A blade crack fault identification method based on adaptive stochastic resonance and its application

Abstract

This invention discloses a method and application for blade crack fault identification based on adaptive stochastic resonance. The blade crack fault identification method includes: acquiring blade shaft vibration signals and calculating signal noise intensity; decomposing the blade vibration signal to obtain several components of different frequency bands, selecting components based on the center frequency band component, and reconstructing the signal; calculating the amplitude of the reconstructed signal; estimating the signal input signal-to-noise ratio (SNR) based on the signal noise intensity and the amplitude of the reconstructed signal; obtaining the optimal distribution circle of the stochastic resonance parameters based on the relationship between the SNR and the optimal stochastic resonance parameters; determining the center value and range of the stochastic resonance parameters of the reconstructed signal; optimizing the parameters using a genetic algorithm and obtaining the optimal stochastic resonance output to achieve fault identification of compressor blade cracks. This invention is simple and easy to implement, and compared with other existing stochastic resonance technologies, it can adaptively and accurately achieve fault identification of compressor blade cracks.

Description

Technical Field

[0001] The invention relates to the technical field of rotating machinery vibration signal analysis, in particular to a blade crack fault identification method based on signal-to-noise ratio estimation and adaptive stochastic resonance. Background Art

[0002] Compressors are widely used in petrochemical, electric power and other fields. Blades, as core components, are prone to crack failures under the action of centrifugal force, friction and unstable airflow loads, which affects the normal operation of the entire compressor. Therefore, timely detection of blade crack failures is of great significance to ensure the safe and stable operation of the compressor. In addition, the vibration signal measured in actual engineering is mixed with strong background noise, which makes the identification of compressor blade crack failure extremely complex and difficult. Therefore, how to effectively overcome noise interference and realize adaptive compressor blade crack fault identification deserves in-depth study.

[0003] Traditional signal processing methods mainly overcome noise interference by suppressing noise, but at the same time, weak fault information will also be interfered with, which will affect the recognition effect. Stochastic resonance can cleverly use noise to enhance the target and has good noise resistance. Traditional stochastic resonance methods require accurate determination of the target through prior knowledge to achieve enhancement, which is difficult to widely use for non-standard parts such as compressor blades. Therefore, there is an urgent need for a compressor blade crack fault feature recognition method that can utilize interference noise and achieve adaptive stochastic resonance output based on the characteristics of the vibration signal. Summary of the invention

[0004] In view of the above-mentioned shortcomings, the present invention proposes a blade crack fault identification method based on signal-to-noise ratio estimation and adaptive stochastic resonance, which overcomes the current situation that the existing features are difficult to handle, the compressor noise interference is large, and the characteristic frequency of the blade crack fault cannot be accurately identified, and effectively realizes the adaptive determination of stochastic resonance parameters and blade crack fault identification.

[0005] To solve the above technical problems, the embodiments of the present invention adopt the following technical solutions:

[0006] A blade crack fault identification method based on adaptive stochastic resonance includes the following steps:

[0007] Step 1: Install an acceleration sensor near the compressor motor shaft, collect vibration signals, and calculate the signal noise intensity using second-order and fourth-order moments;

[0008] Step 2: Use the noise reduction empirical mode decomposition algorithm to decompose the vibration signal to obtain several components in different frequency bands, select the components according to the center frequency band components and reconstruct the signal;

[0009] Step 3: Calculate the reconstructed signal amplitude and estimate the signal-to-noise ratio based on the signal-to-noise intensity and amplitude;

[0010] Step 4: Use genetic algorithm to analyze the relationship between the signal input signal-to-noise ratio and the stochastic resonance optimal parameters, and obtain the stochastic resonance optimal parameter distribution circle;

[0011] Step 5: Determine the center value and range of the stochastic resonance parameter of the recombinant signal according to the input signal-to-noise ratio and the stochastic resonance optimal parameter distribution circle;

[0012] Step 6: Use genetic algorithm to optimize parameters and obtain the optimal stochastic resonance output. Observe whether the stochastic resonance output signal has an obvious peak value different from the rotation frequency, so as to realize the fault identification of compressor blade cracks.

