CN110738115A - propeller cavitation degree identification method based on pulse frequency characteristic pattern identification - Google Patents

propeller cavitation degree identification method based on pulse frequency characteristic pattern identification Download PDF

Info

Publication number
CN110738115A
CN110738115A CN201910864069.1A CN201910864069A CN110738115A CN 110738115 A CN110738115 A CN 110738115A CN 201910864069 A CN201910864069 A CN 201910864069A CN 110738115 A CN110738115 A CN 110738115A
Authority
CN
China
Prior art keywords
spectrum
frequency
pulse frequency
cavitation
signal
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN201910864069.1A
Other languages
Chinese (zh)
Other versions
CN110738115B (en
Inventor
初宁
童威棋
吴大转
曹琳琳
车邦祥
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Zhejiang University ZJU
Original Assignee
Zhejiang University ZJU
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Zhejiang University ZJU filed Critical Zhejiang University ZJU
Priority to CN201910864069.1A priority Critical patent/CN110738115B/en
Publication of CN110738115A publication Critical patent/CN110738115A/en
Priority to JP2020125982A priority patent/JP6826276B1/en
Application granted granted Critical
Publication of CN110738115B publication Critical patent/CN110738115B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/02—Preprocessing
    • G06F2218/04—Denoising
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B63—SHIPS OR OTHER WATERBORNE VESSELS; RELATED EQUIPMENT
    • B63H—MARINE PROPULSION OR STEERING
    • B63H1/00—Propulsive elements directly acting on water
    • B63H1/02—Propulsive elements directly acting on water of rotary type
    • B63H1/12—Propulsive elements directly acting on water of rotary type with rotation axis substantially in propulsive direction
    • B63H1/14—Propellers
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/12—Classification; Matching
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation

Abstract

propeller cavitation degree identification method based on pulse frequency characteristic pattern identification comprises the steps of (1) collecting propeller noise signals, (2) leading the noise signals into a program, calculating by using a rapid circulation stationary characteristic function to obtain a circulation density spectrum, (3) obtaining a circulation coherent spectrum after carrying out reduction , then building an enhanced envelope spectrum under a logarithmic coordinate by steps of integral average, (4) judging characteristic frequencies according to the obtained enhanced envelope spectrum, selecting integral multiples of corresponding time periods, carrying out improved time domain average on source data, (5) carrying out ensemble empirical mode decomposition to obtain corresponding inherent modal functions, (6) detecting the pulse frequency of the statistical inherent modal functions by adopting a constant false alarm rate, (7) taking the pulse frequency as a characteristic matrix, obtaining judgment of cavitation states after training and identification by a BP neural network, and expressing the statistical characteristics of the propeller under different cavitation states by utilizing the method, so that the estimation of the states is more accurate.

