CN109633270B - Fault energy region boundary identification and feature extraction method - Google Patents
Fault energy region boundary identification and feature extraction method Download PDFInfo
- Publication number
- CN109633270B CN109633270B CN201910001146.0A CN201910001146A CN109633270B CN 109633270 B CN109633270 B CN 109633270B CN 201910001146 A CN201910001146 A CN 201910001146A CN 109633270 B CN109633270 B CN 109633270B
- Authority
- CN
- China
- Prior art keywords
- time
- frequency
- fault
- boundary
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R23/00—Arrangements for measuring frequencies; Arrangements for analysing frequency spectra
- G01R23/16—Spectrum analysis; Fourier analysis
Abstract
The invention discloses a fault energy region boundary identification and feature extraction method based on instantaneous frequency spectrum entropy and signal-to-noise energy difference, which can effectively determine a time domain boundary and a frequency domain boundary of an energy concentration region in a time-frequency power density spectrum of a dynamic signal of an underwater robot, takes the total energy in the boundary as fault features, and has unique mapping relation between the extracted time-frequency energy fault features and fault degrees; and moreover, the fault characteristics are adopted to construct a fault sample, and when the fault sample is used for classifying the fault degree of the propeller, the fault degree classification precision of the test sample is 100%.
Description
Technical Field
The invention belongs to the technical field of fault diagnosis of a propeller of an underwater robot, and particularly relates to a fault energy region boundary identification and feature extraction method based on instantaneous frequency spectrum entropy and signal-to-noise energy difference.
Background
The thruster is a key part and a power unit of the underwater robot, has the heaviest load, and is one of main fault sources of the underwater robot, so that the monitoring of the running state of the thruster is very necessary. Fault diagnosis is a common technical means for monitoring the operational state of a propeller. The energy of the singular behavior of the dynamic signal of the underwater robot is an important fault feature. The known signal amplitude square sum method extracts the energy characteristics of the singular behaviors of the signals from the time domain, the known wavelet energy method extracts the energy characteristics of the singular behaviors of the signals from the frequency domain, and the fresh method extracts the energy characteristics of the singular behaviors of the signals from the time domain and the frequency domain.
Disclosure of Invention
The purpose of the invention is as follows: the invention aims to solve the defects in the prior art and provides a fault energy region boundary identification and feature extraction method based on instantaneous spectrum entropy and signal-to-noise energy difference.
The technical scheme is as follows: the invention discloses a fault energy region time domain boundary identification method based on instantaneous spectrum entropy, which comprises the following steps of:
first, obtain length L1The underwater robot speed signal data;
secondly, calculating a time-frequency power density spectrum of the speed signal;
obtaining a smooth pseudo-Wigner-Willi spectrum SPWVD (n, m) of the speed signal by adopting a conventional smooth pseudo-Wigner-Willi distribution algorithm, wherein n is a sequence number on a time axis, and n is 1,2,3, …, L1M is the number on the frequency axis, m is 1,2,3, …, N3，N3Dividing the frequency axis by a number of divisions, e.g. setting N3512; obtaining a speed signal time-frequency power density spectrum SPWVDA (n, m) through the conventional absolute value operation of the SPWVD (n, m);
thirdly, calculating instantaneous spectrum entropy distribution;
in the time-frequency power density spectrum SPWVDA (n, m), at a time shaft serial number n being 1, a probability density function p (1, m) at a time point where the time shaft serial number is located is constructed through a formula (1), and then a Shannon entropy at the time point is calculated through a formula (2), and a calculation result H (1) is an instantaneous spectrum entropy at the time shaft serial number n being 1;
repeating the steps, calculating the time shaft sequence numbers n to be 2,3, … and L respectively1Instantaneous spectrum entropy H (n) of time is obtained to obtain instantaneous spectrum entropy distribution;
wherein SPWVDA (1, m) is a time-frequency power density spectrum SPWVDA (N, m), where the time-axis number N is 1 and the energy in the mth frequency band, p (1, m) is a probability density function, and N is3Dividing interval numbers for a frequency axis, wherein H (1) is instantaneous spectrum entropy at a time axis with the sequence number n being 1;
fourthly, determining a time domain boundary of a fault energy region, namely determining a distortion interval in instantaneous frequency spectrum entropy distribution;
determining the position of the minimum value in the instantaneous frequency spectrum entropy distribution, then extending to the adjacent maximum value points from the position to two sides at the same time, if the maximum value is smaller than the minimum values of other areas, continuing to extend to two sides until the encountered maximum value is larger than the minimum values of other areas, and determining the area between the two final maximum value points as a distortion interval; then, the position of the maximum value point on the left side is taken as the time domain lower boundary N of the fault energy regionDAnd setting the position of the right maximum value point as the upper boundary N of the time domain of the fault energy regionU。
