A method for vibration component decomposition and resonance region identification based on blade tip timing raw data

Abstract

A vibration component decomposition and resonance region identification method based on blade tip timing raw data relates to the field of blade vibration state monitoring. In order to solve the defects of existing blade tip timing test methods that cannot realize online analysis and cannot automatically identify resonance regions, the present invention adopts a blade tip timing probe to obtain blade tip timing data; by subtracting the actual arrival time of the rotating blade from the theoretical time, and combining the rotation speed of the rotating blade tip to obtain the vibration displacement of the rotating blade tip position: construct a blade tip vibration signal model, select the blade tip timing signal for sparse auxiliary decomposition, set optimization parameters and the maximum number of iterations, and obtain the decomposition results of synchronous vibration signal components and asynchronous vibration signal components; perform automatic identification of resonance regions, and obtain the regional identification results of synchronous resonance and asynchronous resonance respectively. The present invention is mainly used for online analysis of blade tip timing raw data and automatic resonance region identification.

Description

Technical Field

[0001] The invention relates to the field of blade vibration state monitoring, and in particular to a vibration component decomposition and resonance region identification method based on blade tip timing raw data. Background Art

[0002] Rotating blades are widely used in major equipment such as aircraft engines, gas turbines, and large compressors, playing an important role in energy conversion. Their working environment is usually harsh, and they are prone to cracks, fractures and other faults. The reliability of the blade structure plays a key role in the safe and stable operation of the equipment. In order to grasp the vibration state of the blades in real time, the online measurement technology of blade vibration has been widely valued and studied. For example, strain testing, blade tip timing testing, blade tip clearance testing methods, etc. Among them, the blade tip timing test method has a non-contact test and can simultaneously monitor the characteristics of all blades of the same level, which makes this method have important application prospects.

[0003] During operation, the rotating blades will be stimulated by the airflow wake, rotor vibration and other factors, which will produce synchronous vibration. If the fluid in the flow channel produces unstable flow or the blade tip rubs or foreign objects invade, it will cause asynchronous vibration of the blade. This will cause abnormal vibration of the blade and easily cause fatigue damage to the blade. Therefore, in view of the characteristics of synchronous vibration and asynchronous vibration of the blade, the development of vibration parameter identification algorithm is also mainly carried out for synchronous resonance and asynchronous resonance. However, the premise of the above analysis operation is that the occurrence of resonance events is known, and experienced engineers are required to analyze the data, determine the resonance area in the blade tip timing data, and then intercept the corresponding data for vibration parameter identification. This often requires the analysis engineer to have more engineering experience, and the entire data analysis process takes a lot of time.

[0004] Only a few scholars have conducted some relevant research on the problem of automatic identification of resonance regions. Most existing methods use the calculation method of cross-correlation of multiple channels for analysis. Although this method has high computational efficiency, it is greatly affected by noise and has great limitations in extracting asynchronous resonance. The existing technology has the following technical problems:

[0005] Decomposition of tip timing test signal: The rotating blade vibration signal obtained by the tip timing method is relatively complex and difficult to analyze online. It usually requires experienced engineers to perform offline analysis to evaluate the blade vibration characteristics.

[0006] Automatic identification of synchronous resonance area: There are few existing studies on the online identification method of synchronous resonance area of ​​blade tip timing signal. Although it has been proposed to use the signals obtained by two sensors, calculate the signal cross-correlation, and identify the synchronous resonance area by the degree of cross-correlation, this method cannot automatically identify.

[0007] Automatic identification of asynchronous resonance regions: Existing research lacks identification methods for asynchronous resonance regions and can only be performed by experienced engineers through offline data analysis and evaluation.

