Stability evaluation method of mixed-flow hydraulic turbine by fusing shock characteristic values of acoustic and vibration signals

Abstract

The application discloses a mixed-flow water turbine stability evaluation method fusing acoustic vibration signal impact characteristic values, and specifically comprises the following steps: step 1, collecting vibration signals of a water turbine movable guide vane and an impeller gap by using a vibration sensor, and collecting noise signals between the movable guide vane and the impeller by using a noise sensor; step 2, filtering the vibration signals collected by the vibration sensor by using maximum correlation kurtosis deconvolution to obtain filtered signals y; step 3, filtering the noise signals collected by the noise sensor by using maximum correlation kurtosis deconvolution to obtain filtered signals v; step 4, extracting energy entropy ratios of the signals y and v respectively; and step 5, obtaining a fusion characteristic value R of the comprehensive vibration signals and the noise signals according to the energy entropy ratios obtained in step 4 y,v The application can quantitatively evaluate the stability of the mixed-flow water turbine when running under complex working conditions.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of renewable energy, and relates to a Francis turbine stability evaluation method fusing sound-vibration signal impact characteristic values. BACKGROUND

[0002] In the field of renewable energy and the like, Francis turbines are widely used due to their excellent performance. When the Francis turbine is running in a stable working condition, the vibration is smooth and the noise is small. However, in complex working conditions such as starting and speed changing, the stability deteriorates, causing severe vibration and strong noise of the turbine shell, and even causing damage to the turbine body components and pipe rupture to cause internal fluid leakage. As can be seen, the running stability of the Francis turbine directly affects the performance of the Francis turbine in vibration noise control and overall safety and stability. Traditional evaluation methods for the stability of Francis turbines mainly focus on signal analysis of a single measuring point, such as analysis of a single signal such as a noise signal or a vibration signal. However, these methods are only suitable for evaluating the stability of the Francis turbine running in a simple working condition, but when facing complex and variable working conditions, the information reflected by the traditional stability evaluation method is one-sided, and the influence of the complex vibration and noise on the stability of the Francis turbine cannot be accurately judged. SUMMARY

[0003] The purpose of the present application is to provide a Francis turbine stability evaluation method fusing sound-vibration signal impact characteristic values, which can quantitatively evaluate the stability of the Francis turbine running in a complex working condition.

[0004] The technical solution adopted by the present application is a Francis turbine stability evaluation method fusing sound-vibration signal impact characteristic values, which specifically comprises the following steps:

[0005] Step 1, using a vibration sensor to collect vibration signals of the turbine moving guide vane and the impeller gap, and using a noise sensor to collect noise signals between the moving guide vane and the impeller;

[0006] Step 2, using maximum correlation kurtosis deconvolution to filter the vibration signals collected by the vibration sensor to obtain filtered signals y;

[0007] Step 3, using maximum correlation kurtosis deconvolution to filter the noise signals collected by the noise sensor to obtain filtered signals v;

[0008] Step 4, extracting the energy entropy ratio of signals y and v, respectively;

[0009] Step 5, obtaining the fusion characteristic value R of the comprehensive vibration signal and the noise signal according to the energy entropy ratio obtained in step 4 y,v .

[0010] The application is also characterized in that:

[0011] The specific process of step 2 is as follows:

[0012] Step 2.1, the correlation kurtosis CK of the vibration signal x is calculated according to the following formula (1): n Mx (T x ):

[0013]

[0014] Wherein, N x is the sample number of the input vibration signal, T x is the deconvolution period of the vibration signal, M x is the filter shift number of the vibration signal, and m x represents the current shift number of the filter of the vibration signal.

[0015] Step 2.2, the correlation kurtosis calculated in step 2.1 is maximized by using the following formula (2):

[0016]

[0017] Wherein, is the filter coefficient of the vibration signal, and k x is the current filter length of the vibration signal.

[0018] Step 2.3, the optimal filter coefficient corresponding to the maximum correlation kurtosis of the vibration signal is calculated by the following formula (3):

[0019]

[0020] Step 2.4, the signal y output by the filter is calculated by the following formula (4):

[0021]

[0022] Wherein, L x is the filter length of the vibration signal.