The invention also discloses a fault energy region frequency domain boundary identification method based on the signal-noise energy difference, which comprises the following steps:
first, obtain length L1The underwater robot speed signal data;
secondly, wavelet decomposition and reconstruction of multiple scales are carried out on the velocity signal by adopting a conventional wavelet transformation method to obtain multiple wavelet approximate components ujA(n), j is the wavelet decomposition scale, j is 0,1,2, …,8, wherein j is 0, which represents that the velocity signal is not wavelet decomposed, n is the number on the time axis, n is 1,2,3, …, L is 1,2,3, …1；
Thirdly, calculating a time-frequency power density spectrum; calculating the approximate component u of the jth scale wavelet by adopting a conventional smooth pseudo-Wigner-Willi distribution algorithm and an absolute value algorithmjA(n) time-frequency Power Density Spectrum SPWVDAj(n,m)；
Fourthly, calculating the signal-to-noise energy difference distribution;
determining time domain lower boundary N of fault energy region based on instantaneous spectrum entropyDAnd upper boundary N in time domainU；
In time frequency power density spectrum SPWVDAj(N, m), the time domain boundary NDAnd NUTime-frequency power density spectrum SPWVDA betweenj(n, m) are summed to obtain the result ESjAs fault energy, the time domain boundary NDAnd NUOutside time-frequency power density spectrum SPWVDAj(n, m) are summed to obtain the result ENjAs noise energy, willEΔSNj＝ESj-ENjAs a signal-to-noise energy difference; calculating the signal-to-noise energy difference E of j equal to 0,1,2, …,8ΔSNjObtaining signal-to-noise energy difference distribution;
fifthly, determining the frequency domain boundary of the fault characteristic region; at the signal-to-noise energy difference EΔSNjIn distribution, determining maximum value of signal-to-noise energy difference, determining wavelet decomposition scale corresponding to the maximum value, determining wavelet approximate component corresponding to the decomposition scale, and determining frequency band [ M ] corresponding to the wavelet approximate componentDMU]I.e. the fault energy region frequency domain boundary, where MDIs the lower boundary of the frequency domain, MUThe upper boundary of the frequency domain.
The invention discloses a fault energy region boundary identification and feature extraction method based on instantaneous spectrum entropy and signal-to-noise energy difference, which comprises the following steps:
first, respectively collecting the lengths L1The underwater robot speed signal and the propeller control voltage change rate signal data;
secondly, determining a time domain lower boundary N of the fault energy region based on the instantaneous spectrum entropyDAnd upper boundary N in time domainU；
Thirdly, determining the lower boundary M of the frequency domain of the fault energy area based on the signal-noise energy differenceDAnd an upper boundary M of the frequency domainU；
Fourthly, extracting time-frequency energy fault characteristics of the propeller;
the time-frequency power density spectrum SPWVDA of the speed signal and the control voltage change rate signal is respectively calculated by adopting a conventional smooth pseudo-Wigner-Willi distribution algorithm and an absolute value algorithmU(n, m) and SPWVDAC(N, m), lower boundary N in time domainDUpper boundary N in time domainUFrequency domain lower boundary MDSum frequency domain upper bound MUIn the enclosed rectangular area, summing the time-frequency power density spectrum to obtain a result FU、FCRespectively as the time-frequency energy fault characteristics of the speed signal and the time-frequency energy fault characteristics of the control signal.
Has the advantages that: the method can effectively determine the time domain boundary and the frequency domain boundary of the energy concentration region in the time-frequency power density spectrum of the dynamic signal of the underwater robot, and takes the total energy in the boundary as the fault characteristic. The mapping relation between the extracted time-frequency energy fault characteristics and the fault degree is unique. Moreover, the failure sample constructed by the failure characteristics can realize accurate classification of failure degrees of the propeller.
Drawings
FIG. 1 is a flow chart of a fault energy region time domain boundary identification method based on transient spectrum entropy according to the present invention;
FIG. 2 is a flowchart of a method for identifying the frequency domain boundary of a fault energy region based on a signal-to-noise energy difference according to the present invention;
FIG. 3 is a flowchart of the present invention;
FIG. 4 is a time domain waveform of the underwater robot speed signal and thruster control signal data;
FIG. 5 is a diagram of an instantaneous frequency spectrum entropy distribution of a speed signal of an underwater robot;
FIG. 6 is a schematic diagram of time domain boundary division of a speed signal fault energy region;
FIG. 7 is a velocity signal to noise energy differential profile;
FIG. 8 is a schematic diagram of the time-frequency boundary division of the fault energy region of the speed signal;
FIG. 9 is a time-frequency energy distribution diagram of a velocity signal without a boundary;
FIG. 10 is a time-frequency energy distribution diagram of a velocity signal with time-frequency boundaries set;
FIG. 11 propeller fault sample distribution plot;
FIG. 12 is a diagram showing the relative distance distribution between different failure degree test samples of the propeller and each hypersphere model.
