kth order residual r k And a kth order modal component IMF k ：
r k ＝<M(r k-1 +β k-1 E k (w (i) ))>
IMF k ＝r k-1 -r k
Wherein r is k Is the k-th order residual; e (E) k () To pass byA kth IMF component after EMD decomposition;
s205: step S204 is repeated until all modal components are obtained, completing the signal decomposition.
In a further technical scheme, in step S3, the calculating of the correlation coefficient r and the variance contribution e (j) of the IMF is specifically:
the correlation coefficient expression:
wherein,is the average value of the signal x; />Is the average of the IMF components; n is the number of data points;
variance contribution rate expression:
wherein c j (i) For the j-th IMF component;
and drawing a graph according to the calculation result of the correlation coefficient and the variance contribution rate, and determining the minimum value point.
The further technical scheme is that the dividing standard in the step S4 is as follows: the IMF corresponding to the minimum value point is a transition signal component, the IMF corresponding to the minimum value point is a high-frequency noise component, and the IMF corresponding to the minimum value point is a useful signal component.
The further technical scheme is that the specific step of denoising the wavelet threshold value of the transition signal component in the step S5 is as follows:
s501: optimizing the wavelet base function and the decomposition layer number, selecting sym3-sym8 wavelet base function, wherein the decomposition layer number is 3-8, the threshold criterion is a Stein unbiased estimation threshold, and the threshold function is a soft threshold;
wherein the soft threshold function used is:
wherein w is j，k As a result of the initial wavelet coefficients,for wavelet coefficients processed by the threshold function, sgn () is a sign function;
s502: calculating Root Mean Square Error (RMSE), smoothness R and correlation coefficient R of a reconstructed signal under different wavelet base functions and decomposition layer numbers, finding the wavelet base function and the decomposition layer number corresponding to the minimum root mean square error and smoothness and the maximum correlation coefficient, and determining the optimal wavelet base and decomposition layer number; wherein,
root mean square error expression:
r(i)＝x(i)-x′(i)
the smoothness expression:
the correlation coefficient expression:
where x is the original signal; x' is the denoised reconstructed signal; n is the number of data points.
S503, performing discrete wavelet decomposition on the signal according to the optimal wavelet basis function and the decomposition layer number, and processing the decomposed wavelet coefficients based on a Stein unbiased estimation threshold criterion and a soft threshold function; and finally, reconstructing the signal to obtain a noise reduction signal.
A further technical solution is that the fourier transform in step S6 is as follows:
wherein,i represents an imaginary unit, X (j) is an input time domain signal, Y (k) is an output frequency domain signal, n is a signal length, j is a jth sample point of the time domain signal, and k is a kth frequency component of the frequency domain signal;
the further technical scheme is that the specific calculation steps of the sample entropy feature value in the step S7 are as follows:
s701: through the data sample { x (i), 1.ltoreq.i.ltoreq.N }, an m-dimensional vector is constructed:
X m (i)＝{x(i)，x(i+1)，x(i+m-1)}，1≤i≤N-m+1
s702: definition vector X m (i) And X is m (j) The absolute value of the maximum difference in the two corresponding elements is the distance d, i.e
d＝max k＝0，1，...，m-1 (|x(i+k)-x(j+k)|)
S703: number B of statistical distances d not greater than similarity threshold r i For 1.ltoreq.i.ltoreq.N-m, define:
s704: calculation of N-mMean>
S705: order theThe dimension is m+1, the steps S701-S704 are repeated, and the calculation is carried outAnd->
S706: calculating sample entropy eigenvalues SampEn (m, r, N), i.e
The invention achieves the beneficial technical effects that:
according to the invention, the method of combining ICEEMDAN and wavelet threshold denoising is utilized to denoise the sound signal of the water flow containing sand flowing through the water turbine, so that the problems of modal aliasing and pseudo-modal are solved, and meanwhile, the detail characteristics of the signal are kept as much as possible, and a good noise reduction effect is obtained; meanwhile, the optimal wavelet base and the decomposition layer number in the wavelet threshold parameters are selected according to the calculation result of the composite evaluation index (root mean square error, smoothness and correlation coefficient), so that the defect that the wavelet base and the decomposition layer number are selected by using artificial experience in the prior art is overcome, and the selection of the wavelet threshold parameters is more basis; finally, through carrying out frequency spectrum and sample entropy analysis on signals under different working conditions, the distribution condition of the particle size of the sediment in the river in the flood season can be judged, and a reference is provided for judging whether the water turbine needs to stop and avoid the sediment; when the actual value of the water turbine is larger than the reference value, the abnormal condition is judged to occur, and shutdown and sand avoidance are carried out.