Description

propeller cavitation degree identification method based on pulse frequency characteristic pattern identification
Technical Field
The invention belongs to the field of signal processing and feature extraction, and particularly relates to propeller cavitation degree identification methods based on pulse frequency feature pattern identification.
Background
Cavitation is a common problem in propeller operation. The generation and development of cavitation not only affect the speed distribution in the flow channel, so that the working condition of the propeller is deteriorated, the efficiency is reduced, but also affect the dynamic response of the propeller, and long-time cavitation can also seriously damage flow passage components such as an impeller and the like. And a proper fault diagnosis and identification method is found, and the method has important significance for effectively controlling the propeller cavitation.
The cavitation state identification is which is the difficulty of propeller state monitoring, the current rotating machinery fault detection method commonly used in the signal processing field mainly comprises Fourier transform, short-time Fourier transform, wavelet transform, second generation wavelet transform, multi-wavelet transform and the like, so to speak, the method is the characteristic waveform basis function signal decomposition based on the inner product principle, and aims to flexibly apply a basis function matched with the characteristic waveform to better process the signal and extract fault characteristics, thereby realizing fault diagnosis.
However, the following disadvantages and shortcomings exist in the prior art:
in addition, , the traditional detection methods are difficult to detect important characteristics of the rotating machinery due to theoretical limitations, such as blade passing frequency BPF, blade specific frequency BRF and the like, and have great limitations.
The behavior of cavitation bubbles under different cavitation conditions can be seen as producing a pulsed signal in different frequency bands, whereas in the time domain the pulses are stacked on top of each other and may be buried in noise.
Therefore, the separation and extraction of the pulse frequency under low signal-to-noise ratio conditions becomes the basis for the identification of the cavitation state of the propeller.
Disclosure of Invention
The invention provides propeller cavitation degree identification methods based on pulse frequency characteristic pattern identification, which can express statistical characteristics of propellers in different cavitation states, obtain more accurate state estimation and have practical guiding significance for signal processing and fault diagnosis of steps.
A propeller cavitation degree identification method based on pulse frequency characteristic pattern identification comprises the following steps:
(1) collecting noise signals of underwater propellers in different cavitation states;
(2) importing the collected noise signals into a program, and calculating by using a rapid circulation stationary characteristic function to obtain a circulation density spectrum;
(3) performing classification on the obtained circulating density spectrum to obtain a circulating coherent spectrum, and then performing steps of integral average to construct an enhanced envelope spectrum under a logarithmic coordinate;
(4) judging characteristic frequency according to the obtained enhanced envelope spectrum, selecting integral multiple of corresponding time period, and performing improved time domain averaging on source data;
(5) performing Ensemble Empirical Mode Decomposition (EEMD) on the improved time domain averaged signal to obtain a corresponding Intrinsic Mode Function (IMF);
(6) detecting and counting the pulse frequency of the inherent mode functions of different cavitation states and different orders by adopting constant false alarm rate;
(7) and (4) taking the pulse frequency as a characteristic matrix, and obtaining the judgment of the cavitation state after training and identification of the BP neural network.
Compared with the traditional method, the time domain averaging effect is more accurate based on the cyclostationary period estimation, and meanwhile, the EEMD result can reduce the stacking of pulses in different frequency bands and more accurately obtain the pulse frequency in different cavitation states.
The method can express the pulse frequency of the propeller in different cavitation states, the obtained state estimation is closer to the essence of the propeller cavitation noise, and preliminary state identification and fault diagnosis can be realized through the obtained pulse frequency.
In the step (2), the cyclostationary feature function is:
Figure BDA0002200722200000031
wherein α is the cycle frequency, f is the frequency spectrum frequency, X is the signal to be measured, X is the frequency spectrum of the signal X*The complex conjugate of X; e (-) is the mathematical expectation; wherein the mathematical expression of the amplitude modulation model of x is:
Figure BDA0002200722200000032
wherein A isiAmplitude corresponding to each characteristic frequency α i2 times the characteristic frequency; t represents time; n represents a number; v (t) represents background noise.
In the step (3), the mathematical expression of the cyclic coherence spectrum is as follows:
Figure BDA0002200722200000033
wherein the content of the first and second substances,
Figure BDA0002200722200000034
the circular coherent spectrum corresponding to the working condition of the bubble,
Figure BDA0002200722200000035
is a circulating density spectrum corresponding to the working condition of bubbles,
Figure BDA0002200722200000036
the circulation density spectrum with the circulation frequency of 0 corresponding to the working condition with bubbles.
The specific steps of constructing the enhancement envelope spectrum under the logarithmic coordinate are as follows:
(3-1) calculating a function value corresponding to each cycle frequency of the enhancement envelope spectrum; the mathematical expression of the enhanced envelope spectrum is as follows:
Figure BDA0002200722200000037
wherein，
Figure BDA0002200722200000038
A cyclic coherence spectrum corresponding to a bubble working condition;
(3-2) calculating the function value by taking 10 logarithms to obtain a sound pressure level, setting a value-taking interval according to the obtained logarithm function value range, and assigning the rest logarithm function values as corresponding most values;
and (3-3) constructing an enhancement envelope spectrum under a logarithmic coordinate according to the corresponding coordinate point and the function value.
In step (4), the formula of the improved time domain average is as follows:
Figure BDA0002200722200000041
where x (N) is a time signal obtained by discrete sampling at a time interval Δ t, N is the number of averaged period segments, M is the number of sampling points in periods, and y (N) is a signal obtained by improved time domain averaging.
In the step (4), the method for judging the characteristic frequency adopts the frequency corresponding to the line spectrum with the highest amplitude in the enhanced envelope spectrum in the step (3), then, the integral multiple of the corresponding time period is selected as the improved time domain average length, the time period length selection needs to meet the requirements of two aspects, is the requirement of the resolution ratio of the cyclic frequency, the value delta α is approximately equal to 1/T, wherein T is M/Fs which is the time period, Fs is the sampling frequency, for example, when the resolution ratio of the cyclic frequency is required to reach 0.1Hz, the required T is approximately 10S, and the requirement of the average segment number is adopted, and the noise reduction is more obvious when N is larger under the condition that the resolution ratio requirement and the calculation efficiency are allowed.
In the step (5), the calculation steps of Ensemble Empirical Mode Decomposition (EEMD) are as follows:
(5-1) adding normal distribution white noise to the signal;
(5-2) finding all maximum value points and minimum value points of a signal sequence x (t) added with normal distribution white noise, fitting the maximum value points and the minimum value points to upper and lower envelope lines of an original sequence by a cubic spline function, wherein the average values of the upper and lower envelope lines are m1(t), subtracting m1(t) from the original data sequence to obtain new sequences h1(t) with low frequency subtracted, and h1(t) is x (t) -m 1 (t);
judging whether to continue decomposing and repeating the above process or not by using the difference value and the IMF condition to obtain th eigenmode function component c1(t) which represents the component of the highest frequency of the signal data sequence;
(5-3) subtracting c1(t) from x (t) to obtain new data sequences r1(t) with high frequency components removed, adding normally distributed white noise to r1(t), and then performing decomposition in the step (5-2) to obtain a second eigenmode function component c2(t), and repeating the steps until the last data sequences rn (t) cannot be decomposed;
(5-4) repeating the steps (5-1) to (5-3) each time a new white noise sequence is added;
and (5-5) integrating the mean value of the intrinsic mode functions IMF obtained each time as a final result.
The IMF conditions were as follows:
1. the number of extreme points (maximum or minimum) of the signal is equal to the number of zero-crossing points or differs from the zero-crossing points by at most;
2. the average of the upper envelope composed of local maxima and the lower envelope composed of local minima is zero.
In the step (6), the constant false alarm rate detection adopts a rising edge counting method, wherein the rising edge judgment condition is that the original state is descending, and for two continuous points of the time sequence, the th point value is smaller than the threshold value and the second point is larger than the threshold value, and the falling edge judgment condition is that the original state is ascending, and for two continuous points of the time sequence, the th point value is larger than the threshold value and the second point is smaller than the threshold value.
The self-adaptive limit detector is used as types of CFAR detection, the principle is to estimate the average power of clutter of a unit to be detected through the unit power of a close distance, and the estimated average power of the clutter is multiplied by coefficients to serve as the detection limit .
Assuming that the noise follows a rayleigh distribution, the probability density function is:
Figure BDA0002200722200000051
wherein, Pn＝2b2Is the average power of the noise.
Then, the false alarm rate is:
wherein, η ═ PnlnPfIs the limit.
Therefore, when the average power of the noise is determined, the limit value obtained by multiplying the noise by coefficients can ensure a constant false alarm rate, so that the statistical detection of low false alarms is adaptively realized.
The rising edge counting method comprises the following specific steps:
(6-1) inputting intrinsic mode function IMF signals of unit time, setting the size of a constant false alarm rate according to the mean value and the variance of the signals, and obtaining a threshold value by a self-adaptive limit detector;
(6-2) sequentially carrying out rising edge judgment and falling edge judgment on the first two points of the signal, and carrying out ' addition' operation on the pulse counter when the rising edge judgment is met;
(6-3) gradually moving back the determined points until all the points of the input signal are traversed.
