CN102928817A - Method for positioning rotor rubbing sound emission source by applying time delay estimation - Google Patents

Abstract

The invention discloses a method for locating a rotor rubbing acoustic emission source by using time delay estimation, comprising steps: 10): establishing a one-dimensional linear positioning model; 20): setting attenuation coefficient estimation

Initial iteration values and time delay estimates for initial iteration value of ; 30): measure

The estimated error between the true value and the estimated value of

;40): According to the estimation error, calculate the next iteration step size; 50): Calculate the time delay estimation and attenuation coefficient estimation after iteration; 60): Calculate the new estimation error

; 70): compare the relative error of the two estimation errors calculated in step 30) and step 60), if it is less than the set value, it will be estimated by the delay time

The position of the rubbing source is measured; otherwise, turn to step 40) to continue iteration. This method considers the problem of signal attenuation and strong noise interference in the propagation process of the rubbing acoustic emission signal of the rotor system, and accurately realizes the location of the rubbing fault source by adding attenuation estimation and variable step steps.

Description

A Method of Rotor Rubbing Acoustic Emission Source Location Using Time Delay Estimation

technical field

The invention relates to a method for locating a rotor rubbing acoustic emission source, in particular to a method for locating a rotor rubbing acoustic emission source by using time delay estimation. the

Background technique

Rubbing of moving and static parts of rotating machinery is a common fault in operation. In particular, the current rotating machinery is developing towards large-scale, high-parameter, and high-efficiency development. The equipment structure is becoming more and more complex, and the dynamic and static gaps are getting smaller and smaller, which makes the rubbing problem of moving and static parts more and more prominent. Acoustic emission technology (Acoustic Emission, AE) is an effective method for diagnosing rubbing faults. It can not only judge the occurrence of rubbing, but also calculate the location of rubbing through positioning technology, which has important application value. Time Delay estimation (TDE) is a method for estimating the time difference between sound source signals arriving at different sensors. At present, such algorithms mainly include Cross-Correlation algorithm and Generalized Cross-Correlation (GCC) algorithm. ) algorithm, recursive least squares (Recursive Least Squares, RLS) algorithm, least mean square (Least Mean Square, LMS) algorithm, among which LMS adaptive delay estimation is the most commonly used. In the LMS algorithm, the time delay is equivalent to a signal passing through a finite impulse response filter, and the time delay is continuously iterated using the gradient descent method according to the difference between the filter output and the reference signal. However, this algorithm only tracks the time-varying signal-to-noise ratio and does not take into account the attenuation of the signal itself. If it is directly applied to the location of the AE source, a large error will occur. the

Contents of the invention

Technical problem: The technical problem to be solved by the present invention is to provide a method for locating the rotor rubbing AE source by using time delay estimation, which takes into account the signal attenuation and strong For the problem of noise interference, by adding attenuation estimation and variable step size steps, the rubbing fault source can be located more accurately. the

Technical scheme: in order to solve the above technical problems, the technical scheme adopted in the present invention is:

A method for locating a rotor rubbing acoustic emission source using time delay estimation, the method comprising the following steps:

10): Establish a one-dimensional linear positioning model: set the first sensor (1) and the second sensor (2) on the rubbing waveguide plate of the rotor test bench, and the rotor rubbing source is located between the two sensors. And the rotor rubbing source and the two sensors are located on the same straight line, and the acoustic emission signal is generated after the rotor rubbing, and the acoustic emission signal is received by the two sensors, and the acoustic emission signal is accompanied by attenuation and noise interference during propagation;

的初始迭代值，最后设定一个时间延迟估计 

的初始迭代值；  20): Set the attenuation coefficient estimation Initial iteration values and time delay estimates for The initial iterative value of : the acoustic emission signal x(t) received by the first sensor (1) and the acoustic emission signal y(t) received by the second sensor (2) are collected by the rubbing acoustic emission test device, and then the two Energy ratio of road acoustic emission signal as attenuation coefficient estimation

The initial iteration value of , and finally set a time delay estimate

The initial iteration value of ;