[0013] Preferably, the specific steps of step 2 of using the noise reduction empirical mode decomposition algorithm to decompose and reorganize the vibration signal are:

[0014] Step 2.1, create a Gaussian white noise with an intensity equal to the noise intensity estimated in step 1;

[0015] Step 2.2, subtracting the created Gaussian white noise from the vibration signal to obtain a residual signal;

[0016] Step 2.3, using the empirical mode decomposition algorithm to process the remaining signal to obtain several components in different frequency bands;

[0017] Step 2.4, repeat steps 2.1, 2.2 and 2.3 10 times to obtain 10 components of several different frequency bands, and perform average processing according to the frequency band to obtain an integrated average component. The expression of the integrated average component is:

[0018]

[0019] Among them, IMF ki represents the i-th component obtained by the k-th empirical mode decomposition, IMF i represents the i-th integrated average component.

[0020] Step 2.5, select all components whose center frequency band is within the estimated fault characteristic frequency range, and perform signal reorganization. The expression of the signal reorganization is:

[0021]

[0022] Among them, xc represents the recombination signal, IMFc i represents the i-th selected component, and m represents the number of selected components.

[0023] Preferably, the expression for estimating the signal-to-noise ratio of the input signal in step 3 is:

[0024]

[0025] Among them, SNR i A represents the signal-to-noise ratio of the signal input. f represents the amplitude of the recombination signal, D f Indicates the signal-to-noise intensity.

[0026] Preferably, the specific steps of establishing the stochastic resonance optimal parameter distribution circle in step 4 are:

[0027] Step 4.1, create a cosine signal with an amplitude of 0.2 and a frequency of 0.01Hz, and add Gaussian white noise of different intensities so that the input signal-to-noise ratio increases from -30dB to -15dB at 1dB intervals, and establish the original signal with different input signal-to-noise ratios. The expression of the original signal is:

[0028] x i (t)=0.2*cos(0.02πt)+δ i (t)

[0029] Among them, x i (t) represents the i-th original signal, δ i (t) represents the i-th Gaussian white noise.

[0030] Step 4.2, use genetic algorithm to search for stochastic resonance parameters a and b, and finally get 200 pairs of stochastic resonance parameters, where the initial population of the genetic algorithm is 200, the hybridization rate is 0.7, the selection rate is 0.5, the mutation rate is 0.001, the binary code length is 22, and the inheritance is 100 generations.

[0031] Step 4.3, with the mean of 200 pairs of stochastic resonance parameters as the center of the circle and the distance from the center to the farthest point as the radius, obtain the stochastic resonance optimal parameter distribution circle.

[0032] Preferably, the central value of the stochastic resonance parameter in step 5 is the center of the stochastic resonance optimal parameter distribution circle corresponding to the input signal-to-noise ratio, and the stochastic resonance parameter range is calculated by the stochastic resonance optimal parameter distribution circle and the expansion multiple, and the expansion multiple is 10 to 20 times the ratio of the estimated fault characteristic frequency range and the cosine signal frequency created in step 4.1. The expression of the stochastic resonance parameter range is:

[0033] a r =k m *(a0±r m )

[0034] b r =k m *(b0±r m )

[0035] Among them, a r and b r represents the range of stochastic resonance parameters, k m represents the expansion factor, a0 and b0 represent the center value of the stochastic resonance parameter, r m represents the radius of the stochastic resonance optimal parameter distribution circle.

[0036] Preferably, the parameter optimization in step 6 is to maximize the amplitude of the maximum frequency of the amplitude different from the rotation frequency as the optimization target, and the frequency different from the rotation frequency is the frequency other than the rotation frequency and its multiples.

[0037] The beneficial effects of the present invention are as follows:

[0038] 1) The present invention provides a method for identifying a crack fault in a compressor blade based on signal-to-noise ratio estimation and adaptive stochastic resonance. The method adaptively determines the stochastic resonance parameters by estimating the signal-to-noise ratio of the input signal, achieves the optimal stochastic resonance output, and can identify a crack fault in the compressor blade based on the output result. This avoids the disadvantage that the parameters of the original stochastic resonance algorithm are difficult to adaptively determine for compressor blades. Compared with the traditional adaptive stochastic resonance method, the method of the present invention can accurately identify the characteristic frequency of blade crack faults in a strong noise environment, and the fault characteristic amplitude and signal-to-noise ratio of the stochastic resonance output are better than those of the traditional adaptive stochastic resonance method, thereby improving the output quality of the stochastic resonance.