[0008] Therefore, there is a need for a vibration component decomposition and resonance area identification method based on the original data of blade tip timing, which can realize online analysis of the original data of blade tip timing and automatic resonance area identification, and effectively improve the efficiency of blade vibration characteristic evaluation. Summary of the invention

[0009] In order to solve the defects of the existing blade tip timing test method that can only test known resonance characteristics, cannot realize prior analysis, and cannot automatically identify the resonance area, the present invention provides a vibration component decomposition and resonance area identification method based on the blade tip timing raw data, which can realize online analysis of the blade tip timing raw data and automatic resonance area identification, and effectively improve the efficiency of blade vibration characteristic evaluation.

[0010] The present invention provides a vibration component decomposition and resonance region identification method based on blade tip timing raw data, comprising the following steps:

[0011] S1, blade tip timing data acquisition;

[0012] Arrange multiple blade tip timing probes in the circumferential direction of the rotating blades to record the moment when the rotating blades reach the blade tip timing probes; collect and record the moment when each blade reaches the probes during the blade rotation process through the blade tip timing probes, and collect each rotation time of the rotating shaft through the phase probes as a reference; and obtain the vibration displacement of the rotating blade tip position by subtracting the actual arrival time of the rotating blade from the theoretical time and combining the rotation speed of the rotating blade tip:

[0013] S2. Construct a blade tip vibration signal model, wherein the blade tip vibration signal model is:

[0014] (2);

[0015] Where y is the original blade tip timing signal, s is the synchronous vibration signal component of the blade, x represents the asynchronous vibration signal component of the blade, and w is the signal residual;

[0016] Among them, the blade tip timing signal Y after filtering out the synchronous vibration signal component s is expressed by the following equation:

[0017] (5);

[0018] In the formula, is the basis vector, c is the short-time Fourier transform coefficient array;

[0019] S3, blade tip vibration signal decomposition;

[0020] Select the blade tip timing signal Y for sparse auxiliary decomposition and set the optimization parameters , and , and the maximum number of iterations N, to obtain the decomposition result of the asynchronous vibration signal component x;

[0021] S4, automatic identification of resonance regions;

[0022] Based on the decomposition results, the resonance region is automatically identified, and the identification results of the synchronous resonance region and the asynchronous resonance region are obtained respectively; for the synchronous vibration signal, formula (21) is used to calculate the short-time energy:

[0023] Assume that the short-time energy of the nth frame signal is The calculation formula is:

[0024] (twenty one);

[0025] in, is the frame length, is the sample point in the frame;

[0026] The short-time zero-crossing rate is a measure of the number of times a signal crosses the zero point in a short period of time, that is, the number of times the signal changes from positive to negative or from negative to positive. Let the short-time zero-crossing rate of the n-th frame speech signal be It means that for asynchronous vibration signals, formula (22) is used to calculate the short-time zero-crossing rate:

[0027] (twenty two);

[0028] Where N represents the number of frames, M represents the number of sample points in each frame, and sgn is the sign function, that is:

[0029] (twenty three).

[0030] Further: in S1, the vibration displacement of the tip position of the rotating blade is:

[0031] (1);

[0032] in, is the moment when the tip of the rotating blade actually reaches the tip timing probe, is the theoretical time when the tip of the rotating blade reaches the tip timing probe, is the angular velocity of the rotating blade, is the radius of the rotating blade tip trajectory.

[0033] Further: in S2, the process of constructing the blade tip vibration signal model includes the following steps:

[0034] Assume that the original blade tip timing signal y obtained by the blade tip timing test is expressed by the following equation:

[0035] (2);

[0036] Among them, s represents the synchronous vibration signal component of the blade, x represents the asynchronous vibration signal component of the blade, and w is the signal residual;

[0037] Using a time-frequency array of size M × K of short-time Fourier transform coefficients, for a blade tip timing signal y of length N, we define and for:

[0038] , (3);

[0039] in, is the basis vector, , is the identity matrix, for The Hermitian transpose of ; c is the short-time Fourier transform coefficient array;

[0040] In order to estimate the synchronous vibration signal component s and the asynchronous vibration signal component x of the signal blade from the original blade tip timing signal y, the asynchronous vibration signal component x is modeled as follows:

[0041] (4);

[0042] Since the fluctuation frequency of the synchronous vibration signal component s is smaller than the fluctuation frequency of the noise and asynchronous vibration signal component x, the SG filter is applied to the blade tip vibration signal model in (2), and we have:

[0043] (5).