30): Calculate the estimation error between the real value and the estimated value of y(t): Assume that in the two acoustic emission signals, the energy of y(t) is smaller than the energy of x(t), and the rubbing source signal reaches the first sensor first (1), and then arrive at the second sensor (2) after a delay time Δτ, then calculate the estimation error between the real value and the estimated value of y(t) according to formula (1):

$e\left(t\right)=\mathrm{the\; y}\left(t\right)-\widehat{\mathrm{\α}}x(t-\mathrm{\Δ}\widehat{\mathrm{\τ}})$ Formula 1)

表示真实衰减系数α的衰减系数估计， 

表示真实时间延迟Δτ的时间延迟估计， 

表示第二传感器(2)接收的信号，此信号是第一传感器(1)接收到的信号x(t)在延时 

之后的信号；  Among them, e(t) represents the estimation error,

represents the decay coefficient estimate of the true decay coefficient α,

represents the time delay estimate of the real time delay Δτ,

Indicates the signal received by the second sensor (2), which is the signal x(t) received by the first sensor (1) after the delay

after the signal;

40): According to the estimated error e(t), calculate the next iteration step size;

50): according to formula (2), measure and calculate the time delay estimation and attenuation coefficient estimation after iteration;

$\left\{\begin{array}{c}\mathrm{\Δ}\widehat{\mathrm{\α}}(t+1)=\mathrm{\Δ}\widehat{\mathrm{\α}}\left(t\right)-{\mathrm{\μ}}_{\mathrm{\α}}e\left(t\right)\frac{d\left(e\left(t\right)\right)}{d\left(\mathrm{\Δ}\widehat{\mathrm{\α}}\left(t\right)\right)}\\ \mathrm{\Δ}\widehat{\mathrm{\τ}}(t+1)=\mathrm{\Δ}\widehat{\mathrm{\τ}}\left(t\right)-{\mathrm{\μ}}_{\mathrm{\Δ\τ}}e\left(t\right)\frac{d\left(e\left(t\right)\right)}{d\left(\mathrm{\Δ}\widehat{\mathrm{\τ}}\left(t\right)\right)}\end{array}\right.$ Formula (2)

表示迭代后的衰减系数估计， 

表示迭代前的衰减系数估计，μ_α表示 

的迭代步长， 

表示对迭代前的衰减系数估计的偏导， 

表示迭代后的时间延迟估计， 表示迭代前的时间延迟估计，μΔ_τ表示 

的迭代步长， 

表示对迭代前的时间延迟估计的偏导；  in,

Denotes the decay coefficient estimate after iteration,

Indicates the attenuation coefficient estimate before iteration, μ _α represents

The iterative step size of

Represents the partial derivative of the decay coefficient estimate before the iteration,

Denotes the time delay estimate after iteration, Indicates the time delay estimate before the iteration, μΔ _τ represents

The iterative step size of

Represents the partial derivative of the time delay estimate before the iteration;

60): Substituting the iterative time delay estimate and the iterative attenuation coefficient estimate calculated in step 50) into formula (1), and calculating a new estimated error e(t+1);

如果相对误差小于1％，则停止测算，将步骤50)测得的时间延迟估计作为真实时间延迟Δτ，依据式(3)，测得碰摩源的位置；如果相对误差大于等于1％，则将步骤60)中测算出新的估计误差e(t+1)取代步骤30)中的e(t)，然后重复步骤40)至步骤70)；  70): compare step 30) and step 60) the relative error of the two estimated errors measured and calculated, the relative error is

If the relative error is less than 1%, stop the calculation, and use the estimated time delay measured in step 50) as the real time delay Δτ, and measure the position of the friction source according to formula (3); if the relative error is greater than or equal to 1%, then The new estimation error e(t+1) measured and calculated in step 60) replaces e(t) in step 30), and then repeats steps 40) to 70);

S ₁ ＝(D-vΔτ)/2 Formula (3)

Among them, _S1 is the distance from the rubbing source to the first sensor (1), D represents the distance between the first sensor (1) and the second sensor (2), v is the propagation velocity of the acoustic emission wave, and Δτ represents the acoustic wave The time difference between the transmitted waves arriving at the two sensors.