[0039] 2) Based on the analysis of the relationship between the input signal-to-noise ratio and the stochastic resonance parameters, the present invention provides a method for determining the central value and range of the stochastic resonance parameters. By analyzing the input signal-to-noise ratio and the stochastic resonance optimal parameter distribution circle, the central value and range of the stochastic resonance parameters can be determined, thereby achieving rapid optimization and selection of the stochastic resonance parameters.

[0040] 3) The present invention provides a noise reduction empirical mode decomposition algorithm, which effectively suppresses the noise and interference components in the components of empirical mode decomposition by subtracting Gaussian white noise with the same noise intensity as the original signal and performing integrated averaging, thereby enhancing the characteristic frequency of crack faults and making the identification of existing compressor blade crack faults more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 is a flowchart of an embodiment of the present invention;

[0042] Figure 2 is the spatial distribution result of the optimal parameter distribution circle of random resonance in the embodiment of the present invention

[0043] Figure 3 The optimal parameter distribution of random resonance in the embodiment of the present invention is the circular plane distribution result

[0044] Figure 4 is a time-frequency diagram of the simulation signal x(t) in an embodiment of the present invention.

[0045] Figure 5 3 is a time-frequency diagram of a simulated signal after adding noise in an embodiment of the present invention.

[0046] Figure 6 It is a time domain diagram of a signal decomposed by a noise reduction empirical mode decomposition algorithm in an embodiment of the present invention.

[0047] Figure 7 It is a frequency domain diagram of a signal decomposed by a noise reduction empirical mode decomposition algorithm in an embodiment of the present invention.

[0048] Figure 8 It is a time-frequency diagram of the signal reconstructed by the noise reduction empirical mode decomposition algorithm in an embodiment of the present invention.

[0049] Fig. 9 It is a time-frequency diagram of the optimal stochastic resonance output signal in an embodiment of the present invention.

[0050] Fig.10 are the stochastic resonance output results of different methods in the embodiments of the present invention; wherein a1 is the stochastic resonance output time domain diagram of the MNT-BSR method, and a2 is the stochastic resonance output frequency domain diagram of the MNT-BSR method; b1 is the stochastic resonance output time domain diagram of the AMSTSR method, and b2 is the stochastic resonance output frequency domain diagram of the AMSTSR method; c1 is the stochastic resonance output time domain diagram of the ASR method, and c2 is the stochastic resonance output frequency domain diagram of the ASR method; d1 is the stochastic resonance output time domain diagram of the ACBSR method, and d2 is the stochastic resonance output frequency domain diagram of the ACBSR method. DETAILED DESCRIPTION

[0051] The following description is merely exemplary in nature and is not intended to limit the present disclosure, application, or uses. It should be understood that throughout the drawings, corresponding reference numerals indicate the same or corresponding components and features.

[0052] like Figure 1 As shown, a compressor blade crack fault identification method based on signal-to-noise ratio estimation and adaptive stochastic resonance according to an embodiment of the present invention comprises the following steps:

[0053] Step 1: Install an acceleration sensor near the compressor motor shaft, collect vibration signals, and calculate the signal noise intensity using second-order and fourth-order moments;

[0054] Step 2: Use the noise reduction empirical mode decomposition algorithm to decompose the vibration signal to obtain several components in different frequency bands, select the components according to the center frequency band components and reconstruct the signal;

[0055] Step 3: Calculate the reconstructed signal amplitude and estimate the signal-to-noise ratio based on the signal-to-noise intensity and amplitude;

[0056] Step 4: Use genetic algorithm to analyze the relationship between the signal input signal-to-noise ratio and the stochastic resonance optimal parameters, and obtain the stochastic resonance optimal parameter distribution circle;

[0057] Step 5: Determine the center value and range of the stochastic resonance parameter of the recombinant signal according to the input signal-to-noise ratio and the stochastic resonance optimal parameter distribution circle;

[0058] Step 6: Use genetic algorithm to optimize parameters and obtain the optimal stochastic resonance output. Observe whether the stochastic resonance output signal has an obvious peak value different from the rotation frequency, so as to realize the fault identification of compressor blade cracks.