Detailed Description
The technical solution of the present invention is described in detail below, but the scope of the present invention is not limited to the embodiments.
As shown in fig. 1, a method for identifying a fault energy region time domain boundary based on instantaneous spectrum entropy of the present invention includes the following steps:
first, obtain length L 1400 underwater robot velocity signal data；
Secondly, calculating a time-frequency power density spectrum of the speed signal;
obtaining a smooth pseudo-Wigner-Willi spectrum SPWVD (n, m) of the speed signal by adopting a conventional smooth pseudo-Wigner-Willi distribution algorithm, wherein n is a sequence number on a time axis, and n is 1,2,3, …, L1，L 1400, m is the number on the frequency axis, 1,2,3, …, N3，N3For dividing the frequency axis by the number of intervals, N in this embodiment3512; obtaining a speed signal time-frequency power density spectrum SPWVDA (n, m) through the conventional absolute value operation of the SPWVD (n, m);
thirdly, calculating instantaneous spectrum entropy distribution;
in the time-frequency power density spectrum SPWVDA (n, m), at a time axis n-1, a probability density function p (1, m) at the time point is constructed by formula (1), and then a shannon entropy at the time point is calculated by formula (2), and a calculation result H (1) is an instantaneous spectrum entropy at the time axis n-1;
repeating the steps, calculating the time shaft sequence numbers n to be 2,3, … and L respectively1Instantaneous spectrum entropy H (n) of time is obtained to obtain instantaneous spectrum entropy distribution;
wherein SPWVDA (1, m) is a time-frequency power density spectrum SPWVDA (N, m), where the time-axis number N is 1 and the energy in the mth frequency band, p (1, m) is a probability density function, and N is3Dividing the frequency axis by a number of intervals, N3H (1) is the instantaneous spectrum entropy at time axis index n-1;
fourthly, determining a time domain boundary of a fault energy region, namely determining a distortion interval in instantaneous frequency spectrum entropy distribution;
determining the position of the minimum value in the instantaneous frequency spectrum entropy distribution, and extending from the position to two sides to the adjacent maximum value point simultaneously ifIf the maximum value is smaller than the minimum values of other areas, continuing extending towards two sides until the encountered maximum value is larger than the minimum values of other areas, and determining the area between two final maximum value points as a distortion interval; then, the position of the maximum value point on the left side is taken as the time domain lower boundary N of the fault energy regionDAnd setting the position of the right maximum value point as the upper boundary N of the time domain of the fault energy regionU。
As shown in fig. 2, a method for identifying a frequency domain boundary of a fault energy region based on a signal-to-noise energy difference according to the present invention includes the following steps:
first, obtain length L 1400 underwater robot velocity signal data;
secondly, wavelet decomposition and reconstruction of multiple scales are carried out on the velocity signal by adopting a conventional wavelet transformation method to obtain multiple wavelet approximate components ujA(n), j is the wavelet decomposition scale, j is 0,1,2, …,8, wherein j is 0, which means that the velocity signal is not wavelet decomposed;
thirdly, calculating a time-frequency power density spectrum; calculating the approximate component u of the jth scale wavelet by adopting a smooth pseudo-Wigner-Willi distribution algorithm and an absolute value algorithmjA(n) time-frequency Power Density Spectrum SPWVDAj(n,m)；
Fourthly, calculating the signal-to-noise energy difference distribution;
determining time domain lower boundary N of fault energy region based on instantaneous spectrum entropyDAnd upper boundary N in time domainU；
In time frequency power density spectrum SPWVDAj(N, m), the time domain boundary NDAnd NUTime-frequency power density spectrum SPWVDA betweenj(n, m) are summed to obtain the result ESjAs fault energy, the time domain boundary NDAnd NUOutside time-frequency power density spectrum SPWVDAj(n, m) are summed to obtain the result ENjAs noise energy, EΔSNj＝ESj-ENjAs a signal-to-noise energy difference; calculating the signal-to-noise energy difference E of j equal to 0,1,2, …,8ΔSNjObtaining signal-to-noise energy difference distribution;
the fifth step, determine the faultA characteristic region frequency domain boundary; at the signal-to-noise energy difference EΔSNjIn distribution, determining maximum value of signal-to-noise energy difference, determining wavelet decomposition scale corresponding to the maximum value, determining wavelet approximate component corresponding to the decomposition scale, and determining frequency band [ M ] corresponding to the wavelet approximate componentDMU]I.e. the fault signature region frequency domain boundary, where MDIs the lower boundary of the frequency domain, MUThe upper boundary of the frequency domain.