In the step (7), the construction and training of the BP neural network are completed by an MATLAB toolkit, the characteristic matrix is subjected to grouping treatment after being divided into training verification and testing, a 3-layer network is adopted, the number of hidden layer nodes is 10 as default, the number of output layer nodes is 4, and the hidden layer nodes correspond to 4 cavitation states, wherein the 4 cavitation states are tip vortex cavitation, sheet cavitation, bubble cavitation and guide vane cavitation respectively.
Compared with the prior art, the invention has the following beneficial effects:
1. the method can extract parts of the periodic envelope information of the signal and eliminate the influence of the periodic envelope information by an improved time domain average method, wherein the periodic envelope information is obtained by cyclostationary analysis and is closer to an actual signal compared with a traditional detection mode, so that the problem of low signal-to-noise ratio is solved.
2. The invention adopts Ensemble Empirical Mode Decomposition (EEMD), adds the steps of white noise and averaging on the basis of EMD, and solves the aliasing problem caused by EMD; then, constant false alarm detection is utilized, pulse frequencies of different IMFs can be counted robustly, characteristics of different cavitation stages are reflected, and finally estimation of the cavitation state is more accurate.
Drawings
FIG. 1 is a schematic flow chart of propeller cavitation degree identification methods based on pulse frequency characteristic pattern identification according to the present invention;
FIG. 2 is a diagram of IMF 2-IMF 11 and residual error of the propeller according to the embodiment of the present invention;
FIG. 3 is a graph of IMF3 pulse frequency for a propeller according to an embodiment of the present invention;
FIG. 4 is a graph of IMF4 pulse frequency for a propeller according to an embodiment of the present invention;
FIG. 5 is a graph of IMF5 pulse frequency for a propeller according to an embodiment of the present invention;
FIG. 6 is a graph of IMF6 pulse frequency for a propeller according to an embodiment of the present invention;
FIG. 7 is a graph of IMF7 pulse frequency for a propeller according to an embodiment of the present invention;
FIG. 8 is a graph of IMF8 pulse frequency for a propeller according to an embodiment of the present invention;
FIG. 9 is a diagram of the training result of the BP neural network according to the embodiment of the present invention.
Detailed Description
The invention will now be described in further detail with reference to the accompanying drawings and examples, which are intended to facilitate an understanding of the invention and are not intended to be limiting in any way.
As shown in FIG. 1, a propeller cavitation degree identification method based on pulse frequency characteristic pattern identification comprises the following steps:
and S01, collecting the noise of the underwater propeller by using a hydrophone.
S02, setting corresponding parameters in the program, introducing the collected signals into the program, and calculating the circulating density spectrum:
wherein α is the cycle frequency, f is the frequency spectrum frequency, X is the signal to be measured, X is the frequency spectrum of the signal X*The complex conjugate of X is shown, and E is the mathematical expectation.
Wherein the mathematical expression of the amplitude modulation model of x is:
Figure BDA0002200722200000072
wherein: a. theiAmplitude corresponding to each characteristic frequency α i2 times the characteristic frequency; t represents time; n represents the number.
S03, calculating a function value corresponding to each cycle frequency of the enhancement envelope spectrum according to the following formula by using the cycle density function calculated by the cycle stationary characteristic function in the S02:
Figure BDA0002200722200000073
and S04, obtaining the sound pressure level by taking 10 logarithm calculation and the like for the function value, setting the most value limit according to the obtained logarithm function value range, and constructing the enhancement envelope spectrum under the logarithmic coordinate according to the corresponding coordinate point and the function value.
And S05, estimating the frequency corresponding to the highest peak as a characteristic frequency according to the obtained enhanced envelope spectrum, and selecting a proper integral multiple by using the corresponding time period to perform improved time domain averaging on the source data. The output of the improved time domain averaging is as follows:
Figure BDA0002200722200000082
where x (N) is a time signal obtained by discrete sampling at a time interval Δ t, N is the number of averaged period segments, M is the number of sampling points in periods, and y (N) is a signal obtained by improved time domain averaging.
The estimation of the time period is derived from the line spectrum corresponding to the emphasis envelope spectrum in step S04, and generally uses the frequency corresponding to the line spectrum with the highest amplitude as the characteristic frequency.
The length of the time period is selected to meet the requirement of the resolution of the cyclic frequency at , the value of the time period is delta α approximately equal to 1/T, wherein T is M/Fs, the time period is Fs, the sampling frequency is Fs, for example, T is about 10S when the resolution of the cyclic frequency is required to reach 0.1Hz, and the requirement of the average segment number is that the noise reduction is more obvious when N is larger under the condition that the resolution requirement and the calculation efficiency are allowed.
And S06, performing Ensemble Empirical Mode Decomposition (EEMD) on the improved time domain averaged signal y (n) to obtain a corresponding Intrinsic Mode Function (IMF).
The IMF conditions were as follows:
1. the number of extreme points (maximum or minimum) of the signal is equal to the number of zero-crossing points or differs from the zero-crossing points by at most;