Beneficial effect: compared with prior art, the present invention has the following advantages:

(1) Accurately realize the location of rubbing fault source. The present invention considers the attenuation factor of the acoustic emission signal in the relaying process of the rotor system, and adopts band attenuation estimation in the positioning process. This can significantly improve the estimation performance of time delay parameters at low signal-to-noise ratios, and significantly improve the positioning accuracy of rubbing sources. the

(2) Accelerate the convergence speed and improve the calculation efficiency. Adopting the variable step size method proposed by the present invention can effectively speed up the convergence speed of LMS, make the estimation system have better robustness, reduce the amount of steady-state misalignment, effectively improve the time-varying tracking ability of the algorithm, and better meet the needs of real-time positioning. It can significantly improve the performance of LMS algorithm at low signal-to-noise ratio, make it have faster convergence speed, and improve the accuracy of time delay estimation. the

Description of drawings

Figure 1 is a schematic diagram of the present invention. the

Fig. 2 is a schematic diagram of the one-dimensional linear positioning model established in step 10) of the present invention. the

Fig. 3 is an AE waveform diagram of a period of continuous rubbing received by the first sensor in the present invention when the rotor speed is 2040r/min. the

Fig. 4 is an AE waveform diagram of a period of continuous rubbing received by the second sensor in the present invention when the rotor speed is 2040r/min. the

FIG. 5 is an expanded view of a cluster of AE waveforms in FIG. 3 . the

FIG. 6 is an expanded view of a cluster of AE waveforms in FIG. 4 . the

Fig. 7 In the experiment of the present invention, when the signal-to-noise ratio is 10 dB, the time delay and attenuation coefficient iteration curves of LMS with attenuation estimation. the

Fig. 8 is an iterative curve diagram of the conventional LMS and the LMS method with attenuation estimation when the signal-to-noise ratio is 0 dB in the experiment of the present invention. the

Fig. 9 is an iterative curve of the conventional LMS and the method of the present invention when the signal-to-noise ratio is 0 dB in the experiment of the present invention. the

Detailed ways

The technical solution of the present invention will be further elaborated below in conjunction with the accompanying drawings and examples. the

As shown in Figures 1 and 2, a method for locating rotor rubbing acoustic emission sources using time delay estimation of the present invention includes the following steps:

10): Establish a one-dimensional linear positioning model: set two sensors, the first sensor 1 and the second sensor 2, on the rubbing waveguide plate of the rotor test bench, the rotor rubbing source is located between the two sensors, and the rotor rubbing The source and the two sensors are located on the same straight line. After the rotor rubs, an acoustic emission signal is generated, which is received by the two sensors. The acoustic emission signal is accompanied by attenuation and noise interference during propagation. the

In step 10), the rotor rubbing acoustic emission signal is obtained by obtaining the rotor rubbing acoustic emission signal on the rotor rubbing test bench. A movable rubbing bracket installed on the base of the rotor table is used to simulate and realize static and dynamic rubbing. The bracket is equipped with retractable rubbing screws. The rubbing screws are located on the side of the disc, facing the center of the rotating shaft along the radial direction of the shaft. , by adjusting the rubbing screw, different degrees of rubbing faults can be simulated. The AE signal generated by the rubbing source is transmitted to the AE sensors on both sides through the waveguide, that is, the first sensor 1 and the second sensor 2 . The first sensor 1 and the second sensor 2 adopt UT-1000 broadband sensor, the signal acquisition system is PCI-2 collector and supporting software, 18-bit A/D resolution, frequency response range 1kHz-3MHz. The wave speed is measured in advance by the simulated rubbing AE waveform in static state. In dynamic mode, every time the rotor rotates, it rubs against the rubbing screw once, forming a cluster of AE waves, so the AE signal is periodic during continuous rubbing. Fig. 3 is the AE waveform of a section of continuous rubbing received by the first sensor 1 when the rotor speed is 2040r/min, and Fig. 4 is the AE waveform of a section of continuous rubbing received by the second sensor 2 when the rotor speed is 2040r/min. 5 is an expanded view of a cluster of AE waveforms in FIG. 3 . FIG. 6 is an expanded view of a cluster of AE waveforms in FIG. 4 . In this experiment, the distance difference between the rubbing source and the two sensors is 184mm. the

的初始迭代值，最后设定一个时间延迟估计 

的初始迭代值。  20): Set the attenuation coefficient estimation Initial iteration values and time delay estimates for The initial iterative value of : Acquire the acoustic emission signal x(t) received by the first sensor 1 and the acoustic emission signal y(t) received by the second sensor 2 through the rubbing acoustic emission test device, and then use the two-way acoustic emission signal The energy ratio of the attenuation coefficient is estimated as

The initial iteration value of , and finally set a time delay estimate

The initial iteration value of .