[0044] Further: In order to estimate the short-time Fourier transform coefficient array c from the given blade tip timing signal Y and minimize the energy of the signal residual w, the following unconstrained optimization problem is proposed:

[0045] (6);

[0046] in," " means "equivalent to", the objective function in formula (6) uses the L1 norm to improve the sparsity of the first-order derivative Dx of the asynchronous vibration signal component x and the short-time Fourier transform coefficient array c, the scalar and is the optimization parameter of regularization; Y is the blade tip timing signal after filtering out the synchronous vibration signal component s.

[0047] Further: in S3, the sparse auxiliary decomposition includes the following steps:

[0048] The variable splitting method is used to decouple the cost function in (6); formula (6) is rewritten as:

[0049] (7);

[0050] In the formula, ;

[0051] Using the scaled augmented Lagrangian for equation (7), minimize (7) through the following iterative process;

[0052] (8a);

[0053] (8b);

[0054] (8c);

[0055] in, is the ADMM convergence parameter, used to control the convergence rate; is the error metric used to judge the degree of convergence;

[0056] Formula (8a) is a standard least squares problem. To find the solution to the standard least squares problem, we need to differentiate Formula (8a) and set the derivative equal to 0 to obtain its normal equation. The display expression is:

[0057] (9);

[0058] Where I is the unit matrix;

[0059] For formula (9) Using the matrix inverse lemma, it can be expanded into:

[0060] (10);

[0061] Substituting formula (10) into formula (9), we get:

[0062] (11);

[0063] Since the first term in formula (11) does not participate in iterative optimization, it is listed separately and formula (11) is rewritten as:

[0064] (12);

[0065] The variable splitting method is also used for formula (8b), and it is rewritten as:

[0066] (13);

[0067] Using the scaled augmented Lagrangian algorithm, equation (13) is minimized through the following iterative process:

[0068] (14a);

[0069] (14b);

[0070] (14c);

[0071] in, It is also an error measure, used to judge the degree of convergence.

[0072] Further: To solve equation (14a), remove the terms that do not depend on the short-time Fourier transform coefficient array c from equation (14a) and make the following replacements:

[0073] (15);

[0074] Therefore, formula (15a) is simplified to:

[0075] (16);

[0076] Among them, p is the intermediate variable, and is the convergence parameter;

[0077] The solution of formula (16) is the solution of the minimum absolute shrinkage and selection operator problem, which is expressed as:

[0078] (17);

[0079] in, is the soft threshold function;

[0080] In order to solve equation (14b), it is simplified to the following form:

[0081] (18);

[0082] in, , , , m, and is the intermediate variable, and Both represent proximal operators;

[0083] In formula (18), the semi-orthogonal linear transformation of the proximal operator is used; the proximal operator in formula (18) is further simplified as:

[0084] (19);

[0085] where tvd(·,·) represents the solution to the total variation denoising problem; therefore, the solution to the optimization problem in (18b) is implemented in two steps:

[0086] (20).

[0087] Further: The short-time zero-crossing rate is used to identify the periodicity of the signal and the oscillatory nature of the waveform.

[0088] The beneficial effects of the present invention are:

[0089] The vibration component decomposition and resonance region identification method based on blade tip timing raw data described in the present invention is a vibration signal decomposition and resonance region identification method based on blade tip timing signals for rotating blades, which can realize online analysis of blade tip timing raw data and automatic resonance region identification, and can effectively improve the efficiency of blade vibration characteristic evaluation, detect blade failure early, and thus reduce the probability of accidents. The present invention can realize the extraction of synchronous resonance components and asynchronous resonance components of blades, and respectively propose online and automatic identification methods for synchronous resonance regions and asynchronous resonance regions. It provides a fast and effective means for online evaluation of blade vibration status.