2. the average of the upper envelope composed of local maxima and the lower envelope composed of local minima is zero.
The EEMD is calculated as follows:
s06-1, adding normal distribution white noise to the signal;
s06-2, finding out all maximum value points and minimum value points of a signal sequence x (t) added with normal distribution white noise, fitting the maximum value points and the minimum value points to the upper envelope line and the lower envelope line of the original sequence respectively by a cubic spline function, wherein the mean value of the upper envelope line and the lower envelope line is m1(t), and subtracting m1(t) from the original data sequence to obtain new sequences with low frequency subtracted, namely h1(t) ═ x (t) -m 1 (t);
judging whether to continue decomposing and repeating the above process by the difference value and the IMF condition, thus obtaining th eigenmode function component c1(t) which represents the component of the highest frequency of the signal data sequence;
s06-3, subtracting c1(t) from x (t) to obtain new data sequences r1(t) with high frequency components removed, decomposing r1 in the above S06-2 to obtain a second eigenmode function component c2(t), and repeating the steps until the last data sequences rn (t) can not be decomposed;
s06-4, repeating the steps S06-1 to S06-3, and adding a new white noise sequence each time;
and S06-5, taking the IMF integrated mean value obtained each time as a final result.
And S07, counting the pulse frequency by adopting a rising edge counting method according to the IMFs of different cavitation states and different orders obtained in the step S06, and comparing to obtain the pulse frequency corresponding to different characteristics.
The self-adaptive limit detector is used as types of CFAR detection, the principle is to estimate the average power of clutter of a unit to be detected through the unit power of a close distance, and the estimated average power of the clutter is multiplied by coefficients to serve as the detection limit .
The rising edge counting method comprises the following steps:
s07-1, inputting an IMF signal in unit time, setting the size of a constant false alarm rate according to the mean value and the variance of the signal, and obtaining a threshold value by a self-adaptive limit detector;
s07-2, sequentially carrying out rising edge and falling edge judgment on the first two points of the signal, and carrying out '' addition operation on a pulse counter when the rising edge judgment is met;
and S07-3, gradually moving the determined points backwards until all the points of the input signal are traversed.
And S08, training the pulse frequency as a characteristic matrix through a BP neural network to obtain judgment of the cavitation state.
The construction and training of the BP Neural network are completed by Neural Net Fitting of MATLAB toolkit, the characteristic matrix is divided into training verification and test, and then the training verification and test are respectively carried out for processing, a 3-layer network is adopted, the number of hidden layer nodes is 10 in default, the number of output layer nodes is 4, corresponding 4 cavitation states are adopted, the default settings are selected for other settings such as training functions, and finally a plurality of signal characteristic verification cavitation types are input.
In order to embody the advantages and the characteristics of the method in the field of propeller noise detection in different cavitation states, ten groups of seven-blade propeller noise in different cavitation states are adopted for analysis.
The rotating speed of the seven-blade paddle is about 21 revolutions per second, IMFs of different orders in a certain state are shown in fig. 2 from top to bottom, and IMFs 2-IMFs 11 (corresponding to c 1-c 10 in the above) and the residual difference are shown, the IMFs of different orders all have pulses of different frequency ranges.
And then, taking the characteristic matrix as input to obtain a classification result of the BP neural network. The grouping and training results for 300 sets of data are shown in fig. 9, where the accuracy of the test set can reach 94%, and it can be considered that the correct classification is substantially achieved.
From fig. 3 to 9, the following conclusions can be drawn:
1. along with the increase of the IMF order, namely the reduction of the frequency, the dominant stage of the pulse frequency gradually moves to the later stage of the cavitation development;
the conclusion of IMF7 is very typical, the overall linear growth, the reduction of the bubble cavitation stage may represent that the sheet cavitation is inhibited and the number of bubble collapse is reduced;
3, testing results of the BP neural network verify the effectiveness of the characteristics on the pulse frequency on the identification of the cavitation state;
4. because of the nature of EEMD, pulses for each IMF order are not determined to have a definite physical significance, and the statistical nature of the response is not necessarily accurate.
Compared with the traditional method, the cyclostationary-based period estimation has the advantages that the time domain averaging effect is more accurate, the integral multiple processing ensures that the resolution is ensured, and meanwhile, the different cavitation states of the propeller can be effectively judged according to typical characteristics of IMF7 and the like in the verification of the processing result.
The embodiments described above are intended to illustrate the technical solutions and advantages of the present invention, and it should be understood that the above-mentioned embodiments are only specific embodiments of the present invention, and are not intended to limit the present invention, and any modifications, additions and equivalents made within the scope of the principles of the present invention should be included in the scope of the present invention.