[0023] In step 2.4, the filter coefficient f kx of the vibration signal is calculated by the following formula (5):

[0024]

[0025] Wherein:

[0026]

[0027] The specific process of step 3 is as follows: ​

[0028] Step 3.1, the correlation kurtosis of the noise signal c is calculated by using the following formula (6) n

[0029]

[0030] where N c is the sample number of the input noise signal, T c is the deconvolution period for the noise signal, M c is the filter shift number for the noise signal, m c represents the current filter shift number for the noise signal;

[0031] Step 3.2, the correlation kurtosis calculated in step 3.1 is maximized by using the following formula (7)

[0032]

[0033] where f is the filter coefficient for the noise signal, k c is the current filter length for the noise signal;

[0034] Step 3.3, the optimal filter coefficient corresponding to the maximum correlation kurtosis of the noise signal is obtained by the following formula (8)

[0035]

[0036] Step 3.4, the signal v output by the filter is calculated by the following formula (9)

[0037]

[0038] where L c is the filter length for the noise signal, k c is the length of the current filter for the noise signal, f is the filter coefficient for the noise signal.

[0039] In step 3.4, the expression of the filter coefficient f kc is shown in formula (10)

[0040]

[0041] where:

[0042]

[0043] The specific process of step 4 is as follows:

[0044] ​Step 4.1, select the Hann window function to frame the signal y and v, the specific calculation formula is as follows:

[0045]

[0046] Where, η y = 0. 5, w v is the window length, A y and A v are sample point indexes in the window, the value is 0 < A y < η y - 1, 0 < A v < η v - 1;

[0047] Step 4.2, for the time series of signal y (o y ) and v (o v ) with length N y and N v , the i-th frame signal obtained after the windowing function δ (u y ) and δ (u v ) are framed is t i (o) and g i (o) respectively, which satisfies the following conditions:

[0048] t i (o) = δ (A y ) y ((i-1) i ny + o y ), 1 ≤ o y ≤ w leny , 1 ≤ i ≤ J ny (13)

[0049] g i (o) = δ (A v ) v ((i-1) i nv + o v ), 1 ≤ o v ≤ w lenv , 1 ≤ i ≤ J nv (14)

[0050] Where, w leny , w lenv are the lengths of a frame of different signals; i ny , i nv are the lengths of a frame of different signals; J ny and J nv are the total number of frames after framing different signals;

[0051] Step 4.3, t i (o) and g i(o) Fourier transform is carried out, and the specific formula is as follows:

[0052]

[0053] Wherein, T i (χ y ) and G i (χ v ) are the frequency spectrum functions after Fourier transform; χ y,v is the independent variable, with Hz as the unit, is a complex function, and the specific expression is as follows:

[0054]

[0055] Step 4.4, the energy spectrum of the frequency component J y and J v of the ξ ξy th and ξ ξv th spectrum line after Fourier transform is Y i (ξ y ) and V i (ξ v ) respectively, and the definition of the normalized spectrum probability density function P i (ξ y ) and Q i (ξ v ) of each frequency component under different signals is as follows:

[0056]

[0057] Wherein, P i (ξ y ) is the probability density corresponding to the J ξy th frequency component J y of the i th frame; Q i (ξ v ) is the probability density corresponding to the J ξv th frequency component J v of the i th frame;

[0058] Step 4.5, the specific expressions of the spectrum entropy H yi of the vibration signal in the i th frame and the spectrum entropy H vi of the noise signal in the i th frame are as follows:

[0059]

[0060]

[0061] Step 4.6, the energy E yi of the vibration signal in the i th frame and the energy E viThe specific expression of is as follows:

[0062]

[0063] Wherein, a is used for adjusting the degree of energy change and helps to distinguish signal mutation;

[0064] Step 4.7, the energy entropy ratio R of the vibration signal in the i-th frame yi And the energy entropy ratio R of the noise signal in the i-th frame vi The specific expression of is as follows:

[0065]

[0066] In step 4.4, Y i (ξ y ) = |J ξy | 2 ; V i (ξ v ) = |J ξv | 2 .