As shown in the figure, the fault energy region boundary identification and feature extraction method based on the instantaneous spectrum entropy and the signal-to-noise energy difference comprises the following steps:
first, respectively obtaining the lengths L 1400 underwater robot speed signal and thruster control voltage change rate signal data;
secondly, determining a time domain lower boundary N of the fault energy region based on the instantaneous spectrum entropyDAnd upper boundary N in time domainU；
Thirdly, determining the lower boundary M of the frequency domain of the fault energy area based on the signal-noise energy differenceDAnd an upper boundary M of the frequency domainU；
Fourthly, extracting time-frequency energy fault characteristics of the propeller;
adopting a smooth pseudo-Wigner-Willi distribution algorithm and an absolute value algorithm to respectively calculate time-frequency power density spectrums SPWVDA of speed signals and control voltage change rate signalsU(n, m) and SPWVDAC(N, m), lower boundary N in time domainDUpper boundary N in time domainUFrequency domain lower boundary MDSum frequency domain upper bound MUIn the enclosed rectangular area, summing the time-frequency power density spectrum to obtain a result FU、FCRespectively as the time-frequency energy fault characteristics of the speed signal and the time-frequency energy fault characteristics of the control signal.
Example (b):
as shown in fig. 4, the longitudinal target speed of the underwater robot is 0.3m/s, the underwater robot starts to operate from 0 moment, the longitudinal speed is gradually increased, after 100 beats, the underwater robot starts to operate at a steady-state speed of 0.3m/s, and from the 250 th time beat, the propeller has thrust loss faults, and the fault degrees are respectively 0%, 10%, 20%, 30% and 40% until the test is finished. The control signals for 10%, 20%, 30%, 40% propeller failure levels gradually increased from beat 250 and then fluctuated around a steady value. The velocity signals with 10%, 20%, 30%, 40% failure levels formed singular behaviors that declined first and then risen in beats 250 to 350, as shown by the oval circles in fig. 4.
The length of a group of experimental data of the underwater robot in the embodiment is 600, wherein the length of the first 100 beats is that the underwater robot has not reached a steady state, so that the data cannot be used, and therefore, only the data of the last 500 beats can be used for experimental analysis. To obtain 50 fault samples, the time window is slid 50 times, the length of the time window L1Designated as 400. Initially, length L is used1The experimental data of 101 ~ 500 beats are intercepted to the time window of 400 for experimental analysis, then slide this time window a time beat to the right, when sliding 50 time beats, the experimental data of time window interception 151 ~ 550 is used for experimental analysis.
By length L1The speed signals of the 101 th to 500 th beats in fig. 4 are intercepted by the time window function of 400, and the instantaneous spectrum entropies of the time-frequency power density spectrums of the speed signals corresponding to different failure degrees of the propeller are respectively calculated, and the result is shown in fig. 5. In FIG. 5, the minimum value 3.805 of the instantaneous spectrum entropy curve of the speed signal corresponding to the propeller failure degree of 40% is at the 279 th beat, and then the position extends to the adjacent maximum value point from the minimum value 3.805 to the two sides, the left side extends to the 192 th beat maximum value 5.371, the right side extends to the 387 beat maximum value 5.352, and the distortion interval of the instantaneous spectrum entropy curve is [ 192387 ] as shown by the corresponding vertical lines]The time beat, namely the time domain boundary of the fault energy region in the time-frequency power density spectrum corresponding to the fault degree of 40 percent is [ 192387 ]]The time beat. Similarly, the time domain boundaries of the fault energy regions in the time-frequency power density spectrum corresponding to the propeller fault degrees of 0%, 10%, 20% and 30% are [ 100239%]、[226 332]、[174 314]、[211 337]The time beat.
As shown in fig. 6, the time domain boundary of the fault energy region is divided in the time-frequency power density spectrum of the speed signal shown in fig. 6 according to the time domain boundary shown in fig. 5, and it can be seen from the result of the time domain boundary division that the time domain boundary divided in fig. 5 is consistent with the actual time domain boundary of the energy concentration region, which shows that the fault energy region time domain boundary identification method based on the instantaneous frequency spectrum entropy of the present invention is effective.
As shown in fig. 7, the maximum value of the signal-to-noise energy difference corresponding to the propeller failure degree of 40% is 0.0199, the corresponding wavelet decomposition scale is 5, and similarly, the maximum values of the signal-to-noise energy difference corresponding to the propeller failure degrees of 0%, 10%, 20%, and 30% are 0.0001, 0.0006, 0.0020, and 0.0111, and the corresponding wavelet decomposition scales are 6, 5, and 5, respectively.
As shown in fig. 8, the frequency domain boundary of the fault energy region is divided according to the wavelet approximate component corresponding to the wavelet decomposition scale determined in fig. 7 and the frequency band corresponding to the wavelet approximate component, and as can be seen from the result of frequency domain boundary division, the frequency domain boundary divided according to fig. 7 is consistent with the actual frequency domain boundary of the energy concentration region, which illustrates that the fault energy region frequency domain boundary identification method based on the signal-to-noise energy difference of the present invention is effective.