的初始迭代值的测算是：在一个相同的时间段内，测算x(t)的能量和，以及y(t)的能量和，用其中能量和较小的值除以较大的值，得出二者的比值，即得到两个传感器的能量比，该能量比小于1，该能量比即为衰减系数估计 的初始迭代值。同时，给定一个时间延迟值 

初始迭代值。因为采集信号为离散时间序列，因此给定的时间延迟值为某一个数据点值。  In step 20), the attenuation coefficient estimation

The calculation of the initial iterative value of is: in the same time period, measure the energy sum of x(t) and the energy sum of y(t), and divide the smaller value of the energy sum by the larger value to get The ratio of the two is obtained, that is, the energy ratio of the two sensors is obtained. The energy ratio is less than 1, and the energy ratio is the estimation of the attenuation coefficient. The initial iteration value of . At the same time, given a time delay value

Initial iteration value. Because the acquisition signal is a discrete time series, the given time delay value is a certain data point value.

30): Calculate the estimation error between the real value and the estimated value of y(t): Assume that in the two acoustic emission signals, the energy of y(t) is smaller than the energy of x(t), and the rubbing source signal reaches the first sensor first 1, and then arrive at the second sensor 2 after a delay time Δτ, then calculate the estimation error between the real value and the estimated value of y(t) according to formula (1):

$e\left(t\right)=\mathrm{the\; y}\left(t\right)-\widehat{\mathrm{\α}}x(t-\mathrm{\Δ}\widehat{\mathrm{\τ}})$ Formula 1)

表示真实衰减系数α的衰减系数估计， 

表 示真实时间延迟Δτ的时间延迟估计， 表示第二传感器(2)接收的信号，此信号是第一传感器(1)接收到的信号x(t)在延时 之后的信号。  Among them, e(t) represents the estimation error,

represents the decay coefficient estimate of the true decay coefficient α,

Denotes the time delay estimate of the real time delay Δτ, Indicates the signal received by the second sensor (2), which is the signal x(t) received by the first sensor (1) after the delay after the signal.

40): According to the estimated error e(t), calculate the next iteration step size. the

In step 40), when the estimated error e (t) calculated in step 30) is greater than 0.1, the iteration step size is calculated according to formula (4),

$\left\{\begin{array}{c}{\mathrm{\μ}}_{\mathrm{\α}}(t+1)=b_{\mathrm{\α}}[1-\exp (-a_{\mathrm{\α}}|{\▿}_{\mathrm{\α}}\left|\right)]\\ {\mathrm{\μ}}_{\mathrm{\Δ\τ}}(t+1)=b_{\mathrm{\Δ\τ}}[1-\exp \left({-a}_{\mathrm{\Δ\τ}}\right|{\▿}_{\▿\mathrm{\τ}}\left|\right)]\end{array}\right.$ Formula (4)

When the estimated error e(t) is less than or equal to 0.1, the iteration step size is calculated according to formula (5),

$\left\{\begin{array}{c}{\mathrm{\μ}}_{\mathrm{\α}}(t+1)=b_{\mathrm{\α}}\{1-\exp [-a_{\mathrm{\α}}\mathrm{\ϵ}\left(t\right)\left]\right\}\\ {\mathrm{\μ}}_{\mathrm{\Δ\τ}}(t+1)=b_{\mathrm{\Δ\τ}}\{1-\exp [-a_{\mathrm{\Δ\τ}}\mathrm{\ϵ}\left(t\right)\left]\right\}\end{array}\right.$ Formula (5)