Drawings
FIG. 1 is a schematic block diagram of the present invention;
FIG. 2 is a schematic diagram of a data acquisition platform in an embodiment;
FIG. 3 is a spectrum of a simulated signal and a clean signal based on a formula;
FIG. 4 is a three-dimensional waveform diagram of simulated signals decomposed by ICEEMDAN;
FIG. 5 is a graph showing the correlation coefficient and variance contribution rate calculation results of IMFs after the simulation signal is decomposed by ICEEMDAN;
FIG. 6 is a graph of the root mean square error, smoothness and correlation coefficient calculation results for a simulated signal;
FIG. 7 is a graph of the spectrum after noise reduction of the simulated signal;
FIG. 8 is a graph of sound signals based on three conditions measured by the experimental platform;
FIG. 9 is an exploded three-dimensional view of the non-sanded condition acoustic signal ICEEMDAN;
FIG. 10 is an exploded three-dimensional view of the 1-3mm silt condition acoustic signal ICEEMDAN;
FIG. 11 is an exploded three-dimensional view of the 2-4mm silt condition acoustic signal ICEEMDAN;
FIG. 12 is a graph showing the correlation coefficient and variance contribution of IMF after decomposition of the measured acoustic signal ICEEMDAN;
FIG. 13 is a time domain diagram of an actual measurement of the noise reduction of three operating mode sound signals through ICEEMDAN-wavelet thresholds;
FIG. 14 is a graph of the spectrum after noise reduction of the measured three operating mode signals;
FIG. 15 shows sample entropy features of the measured three operating mode signals.
Detailed Description
The invention will be further described with reference to the drawings and detailed description.
Example 1
The simulation part is only used for searching the optimal wavelet base and the decomposition layer number in the wavelet threshold parameter, and comparing the denoising effect of the ICEEMDAN, the wavelet threshold and the ICEEMDAN-wavelet threshold. Thus comprising in particular the following steps:
s1: simulating an original sound signal based on a simulation platform;
s2: ICEEMDAN decomposition is carried out on the original sound signal to obtain a series of Intrinsic Mode Functions (IMFs);
s3: calculating the correlation coefficient and variance contribution rate of the IMF, and finding out a minimum value point;
s4: dividing the IMF into a high-frequency noise component, a transition signal component and a useful signal component by a minimum value point;
s5: carrying out wavelet threshold denoising on the transition signal component, and reconstructing the transition signal component and the useful component to obtain a noise reduction signal;
s6: and carrying out Fourier transform on the time domain signal after noise reduction to obtain a spectrogram.
Further, the simulation of the original signal in step S1 is based on the following formula:
y 1 (t)＝10sin(4πt)+7sin(8πt)+5sin(12πt)
y(t)＝y 1 (t)+α(t)＝10sin(4πt)+7sin(8πt)+5sin(12πt)+α(t)
where t is time, sampling frequency 500HZ, 2500 sample data, y 1 (t) is a clean signal, alpha (t) is Gaussian white noise, and y (t) is a simulation signal. The spectrum of the clean signal, the simulation signal and the clean signal is shown in fig. 3.