As shown in fig. 9 and 10, in fig. 9, in the time-frequency power density spectrum of the speed signal corresponding to a certain failure degree of the propeller, no boundary is set, all values in the time-frequency power density spectrum are directly summed, and the obtained result is used as a time-frequency energy failure feature of the propeller, where the mapping relationship between the failure feature obtained by the method and the failure degree is not unique, that is, one failure feature value may correspond to multiple failure degrees, for example, the failure feature 0.060 corresponds to four failure degrees, 4.9%, 13.7%, 26.6%, and 32.7%. In fig. 10, in a time-frequency power density spectrum of a speed signal corresponding to a certain failure degree of a propeller, a time-domain boundary and a frequency-domain boundary of a failure energy region are identified according to the method of the present invention, then time-frequency power densities within the boundaries are summed, and the obtained result is used as a time-frequency energy failure feature of the propeller, where the mapping relationship between the failure features and the failure degrees obtained by the method is unique, that is, one failure feature value corresponds to only one failure degree, and if the failure feature value is 0.010, only one failure degree is 27.2%.
The method comprises the steps of intercepting speed signal data and control signal data from 101 th beat to 500 th beat in the figure 4 by adopting a time window function with the length of 400, corresponding to each fault degree of a propeller, extracting time-frequency energy fault characteristics from the data by adopting the characteristic boundary identification method of the patent, constructing a fault sample, then moving the time window to the right for a time beat, extracting the fault characteristics again, constructing a fault sample, repeating the process, moving the time window for 50 beats, constructing 50 fault samples, and distributing the obtained fault samples as shown in the figure 11. In fig. 11, the fault samples of the same fault degree are gathered together, and the distance between the fault samples corresponding to different fault degrees is larger, which is beneficial to the classification of the fault degrees of the propeller.
In fig. 11, for each failure degree, 25 samples are randomly selected as training samples, and the remaining 25 samples are used as test samples. A classification model of the failure degree of the propeller is established by adopting a training sample based on a conventional support vector field description method, a test sample is brought into the classification model, and the relative distance between the test sample and each hypersphere model is calculated, wherein the results are shown in FIG. 12, wherein the failure degrees corresponding to hypersphere 1, hypersphere 2, hypersphere 3, hypersphere 4 and hypersphere 5 are respectively 0%, 10%, 20%, 30% and 40%.
In fig. 12, the degree of failure corresponding to the super ball model with the smallest relative distance to the test sample is used as the result of classifying the degree of failure of the test sample, for example, in fig. 12(a), if the relative distance between the test sample with the degree of failure 0% and the super ball 1 is smallest, the degree of failure 0% corresponding to the super ball 1 will be used as the result of classifying the degree of failure of the test sample. The classification accuracy of the fault degrees of all the test samples in fig. 12 is counted, and the result is 100%, which shows that the fault characteristics obtained by the fault energy region boundary identification and characteristic extraction method based on the instantaneous frequency entropy and the signal-to-noise energy difference are beneficial to classification of the fault degrees of the propeller, and the classification accuracy is 100%.
Claims (3)
1. A fault energy region time domain boundary identification method based on instantaneous frequency spectrum entropy is characterized in that: the method comprises the following steps:
in the first step, the first step is that,obtaining a length L1The underwater robot speed signal;
secondly, calculating a time-frequency power density spectrum of the speed signal; obtaining a smooth pseudo-Wigner-Willi spectrum SPWVD (n, m) of the speed signal by adopting a conventional smooth pseudo-Wigner-Willi distribution algorithm, wherein n is a sequence number on a time axis, and n is 1,2,3, …, L1M is the number on the frequency axis, m is 1,2,3, …, N3，N3Dividing interval numbers for a frequency axis; and obtaining a speed signal time-frequency power density spectrum SPWVDA (n, m) through the conventional absolute value operation of the SPWVD (n, m);
thirdly, calculating instantaneous spectrum entropy distribution; in the time-frequency power density spectrum SPWVDA (n, m), at a time shaft serial number n being 1, a probability density function p (1, m) at a time point where the time shaft serial number is located is constructed through a formula (1), and then a Shannon entropy at the time point is calculated through a formula (2), and a calculation result H (1) is an instantaneous spectrum entropy at the time shaft serial number n being 1;
repeating the steps, calculating the time shaft sequence numbers n to be 2,3, … and L respectively1Instantaneous spectrum entropy H (n) of time is obtained to obtain instantaneous spectrum entropy distribution;