[0090] The blade tip timing signal decomposition method based on sparse optimization and the resonance area automatic identification method proposed in the present invention only rely on the signal of one blade tip timing sensor for analysis, which can effectively solve the problem of difficult online evaluation of blade vibration state, realize online identification of blade resonance area, and provide effective support for fault warning of rotating blades. BRIEF DESCRIPTION OF THE DRAWINGS

[0091] Figure 1 It is the waveform diagram of the timing signal of the tip of different blades;

[0092] Figure 2 It is a waveform diagram of the original data of blade tip timing in the sparse assisted decomposition result of blade tip timing signal;

[0093] Figure 3 is a waveform diagram of the synchronous vibration signal component in the sparse-assisted decomposition result of the blade tip timing signal;

[0094] Figure 4 It is a waveform diagram of the asynchronous vibration signal component in the sparse-assisted decomposition result of the blade tip timing signal;

[0095] Figure 5 It is the amplitude waveform of different shaft frequencies in the synchronous resonance region identification result;

[0096] Figure 6 It is the waveform of short-time energy in the synchronous resonance region identification result;

[0097] Figure 7 This is the amplitude waveform of the resonance mark in the synchronous resonance region identification result

[0098] Figure 8 It is the amplitude waveform of different shaft frequencies in the asynchronous resonance region identification result;

[0099] Fig. 9 It is the amplitude waveform of the short-time zero-crossing rate in the asynchronous resonance region identification result;

[0100] Fig.10 It is the amplitude waveform of the resonance mark in the asynchronous resonance region identification result. DETAILED DESCRIPTION

[0101] The following are only preferred specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by a technician familiar with the technical field within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. The embodiments described below are only used to explain the present invention and cannot be interpreted as limitations on the present invention. The protection scope of the present invention should be based on the protection scope of the claims. The embodiments of the present invention are described in detail below. In order to facilitate the description of the present invention and simplify the description, the technical terms used in the specification of the present invention should be interpreted in a broad sense, including but not limited to conventional replacement schemes not mentioned in this application, and also including direct implementation methods and indirect implementation methods.

[0102] Example 1

[0103] Combination Figure 1-Figure 10 The present embodiment describes a vibration component decomposition and resonance region identification method based on blade tip timing raw data, a blade tip timing raw data decomposition method based on sparse assisted optimization, and a synchronous resonance region and asynchronous resonance region identification method based on instantaneous energy and short-time zero-crossing rate, including the following steps:

[0104] Step 1: blade tip timing data acquisition;

[0105] The tip timing test is an effective means of analyzing the vibration of rotating blades. By installing a tip timing probe on the casing or housing, the time when each blade reaches the probe during the rotation process is recorded, and a key phase probe is installed near the shaft as the time reference for each revolution. Based on the tip timing test principle, multiple tip timing probes are arranged in the circumference of the rotating blade to record the time when the rotating blade reaches the tip timing probe; the vibration displacement d of the tip position of the rotating blade is obtained by subtracting the actual arrival time of the rotating blade from the theoretical time and combining the rotation speed of the tip of the rotating blade:

[0106] (1);

[0107] in, is the moment when the tip of the rotating blade actually reaches the tip timing probe, is the theoretical time when the tip of the rotating blade reaches the tip timing probe, is the angular velocity of the rotating blade, is the radius of the rotating blade tip trajectory;

[0108] Step 2: blade tip vibration signal description;

[0109] Assume that the blade tip timing signal y can be expressed by the following equation

[0110] (2);

[0111] Among them, y is the original blade tip timing signal, s is the synchronous vibration signal component of the blade, x is the asynchronous vibration signal component of the blade, that is, the sparse oscillation component caused by the asynchronous resonance of the blade, and w is the signal residual.