Further, the ICEEMDAN algorithm in step S2 specifically includes:
s201: there is an original signal x to be decomposed, to which i sets of white noise w are added (i) To obtain
x (i) ＝x+β 0 E 1 (w (i) )
Wherein beta is 0 To add the ratio of the signal-to-noise ratio of noise relative to the original signal to the standard deviation of the added noise; w (w) (i) (i=1, 2,3, … …, i) is a series of gaussian white noise with a mean of 0 and a unit variance of 1;
s202: using EMD for signal x (i) Decomposing, calculating local mean value of the signal, thereby obtaining first order residual error r 1 And a first order modal component IMF 1 ：
r 1 ＝<M(x (i) )>
IMF 1 ＝x-r 1
Wherein M () is the local mean of the decomposed signal; the average value of all iteration times is calculated;
s203: adding white noise into the first order residual error, and calculating a second order residual error r 2 And a second order modal component IMF 2 ：
r 2 ＝<M(r 1 +β 1 E 2 (w (i) ))>
IMF 2 ＝r 1 -r 2 ＝r 1 -<M(r 1 +β 1 E 2 (w (i) ))>
S204: and so on, calculate the kth order residual r k And a kth order modal component IMF k ：
r k ＝<M(r k-1 +β k-1 E k (w (i) ))>
IMF k ＝r k-1 -r k
Wherein r is k Is the k-th order residual; e (E) k () Is the kth IMF component after EMD decomposition;
s205: and (4) repeating the step until all the modal components are obtained, and completing signal decomposition.
Further, after the simulated original signal is decomposed by ICEEMDAN, 8 IMF components are obtained, and a three-dimensional diagram waveform diagram of the simulated original signal is shown in FIG. 4.
Further, in step S3, the calculation of the correlation coefficient (r) and the variance contribution ratio (e (j)) of the IMF is specifically:
the correlation coefficient expression:
wherein,is the average value of the signal x; />Is the average of the IMF components; n is the number of data points.
Variance contribution rate expression:
wherein c j (i) Is the j-th IMF component.
Further, the correlation coefficient and variance contribution rate calculation result of each IMF after the simulation signal is decomposed by ICEEMDAN is shown in fig. 5, the minimum value appearing at the IMF3 can be determined, the minimum value is regarded as a transition signal component, wavelet threshold denoising is carried out, IMF1 and IMF2 are used as high-frequency noise components, the high-frequency noise components are discarded, and IMF4-IMF8 are useful signal components.
In step S5, when the wavelet threshold is denoising, in order to achieve the ideal noise reduction effect, optimizing the wavelet basis function and the decomposition layer number, selecting sym3-sym8 wavelet basis function, wherein the decomposition layer number is 3-8, the threshold criterion is a Stein unbiased estimation threshold, and the threshold function is a soft threshold; and calculating Root Mean Square Error (RMSE), smoothness (R) and correlation coefficient (R) under different wavelet base functions and decomposition layer numbers, finding out the wavelet base functions and the decomposition layer numbers corresponding to the minimum RMSE and the minimum Rmax and further determining the optimal wavelet base and the optimal decomposition layer number.
Further, the soft threshold function used is:
wherein w is j，k As a result of the initial wavelet coefficients,for wavelet coefficients subjected to threshold function processing, sgn () is a sign function.
Further, the root mean square error, smoothness and correlation coefficient are specifically:
the root mean square error expression is:
T(i)＝x(i)-x′(i)
the smoothness expression is:
the correlation coefficient expression is:
where x is the original signal; x' is the denoised reconstructed signal; n is the number of data points.
Further, the Root Mean Square Error (RMSE), smoothness (R) and correlation coefficient (R) of the reconstructed signal at different wavelet base functions and decomposition levels are calculated as shown in fig. 6.
Further, when the wavelet base function is sym6 and the decomposition layer number is 5, the Root Mean Square Error (RMSE) and the value of the smoothness (R) are minimum, and the correlation coefficient (R) is maximum, so that the determined optimal wavelet base function and decomposition layer number are sym6 and 5 layers, respectively.
Further, in order to show the superiority of ICEENMDAN-wavelet threshold joint noise reduction, ICEEMDAN denoising, wavelet threshold denoising and ICEEMDAN-wavelet threshold joint denoising are carried out on the original simulation signals, a wavelet base in wavelet threshold parameters is selected to be sym6, and the decomposition layer number is selected to be 5. The root mean square error, smoothness and correlation coefficient of the reconstructed signal after noise reduction by three methods were calculated as shown in table 1 below:
TABLE 1
| Root mean square error | Smoothness degree | Correlation coefficient | |
| ICEEMDAN denoising | 1.6809 | 1.4121 | 0.9838 |
| Wavelet threshold denoising | 1.0817 | 1.1377 | 0.9939 |
| ICEEMDAN-wavelet threshold denoising | 0.8554 | 1.1114 | 0.9958 |
Further, as can be seen from table 1, the root mean square error and smoothness of the combined noise reduction of the icemdan-wavelet threshold are the smallest, and the correlation coefficient is the largest, which indicates that the combined noise reduction effect is superior to that of either alone.