wherein SPWVDA (1, m) is a time-frequency power density spectrum SPWVDA (N, m), where the time-axis number N is 1 and the energy in the mth frequency band, p (1, m) is a probability density function, and N is3Dividing interval numbers for a frequency axis, wherein H (1) is instantaneous spectrum entropy at a time axis with the sequence number n being 1;
fourthly, determining a time domain boundary of a fault characteristic region, namely determining a distortion interval in instantaneous frequency spectrum entropy distribution;
determining the position of the minimum value in the instantaneous frequency spectrum entropy distribution, then extending to the adjacent maximum value points from the position to two sides, if the maximum value is smaller than the minimum values of other regions, continuing to extend to two sidesExtending the edges until the encountered maximum value is larger than the minimum values of other areas, and determining the area between the two final maximum value points as a distortion interval; then, the position of the maximum value point on the left side is taken as the time domain lower boundary N of the fault energy regionDAnd setting the position of the right maximum value point as the upper boundary N of the time domain of the fault energy regionU。
2. A fault energy region frequency domain boundary identification method based on signal-to-noise energy difference is characterized in that: the method comprises the following steps:
first, obtain length L1The underwater robot speed signal data;
secondly, wavelet decomposition and reconstruction of multiple scales are carried out on the velocity signal by adopting a conventional wavelet transformation method to obtain multiple wavelet approximate components ujA(n), j is the wavelet decomposition scale, j is 0,1,2, …,8, wherein j is 0, which means that the velocity signal is not wavelet decomposed;
thirdly, calculating a time-frequency power density spectrum; calculating the approximate component u of the jth scale wavelet by adopting a conventional smooth pseudo-Wigner-Willi distribution algorithm and an absolute value algorithmjA(n) time-frequency Power Density Spectrum SPWVDAj(n,m)；
Fourthly, calculating the signal-to-noise energy difference distribution;
determining time domain lower boundary N of fault energy region based on instantaneous spectrum entropyDAnd upper boundary N in time domainUThe process comprises calculating a transient spectral entropy distribution; in the time-frequency power density spectrum SPWVDA (n, m), at a time shaft serial number n being 1, a probability density function p (1, m) at a time point where the time shaft serial number is located is constructed through a formula (1), and then a Shannon entropy at the time point is calculated through a formula (2), and a calculation result H (1) is an instantaneous spectrum entropy at the time shaft serial number n being 1;
repeating the steps, calculating the time shaft sequence numbers n to be 2,3, … and L respectively1Instantaneous spectrum entropy H (n) of time is obtained to obtain instantaneous spectrum entropy distribution;
wherein SPWVDA (1, m) is a time-frequency power density spectrum SPWVDA (N, m), where the time-axis number N is 1 and the energy in the mth frequency band, p (1, m) is a probability density function, and N is3Dividing interval numbers for a frequency axis, wherein H (1) is instantaneous spectrum entropy at a time axis with the sequence number n being 1;
determining a time domain boundary of a fault characteristic region, namely determining a distortion interval in instantaneous frequency spectrum entropy distribution;
determining the position of the minimum value in the instantaneous frequency spectrum entropy distribution, then extending to the adjacent maximum value points from the position to two sides at the same time, if the maximum value is smaller than the minimum values of other areas, continuing to extend to two sides until the encountered maximum value is larger than the minimum values of other areas, and determining the area between the two final maximum value points as a distortion interval; then, the position of the maximum value point on the left side is taken as the time domain lower boundary N of the fault energy regionDAnd setting the position of the right maximum value point as the upper boundary N of the time domain of the fault energy regionU；
In time frequency power density spectrum SPWVDAj(N, m), the time domain boundary NDAnd NUTime-frequency power density spectrum SPWVDA betweenj(n, m) are summed to obtain the result ESjAs fault energy, the time domain boundary NDAnd NUOutside time-frequency power density spectrum SPWVDAj(n, m) are summed to obtain the result ENjAs noise energy, EΔSNj＝ESj-ENjAs a signal-to-noise energy difference; calculating the signal-to-noise energy difference E of j equal to 0,1,2, …,8ΔSNjObtaining signal-to-noise energy difference distribution;
fifthly, determining the frequency domain boundary of the fault energy area; at the signal-to-noise energy difference EΔSNjIn distribution, determining maximum value of signal-to-noise energy difference, determining wavelet decomposition scale corresponding to the maximum value, determining wavelet approximate component corresponding to the decomposition scale, and determining frequency band [ M ] corresponding to the wavelet approximate componentDMU]Is that it isEnergy-barrier region frequency domain boundary, where MDIs the lower boundary of the frequency domain, MUThe upper boundary of the frequency domain.