[0067] The specific process of step 5 is: the fusion feature value R y,v of the integrated vibration signal and the noise signal is calculated through formula (27):

[0068]

[0069] The beneficial effects of the present application are: based on the time domain signals of two measuring points, the fusion impulsive feature value of the mixed flow turbine stability evaluation method of the sound and vibration signals improves the precision and reliability of the detection of the impulsive components in the noise and vibration signals. The mixed flow turbine stability evaluation method of the fusion impulsive feature value of the sound and vibration signals quantitatively evaluates the operation stability of the mixed flow turbine, provides reliable diagnostic basis, helps to find potential problems in time and perform maintenance and optimization. The mixed flow turbine stability evaluation method of the fusion impulsive feature value of the sound and vibration signals is not only suitable for the stability detection of the hydraulic mechanical system, but also can be popularized and applied to the signal analysis of other complex mechanical systems, has strong universality and practicality, thereby improving the operation stability, safety and service life of the equipment, and reducing the maintenance cost. BRIEF DESCRIPTION OF DRAWINGS

[0070] Figure 1 is the structural schematic diagram of the mixed flow turbine vibration and noise data acquisition test system adopted in the mixed flow turbine stability evaluation method of the fusion impulsive feature value of the sound and vibration signals of the present application;

[0071] Fig. 2(a) is the vibration signal collected by the vibration sensor in the mixed flow turbine stability evaluation method of the fusion impulsive feature value of the sound and vibration signals of the present application;

[0072] Fig. 2(b) is the noise signal collected by the noise sensor in the stability evaluation method of the mixed-flow hydraulic turbine of the application fusing the impact characteristic values of the acoustic and vibration signals;

[0073] Fig. 3(a) is the vibration signal filtered by the maximum correlation kurtosis deconvolution in the stability evaluation method of the mixed-flow hydraulic turbine of the application fusing the impact characteristic values of the acoustic and vibration signals;

[0074] Fig. 3(b) is the noise signal filtered by the maximum correlation kurtosis deconvolution in the stability evaluation method of the mixed-flow hydraulic turbine of the application fusing the impact characteristic values of the acoustic and vibration signals;

[0075] Figure 4 is the fusion characteristic value result in the stability evaluation method of the mixed-flow hydraulic turbine of the application fusing the impact characteristic values of the acoustic and vibration signals.

[0076] In the figure, 1. spiral case, 2. movable guide vane, 3. impeller, 4. draft tube, 5. vibration sensor, 6. noise sensor, 7. data acquisition card, 8. computer. DETAILED DESCRIPTION

[0077] The application will be described in detail below in combination with the drawings and specific embodiments.

[0078] Example 1

[0079] The stability evaluation method of the mixed-flow hydraulic turbine of the application fusing the impact characteristic values of the acoustic and vibration signals, as shown in Figure 1 , analyzes by collecting the real-time vibration signal and noise signal of the mixed-flow hydraulic turbine respectively in the mixed-flow hydraulic turbine test system. The mixed-flow hydraulic turbine includes, in order according to the water flow direction: spiral case 1, movable guide vane 2, impeller 3, and draft tube 4. For any mixed-flow hydraulic turbine device containing spiral case 1, movable guide vane 2, impeller 3, and draft tube 4, the vibration and noise data of the gap between the movable guide vane 2 and the impeller 3 of the mixed-flow hydraulic turbine are measured by the vibration sensor 5 and the noise sensor 6. The vibration sensor 5 and the noise sensor 6 are magnetic attraction type sensors that can be directly attached to the measurement point shell. The vibration sensor 5 and the noise sensor 6 are both connected with the data acquisition card 7 through the data transmission line, and are transmitted to the computer 8 through the data line.

[0080] Example 2

[0081] The stability evaluation method of the mixed-flow hydraulic turbine of the application fusing the impact characteristic values of the acoustic and vibration signals is implemented according to the following steps:

[0082] Step 1, start the mixed-flow hydraulic turbine vibration and noise data acquisition system, as Figure 1As shown, the water flows sequentially through the volute 1, the movable guide vane 2, the impeller 3, and the tailrace pipe 4. A vibration sensor 5 is used to collect the vibration signal between the movable guide vane 2 and the impeller 3, while a noise sensor 6 is used to collect the noise signal between the movable guide vane 2 and the impeller 3. All collected signals are acquired through a data acquisition card 7 and transmitted to a computer 8 via a data cable, resulting in a vibration signal component diagram with n signal points as shown in Figure 2(a) and a noise signal component diagram with n signal points as shown in Figure 2(b).