Further, in step S6, the fourier transform is as follows:
wherein,i represents an imaginary unit, X (j) is an input time domain signal, Y (k) is an output frequency domain signal, n is a signal length, j is a jth sample point of the time domain signal, and k is a kth frequency component of the frequency domain signal;
further, the transition signal component is subjected to ICEEMDAN-wavelet threshold denoising, and the noise reduction signal spectrum is obtained through Fourier transform, as shown in FIG. 7. As can be seen from fig. 7, the noise signal with frequency higher than 20Hz can be effectively filtered, which illustrates that the icemdan-wavelet threshold joint noise reduction method is effective in the field of noise removal of the sound signal.
Example 2
The effectiveness of ICEEMDAN-wavelet threshold combined noise reduction is verified based on the simulation signals, and sound signals of the non-sand-containing working condition, the 1-3mm sediment working condition and the 2-4mm sediment working condition, which are measured through the experimental platform of fig. 2, are shown in fig. 8.
Further, when the above experimental platform is used to collect sound data: the hydroelectric generating set is 5kW, the particle size of sediment in water is 1-3mm and 2-4mm respectively, and the measuring frequency range of the sensor is 10-20000Hz.
Further, ICEEMDAN decomposition is performed on the sound signals under three working conditions, and 10 IMF components are obtained for each working condition, and the first eighth-order IMF components are shown in FIG. 9, FIG. 10 and FIG. 11.
Further, the calculation results of the correlation coefficient and variance contribution rate of each IMF after the actual measurement signal is decomposed by the icemdan are shown in fig. 12, it can be determined that an extremely small value appears at the IMF4, the extremely small value is regarded as a transition signal component, wavelet threshold denoising is performed, IMFs 1 to 3 are used as high-frequency noise components, and IMFs 5 to 8 are used as useful signal components.
Further, the transition signal component is subjected to sym6 and 5-layer wavelet threshold denoising, and reconstructed together with dominant mode components IMF5 and IMF6 in the useful signal component, so as to obtain a noise reduction signal, as shown in fig. 13.
Further, fourier transforming the noise reduction signal to obtain spectrums of three working conditions is shown in fig. 14. As can be seen from fig. 14, the high frequency part of the signal is smoother, which means that the high frequency background noise caused by the pressure pump, the submersible pump and the like can be effectively filtered, and the feasibility of the ICEEMDAN-wavelet threshold combined noise reduction method is verified.
Further, as can be seen from fig. 14, in the three working conditions, there is a frequency component of 59.39Hz, which indicates that a natural frequency is generated when water flows through the hydroelectric generating set.
Furthermore, in the working condition without sand, the water flow motion state is stable, so the sound signal mainly comprises the vibration frequency 59.39Hz of the water flow impacting the runner blade; in the working condition of 1-3mm sediment, main frequency components higher than 59.39Hz appear, because sediment contained in water can impact the runner blades along with the water flowing, and then sound signals with higher frequency are excited; compared with the 1-3mm sediment working condition, more high-frequency components appear in the sound signal of the 2-4mm sediment working condition, because the inertia of the large-particle-size sediment particles is larger, the impact force generated on the blades along with the water flow is larger, and therefore the sound signal of more high-frequency components is formed.
Further, based on the analysis, when the water flow containing silt with different particle sizes impacts the turbine runner blade, the generated sound signal spectrum characteristics can change along with the difference of the sand-carrying particle sizes. The sand-containing water flow can generate more high-frequency components relative to the water flow without sand, and the larger the particle size of the sand, the more high-frequency components are formed, and according to the characteristics, the distribution of the particle size of the sand in the sand-containing water flow can be judged, and the reference is provided for stopping the water turbine in flood season or avoiding sand.