Claims (9)

1, propeller cavitation degree identification method based on pulse frequency characteristic pattern identification, which is characterized by comprising the following steps:
(1) collecting noise signals of underwater propellers in different cavitation states;
(2) importing the collected noise signals into a program, and calculating by using a rapid circulation stationary characteristic function to obtain a circulation density spectrum;
(3) performing classification on the obtained circulating density spectrum to obtain a circulating coherent spectrum, and then performing steps of integral average to construct an enhanced envelope spectrum under a logarithmic coordinate;
(4) judging characteristic frequency according to the obtained enhanced envelope spectrum, selecting integral multiple of corresponding time period, and performing improved time domain averaging on source data;
(5) performing Ensemble Empirical Mode Decomposition (EEMD) on the improved time domain averaged signal to obtain a corresponding Intrinsic Mode Function (IMF);
(6) detecting and counting the pulse frequency of the inherent mode functions of different cavitation states and different orders by adopting constant false alarm rate;
(7) and (4) taking the pulse frequency as a characteristic matrix, and obtaining the judgment of the cavitation state after training and identification of the BP neural network.
2. The propeller cavitation degree identification method based on pulse frequency characteristic pattern identification as claimed in claim 1, wherein in the step (2), the cyclostationary characteristic function is:
Figure FDA0002200722190000011
wherein α is the cycle frequency, f is the frequency spectrum frequency, X is the signal to be measured, X is the frequency spectrum of the signal X*The complex conjugate of X; e (-) is the mathematical expectation; wherein the mathematical expression of the amplitude modulation model of x is:
Figure FDA0002200722190000012
wherein A isiAmplitude corresponding to each characteristic frequency αi2 times the characteristic frequency; t represents time; n represents a number; v (t) represents background noise.
3. The propeller cavitation degree identification method based on pulse frequency characteristic pattern identification as claimed in claim 1, wherein in the step (3), the mathematical expression of the cyclic coherence spectrum is as follows:
Figure FDA0002200722190000021
wherein the content of the first and second substances,
Figure FDA0002200722190000022
the circular coherent spectrum corresponding to the working condition of the bubble,
Figure FDA0002200722190000023
is a circulating density spectrum corresponding to the working condition of bubbles,
Figure FDA0002200722190000024
the circulation density spectrum with the circulation frequency of 0 corresponding to the working condition with bubbles.
4. The propeller cavitation degree identification method based on pulse frequency characteristic pattern identification as claimed in claim 1, wherein in the step (3), the specific steps of constructing the enhancement envelope spectrum under logarithmic coordinates are as follows:
(3-1) calculating a function value corresponding to each cycle frequency of the enhancement envelope spectrum; the mathematical expression of the enhanced envelope spectrum is as follows:
Figure FDA0002200722190000025
wherein the content of the first and second substances,
Figure FDA0002200722190000026
a cyclic coherence spectrum corresponding to a bubble working condition;
(3-2) calculating the function value by taking 10 logarithms to obtain a sound pressure level, setting a value-taking interval according to the obtained logarithm function value range, and assigning the rest logarithm function values as corresponding most values;
and (3-3) constructing an enhancement envelope spectrum under a logarithmic coordinate according to the corresponding coordinate point and the function value.
5. The propeller cavitation degree identification method based on pulse frequency characteristic pattern identification as claimed in claim 1, wherein in the step (4), the formula of the improved time domain average is as follows:
Figure FDA0002200722190000027
where x (N) is a time signal obtained by discrete sampling at a time interval Δ t, N is the number of averaged period segments, M is the number of sampling points in periods, and y (N) is a signal obtained by improved time domain averaging.
6. The propeller cavitation degree identification method based on pulse frequency characteristic pattern identification as claimed in claim 1, wherein in the step (4), the method for judging the characteristic frequency adopts the frequency corresponding to the line spectrum with the highest amplitude in the enhanced envelope spectrum; the integral multiple of the corresponding time period is selected to meet the requirements of the resolution of the cycle frequency and the average segment number.
7. The propeller cavitation level identification method based on pulse frequency characteristic pattern identification as claimed in claim 1, wherein in the step (5), the calculation steps of the Ensemble Empirical Mode Decomposition (EEMD) are as follows:
(5-1) adding normal distribution white noise to the signal;
(5-2) finding all maximum value points and minimum value points of a signal sequence x (t) added with normal distribution white noise, fitting the maximum value points and the minimum value points to upper and lower envelope lines of an original sequence by a cubic spline function, wherein the average values of the upper and lower envelope lines are m1(t), subtracting m1(t) from the original data sequence to obtain new sequences h1(t) with low frequency subtracted, and h1(t) is x (t) -m 1 (t);
judging whether to continue decomposing and repeating the above process or not by using the difference value and the IMF condition to obtain th eigenmode function component c1(t) which represents the component of the highest frequency of the signal data sequence;
(5-3) subtracting c1(t) from x (t) to obtain new data sequences r1(t) with high frequency components removed, adding normally distributed white noise to r1(t), and then performing decomposition in the step (5-2) to obtain a second eigenmode function component c2(t), and repeating the steps until the last data sequences rn (t) cannot be decomposed;
(5-4) repeating the steps (5-1) to (5-3) each time a new white noise sequence is added;
and (5-5) integrating the mean value of the intrinsic mode functions IMF obtained each time as a final result.
8. The propeller cavitation degree identification method based on pulse frequency characteristic pattern identification as claimed in claim 1, wherein in the step (6), the constant false alarm rate detection adopts a rising edge counting method, and the specific steps are as follows:
(6-1) inputting intrinsic mode function IMF signals of unit time, setting the size of a constant false alarm rate according to the mean value and the variance of the signals, and obtaining a threshold value by a self-adaptive limit detector;
(6-2) sequentially carrying out rising edge judgment and falling edge judgment on the first two points of the signal, and carrying out ' addition' operation on the pulse counter when the rising edge judgment is met;
(6-3) gradually moving back the determined points until all the points of the input signal are traversed.
9. The propeller cavitation degree identification method based on pulse frequency characteristic pattern identification according to claim 1 is characterized in that in the step (7), the construction and training of the BP neural network are completed by a MATLAB toolkit, a characteristic matrix is divided into training verification and testing, and then classification into is respectively carried out, a 3-layer network is adopted, the number of hidden layer nodes is 10 as default, the number of output layer nodes is 4, and 4 cavitation states are corresponding.
CN201910864069.1A 2019-09-12 2019-09-12 Propeller cavitation degree identification method based on pulse frequency characteristic pattern identification Active CN110738115B (en)