[0059] In the above embodiment, the vibration signal of the compressor is collected by an acceleration sensor, and the signal noise intensity is calculated using the second-order and fourth-order moments. Then, the amplitude of the signal is obtained by using the noise reduction empirical mode decomposition, and the signal input signal-to-noise ratio is estimated. Then, the central value and range of the random resonance parameters are determined based on the relationship between the input signal-to-noise ratio and the random resonance optimal parameter distribution circle. Finally, a genetic algorithm is combined to achieve the optimal random resonance output and realize the identification of compressor blade crack faults.

[0060] It should be noted that the expressions for calculating the signal noise intensity by the second-order and fourth-order moments are:

[0061]

[0062] Among them, P n represents the signal noise intensity, m2 and m4 represent the second and fourth order moments, k y and k n Represents the relevant parameters of the signal. For the crack fault vibration signal, k is taken y is 1, k n is 2.

[0063] Step 2: The specific steps of decomposing and recombining the vibration signal using the noise reduction empirical mode decomposition algorithm are as follows:

[0064] Step 2.1, create a Gaussian white noise with an intensity equal to the noise intensity estimated in step 1;

[0065] Step 2.2, subtracting the created Gaussian white noise from the vibration signal to obtain a residual signal;

[0066] Step 2.3, using the empirical mode decomposition algorithm to process the remaining signal to obtain several components in different frequency bands;

[0067] Step 2.4, repeat steps 2.1, 2.2 and 2.3 10 times to obtain 10 components of several different frequency bands, and perform average processing according to the frequency band to obtain an integrated average component. The expression of the integrated average component is:

[0068]

[0069] Among them, IMF ki represents the i-th component obtained by the k-th empirical mode decomposition, IMF i represents the i-th integrated average component.

[0070] Step 2.5, select all components whose center frequency band is within the estimated fault characteristic frequency range, and perform signal reorganization. The expression of the signal reorganization is:

[0071]

[0072] Among them, xc represents the recombination signal, IMFc i represents the i-th selected component, and m represents the number of selected components. The estimated fault characteristic frequency range can be roughly estimated based on the compressor experiment and experience, which is generally in the tens of Hz, so the range of 10Hz-100Hz can be taken.

[0073] It should be noted that due to the randomness of Gaussian white noise, there will be certain differences in the results of decomposing and recombining the vibration signal using the noise reduction empirical mode decomposition algorithm. However, due to the integrated average of 10 results, the difference is small and does not affect the subsequent processing.

[0074] Preferably, the expression for estimating the signal-to-noise ratio of the input signal in step 3 is:

[0075]

[0076] Among them, SNR i A represents the signal-to-noise ratio of the signal input. f represents the amplitude of the recombination signal, D f Indicates the signal-to-noise intensity.

[0077] Preferably, the specific steps of establishing the stochastic resonance optimal parameter distribution circle in step 4 are:

[0078] Step 4.1, create a cosine signal with an amplitude of 0.2 and a frequency of 0.01Hz, and add Gaussian white noise of different intensities so that the input signal-to-noise ratio increases from -30dB to -15dB at 1dB intervals, and establish the original signal with different input signal-to-noise ratios. The expression of the original signal is:

[0079] x i (t)=0.2*cos(0.02πt)+δ i (t)

[0080] Among them, x i (t) represents the i-th original signal, δ i (t) represents the i-th Gaussian white noise.

[0081] Step 4.2, use genetic algorithm to search for stochastic resonance parameters a and b, and finally get 200 pairs of stochastic resonance parameters, where the initial population of the genetic algorithm is 200, the hybridization rate is 0.7, the selection rate is 0.5, the mutation rate is 0.001, the binary code length is 22, and the inheritance is 100 generations.

[0082] Step 4.3, take the mean of 200 pairs of stochastic resonance parameters as the center of the circle, and the distance from the center to the farthest point as the radius to obtain the stochastic resonance optimal parameter distribution circle. The distribution circle result is as follows: Figure 2 and Figure 3 The specific results are shown in Table 1.

[0083] Table 1 Distribution circles of optimal parameters of stochastic resonance under different input signal-to-noise ratios

[0084]

[0085]

[0086] It should be noted that although the center value changes significantly under different input signal-to-noise ratios, the radius of the distribution circle is stable and less than 0.4. The higher aggregation indicates that the distribution circle of the stochastic resonance optimal parameters is compact and the range of the stochastic resonance optimal parameters can be determined.