的迭代步长，μ_Δτ(t+1)表示新的 

的迭代步长，a_α、b_α、a_Δτ和b_Δτ均为步长调整因子，ε(t)表示对误差矩阵进行线性加权后得到的误差， 

表示估计误差e(t)对于衰减系数的梯度， 表示估计误差e(t)对于时间延迟的梯度。  Among them, μ _α (t+1) represents the new

The iteration step size of , μ _Δτ (t+1) represents the new

The iterative step size of , a _α , b _α , a _Δτ and b _Δτ are all step size adjustment factors, ε(t) represents the error obtained by linearly weighting the error matrix,

Represents the gradient of the estimated error e(t) for the attenuation coefficient, Indicates the gradient of the estimated error e(t) with respect to the time delay.

The iteration step size determines the convergence speed of the calculation to a certain extent. In order to speed up the convergence speed of the calculation, it is hoped that when the error is large, a larger iterative step size is used so that the calculation can quickly approach the real value; when the error is small, a smaller step size is used to reduce the oscillation of the estimated value at the real value and reduce the convergence time error. At present, the variable step size method generally adjusts the step size directly according to the size of the error, or uses the number of iterations to adjust the step size. However, these methods are greatly affected when the noise changes strongly, and the convergence speed is slow. The invention proposes to use the gradient of the error to adjust the step size when the estimation error is large, and introduces the forgetting factor to weight the error at the previous time when the error is small, and uses the weighted error to adjust the step size. the

表示估计误差e(t)对于衰减系数的梯度，  表示估计误差e(t)对于时间延迟的梯度， 则迭代步长如上述式(4)所示。  When the estimation error is greater than 0.1, the adjustment method of the iteration step size is as follows:

Represents the gradient of the estimated error e(t) for the attenuation coefficient, Represents the gradient of the estimated error e(t) for the time delay, Then the iteration step size is shown in the above formula (4).

When the estimation error is less than or equal to 0.1, the adjustment method of the iterative step is as follows: define the forgetting matrix λ=[λ _L λ _L-1 ...λ ₀ ], where λ _i is the forgetting factor, 0≤i≤L. The function of the forgetting matrix is to carry out non-linear weighting on all the errors of the previous time length L. The closer to the current moment, the larger the forgetting factor and the greater the impact on the weighted errors. The forgetting factor adopts exponential decay form, λ _i = ^ci , c is a constant close to 1, 0<c<1. The weighted error is denoted as ε(t), ε(t)=e(t)λ ^T , where e(t) is an error matrix with a duration of L before time t, where e(t)=[ e(t-L+1)L e(t-1) e(t)], the step size adjustment method is shown in formula (5).

50): according to formula (2), measure and calculate the time delay estimation and attenuation coefficient estimation after iteration;

$\left\{\begin{array}{c}\mathrm{\&Delta;}\widehat{\mathrm{\&alpha;}}(t+1)=\mathrm{\&Delta;}\widehat{\mathrm{\&alpha;}}\left(t\right)-{\mathrm{\&mu;}}_{\mathrm{\&alpha;}}e\left(t\right)\frac{d\left(e\left(t\right)\right)}{d\left(\mathrm{\&Delta;}\widehat{\mathrm{\&alpha;}}\left(t\right)\right)}\\ \mathrm{\&Delta;}\widehat{\mathrm{\&tau;}}(t+1)=\mathrm{\&Delta;}\widehat{\mathrm{\&tau;}}\left(t\right)-{\mathrm{\&mu;}}_{\mathrm{\&Delta;\&tau;}}e\left(t\right)\frac{d\left(e\left(t\right)\right)}{d\left(\mathrm{\&Delta;}\widehat{\mathrm{\&tau;}}\left(t\right)\right)}\end{array}\right.$ Formula (2)

表示迭代前的衰减系数估计，μ_α表示 

的迭代步长， 

表示对迭代前的衰减系数估计的偏导， 

表示迭代后的时间延迟估计， 

表示迭代前的时间延迟估计，μ_Δτ表示 

的迭代步长， 

表示对迭代前的时间延迟估计的偏导。  in, Denotes the decay coefficient estimate after iteration,

Indicates the attenuation coefficient estimate before iteration, μ _α represents

The iterative step size of

Represents the partial derivative of the decay coefficient estimate before the iteration,

Denotes the time delay estimate after iteration,

Indicates the time delay estimate before the iteration, μ _Δτ represents

The iterative step size of

Represents the partial derivative of the time delay estimate before the iteration.