Further, the sample entropy feature value of the noise reduction signal is calculated as shown in fig. 15 as a reference value. The specific calculation steps are as follows:
s701: through the data sample { x (i), 1.ltoreq.i.ltoreq.N }, an m-dimensional vector is constructed:
X m (i)＝{x(i)，x(i+1)，x(i+m-1)}，1≤i≤N-m+1
s702: definition vector X m (i) And X is m (j) The absolute value of the maximum difference in the two corresponding elements is the distance d, i.e
d＝max k＝0，1，...，m-1 (|x(i+k)-x(j+k)|)
S703: number B of statistical distances d not greater than similarity threshold r i For 1.ltoreq.i.ltoreq.N-m, define:
s704: calculation of N-mMean>
S705: let the dimension be m+1, repeat steps S701-S704, calculateAnd->
S706: calculating sample entropy eigenvalues SampEn (m, r, N), i.e
As can be seen from fig. 15, the sample entropy is the smallest in the non-sandy condition, and increases from the 1-3mm silt condition to the 2-4mm silt condition. This is because the motion of the water flow without sand is relatively simple and the complexity of generating sound signals when flowing through the turbine is low; and the sediment particles in the sand-containing water flow impact the runner blades to form sound signals of high-frequency components, and along with the increase of the diameter of the sediment particles, the inertia of the sediment particles is enhanced, the impact force on the blades is larger, so that the formed high-frequency sound components are more, and the complexity of the sound signals is increased by the high-frequency components.
Further, based on the above, the sample entropy values of the sound signals are different when the water turbine runs under three working conditions, and the complexity of the water flow is judged according to the sample entropy, so that the analysis of the distribution condition of the particle size of the sediment in the sediment-containing water flow is facilitated, and further the serious abrasion of the sediment with large particle size to the water turbine in flood season is avoided. Specifically, the sound signals of the runner blades during the running of the water turbine unit are collected, the sample entropy characteristic values of the signals after noise reduction in the running state are calculated after the sound signals are processed according to the steps, the actual values are obtained, the actual values are compared with the reference values, and when the actual values are larger than the reference values, the abnormal shutdown and sand avoidance are judged.
The description of the above embodiments is only intended to aid in the understanding of the invention and is not intended to be limiting. It should be noted that it will be apparent to those skilled in the art that the present invention may be modified and practiced without departing from the spirit of the present invention.
Claims (8)
1. The sound signal processing method of the sand-containing water flow impulse turbine is characterized by comprising the following steps of:
s1: collecting sound signals at the rotating wheel blades of the water turbine set under different working conditions;
s2: ICEEMDAN decomposition is carried out on the original sound signal to obtain a series of natural mode functions;
s3: calculating the correlation coefficient and variance contribution rate of the IMF, and finding out a minimum value point;
s4: dividing the IMF into a high-frequency noise component, a transition signal component and a useful signal component by a minimum value point;
s5: carrying out wavelet threshold denoising on the transition signal component, and reconstructing the transition signal component and the useful signal component to obtain a noise reduction signal;
s6: performing Fourier transform on the noise reduction signal to obtain a spectrogram;
s7: calculating sample entropy characteristic values of noise reduction signals under different working conditions to obtain a reference value;
s8: and acquiring sound signals at the rotating wheel blades when the water turbine unit operates, calculating a sample entropy characteristic value after noise reduction of the signals in an operating state according to the processing of the S2-S6, obtaining an actual value, comparing the actual value with a reference value, and judging that abnormal shutdown and sand avoidance occur when the actual value is larger than the reference value.
2. The method for processing the acoustic signal of the sandy water flow impulse turbine as claimed in claim 1, wherein: in the step S1, the working conditions are respectively three types of sediment free, sediment particle size of 1-3mm and sediment particle size of 2-4mm, and sampling data are randomly acquired under different working conditions.