[0087] In step 5, the center value of the stochastic resonance parameter is the center of the stochastic resonance optimal parameter distribution circle corresponding to the input signal-to-noise ratio. The stochastic resonance parameter range is calculated by the stochastic resonance optimal parameter distribution circle and the expansion multiple. The expansion multiple is 10 to 20 times the ratio of the estimated fault characteristic frequency range and the cosine signal frequency created in step 4.1. The expression of the stochastic resonance parameter range is:

[0088] a r =k m *(a0±r m )

[0089] b r =k m *(b0±r m )

[0090] Among them, a r and b r represents the range of stochastic resonance parameters, k m represents the expansion factor, a0 and b0 represent the center value of the stochastic resonance parameter, r m represents the radius of the stochastic resonance optimal parameter distribution circle.

[0091] In step 6, the parameter optimization is to maximize the amplitude of the maximum frequency other than the rotation frequency as the optimization target, where the other than the rotation frequency is the frequency other than the rotation frequency and its multiples, and the optimization target is as follows. Parameter optimization is to use genetic algorithm to search for parameter r m and k m The optimized stochastic resonance parameters are applied to the recombined signal to realize stochastic resonance output, and finally the amplitude and frequency of the maximum frequency of the amplitude different from the rotation frequency are obtained.

[0092] OT=MAX{Af tn}

[0093] Among them, OT represents the optimization target, Af tn Indicates the amplitude of the maximum frequency that differs from the rotation frequency.

[0094] The solution of this application is further described with the following specific embodiments:

[0095] Construct a frequency modulation simulation signal - a simulation signal, and perform Fourier transformation on it to obtain the time domain and frequency domain waveforms, such as Figure 4 As shown. The simulation signal is composed of the fault characteristic frequency f c , shaft frequency f r and the pass frequency f m The three-part blend consists of c 、f r and f m 35Hz, 20Hz and 240Hz respectively (simulating 12 blades):

[0096] x(t)=0.2(1+0.1cos(2πf c t)+0.3sin(2πf r t))*sin(2πf m t)

[0097] At the same time, 0.5dB Gaussian white noise is introduced to simulate the effect of noise on the signal. The results are as follows: Figure 5 As shown in the figure, the sampling frequency of the simulation signal x(t) is 5120 Hz and the number of sampling points is 5120. Figure 6-Figure 8 The above simulation signal is decomposed and reconstructed by using the noise reduction empirical mode decomposition algorithm of the method of the present application. On this basis, the central values ​​of the stochastic resonance parameters are adaptively determined by the method of the present application, 1.05×15×5000 and 1.03×15×5000, with a range of (1.05±0.06)×15×5000 and (1.03±0.06)×15×5000. The optimal stochastic resonance parameters are determined to be 1×15.38×5000 and 1.02×15.38×5000 by genetic algorithm optimization. The optimal stochastic resonance output is as follows: Fig. 9It can be found that in addition to the shaft frequency, the fault characteristic frequency of 35Hz is accurately identified, and the amplitude of the fault characteristic frequency is significantly improved after optimization, which verifies the reliability of the method of this application.

[0098] In order to verify the effectiveness of the compressor blade crack fault identification method based on signal-to-noise ratio estimation and adaptive stochastic resonance of the present invention, it is compared with several adaptive stochastic resonance methods, including an adaptive stochastic resonance method based on multi-scale noise tuning (MNT-BSR13), an adaptive stochastic resonance method based on grid search (ASR), an adaptive stochastic resonance method based on coupled bistable system (ACBSR) and an adaptive stochastic resonance method based on adaptive multi-scale noise tuning (AMSTSR). Among them, the MNT-BSR13 method performs multi-scale tuning of noise to achieve adaptive stochastic resonance output; ASR performs grid search with the highest output signal-to-noise ratio as the objective function to achieve adaptive stochastic resonance output; ACBSR improves the potential function and achieves adaptive stochastic resonance output with the weighted spectrum peak signal-to-noise ratio as the target; AMSTSR uses the weighted power spectrum kurtosis as the optimization index to achieve adaptive stochastic resonance output after tuning the multi-scale noise. The random resonance output results of the comparison methods are shown in the figure. Fig.10 And as shown in Table 2.