60): Substituting the iterative time delay estimate and the iterative attenuation coefficient estimate calculated in step 50) into formula (1), and calculating a new estimated error e(t+1). the

如果相对误差小于1％，则停止测算，将步骤50)测得的时间延迟估计作为真实时间延迟Δτ，依据式(3)，测得碰摩源的位置；如果 相对误差大于等于1％，则将步骤60)中测算出新的估计误差e(t+1)取代步骤30)中的e(t)，然后重复步骤40)至步骤70)；  70): compare step 30) and step 60) the relative error of the two estimated errors measured and calculated, the relative error is

If the relative error is less than 1%, stop the calculation, and use the estimated time delay measured in step 50) as the real time delay Δτ, and measure the position of the friction source according to formula (3); if the relative error is greater than or equal to 1%, then The new estimation error e(t+1) measured and calculated in step 60) replaces e(t) in step 30), and then repeats steps 40) to 70);

S ₁ ＝(D-vΔτ)/2 Formula (3)

Among them, _S1 is the distance from the rubbing source to the first sensor (1), D represents the distance between the first sensor (1) and the second sensor (2), v is the propagation velocity of the acoustic emission wave, and Δτ represents the acoustic wave The time difference between the transmitted waves arriving at the two sensors.

The above method adopts least mean square adaptive time delay estimation with attenuation estimation and variable step size to calculate the time delay, which improves the accuracy of rubbing source location. The following is demonstrated through experiments. the

When the step size is fixed, the actual effect of LMS estimation with attenuation:

Gaussian white noise with different signal-to-noise ratios is added to the waveforms in Figure 3 and Figure 4 to simulate the actual noisy AE signal on site. Use the LMS algorithm with attenuation estimation when the step size is fixed to estimate the time delay of the noisy AE signal. The experimental parameters are as follows:

Signal sampling frequency: 512kHz;

Delay iteration step size: 0.05;

Attenuation coefficient iteration step size: 0.01;

Delay initial iteration value: 100;

Signal-to-noise ratio: 10dB. the

The real time delay is calculated by measuring the distance difference between the rubbing source and the two sensors and the wave velocity, which is 575 time delay points. The attenuation coefficient of the signal is obtained by comparing the total energy of the two signal segments within the same time period. Figure 7 shows the iterative results. After 1200 iterations, although the attenuation coefficient has not converged to the real value, the estimation of the time delay has converged. LMS is essentially a noise cancellation algorithm. In practical applications, it is possible to cancel a part of the signal, or there is still a certain amount of noise after cancellation, which is the reason for the estimation error of the attenuation coefficient. the

Figure 8 is a comparison diagram of the iterative curves of conventional LMS and LMS with attenuation estimation when 0dB signal-to-noise ratio noise is added to the two signals. It can be seen that as the signal-to-noise ratio decreases, the conventional LMS cannot converge to the true value, but the method with attenuation estimation of the present invention can converge to the true value. When the same source rubbing AE signal is transmitted to different sensors, due to the distance difference, the signal received by the far sensor must be attenuated, but the noise received by each sensor is at the same energy level, and there is no attenuation between each other , it is estimated that the function of the attenuation is to enable the LMS to better offset the noise of the same energy level, thereby improving the ability to resist noise interference during the positioning process. the

The actual effect of LMS with attenuation estimation with variable step size: when the estimation error is greater than 0.1, the gradient of the error is used to determine the step size, and when it is less than 0.1, the forgetting factor is introduced to weight the error to determine the step size. Take the forgetting factor c as 0.95, and the length L of the forgetting matrix as 20. It can be seen from Figure 9 that the LMS with variable step size converges after 300 iterations, while the method without variable step size converges after 1200 steps, and the tracking ability of the measurement is significantly improved. the

Claims (2)

1. A method for rotor rub-impact acoustic emission source localization using time delay estimation, comprising the steps of:

   10): establishing a one-dimensional linear positioning model: the method comprises the following steps that a first sensor (1) and a second sensor (2) are arranged on a rub-impact wave guide plate of a rotor test bed, a rotor rub-impact source is located between the two sensors, the rotor rub-impact source and the two sensors are located on the same straight line, an acoustic emission signal is generated after the rotor rub-impact, the acoustic emission signal is received by the two sensors, and the acoustic emission signal is accompanied with attenuation and noise interference in the transmission process;

   20): setting attenuation coefficient estimation

   

   Initial iteration value and time delay estimation of

   

   Initial iteration value of (a): collecting acoustic emission signals x (t) received by a first sensor (1) and acoustic emission signals y (t) received by a second sensor (2) by using a rub-impact acoustic emission testing device, and then taking the energy ratio of the two acoustic emission signals as an attenuation coefficient to estimate

   

   Initial iteration value of, finally, setting a time delay estimate

   

   The initial iteration value of (a);

   30): measuring and calculating the estimation error between the true value and the estimated value of y (t): assuming that the energy of y (t) is less than that of x (t) in the two acoustic emission signals, the rub-impact source signal reaches the first sensor (1) first, and then reaches the second sensor (2) after a delay time delta tau, and then the estimation error of the true value and the estimated value of y (t) is measured according to the formula (1):

   

   formula (1)

   Wherein e (t) represents an estimation error,

   

   an attenuation coefficient estimate representing the true attenuation coefficient alpha,

   

   a time delay estimate representing the true time delay deltatau,

   

   representing the signal received by the second sensor (2), the signal x (t) received by the first sensor (1) being delayed in time

   

   The subsequent signal;

   40): according to the estimation error e (t), calculating the step length of the next iteration;

   50): measuring and calculating time delay estimation and attenuation coefficient estimation after iteration according to the formula (2);

   

   formula (2)

   Wherein,

   

   representing the estimate of the attenuation coefficient after the iteration,

   

   representing the attenuation coefficient estimate, mu, before iteration_αTo represent

   

   The step size of the iteration of (a),

   

   representing the partial derivatives of the attenuation coefficient estimates before iteration,

   

   representing the time delay estimate after the iteration,

   

   representing the time delay estimate, mu, before the iteration_ΔτTo represent

   

   The step size of the iteration of (a),

   

   representing a partial derivative of an estimate of the time delay before the iteration;

   60): substituting the time delay estimation after iteration and the attenuation coefficient estimation after iteration, which are measured and calculated in the step 50), into an equation (1), and measuring and calculating a new estimation error e (t + 1);

   70): comparing the relative error of the two estimation errors measured in the step 30) and the step 60), wherein the relative error is

   

   If the relative error is less than 1%, stopping measuring and calculating, taking the time delay estimation measured in the step 50) as the real time delay delta tau, and measuring the position of the rub-impact source according to the formula (3); if the relative error is more than or equal to 1%, replacing e (t) in the step 30) with a new estimation error e (t +1) calculated in the step 60), and then repeating the step 40) to the step 70);

   S₁= D (D-v. DELTA. tau)/2 formula (3)

   Wherein S is₁D represents the distance between the first sensor (1) and the second sensor (2), v represents the propagation speed of the acoustic emission wave, and delta tau represents the time difference of the acoustic emission wave reaching the two sensors.
2. 2. The method for rotor rub-impact acoustic emission source localization according to claim 1, wherein in said step 40), when the estimation error e (t) calculated in step 30) is greater than 0.1, the iterative step is calculated according to equation (4),

   

   formula (4)

   When the estimation error e (t) is less than or equal to 0.1, the iteration step is calculated according to the formula (5),

   

   formula (5)

   Wherein, mu_α(t +1) represents a novel

   

   Iteration step size, μ_Δτ(t +1) represents a novel

   

   Iteration step size of a_α、b_α、a_△τAnd b_△τAll are step size adjustment factors, epsilon (t) represents the error obtained after linear weighting of the error matrix,

   

   representing the gradient of the estimation error e (t) for the attenuation coefficient,

   

   representing the gradient of the estimation error e (t) with respect to the time delay.

Images