3. A fault energy region boundary identification and feature extraction method based on instantaneous spectrum entropy and signal-to-noise energy difference is characterized by comprising the following steps: the method comprises the following steps:
first, respectively obtaining the lengths L1The underwater robot speed signal and the propeller control voltage change rate signal data;
secondly, determining a time domain lower boundary N of the fault energy region based on the instantaneous spectrum entropyDAnd upper boundary N in time domainU(ii) a The method comprises the following steps: wavelet decomposition and reconstruction of multiple scales are carried out on the velocity signal by adopting a conventional wavelet transformation method to obtain multiple wavelet approximate components ujA(n), j is the wavelet decomposition scale, j is 0,1,2, …,8, wherein j is 0, which means that the velocity signal is not wavelet decomposed;
calculating a time-frequency power density spectrum; calculating the approximate component u of the jth scale wavelet by adopting a conventional smooth pseudo-Wigner-Willi distribution algorithm and an absolute value algorithmjA(n) time-frequency Power Density Spectrum SPWVDAj(n,m)；
Calculating signal-to-noise energy difference distribution;
determining time domain lower boundary N of fault energy region based on instantaneous spectrum entropyDAnd upper boundary N in time domainUThe process comprises calculating a transient spectral entropy distribution; in the time-frequency power density spectrum SPWVDA (n, m), at a time shaft serial number n being 1, a probability density function p (1, m) at a time point where the time shaft serial number is located is constructed through a formula (1), and then a Shannon entropy at the time point is calculated through a formula (2), and a calculation result H (1) is an instantaneous spectrum entropy at the time shaft serial number n being 1;
repeating the steps, calculating the time shaft sequence numbers n to be 2,3, … and L respectively1Instantaneous spectrum entropy H (n) of time is obtained to obtain instantaneous spectrum entropy distribution;
wherein SPWVDA (1, m) is a time-frequency power density spectrum SPWVDA (N, m), where the time-axis number N is 1 and the energy in the mth frequency band, p (1, m) is a probability density function, and N is3Dividing interval numbers for a frequency axis, wherein H (1) is instantaneous spectrum entropy at a time axis with the sequence number n being 1;
determining a time domain boundary of a fault characteristic region, namely determining a distortion interval in instantaneous frequency spectrum entropy distribution;
determining the position of the minimum value in the instantaneous frequency spectrum entropy distribution, then extending to the adjacent maximum value points from the position to two sides at the same time, if the maximum value is smaller than the minimum values of other areas, continuing to extend to two sides until the encountered maximum value is larger than the minimum values of other areas, and determining the area between the two final maximum value points as a distortion interval; then, the position of the maximum value point on the left side is taken as the time domain lower boundary N of the fault energy regionDAnd setting the position of the right maximum value point as the upper boundary N of the time domain of the fault energy regionU(ii) a Thirdly, determining the lower boundary M of the frequency domain of the fault energy area based on the signal-noise energy differenceDAnd an upper boundary M of the frequency domainU；
In time frequency power density spectrum SPWVDAj(N, m), the time domain boundary NDAnd NUTime-frequency power density spectrum SPWVDA betweenj(n, m) are summed to obtain the result ESjAs fault energy, the time domain boundary NDAnd NUOutside time-frequency power density spectrum SPWVDAj(n, m) are summed to obtain the result ENjAs noise energy, EΔSNj＝ESj-ENjAs a signal-to-noise energy difference; calculating the signal-to-noise energy difference E of j equal to 0,1,2, …,8ΔSNjObtaining signal-to-noise energy difference distribution;
determining the frequency domain boundary of the fault energy region; at the signal-to-noise energy difference EΔSNjIn distribution, determining maximum value of signal-to-noise energy difference, determining wavelet decomposition scale corresponding to the maximum value, and determining wavelet approximate component corresponding to the decomposition scaleFrequency band [ M ] corresponding to approximate componentDMU]I.e. the fault energy region frequency domain boundary, wherein
MDIs the lower boundary of the frequency domain, MUIs the upper boundary of the frequency domain;
fourthly, extracting time-frequency energy fault characteristics of the propeller;
adopting a smooth pseudo-Wigner-Willi distribution algorithm and an absolute value algorithm to respectively calculate time-frequency power density spectrums SPWVDA of speed signals and control voltage change rate signalsU(n, m) and SPWVDAC(N, m), lower boundary N in time domainDUpper boundary N in time domainUFrequency domain lower boundary MDSum frequency domain upper bound MUWithin the enclosed rectangular area, the power density spectrums of the respective time frequencies are summed to obtain a result FU、FCRespectively as the time-frequency energy fault characteristics of the speed signal and the time-frequency energy fault characteristics of the control signal.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910001146.0A CN109633270B (en) | 2019-01-02 | 2019-01-02 | Fault energy region boundary identification and feature extraction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910001146.0A CN109633270B (en) | 2019-01-02 | 2019-01-02 | Fault energy region boundary identification and feature extraction method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109633270A CN109633270A (en) | 2019-04-16 |