[0112] Signals that allow for sparse representation can be described by using a small number of coefficients of an appropriate transform. Such representations, when they exist, can explain most of the energy contained in the signal. Signals consisting of oscillatory components can be sparsely represented using the STFT (Short-time Fourier Transform). The STFT requires the specification of several parameters, such as the window length, the overlap factor, and the Discrete Fourier Transform (DFT) length. Specifically, using a time-frequency array of size M × K of the STFT coefficients, for a blade tip timing signal y of length N, we define and for

[0113] , , (3);

[0114] in, is the basis vector, , is the identity matrix, for The Hermitian transpose of ; c is the short-time Fourier transform (STFT) coefficient array;

[0115] In order to evaluate the synchronous vibration signal component s and the asynchronous vibration signal component x of the signal blade from the original BTT signal (i.e., the blade tip timing signal) y, the asynchronous vibration signal component x can be modeled as:

[0116] (4);

[0117] Since the fluctuation frequency of the synchronous vibration signal component s is smaller than the fluctuation frequency of the noise and asynchronous vibration signal component x, the SG filter (i.e., Savitzky-Golay filter, a digital filter widely used in data smoothing and fitting) is considered to be applied to the signal model in (2), which is:

[0118] (5);

[0119] In order to estimate the short-time Fourier transform coefficient array c from the given blade tip timing signal Y and minimize the energy of the signal residual w, the following unconstrained optimization problem is proposed:

[0120] (6);

[0121] in," ” means “equivalent to”. The objective function in Equation (6) uses the L1 norm (also known as the Manhattan distance or absolute value norm) to improve the sparsity of the first-order derivative Dx of the signal x and the short-time Fourier transform coefficient array c. The scalar and is the regularization parameter, and Y is the blade tip timing signal after filtering out the synchronous vibration signal component s.

[0122] Step 3: blade tip vibration signal decomposition;

[0123] The Douglas-Rachford Splitting Method (DRSM or DR method) and the Alternating Direction Method of Multipliers (ADMM) are applied to solve optimization problems with separable and structured objective functions. The basic idea is to decompose the complex optimization problem into simpler subproblems, solve these subproblems iteratively, and update the Lagrange multipliers to ensure convergence. First, the "variable splitting" method is used to decouple the cost function in (6). Formula (6) can be rewritten as:

[0124] (7);

[0125] Using the scaled augmented Lagrangian for equation (7), (7) can be minimized by the following iterative process.

[0126] (8a);

[0127] (8b);

[0128] (8c);

[0129] in, is the ADMM convergence parameter, which is used to control the convergence rate.

[0130] Formula (8a) is a standard least squares problem. To find the solution to the standard least squares problem, we can derive Formula (8a) and set the derivative equal to 0 to obtain its normal equation. The display expression is

[0131] (9);

[0132] For formula (9) Using the matrix inverse lemma, it can be expanded into

[0133] (10);

[0134] Substituting equation (10) into equation (9), we can obtain

[0135] (11);

[0136] Since the first term in formula (11) does not participate in iterative optimization, it is listed separately. Formula (11) can be rewritten as:

[0137] (12);

[0138] The same variable splitting method is used for formula (8b) and it can be rewritten as

[0139] (13);

[0140] Using the scaled augmented Lagrangian algorithm, we minimize equation (13) through the following iterative process:

[0141] (14a);

[0142] (14b);

[0143] (14c);

[0144] To solve (14a), remove the terms that do not depend on c from (14a) and make the following substitutions

[0145] (15);

[0146] Therefore, (15a) can be simplified to

[0147] (16);

[0148] The solution of formula (16) is the solution of the least absolute shrinkage and selection operator (LASSO, full name Least Absolute Shrinkage and Selection Operator, that is, the least absolute shrinkage and selection algorithm, which is a linear regression method using L1 regularization) problem, expressed as

[0149] (17);

[0150] In order to solve equation (14b), it is simplified to the following form

[0151] (18);

[0152] in, , , , m, and is the intermediate variable, and Both represent proximal operators;

[0153] In (18), a semi-orthogonal linear transformation of the proximal operator is used. The proximal operator in (18) can be further simplified to

[0154] (19);

[0155] Here, tvd(·,·) represents the solution to the total variation denoising problem. Therefore, the solution to the optimization problem in (18b) can be implemented in two steps:

[0156] (20).