[0083] Step 2: Use maximum correlation kurtosis deconvolution to process the vibration signal x acquired by vibration sensor 5 as shown in Figure 2(a). n Filtering is performed to make the impulse characteristics in the signal easier to detect. The filtered signal is signal y in Figure 3(a).

[0084] Step 3, use maximum correlation kurtosis deconvolution to deconvolve the noise signal c collected by the noise sensor as shown in Figure 2(b). n Filtering is performed to make the impulse features in the signal easier to detect, and the signal is converted into signal v as shown in Figure 3(b).

[0085] Step 4: Extract the energy entropy ratio of signals y and v respectively;

[0086] Step 5: Calculate the fusion characteristic value R of the combined vibration signal and noise signal based on the energy entropy ratio obtained in Step 4. y,v .

[0087] Example 3

[0088] The specific process of step 2 is as follows:

[0089] Step 2.1, vibration signal x n (n = 1, 2, ..., N) x ) related kurtosis The calculation formula is as follows:

[0090]

[0091] Where, N x The number of samples for the input vibration signal is 51200; T x The deconvolution period for the vibration signal is set to 10⁵; M x The filter shift value for the vibration signal is 4; m x (m=0,1,…,M x ) represents the current shift value for the vibration signal filter.

[0092] Step 2.2, maximize the relevant kurtosis, as shown in equation (2):

[0093]

[0094] in, For the filter coefficients of the vibration signal, k x (k x =1,2,…,L x ) represents the current filter length for the vibration signal.

[0095] Step 2.3: Calculate the optimal filter coefficients corresponding to the maximum correlation kurtosis for the vibration signal. It can be calculated using equation (3):

[0096]

[0097] Step 2.4, the formula for calculating the output signal y of the filter is as follows:

[0098]

[0099] Among them, L x This is the filter length for vibration signals, and its value is 100.

[0100] Filter coefficients The specific expression for the solution is shown in equation (5):

[0101]

[0102] in:

[0103]

[0104] Example 4

[0105] The specific process of step 3 is as follows:

[0106] Step 3.1, noise signal c n (n = 1, 2, ..., N) c ) related kurtosis The calculation formula is as follows:

[0107]

[0108] Where, N c The number of samples for the input noise signal, with a value of 51200; T c The deconvolution period for the noisy signal is set to 10⁵; M c The filter shift value is 4 for noise signals; m c (0,1,…,M c ) represents the current shift value of the filter for the noisy signal.

[0109] Step 3.2, maximize the relevant kurtosis, as shown in equation (7):

[0110]

[0111] in, For the filter coefficients of the noise signal, k c (k c =1,2,…,L c ) represents the current filter length for the noise signal.

[0112] Step 3.3: Calculate the optimal filter coefficients corresponding to the maximum correlation kurtosis for the noise signal. It can be calculated using equation (8):

[0113]

[0114] Step 3.4, the formula for calculating the output signal v of the filter is as follows:

[0115]

[0116] Among them, L c This is the filter length for noise signals, with a value of 100. k c (k c =1,2,…,L c ) represents the length of the current filter for the noise signal. These are the filter coefficients for noise signals.

[0117] Filter coefficients The solution is expressed as shown in equation (10):

[0118]

[0119] in:

[0120]

[0121] Example 5

[0122] The specific process of step 4 is as follows:

[0123] Step 4.1: Select the Hanning window function to frame the signals y and v. The specific calculation formula is as follows:

[0124]

[0125] Where, η y and η v Let A be the window length, with a value of 1024. y With A vIt is the index of the sample point in the window, and its value is 0. y <η y -1, 0 v <η v -1.