| CN109633270B true CN109633270B (en) | 2020-10-02 |
Family
ID=66056466
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910001146.0A Active CN109633270B (en) | 2019-01-02 | 2019-01-02 | Fault energy region boundary identification and feature extraction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109633270B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110739984B (en) * | 2019-11-08 | 2021-07-02 | 江苏科技大学 | Camouflage communication method based on wavelet analysis |
| CN112001314A (en) * | 2020-08-25 | 2020-11-27 | 江苏师范大学 | Early fault detection method for variable speed hoist |
| CN112051836B (en) * | 2020-09-11 | 2021-09-24 | 江苏科技大学 | Underwater robot propeller state monitoring method based on multi-core model |
| CN113655778B (en) * | 2021-08-16 | 2024-03-26 | 江苏科技大学 | Underwater propeller fault diagnosis system and method based on time-frequency energy |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101672873A (en) * | 2009-10-20 | 2010-03-17 | 哈尔滨工业大学 | Detection method of transient harmonic signals of power system based on combination of Tsallis wavelet singular entropy and FFT computation |
| JP4743914B2 (en) * | 2009-06-17 | 2011-08-10 | 有限会社諏訪トラスト | Time series data analysis system and time series data analysis program |
| CN103344875A (en) * | 2013-07-02 | 2013-10-09 | 福州大学 | Classification line selection method for single-phase earth fault of resonance earthing system |
| CN109033965A (en) * | 2018-06-22 | 2018-12-18 | 江苏科技大学 | A kind of underwater robot propeller failure time-frequency characteristics Enhancement Method |
-
2019
- 2019-01-02 CN CN201910001146.0A patent/CN109633270B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP4743914B2 (en) * | 2009-06-17 | 2011-08-10 | 有限会社諏訪トラスト | Time series data analysis system and time series data analysis program |
| CN101672873A (en) * | 2009-10-20 | 2010-03-17 | 哈尔滨工业大学 | Detection method of transient harmonic signals of power system based on combination of Tsallis wavelet singular entropy and FFT computation |
| CN103344875A (en) * | 2013-07-02 | 2013-10-09 | 福州大学 | Classification line selection method for single-phase earth fault of resonance earthing system |
| CN109033965A (en) * | 2018-06-22 | 2018-12-18 | 江苏科技大学 | A kind of underwater robot propeller failure time-frequency characteristics Enhancement Method |
Non-Patent Citations (1)
| Title |
|---|
| 自主式水下机器人故障特征增强方法;张铭钧等;《哈尔滨工程大学学报》;20140930;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109633270A (en) | 2019-04-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109633270B (en) | Fault energy region boundary identification and feature extraction method | |
| CN110261109B (en) | Rolling bearing fault diagnosis method based on bidirectional memory cyclic neural network | |
| CN107860548B (en) | Approximate online flutter boundary prediction method | |
| CN106778694A (en) | A kind of gear transmission noises analysis method based on set empirical mode decomposition and SVMs | |
| CN101900789B (en) | Tolerance analog circuit fault diagnosing method based on wavelet transform and fractal dimension | |
| CN104198184A (en) | Bearing fault diagnostic method based on second generation wavelet transform and BP neural network | |
| CN107958267B (en) | Oil product property prediction method based on spectral linear representation | |
| CN108875706B (en) | Ocean structure time-frequency analysis method based on moving average and energy collection | |
| CN104215456B (en) | Plane clustering and frequency-domain compressed sensing reconstruction based mechanical fault diagnosis method | |
| CN113375939B (en) | Mechanical part fault diagnosis method based on SVD and VMD | |
| CN111367959B (en) | Zero-time-lag nonlinear expansion Granger causal analysis method | |
| CN110705456A (en) | Micro motor abnormity detection method based on transfer learning | |
| CN105137373B (en) | A kind of denoising method of exponential signal | |
| CN109696906B (en) | Underwater robot propeller fault diagnosis method based on wavelet correction Bayes convolution energy | |
| CN108645620A (en) | A kind of Fault Diagnosis of Rolling Element Bearings method based on comentropy and Multiscale Morphological | |
| CN109813547A (en) | Rotating machinery local type method for diagnosing faults based on sparse decomposition optimization algorithm | |
| CN115954017A (en) | HHT-based engine small sample sound abnormal fault identification method and system | |
| CN103926097B (en) | Method for collecting and extracting fault feature information of low-speed and heavy-load device | |
| CN103852525A (en) | Acoustic emission signal identification method based on AR-HMM | |
| CN109062051B (en) | Method for improving robot dynamics parameter identification precision | |
| CN110057918A (en) | Damage of composite materials quantitative identification method and system under strong noise background | |
| CN112016471B (en) | Rolling bearing fault diagnosis method under incomplete sample condition | |
| CN113642417A (en) | Improved wavelet algorithm-based denoising method for partial discharge signals of insulated overhead conductor | |
| CN113343869A (en) | Electroencephalogram signal automatic classification and identification method based on NTFT and CNN | |
| CN114813129B (en) | Rolling bearing acoustic signal fault diagnosis method based on WPE and EMD |
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 |