[0099] Table 2 Stochastic resonance output results of different methods

[0100]

[0101] exist Fig.10 In the study, the MNT-BSR method combines multi-scale noise tuning, but the recognition and output effects are poor for signals with high noise intensity and weak target amplitude. The AMSTSR method obtains 63Hz noise as the highest frequency. Although the signal-to-noise ratio is large, it achieves the wrong target. The ASR and ACBSR methods also make mistakes in recognition, and the signal-to-noise ratio is too small for fault diagnosis. By comparing with several traditional methods, the method of the present invention can accurately identify the fault characteristic frequency of 35Hz, while other methods cannot accurately identify the target frequency and the output effect is poor. In addition, the maximum signal-to-noise ratio of the method of the present invention is 25.67dB, which shows that SEASR can identify fault characteristics and realize fault diagnosis under large noise.

[0102] Compared with the prior art, the compressor blade crack fault identification method based on signal-to-noise ratio estimation and adaptive stochastic resonance of the present invention achieves optimal stochastic resonance output by adaptively determining stochastic resonance parameters by estimating the signal-to-noise ratio of the input signal, and can realize compressor blade crack fault identification based on the output result, thereby avoiding the deficiency of the original stochastic resonance algorithm parameters that are difficult to adaptively determine for the compressor blades and improving the output quality of stochastic resonance.

[0103] The specific implementation cases described in the present invention are only preferred implementation cases of the present invention and are not intended to limit the scope of implementation of the present invention. That is, all equivalent changes and modifications made according to the content of the patent application scope of the present invention should be regarded as the technical scope of the present invention.

Claims

1. A blade crack fault identification method based on adaptive stochastic resonance, characterized in that: The following steps are involved: Step 1: Collect blade shaft vibration signal and calculate signal noise intensity; Step 2: Use the noise reduction empirical mode decomposition algorithm to decompose the blade vibration signal to obtain several components in different frequency bands, select the components according to the center frequency band components and reconstruct the signal; calculate the amplitude of the reconstructed signal; Step 3: Estimate the signal-to-noise ratio of the signal input based on the signal-to-noise intensity and the amplitude of the reconstructed signal; Step 4: Use genetic algorithm to analyze the relationship between the signal input signal-to-noise ratio and the stochastic resonance optimal parameters, and obtain the stochastic resonance optimal parameter distribution circle; Step 5: Determine the center value and range of the stochastic resonance parameter of the recombinant signal according to the input signal-to-noise ratio and the stochastic resonance optimal parameter distribution circle; Step 6: Use genetic algorithm to optimize parameters and obtain the optimal stochastic resonance output; Step 7: According to the output optimal stochastic resonance output signal, the fault identification of the compressor blade crack is realized.

2. A blade crack fault identification method based on adaptive stochastic resonance according to claim 1, characterized in that: In step 1, the method for calculating the signal noise intensity is the second-order and fourth-order moments, the recursive averaging method, the minimum tracking method or the histogram statistics method.

3. A blade crack fault identification method based on adaptive stochastic resonance according to claim 1, characterized in that: The step 2 uses a noise reduction empirical mode decomposition algorithm to decompose the blade vibration signal to obtain components of several different frequency bands, selects components according to the center frequency band components and recombines the signals; and calculates the amplitude of the recombined signal, including: Step 2.1, create a Gaussian white noise with an intensity equal to the noise intensity estimated in step 1; Step 2.2, subtracting the created Gaussian white noise from the vibration signal to obtain a residual signal; Step 2.3, using the empirical mode decomposition algorithm to process the remaining signal to obtain several components in different frequency bands; Step 2.4, repeat steps 2.1, 2.2 and 2.3 10 times to obtain 10 components of several different frequency bands, and perform average processing according to the frequency band to obtain an integrated average component. The expression of the integrated average component is: Among them, IMF ki represents the i-th component obtained by the k-th empirical mode decomposition, IMF i represents the i-th integrated average component; Step 2.5, select all components whose center frequency band is within the estimated fault characteristic frequency range, and perform signal reorganization. The expression of the signal reorganization is: Among them, xc represents the recombination signal, IMFc i represents the i-th selected component, and m represents the number of selected components; Step 2.6, calculate the amplitude of the recombination signal: Among them, A f represents the amplitude of the recombination signal, N represents the length of the recombination signal, xc i Represents the amplitude of the recombined signal at the i-th moment.