[0157] Step 4: Automatic identification of resonance regions;

[0158] The blade tip synchronous resonance signal and blade tip asynchronous resonance signal can be obtained by the sparse auxiliary decomposition method of the blade tip timing signal introduced in step 2. In order to realize the automatic identification of the resonance area, a synchronous resonance area identification method based on short-time energy and an asynchronous resonance area identification method based on short-time zero-crossing rate are proposed respectively. Short-time energy (STE) is used to measure the amplitude change of the signal in a short time, so as to capture the instantaneous characteristics of the signal. Suppose the short-time energy of the n-th frame signal is expressed as The calculation formula is

[0159] (twenty one);

[0160] in, is the frame length, is the sample point in the frame. For a signal with higher energy in a short time, its STE value is larger; and for a signal with lower energy in a short time, its STE value is smaller.

[0161] The short-time zero-crossing rate (STZCR) is a measure of the number of times a signal crosses the zero point in a short period of time, that is, the number of times the signal changes from positive to negative or from negative to positive. The calculation formula is

[0162] (twenty two);

[0163] Where N represents the number of frames, M represents the number of sample points in each frame, and sgn is the sign function, that is,

[0164] (twenty three);

[0165] STZCR can be used to identify the periodicity of a signal and the oscillatory nature of a waveform. For a periodic signal, STZCR may be high because the signal frequently crosses zero. For a stationary signal or a signal with a large DC component, STZCR may be low because the signal maintains the same sign for a longer period of time.

[0166] Example 2

[0167] This embodiment is described in conjunction with Example 1. This embodiment discloses a method for vibration component decomposition and resonance region identification based on blade tip timing raw data, which specifically includes the following steps:

[0168] Step 1: blade tip timing data acquisition;

[0169] Four blade tip timing probes (TIP1, TIP2, TIP3, TIP4) are installed along the circumferential direction of the blade tip. The relative installation angles of the blade tip probes to the probe TIP1 are 0°, 35°, 45°, and 54°, respectively. The vibration displacement of the blade tip is collected using the key phase method. In order to stimulate the asynchronous resonance of the blade, additional friction is applied to excite the blade under gas excitation. The operating frequency range of the blade disk is set to [33Hz, 50Hz], and the following is obtained based on the blade tip timing test principle: Figure 1 Tip vibration displacement of different blades shown.

[0170] Step 2: Construction of blade tip timing signal model;

[0171] Assume that the blade tip timing signal y can be expressed by the following equation:

[0172] (26);

[0173] Where s represents the synchronous resonance of the blade, x is the sparse oscillation component caused by the asynchronous resonance of the blade, and w is the signal residual.

[0174] In order to evaluate the synchronous resonance component s and asynchronous resonance component x of the signal blade from the original BTT signal y, the signal component x can be modeled as:

[0175] (27);

[0176] Since the fluctuation frequency of the synchronous resonance component is smaller than the fluctuation frequency of the noise and the asynchronous resonance, considering applying the SG filter to the signal model in (6), we have:

[0177] (28);

[0178] In order to estimate the coefficient array c from the given blade tip timing signal Y and minimize the energy of the residual w, the following unconstrained optimization problem is proposed:

[0179] (29);

[0180] The objective function in (10) uses the L1 norm to improve the sparsity of the first-order derivative Dx of the signal x and the short-time Fourier transform coefficient array c. and is the regularization parameter.