[0126] Step 4.2, for signal y(o) y ) and v(o v The lengths of the two signals are N. y and N v Time series (o y,v =1,2,…,N y,v The windowed function δ(u) will be applied. y ) and δ(u v The i-th frame signal obtained after frame segmentation is t i (o) and g i (o), then it satisfies:

[0127]

[0128]

[0129] Among them, w leny w lenv The length of a frame for different signals; i ny i nv J represents the length of a frame that different signals move; ny and J nv The total number of frames after different signals are framed.

[0130] Step 4.3, for t i (o) and g i (o) Perform a Fourier transform, the specific formula of which is shown below:

[0131]

[0132] Among them, T i (χ y ) and G i (χ v ) represents the spectrum function after Fourier transform; χ y,v Its independent variable is Hz. It is a complex function, and its specific expression is shown below:

[0133]

[0134] Step 4.4, for the ξ-th digit after Fourier transform y and ξ v Spectral line frequency component J ξy and J ξv The energy spectra are Y​​i (ξ y ) and V i (ξ v The calculation formula for Y is... i (ξ y )=|J ξy | 2 V i (ξ v )=|J ξv | 2 For each frequency component of different signals, the normalized spectral probability density function P... i (ξ y ) and Q i (ξ v The definitions of ) are as follows:

[0135]

[0136] Among them, P i (ξ y ) represents the ξth frame of the i-th frame. y One frequency component J ξy The corresponding probability density; Q i (ξ v ) represents the ξth frame of the i-th frame. v One frequency component J ξv The corresponding probability density.

[0137] Step 4.5, the spectral entropy H of the vibration signal in the i-th frame. yi The spectral entropy H of the noise signal in the i-th frame vi The specific expression is as follows:

[0138]

[0139] Step 4.6, the energy E of the vibration signal in the i-th frame. yi The energy E of the noise signal in the i-th frame vi The specific expression is as follows:

[0140]

[0141] Here, 'a' is used to adjust the degree of energy change and helps to distinguish signal abrupt changes, and its value is 1.5.

[0142] Step 4.7, the energy entropy ratio R of the vibration signal in the i-th frame. yi The ratio of the energy entropy of the noise signal to that of the i-th frame, R vi The specific expression is as follows:

[0143]

[0144] Step 5, based on the energy entropy ratio R of signals y and v extracted in Step 4. yi and R vi The fusion characteristic value R of the combined vibration signal and noise signal is calculated using equation (27). y,v :

[0145]

[0146] Example 6

[0147] Input vibration signal x n and noise signal c n After calculations in steps 2-5, the fused characteristic value R, which comprehensively considers both vibration and noise signals, is obtained. y,v , drawing as Figure 4 Fusion eigenvalues ​​R y,v —The larger the value in the time-domain waveform, the more complex the impact characteristics of the signal in the time domain. Therefore, the fusion characteristic value R of the acoustic-vibration signal of a mixed-flow turbine is... y,v This indicates that the water turbine is less stable.