3. The method for processing the acoustic signal of the sandy water flow impulse turbine as claimed in claim 1, wherein: in step S2, the icemdan decomposition is specifically:
s201: there is an original signal x to be decomposed, to which i sets of white noise w are added (i) To obtain
x (i) ＝x+β 0 E 1 (w (i) )
Wherein beta is 0 To add the ratio of the signal-to-noise ratio of noise relative to the original signal to the standard deviation of the added noise; w (w) (i) (i=1, 2,3, … …, i) is a series of gaussian white noise with a mean of 0 and a unit variance of 1;
s202: using EMD for signal x (i) Decomposing, calculating local mean value of the signal, thereby obtaining first order residual error r 1 And a first order modal component IMF 1 ：
r 1 ＝<M(x (i) )>
IMF 1 ＝x-r 1
Wherein M () is the local mean of the decomposed signal; the < > is to calculate the average value of all iteration times;
s203: adding white noise w to the first order residual (i) Calculate the second order residual r 2 And a second order modal component IMF 2 ：
r 2 ＝<M(r 1 +β 1 E 2 (w (i) ))>
IMF 2 ＝r 1 -r 2 ＝r 1 -<M(r 1 +β 1 E 2 (w (i) ))>
Wherein, gamma 1 To add the ratio of the signal-to-noise ratio of noise relative to the original signal to the standard deviation of the added noise;
s204: and so on, calculate the kth order residual r k And a kth order modal component IMF k ：
r k ＝<M(r k-1 +β k-1 E k (w (i) ))>
IMF k ＝r k-1 -r k
Wherein r is k Is the k-th order residual; e (E) k () Is the kth IMF component after EMD decomposition;
s205: step S204 is repeated until all modal components are obtained, completing the signal decomposition.
4. The method for processing the acoustic signal of the sandy water flow impulse turbine as claimed in claim 1, wherein: in step S3, the calculation of the correlation coefficient r and the variance contribution e (j) of the IMF is specifically:
the correlation coefficient expression:
wherein,is the average value of the signal x; />Is the average of the IMF components; n is the number of data points;
variance contribution rate expression:
wherein c j (i) For the j-th IMF component;
and drawing a graph according to the calculation result of the correlation coefficient and the variance contribution rate, and determining the minimum value point.
5. The method for processing the acoustic signal of the sandy water flow impulse turbine as claimed in claim 1, wherein: the division criteria in step S4 are: the IMF corresponding to the minimum value point is a transition signal component, the IMF corresponding to the minimum value point is a high-frequency noise component, and the IMF corresponding to the minimum value point is a useful signal component.
6. The method for processing the acoustic signal of the sandy water flow impulse turbine as claimed in claim 1, wherein: the specific step of wavelet threshold denoising for the transition signal component in step S5 is as follows:
s501: optimizing the wavelet base function and the decomposition layer number, selecting sym3-sym8 wavelet base function, wherein the decomposition layer number is 3-8, the threshold criterion is a Stein unbiased estimation threshold, and the threshold function is a soft threshold;
wherein the soft threshold function used is:
wherein w is j,k As a result of the initial wavelet coefficients,for wavelet coefficients processed by the threshold function, sgn () is a sign function;
s502: calculating Root Mean Square Error (RMSE), smoothness R and correlation coefficient R of a reconstructed signal under different wavelet base functions and decomposition layer numbers, finding the wavelet base function and the decomposition layer number corresponding to the minimum root mean square error and smoothness and the maximum correlation coefficient, and determining the optimal wavelet base function and the optimal decomposition layer number; wherein,
root mean square error expression:
T(i)＝x(i)-x′(i)
the smoothness expression:
the correlation coefficient expression:
where x is the original signal; x' is the denoised reconstructed signal; n is the number of data points.
S503, performing discrete wavelet decomposition on the signal according to the optimal wavelet basis function and the number of decomposition layers; processing the decomposed wavelet coefficients based on a Stein unbiased estimation threshold criterion and a soft threshold function; and finally, reconstructing the signal to obtain a noise reduction signal.
7. The method for processing the acoustic signal of the sandy water flow impulse turbine as claimed in claim 1, wherein: the fourier transform in step S6 is as follows:
wherein,i represents an imaginary unit, X (j) is an input time domain signal, Y (k) is an output frequency domain signal, n is a signal length, j is a jth sample point of the time domain signal, and k is a kth frequency component of the frequency domain signal.