[0181] Step 3: blade tip vibration signal decomposition;

[0182] Select the tip timing data of blade 4# for sparse auxiliary decomposition and set the optimization parameters , , , , maximum number of iterations . Get Figure 2-Figure 4 The decomposition results of the synchronous vibration signal component and the asynchronous vibration signal component are shown.

[0183] Step 4: Automatic identification of resonance regions;

[0184] based on Figure 2 The decomposition results Figure 3 and Figure 4 Automatically identify the resonance area. For synchronous vibration signals, use formula (21) to calculate the short-time energy, and for asynchronous vibration signals, use formula (22) to calculate the short-time zero-crossing rate. Among them, the data window length is 20, the window overlap rate is 50%, and the threshold , , respectively, as Figure 5-Figure 10 The synchronous resonance and asynchronous resonance region identification results are shown in Fig. Figure 6 In the specification, STE refers to Short-Time Energy, usually abbreviated as STE, and STZCR refers to Short-Time Zero-Crossing Rate, usually abbreviated as STZCR.

[0185] Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are only used to illustrate the technical solutions of the present invention and cannot be understood as limitations of the present invention. Ordinary technicians in the field can modify and replace the above embodiments within the scope of the present invention without departing from the principles and purpose of the present invention.

Claims

1. A vibration component decomposition and resonance region identification method based on blade tip timing raw data, characterized in that: The steps include: S1, blade tip timing data acquisition; A plurality of blade tip timing probes are arranged in the circumferential direction of the rotating blades to record the time when the rotating blades reach the blade tip timing probes; the time when each blade reaches the probes during the blade rotation process is collected and recorded by the blade tip timing probes, and each rotation time of the rotating shaft is collected by the key phase probes as a reference; the vibration displacement of the rotating blade tip position is obtained by subtracting the actual arrival time of the rotating blade from the theoretical time and combining the rotation speed of the rotating blade tip; S2. Construct a blade tip vibration signal model, wherein the blade tip vibration signal model is: (2)； Where y is the original blade tip timing signal, s is the synchronous vibration signal component of the blade, x is the asynchronous vibration signal component of the blade, and w is the signal residual; Among them, the blade tip timing signal Y after filtering out the synchronous vibration signal component s is expressed by the following equation: (5)； In the formula, is the basis vector, c is the short-time Fourier transform coefficient array; In order to estimate the short-time Fourier transform coefficient array c from the given blade tip timing signal Y and minimize the energy of the signal residual w, the following unconstrained optimization problem is proposed: (6)； in," " means "equivalent to", the objective function in formula (6) uses the L1 norm to improve the sparsity of the first-order derivative Dx of the asynchronous vibration signal component x and the short-time Fourier transform coefficient array c, the scalar and is the optimization parameter of regularization; Y is the blade tip timing signal after filtering out the synchronous vibration signal component s; S3, blade tip vibration signal decomposition; Select the blade tip timing signal Y for sparse auxiliary decomposition and set the optimization parameters , and , and the maximum number of iterations N, to obtain the decomposition result of the asynchronous vibration signal component x of the blade; In S3, the sparse auxiliary decomposition includes the following steps: The variable splitting method is used to solve the cost function in formula (6); formula (6) is rewritten as: (7)； In the formula, is the expected value of c; Using the scaled augmented Lagrangian for formula (7), minimize formula (7) through the following iterative process; (8a)； (8b)； (8c)； in, is the ADMM convergence parameter, used to control the convergence rate; is the error metric used to judge the degree of convergence; Formula (8a) is a standard least squares problem. To find the solution to the standard least squares problem, we need to differentiate Formula (8a) and set the derivative equal to 0 to obtain its normal equation. The display expression is: (9)； Where I is the unit matrix; For formula (9) Using the matrix inverse lemma, it can be expanded into: (10)； Substituting formula (10) into formula (9), we obtain: (11)； Since the first term in formula (11) does not participate in iterative optimization, it is listed separately and formula (11) is rewritten as: (12)； The variable splitting method is also used for formula (8b), and it is rewritten as: (13)； in, is the expected value of c, and the scaled augmented Lagrangian algorithm is used to minimize formula (13) through the following iterative process: (14a)； (14b)； (14c)； in, It is also the error measure, used to judge the degree of convergence; S4, automatic identification of resonance regions; Based on the decomposition results, the resonance region is automatically identified, and the identification results of the synchronous resonance region and the asynchronous resonance region are obtained respectively; for the synchronous vibration signal, formula (21) is used to calculate the short-time energy: Assume that the short-time energy of the nth frame signal is The calculation formula is: (21)； in, is the frame length, is the sample point in the frame; The short-time zero-crossing rate is a measure of the number of times a signal crosses the zero point in a short period of time, that is, the number of times the signal changes from positive to negative or from negative to positive. Let the short-time zero-crossing rate of the n-th frame speech signal be It means that for asynchronous vibration signals, formula (22) is used to calculate the short-time zero-crossing rate: (22)； Where N represents the number of frames, M represents the number of sample points in each frame, and sgn is the sign function, that is: (23)。 2. The vibration component decomposition and resonance region identification method based on blade tip timing raw data according to claim 1, characterized in that: In S1, the vibration displacement of the rotating blade tip position is: (1)； in, is the moment when the tip of the rotating blade actually reaches the tip timing probe, is the theoretical time when the tip of the rotating blade reaches the tip timing probe, is the angular velocity of the rotating blade, is the radius of the rotating blade tip trajectory.