Claims

1. A method for evaluating the stability of a Francis turbine by fusing the impact characteristic values of acoustic vibration signals, characterized in that: Specifically comprising the following steps: Step 1, using vibration sensor (5) to collect the vibration signal of the gap between the runner moving vane (2) and the runner (3) x n , using noise sensor (6) to collect the noise signal between the runner moving vane (2) and the runner (3) c n ; Step 2, filter the vibration signal collected by the vibration sensor (5) using maximum correlation kurtosis deconvolution to obtain a filtered signal y ; Step 3, filter the noise signal collected by the noise sensor (6) using maximum correlation kurtosis deconvolution to obtain a filtered signal v ; Step 4, extracting the signal y and v the energy-entropy ratio; the specific process of Step 4 is: Step 4.1, Selecting the Hann window function for the signal y and v Frame the signal, the specific formula is as follows: wherein, η y and η v is the window length, A y and A v is the index of the sample point in the window, taking values from 0 A y η y -1, 0 A v η v -1;​​ Step 4.2, regarding the signal y ( o y )and v ( o v The lengths of the two signals are respectively N y and N v The time series will be processed by a windowing function. δ ( u y )and δ ( u v The first frame obtained after frame processing i The frame signals are respectively t i ( o )and g i ( o ), satisfying the following conditions: wherein, w leny , w lenv is the length of one frame of the different signal; i ny , i nv is the length of one frame movement of the different signal; J ny and J nv is the total number of frames after the different signal is framed. Step 4.3, Fourier transform of t i ( o ) and g i ( o ) is shown below. wherein T i ( χ y ) and G i ( χ v ) is a Fourier transformed spectral function; χ y,v is an independent variable in Hz, is a complex function, the specific expression is as follows: （7） （8） Step 4.4, for the Fourier transform after the ξ y and ξ v spectrum line frequency components J ξy and J ξv the energy spectrum of Y i ( ξ y ) and V i ( ξ v ), the normalized spectral probability density function of each frequency component under different signals P i ( ξ y ) and Q i ( ξ v ) are defined as: wherein P i ( ξ y ) is the i frame the ξ y th frequency component J ξy corresponding probability density; Q i ( ξ v ) is the i frame the ξ v th frequency component J ξv corresponding probability density; Step 4.5, the spectral entropy of the vibration signal at the i frame H yi and the spectral entropy of the noise signal at the i frame H vi The specific expression is as follows: Step 4.6, the energy of the vibration signal in the i frame E yi and the energy of the noise signal in the i frame E vi The specific expression is as follows: wherein, a for adjusting the degree of energy variation and helping to distinguish signal mutations; Step 4.7, the energy-entropy ratio of the vibration signal in the i frame R yi frame i frame R vi The specific expression is as follows: Step 5, the fusion feature value of the comprehensive vibration signal and the noise signal is calculated according to the energy-entropy ratio obtained in step 4 R y,v .

2. The mixed-flow hydraulic turbine stability evaluation method of claim 1, wherein: The specific process of the step 2 is as follows: Step 2.

1. The vibration signal is calculated according to the following equation (17) x n of the correlation kurtosis : wherein, N x is the number of samples of the input vibration signal, T x is the number of deconvolution periods for the vibration signal, M x is the number of filter shifts for the vibration signal, m x denotes the current shift number for the vibration signal filter; Step 2.2, the correlation kurtosis calculated in step 2.1 is maximized by using the following formula (18): wherein is a filter coefficient for the vibration signal, k x is a current filter length for the vibration signal; Step 2.3, the optimal filter coefficients corresponding to the maximum correlation kurtosis of the vibration signal are solved by the following formula (19) : Step 2.

4. Calculate the signal at the filter output by the equation (20) as follows y : wherein, L x is the filter length for the vibration signal.

3. The mixed-flow hydraulic turbine stability evaluation method of claim 2, wherein: In step 2.4, the filter coefficients is obtained by the following equation (21). （21） Wherein: = ； ； 。 4. The mixed-flow hydraulic turbine stability evaluation method of claim 3, wherein: The specific process of the step 3 is as follows: Step 3.1, the noise signal is calculated using equation (22) below c n the correlation kurtosis of : wherein, N c is the number of samples of the input noise signal, T c is the deconvolution period for the noise signal, M c is the filter shift number for the noise signal, m c denotes the current filter shift number for the noise signal; Step 3.2, the correlation kurtosis calculated in step 3.1 is maximized by using the following formula (23): wherein is a filter coefficient for the noise signal, k c is a current filter length for the noise signal; Step 3.3, find the optimal filter coefficients corresponding to the maximum correlation kurtosis of the noise signal by the following equation (24) : Step 3.4, the signal at the filter output is calculated by the equation (25) as follows v : wherein, L c is a length of the filter for the noise signal, k c is a length of the current filter for the noise signal, is a filter coefficient for the noise signal.

5. The mixed-flow hydraulic turbine stability evaluation method of claim 4, wherein: In step 3.4, the filter coefficients The expression for the solution is given by equation (10): Wherein: = ； ； 。 6. The mixed-flow hydraulic turbine stability evaluation method of fused acoustic-vibrational signal impact characteristic values according to claim 5, characterized in that: In step 4.4, Y i ( ξ y )=| J ξy | 2 ; V i ( ξ v )=| J ξv | 2 .

7. The mixed-flow hydraulic turbine stability evaluation method of claim 6, wherein: The specific process of the step 5 is: calculating the fusion characteristic value of the comprehensive vibration signal and the noise signal through formula (27) R y,v : （27）。

Images