8. The method for processing the acoustic signal of the sandy water flow impulse turbine as claimed in claim 1, wherein: the specific calculation steps of the sample entropy feature value in step S7 are as follows:
s701: through the data sample { x (i), 1.ltoreq.i.ltoreq.N }, an m-dimensional vector is constructed:
Xm(i)＝{x(i),x(i+1),x(i+m-1)},1≤i≤N-m+1
s702: definition vector X m (i) And X is m (j) The absolute value of the maximum difference in the two corresponding elements is the distance d, i.e
d＝max k＝0,1,…,m-1 (|x(i+k)-x(j+k)|)
S703: number B of statistical distances d not greater than similarity threshold r i For 1.ltoreq.i.ltoreq.N-m, define:
s704: calculation of N-mMean>
S705: let the dimension be m+1, repeat steps S701-S704, calculateAnd->
S706: calculating sample entropy eigenvalues SampEn (m, r, N), i.e
。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311055700.6A CN117116291A (en) | 2023-08-22 | 2023-08-22 | Sound signal processing method of sand-containing water flow impulse turbine |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311055700.6A CN117116291A (en) | 2023-08-22 | 2023-08-22 | Sound signal processing method of sand-containing water flow impulse turbine |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117116291A true CN117116291A (en) | 2023-11-24 |
Family
ID=88795960
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311055700.6A Pending CN117116291A (en) | 2023-08-22 | 2023-08-22 | Sound signal processing method of sand-containing water flow impulse turbine |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117116291A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117373471A (en) * | 2023-12-05 | 2024-01-09 | 鸿福泰电子科技（深圳）有限公司 | Audio data optimization noise reduction method and system |
| CN117744893A (en) * | 2024-02-19 | 2024-03-22 | 西安热工研究院有限公司 | Wind speed prediction method and system for energy storage auxiliary black start |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN203705398U (en) * | 2014-02-25 | 2014-07-09 | 三峡大学 | Water turbine abnormity early warning system based on ultrasonic detection |
| CN104316317A (en) * | 2014-10-08 | 2015-01-28 | 西北工业大学 | Gear system multi-fault diagnosis method based on COM assemblies |
| CN104636609A (en) * | 2015-01-30 | 2015-05-20 | 电子科技大学 | Signal combined denoising method based on empirical mode decomposition (EMD) and wavelet analysis |
| CN106706122A (en) * | 2017-01-24 | 2017-05-24 | 东南大学 | Correlation coefficient and EMD (Empirical Mode Decomposition) filtering characteristic-based rub-impact acoustic emission signal noise reduction method |
| CN113470694A (en) * | 2021-04-25 | 2021-10-01 | 重庆市科源能源技术发展有限公司 | Remote listening monitoring method, device and system for hydraulic turbine set |
| CN115371988A (en) * | 2022-10-27 | 2022-11-22 | 北谷电子有限公司 | Engineering machinery fault diagnosis method and system based on multi-feature fusion |
| CN116304570A (en) * | 2023-03-23 | 2023-06-23 | 昆明理工大学 | Hydraulic turbine fault signal denoising method based on EEMD combined Chebyshev filtering |
| CN116453526A (en) * | 2023-04-24 | 2023-07-18 | 中国长江三峡集团有限公司 | Multi-working-condition abnormality monitoring method and device for hydroelectric generating set based on voice recognition |
-
2023
- 2023-08-22 CN CN202311055700.6A patent/CN117116291A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN203705398U (en) * | 2014-02-25 | 2014-07-09 | 三峡大学 | Water turbine abnormity early warning system based on ultrasonic detection |
| CN104316317A (en) * | 2014-10-08 | 2015-01-28 | 西北工业大学 | Gear system multi-fault diagnosis method based on COM assemblies |
| CN104636609A (en) * | 2015-01-30 | 2015-05-20 | 电子科技大学 | Signal combined denoising method based on empirical mode decomposition (EMD) and wavelet analysis |
| CN106706122A (en) * | 2017-01-24 | 2017-05-24 | 东南大学 | Correlation coefficient and EMD (Empirical Mode Decomposition) filtering characteristic-based rub-impact acoustic emission signal noise reduction method |
| CN113470694A (en) * | 2021-04-25 | 2021-10-01 | 重庆市科源能源技术发展有限公司 | Remote listening monitoring method, device and system for hydraulic turbine set |
| CN115371988A (en) * | 2022-10-27 | 2022-11-22 | 北谷电子有限公司 | Engineering machinery fault diagnosis method and system based on multi-feature fusion |