Priority Applications (2)

Application Number Priority Date Filing Date Title
CN201910864069.1A CN110738115B (en) 2019-09-12 2019-09-12 Propeller cavitation degree identification method based on pulse frequency characteristic pattern identification
JP2020125982A JP6826276B1 (en) 2019-09-12 2020-07-24 Method of identifying the degree of propeller cavitation based on the identification of the characteristic pattern of pulse frequency

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910864069.1A CN110738115B (en) 2019-09-12 2019-09-12 Propeller cavitation degree identification method based on pulse frequency characteristic pattern identification

Publications (2)

Publication Number Publication Date
CN110738115A true CN110738115A (en) 2020-01-31
CN110738115B CN110738115B (en) 2022-11-15

Family

ID=69267624

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910864069.1A Active CN110738115B (en) 2019-09-12 2019-09-12 Propeller cavitation degree identification method based on pulse frequency characteristic pattern identification

Country Status (2)

Country Link
JP (1) JP6826276B1 (en)
CN (1) CN110738115B (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113776834A (en) * 2021-10-11 2021-12-10 山东大学 Rolling bearing fault diagnosis method and system based on discrete cosine cycle spectrum coherence
CN117476039A (en) * 2023-12-25 2024-01-30 西安理工大学 Acoustic signal-based primary cavitation early warning method for water turbine

Families Citing this family (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113435246B (en) * 2021-05-18 2024-04-05 西安电子科技大学 Intelligent radiation source individual identification method, system and terminal
CN113392368B (en) * 2021-05-28 2023-08-11 浙江大学 Fluid machinery non-uniform cavitation state characterization method based on absolute modulation intensity spectrum
CN113821888B (en) * 2021-09-23 2024-02-27 西安热工研究院有限公司 Vibration data fault diagnosis method based on periodic impact feature extraction and echo state network
CN114018571A (en) * 2021-10-28 2022-02-08 国能太仓发电有限公司 Gear box fault diagnosis method based on frequency-induced variation modal decomposition
CN114492538B (en) * 2022-02-16 2023-09-05 国网江苏省电力有限公司宿迁供电分公司 Urban medium-voltage distribution cable partial discharge signal denoising method
CN114742111B (en) * 2022-05-24 2023-04-07 南京林业大学 Fault diagnosis method and system based on parameter adaptive characteristic modal decomposition
CN116304588B (en) * 2022-09-07 2023-08-18 浙江大学 Compensation processing method for unstable propeller cavitation pulse pressure signal
CN117009858A (en) * 2023-08-22 2023-11-07 哈尔滨工业大学 Synchronous classification method for redundant detection signals of aerospace sealed electronic components