3. The vibration component decomposition and resonance region identification method based on blade tip timing raw data according to claim 1, characterized in that: In S2, the process of constructing the blade tip vibration signal model includes the following steps: Assume that the original blade tip timing signal y obtained by the blade tip timing test is expressed by the following equation: (2)； Among them, s represents the synchronous vibration signal component of the blade, x represents the asynchronous vibration signal component of the blade, and w is the signal residual; Using a time-frequency array of short-time Fourier transform coefficients of size M × K, for a blade tip timing signal y of length N, we define and for ， (3)； in, is the basis vector, , is the identity matrix, for The Hermitian transpose of ; c is the short-time Fourier transform coefficient array; In order to estimate the synchronous vibration signal component s and the asynchronous vibration signal component x of the signal blade from the original blade tip timing signal y, the asynchronous vibration signal component x is modeled as follows: (4)； Since the fluctuation frequency of the synchronous vibration signal component s is smaller than the fluctuation frequency of the noise and asynchronous vibration signal component x, the SG filter is applied to the blade tip vibration signal model in formula (2) to obtain the blade tip timing signal Y with the synchronous resonance component filtered out, which is: (5)。 4. The vibration component decomposition and resonance region identification method based on blade tip timing raw data according to claim 1, characterized in that: To solve equation (14a), remove the terms that do not depend on the short-time Fourier transform coefficient array c from equation (14a) and make the following substitutions: (15)； Therefore, formula (15a) is simplified to: (16)； Among them, p is the intermediate variable, and is the convergence parameter; The solution of formula (16) is the solution of the minimum absolute shrinkage and selection operator problem, expressed as: (17)； in, is the soft threshold function; To solve equation (14b), it is simplified to the following form: (18)； in, , , , m, and is the intermediate variable, and Both represent proximal operators; In formula (18), the semi-orthogonal linear transformation of the proximal operator is used; the proximal operator in formula (18) is further simplified as: (19)； Here, tvd(·,·) represents the solution to the total variation denoising problem; therefore, the solution to the optimization problem in formula (18b) is implemented in two steps: (20)。 5. The vibration component decomposition and resonance region identification method based on blade tip timing raw data according to claim 4, characterized in that: The short-time zero-crossing rate is used to identify the periodicity of the signal and the oscillatory nature of the waveform.

Images