| CN116304570A (en) * | 2023-03-23 | 2023-06-23 | 昆明理工大学 | Hydraulic turbine fault signal denoising method based on EEMD combined Chebyshev filtering |
| CN116453526A (en) * | 2023-04-24 | 2023-07-18 | 中国长江三峡集团有限公司 | Multi-working-condition abnormality monitoring method and device for hydroelectric generating set based on voice recognition |
Non-Patent Citations (2)
| Title |
|---|
| 王季等: "EMD样本熵在滚动轴承信号复杂性度量中的应用", 《中国测试》, vol. 40, no. 1, 30 December 2014 (2014-12-30), pages 45 - 48 * |
| 肖茂华等: "基于ICEEMDAN和小波阈值的滚动轴承故障特征提取方法", 《南京农业大学学报》, vol. 41, no. 4, 31 May 2018 (2018-05-31), pages 767 - 774 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117373471A (en) * | 2023-12-05 | 2024-01-09 | 鸿福泰电子科技（深圳）有限公司 | Audio data optimization noise reduction method and system |
| CN117373471B (en) * | 2023-12-05 | 2024-02-27 | 鸿福泰电子科技（深圳）有限公司 | Audio data optimization noise reduction method and system |
| CN117744893A (en) * | 2024-02-19 | 2024-03-22 | 西安热工研究院有限公司 | Wind speed prediction method and system for energy storage auxiliary black start |
| CN117744893B (en) * | 2024-02-19 | 2024-05-17 | 西安热工研究院有限公司 | Wind speed prediction method and system for energy storage auxiliary black start |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN117116291A (en) | Sound signal processing method of sand-containing water flow impulse turbine | |
| Qin et al. | Transient feature extraction by the improved orthogonal matching pursuit and K-SVD algorithm with adaptive transient dictionary | |
| CN109297713B (en) | Steam turbine fault diagnosis method based on stable and non-stable vibration signal characteristic selection | |
| Dutta et al. | Centrifugal pump cavitation detection using machine learning algorithm technique | |
| Thapa et al. | Optimizing runner blade profile of Francis turbine to minimize sediment erosion | |
| CN108507789B (en) | Rolling bearing fault sparse diagnosis method | |
| CN113420691A (en) | Mixed domain characteristic bearing fault diagnosis method based on Pearson correlation coefficient | |
| Dao et al. | A novel denoising method of the hydro-turbine runner for fault signal based on WT-EEMD | |
| Zeng et al. | An adaptive fractional stochastic resonance method based on weighted correctional signal-to-noise ratio and its application in fault feature enhancement of wind turbine | |
| JP2021018818A (en) | Propeller cavitation state detection method based on wavelet and principal component analysis | |
| CN116304570B (en) | Hydraulic turbine fault signal denoising method based on EEMD combined Chebyshev filtering | |
| CN109342018A (en) | A kind of Turbine Cavitation Testing state monitoring method | |
| CN104111109B (en) | A kind of vibration condition recognition methods based on different order statistic and support vector machine | |
| CN109580224A (en) | Rolling bearing fault method of real-time | |
| Zhang et al. | Study on the improvement of the application of complete ensemble empirical mode decomposition with adaptive noise in hydrology based on RBFNN data extension technology | |
| CN113221986B (en) | Method for separating vibration signals of through-flow turbine | |
| CN108106717B (en) | A method of set state is identified based on voice signal | |
| Xu et al. | Online detection method for variable load conditions and anomalous sound of hydro turbines using correlation analysis and PCA-adaptive-K-means | |
| CN113919525A (en) | Power station fan state early warning method, system and application thereof | |
| Prasetyowati et al. | Comparison Accuracy W-NN and WD-SVM Method In Predicted Wind Power Model on Wind Farm Pandansimo | |
| CN115291103A (en) | Motor fault diagnosis method based on GR-SWPT wavelet packet algorithm embedded with HD-RCF | |
| CN111461461B (en) | Hydraulic engineering abnormity detection method and system | |
| CN112287835B (en) | Blade acoustic signal denoising method based on EWT-SE and wavelet threshold | |
| CN117476039B (en) | Acoustic signal-based primary cavitation early warning method for water turbine | |
| CN113447267A (en) | Gear box complete machine state evaluation method and system based on vibration signal 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 |