4. A blade crack fault identification method based on adaptive stochastic resonance according to claim 3, characterized in that: In step 2.3, the process of the empirical mode decomposition algorithm is: a. Connect the local maximum points through the cubic spline interpolation curve to form the upper envelope and connect the local minimum points to form the lower envelope, and calculate the mean m1 of the upper and lower envelopes; b. Subtract the mean m1 of the upper and lower envelopes from the input signal, as shown below: x1(t)=x(t)-m1 Where x(t) represents the input signal, m1 represents the mean of the first upper and lower envelopes, and x1(t) represents the signal obtained after the first decomposition; c. Take the x1(t) signal as the new input signal and repeat step b above until the stopping rule is satisfied, that is, the number of local extreme points and zero crossing points of the function in the entire time range must be equal or differ by at most one; at any time point, the mean of the upper envelope of the local maximum value and the lower envelope of the local minimum value must be zero; d. The signal that satisfies the stopping rule is the first component IMF of the first decomposition 11 , subtract the IMF from the original input signal x(t) 11 , repeat the above step c to obtain the first several components and the first residual, as shown in the following formula Among them, IMF 1i represents the i-th component obtained by the first empirical mode decomposition, R 1n represents the first order residual.

5. A blade crack fault identification method based on adaptive stochastic resonance according to claim 2, characterized in that: In step 3, the estimated signal input signal-to-noise ratio is: Among them, SNR i A represents the signal-to-noise ratio of the signal input. f represents the amplitude of the recombination signal, D f Indicates the signal-to-noise intensity.

6. A blade crack fault identification method based on adaptive stochastic resonance according to claim 1, characterized in that: The step 4, using a genetic algorithm to analyze the relationship between the signal input signal-to-noise ratio and the stochastic resonance optimal parameters, to obtain the stochastic resonance optimal parameter distribution circle, includes: Step 4.1, create a cosine signal with amplitude A and frequency f, and add Gaussian white noise of different intensities so that the input signal-to-noise ratio increases from -30dB to -15dB at 1dB intervals, and establish the original signal with different input signal-to-noise ratios. The expression of the original signal is: x i (t)=A*cos(2πft)+δ i (t) Among them, x i (t) represents the i-th original signal, δ i (t) represents the i-th Gaussian white noise; Step 4.2, use genetic algorithm to search for stochastic resonance parameters a and b, and finally obtain p pairs of stochastic resonance parameters, where the initial population of the genetic algorithm is p, the hybridization rate is k1, the selection rate is k2, the mutation rate is k3, and the binary code length is L m , genetic algebra L N ; Step 4.3, taking the mean of the stochastic resonance parameters of p pairs as the center of the circle and the distance from the center to the farthest point as the radius, obtain the stochastic resonance optimal parameter distribution circle; The amplitude A ranges from 0.1 to 0.5, the frequency f ranges from 0.005 to 0.02 Hz, the initial population m ranges from 100 to 500, the hybridization rate k1 ranges from 0 to 1, the selection rate k2 ranges from 0 to 1, the mutation rate k3 ranges from 0 to 0.01, and the binary code length L m The range is 10-100, the genetic number L N The range is 1-1000 generations.

7. A blade crack fault identification method based on adaptive stochastic resonance according to claim 6, characterized in that: In step 5, the central value and range of the stochastic resonance parameter of the recombinant signal determined according to the input signal-to-noise ratio and the stochastic resonance optimal parameter distribution circle are: a r =k m *(a0±r m ) b r =k m *(b0±r m ) Among them, a r and b r represents the range of stochastic resonance parameters, k m represents the expansion factor, a0 and b0 represent the center value of the stochastic resonance parameter, r m represents the radius of the stochastic resonance optimal parameter distribution circle.

8. A blade crack fault identification method based on adaptive stochastic resonance according to claim 7, characterized in that: Enlargement factor k m It is 10 to 20 times the ratio of the estimated fault characteristic frequency range and the cosine signal frequency created in step 4.

1.

9. A blade crack fault identification method based on adaptive stochastic resonance according to claim 7, characterized in that: In step 6, the parameter optimization is to maximize the amplitude of the maximum frequency of the amplitude different from the rotation frequency as the optimization target, and the frequency different from the rotation frequency is the frequency other than the rotation frequency and its multiples.

10. A blade crack fault identification method based on the adaptive stochastic resonance according to any one of claims 1 to 9 is applied to a compressor to identify crack faults in compressor blades.

Images