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103412298A (en) * 2013-08-12 2013-11-27 中国兵器科学研究院 Method capable of automatically acquiring variable speed rotation time interval of ship propeller
CN104091085A (en) * 2014-07-18 2014-10-08 安徽工业大学 Cavitation noise feature estimation method based on propeller wake flow pressure fluctuation computing
CN104899656A (en) * 2015-06-05 2015-09-09 三峡大学 Wind power combined predication method based on ensemble average empirical mode decomposition and improved Elman neural network
CN107609488A (en) * 2017-08-21 2018-01-19 哈尔滨工程大学 A kind of ship noise method for identifying and classifying based on depth convolutional network
CN109840495A (en) * 2019-01-29 2019-06-04 浙江大学 A kind of axis frequency line enhancement method under Low SNR

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103412298A (en) * 2013-08-12 2013-11-27 中国兵器科学研究院 Method capable of automatically acquiring variable speed rotation time interval of ship propeller
CN104091085A (en) * 2014-07-18 2014-10-08 安徽工业大学 Cavitation noise feature estimation method based on propeller wake flow pressure fluctuation computing
CN104899656A (en) * 2015-06-05 2015-09-09 三峡大学 Wind power combined predication method based on ensemble average empirical mode decomposition and improved Elman neural network
CN107609488A (en) * 2017-08-21 2018-01-19 哈尔滨工程大学 A kind of ship noise method for identifying and classifying based on depth convolutional network
CN109840495A (en) * 2019-01-29 2019-06-04 浙江大学 A kind of axis frequency line enhancement method under Low SNR

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
杨宏: "经验模态分解及其在水声信号处理中的应用", 《中国博士学位论文全文数据库 (信息科技辑)》 *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113776834A (en) * 2021-10-11 2021-12-10 山东大学 Rolling bearing fault diagnosis method and system based on discrete cosine cycle spectrum coherence
CN117476039A (en) * 2023-12-25 2024-01-30 西安理工大学 Acoustic signal-based primary cavitation early warning method for water turbine
CN117476039B (en) * 2023-12-25 2024-03-08 西安理工大学 Acoustic signal-based primary cavitation early warning method for water turbine

Also Published As

Publication number Publication date
CN110738115B (en) 2022-11-15
JP6826276B1 (en) 2021-02-03
JP2021041915A (en) 2021-03-18

Similar Documents

Publication Publication Date Title
CN110738115B (en) Propeller cavitation degree identification method based on pulse frequency characteristic pattern identification
CN109840495B (en) Axial frequency line spectrum enhancement method under low signal-to-noise ratio condition
CN101813568B (en) Judging method for determining cavitation inception
CN103902844A (en) Transformer vibration signal de-noising method based on EEMD kurtosis threshold value
CN103034757B (en) Wind energy turbine set time-frequency domain modeling method based on empirical mode decomposition
CN111693775A (en) Harmonic detection method, device and medium for power transmission network
CN106771598A (en) A kind of Adaptive spectra kurtosis signal processing method
CN116626408B (en) Power supply ripple noise detection method based on machine learning
CN116865269A (en) Wind turbine generator system high harmonic compensation method and system
CN109342018A (en) A kind of Turbine Cavitation Testing state monitoring method
Spinosa et al. A noise reduction method for force measurements in water entry experiments based on the Ensemble Empirical Mode Decomposition
CN107898443A (en) Dicrotic pulse wave detecting method, device and computer-readable storage medium
CN113389541B (en) High-precision extraction method for oil well working fluid level signal
Feng et al. Experimental investigation on characteristics of cavitation-induced vibration on the runner of a bulb turbine
CN114117912A (en) Sea clutter modeling and inhibiting method under data model dual drive
JP2017198620A (en) Abnormality diagnosis device and abnormality diagnosis method
CN110489902B (en) Multi-element statistical modeling method for fine characteristics of cavitation wake flow of propeller
CN103915102A (en) Method for noise abatement of LFM underwater sound multi-path signals
CN114781466B (en) Fault diagnosis method and system based on harmonic fundamental frequency of rotary mechanical vibration signal
CN115014765B (en) Method for extracting fault characteristics of rolling bearing retainer through acoustic signals
CN115630271A (en) Signal frequency estimation method, device, equipment and storage medium
CN109085763B (en) Extraction of fan vibration characteristics by using circulation stabilization method based on complex amplitude modulation model
Liu et al. A novel random spectral similar component decomposition method and its application to gear fault diagnosis
CN108106717A (en) A kind of method based on voice signal identification set state
Lin et al. Denoising of wind speed data by wavelet thresholding

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant