A method for locating shaft fatigue cracks driven by superharmonic dynamics mode
Through the ultra-harmonic dynamic mode driving method, the fusion of multi-sensor information and multi-scale ultra-harmonic dynamic mode is used to construct damage factors, solving the efficient positioning of the shaft fatigue crack in the operating state, and achieving fast and accurate crack positioning and noise robustness.
Patent Information
- Application Number
- CN202210641952.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-06-07
- Publication Date
- 2024-11-01
- Estimated Expiration
- 2042-06-07
AI Technical Summary
The prior art is difficult to efficiently locate the shaft fatigue cracks in operation, especially in the absence of strong noise interference and healthy reference models, and it is difficult to distinguish cracks from other factors such as ladders and grooves.
The shaft fatigue crack positioning method driven by ultra-harmonic dynamics mode is adopted. The response matrix is constructed by multi-sensor information, and the dynamics mode is decomposed and extracted. The damage factor is realized by fusion of multi-scale ultra-harmonic dynamics mode information. Crack positioning can eliminate interference without reference to healthy models.
It realizes rapid positioning of multiple cracks in the rotation shaft in operation, has strong noise robustness, can accurately eliminate interference from factors such as ladders and grooves, and provides accurate crack position information.
Smart Images

Figure CN114912494B_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the field of rotating shaft crack diagnosis, and in particular relates to a rotating shaft fatigue crack positioning method driven by a superharmonic dynamics mode. Technical Background
[0002] The shaft is an important component for transmitting motion in rotating equipment. Due to long-term complex alternating or even random loads, fatigue cracks are prone to occur in the shaft. If not discovered in time, it may lead to sudden serious consequences. Crack location can not only detect the presence of cracks, but also provide information on the location of cracks, which is of great significance for ensuring the safety of equipment operation and avoiding serious accidents.
[0003] Nondestructive testing methods such as X-ray, magnetic particle testing, eddy current testing, and ultrasonic testing can be used to detect shaft cracks. These methods have high detection accuracy, but they require shutdown and low detection efficiency, making it difficult to achieve online detection. Vibration signals are widely used in state monitoring and fault diagnosis because they are easy to obtain and low in cost. Therefore, many methods attempt to detect whether there are cracks in the shaft through the vibration signals generated by the running equipment. For shaft crack detection methods based on vibration signals, crack detection is generally achieved by extracting crack response characteristics from vibration signals. However, it is still a recognized problem to achieve crack location under running conditions. In order to achieve crack location, sufficient spatial information must be introduced. On the one hand, spatial information can be provided by accurate models, but it is often very difficult to obtain accurate models for complex systems; on the other hand, spatial information can also be introduced through multiple sensors, but current crack location methods often require the reference of a healthy shaft system, and the location of the crack is determined by comparing the response characteristics to be tested under the same working conditions with the response characteristics under a healthy state. However, cracks in the shaft or changes in its local stiffness (such as steps and back-off grooves in the design) will cause distortion in the system's dynamic model. Further breakthroughs are needed to achieve efficient positioning of cracks in the operating shaft in the presence of strong noise interference and the absence of a healthy reference model. Summary of the invention
[0004] Based on the above background, the present invention proposes a method for locating fatigue cracks of a rotating shaft driven by a superharmonic dynamic mode. The method makes full use of multi-sensor information, extracts the superharmonic dynamic mode caused by cracks contained in the multi-sensor information through dynamic mode decomposition, and uses multi-scale superharmonic dynamic mode information fusion to construct damage factors to achieve crack location. The method can realize rapid positioning of multiple cracks of the rotating shaft under running conditions, has strong noise robustness, and the distortion in the superharmonic dynamic mode is only caused by the local nonlinearity of the crack, while the local stiffness reduction caused by linear factors such as steps and grooves has no contribution to it, and the global nonlinearity caused by misalignment has no effect on it. Therefore, the interference of factors such as misalignment and steps can be eliminated without the reference of a healthy model.
[0005] The present invention provides the following technical solutions to achieve the above technical functions:
[0006] A method for locating a shaft fatigue crack driven by a superharmonic dynamics mode, the method comprising the following steps:
[0007] S1: Non-contact test of the lateral vibration response of the rotating shaft at multiple measuring points under stable working conditions, such as displacement, velocity or acceleration response;
[0008] S2: Construct response matrix
[0009] Specific operation: The vibration time domain responses of n measuring points are combined into a response matrix Y, where y i is the response measured at measuring point i, and m is the number of points in the response data.
[0010]
[0011] S3: Extraction of superharmonic dynamical modes (SDM) φ(x) by dynamical mode decomposition
[0012] Specific operations:
[0013] S3.1: Take the first l data points in the response matrix to form the response matrix Y1 and its time shift matrix Y2
[0014] S3.2: Assume that there is a linear transformation between Y1 and Y2 such that:
[0015] Y2=AY1 (2)
[0016] Then the transformation matrix in represents the pseudo-inverse of Y1. The eigenvectors and eigenvalues of the transformation matrix A are the dynamic decomposition patterns and eigenvalues corresponding to the system response matrix. Due to the high dimension of A and the interference of system noise, it is necessary to reduce the dimension of the transformation matrix A.
[0017] S3.3: Perform singular value decomposition on Y1, and we have:
[0018] Y1=UΣV * (3)
[0019] In the formula, * represents the transpose of the matrix, then we have therefore,
[0020] A=Y2VΣ -1 U * (4)
[0021] S3.4: Use eigenmode decomposition to reduce the dimension of A and perform r-order rank truncation. The approximate matrix after dimensionality reduction is Right now:
[0022]
[0023] S3.5: Solution The eigenvalues and eigenvectors of can be used to obtain the eigenvalues and dynamic modes of the system, that is,
[0024]
[0025] Where μ is the eigenvalue of the kinetic mode decomposition and φ is the corresponding kinetic mode.
[0026] The fatigue crack rotating shaft under steady-state operation will introduce local nonlinearity due to the breathing effect of the crack. The response of each measuring point in the system will show superharmonic components with the rotation frequency as the fundamental frequency, and the superharmonic frequency components will be reflected in μ. The dynamic mode corresponding to the superharmonic frequency in μ is the superharmonic dynamic mode.
[0027] S3.6: According to the rotation speed of the shaft, the superharmonic dynamic mode corresponding to the high-order component of the rotation frequency, i.e., φ(x), is extracted.
[0028] S4: Fusion of structural damage indicators using multi-scale superharmonic pattern evidence
[0029] Specific operations:
[0030] S4.1: The extracted superharmonic dynamics mode φ(x) is decomposed into multiple scales using the Gaussian multiscale decomposition method:
[0031]
[0032] in is the Gaussian kernel function
[0033] S4.2: Calculate the Teager energy of the superharmonic dynamics mode at each scale:
[0034] T(x,σ)=L 2 (x,σ)-L(x-1,σ)L(x+1,σ) (8)
[0035] S4.3: Data fusion of Teager energy superharmonic dynamics models at each scale is performed as follows:
[0036] At the kth scale, the crack appears at position x i The probability at (i=1,2…,n) is expressed as:
[0037]
[0038] Therefore, at the kth scale, the crack appears between x1 and x n The probability vector of the position
[0039] By using the DS evidence fusion rule, the basic probability distribution functions at each scale are fused to obtain the probability of cracks at each location. The data source corresponding to the Teager energy superharmonic dynamics mode at each scale can be obtained from S1 to S M According to the fusion rule, the result of the fusion of the first two scales is expressed as:
[0040]
[0041] Where K1 represents the conflict weight between evidence 1 and evidence 2:
[0042]
[0043] When K1=1, Does not exist, indicating that the two pieces of evidence are completely contradictory, which is also a defect of the classical DS evidence theory, which cannot handle completely contradictory evidence. However, in the present invention, due to the similarity of each scale, the problem of completely contradictory evidence will not occur.
[0044] According to this rule, the remaining scales are fused in order, and finally we get:
[0045]
[0046] In the formula, Final income That is the damage factor.
[0047] S5: Determine the damage index distribution of each measuring point. If the distribution is relatively uniform, there is no crack.
[0048] S6: Otherwise, there is a crack in the shaft, and the crack position corresponds to the peak position of the damage index.
[0049] The present invention adopts the above technical solution to bring the following beneficial effects:
[0050] The present invention applies the dynamic mode decomposition method to the location of shaft fatigue cracks for the first time, extracts the corresponding superharmonic dynamic mode of the superharmonic vibration signal generated by the nonlinearity of fatigue cracks, and constructs a new crack damage index on this basis. The method only needs to test the vibration response of multiple measuring points of the shaft system under the running state to realize the location of single and multiple cracks in the shaft under the running state. No model reference of the healthy shaft is required, and no vibration information under the healthy state needs to be tested. The method can avoid the interference of common steps and grooves on the shaft and rotor misalignment that is difficult to distinguish from crack characteristics, and has the advantages of good noise robustness, fast and reliable. BRIEF DESCRIPTION OF THE DRAWINGS
[0051] Figure 1 To locate the flow chart;
[0052] Figure 2 This is a schematic diagram of the cracked rotor and measurement point arrangement;
[0053] Figure 3 It is the time domain diagram of the displacement response with or without noise at the measuring point near the bearing;
[0054] Figure 4 The second-order superharmonic dynamic modes extracted with and without noise;
[0055] Figure 5 is the corresponding second-order multiscale superharmonic dynamics mode with and without noise;
[0056] Figure 6 The corresponding second-order multiscale Teager energy superharmonic dynamics mode with and without noise;
[0057] Figure 7 Comparison of positioning results with and without noise. DETAILED DESCRIPTION
[0058] In order to make the technical solution of the present invention more clear, the invention will be further described below in conjunction with the accompanying drawings and embodiments thereof. The specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.
[0059] like Figure 1-7 As shown, the shaft fatigue crack location process of the present invention is as follows Figure 1 .
[0060] The present application provides a method for locating a shaft fatigue crack driven by a superharmonic dynamics mode, the method comprising the following steps:
[0061] S1: Non-contact test of the lateral vibration displacement response of the rotating shaft at multiple measuring points under stable working conditions. For details, please refer to Figure 2 As shown;
[0062] The strain energy release rate theory of linear fracture mechanics and the finite element method were used to establish Figure 2 The dynamic model of the breathing crack rotor bearing system with stepped shaft segments is shown. A sensor measuring point is arranged every two nodes, with a total of 22 measuring points. Two fatigue cracks are inserted between measuring points 9 and 10, and measuring points 16 and 17. The crack depth is 15% of the shaft diameter, which is equal to 1.5mm. The sampling frequency is set to 5000Hz. The first-order critical speed of the system is 42Hz. The operating speed is set to 1 / 3 of the critical speed, that is, 840r / min. In order to simulate the interference of test noise, 10% Gaussian white noise is added to the theoretical response. Figure 3 The typical vibration displacement response obtained when there is noise or not at measuring point 10.
[0063] S2: Construct response matrix
[0064] The vibration time domain responses of 22 (i.e., n=22) measuring points are combined into a response matrix Y, where y i is the response measured at measuring point i, m is the number of points in the response data, and m = 10000, we can get:
[0065]
[0066] S3: Extraction of superharmonic dynamical modes (SDM) φ(x) by dynamical mode decomposition
[0067] Specific operations:
[0068] S3.1: Take the first l data points in the response matrix to form the response matrix Y1 and its time-shift matrix Y2, l = 5000,
[0069] S3.2: Assume that there is a linear transformation between Y1 and Y2 such that:
[0070] Y2=AY1 (2)
[0071] Then the transformation matrix in represents the pseudo-inverse of Y1. The eigenvectors and eigenvalues of the transformation matrix A are the dynamic decomposition patterns and eigenvalues corresponding to the system response matrix. Due to the high dimension of A and the interference of system noise, it is necessary to reduce the dimension of the transformation matrix A.
[0072] S3.3: Perform singular value decomposition on Y1, and we have:
[0073] Y1=UΣV * (3)
[0074] In the formula, * represents the transpose of the matrix, then we have therefore,
[0075] A=Y2VΣ -1 U* (4)
[0076] S3.4: Use eigenmode decomposition to reduce the dimension of A and perform r-order rank truncation. The approximate matrix after dimensionality reduction is Right now:
[0077]
[0078] S3.5: Solution The eigenvalues and eigenvectors of can be used to obtain the eigenvalues and dynamic modes of the system, that is,
[0079]
[0080] Where μ is the eigenvalue of the kinetic mode decomposition and φ is the corresponding kinetic mode.
[0081] The fatigue crack shaft under steady-state operation will introduce local nonlinearity due to the breathing effect of the crack, and the responses of each measuring point in the system will show superharmonic components with the rotation frequency as the fundamental frequency, mainly the first three order frequency components. The superharmonic frequency components will be reflected in μ, and the dynamic mode corresponding to the superharmonic frequency in μ is the superharmonic dynamic mode. The response of each measuring point in the system will show superharmonic components with the rotation frequency as the fundamental frequency, and the dynamic mode corresponding to the superharmonic frequency components is the superharmonic dynamic mode. In the present invention, the second-order superharmonic dynamic mode is selected for crack location because at the selected speed, the first-order and third-order dynamic modes correspond to the first-order vibration mode of the rotor system, so the distortion interference caused by the local stiffness reduction cannot be avoided. Figure 3 is the corresponding second-order superharmonic dynamic mode with and without noise.
[0082] S3.6: According to the rotation speed of the shaft, the superharmonic dynamic mode corresponding to the second-order component of the rotation frequency, i.e., φ(x), is extracted.
[0083] S4: Fusion of structural damage indicators using multi-scale superharmonic pattern evidence
[0084] Specific operations:
[0085] S4.1: The extracted superharmonic dynamics mode φ(x) is decomposed into multiple scales using the Gaussian multiscale decomposition method:
[0086]
[0087] in is the Gaussian kernel function.
[0088] Figure 4 It is the corresponding multi-scale second-order superharmonic dynamic mode with and without noise.
[0089] S4.2: Calculate the Teager energy of the superharmonic dynamics mode at each scale:
[0090] T(x,σ)=L 2 (x,σ)-L(x-1,σ)L(x+1,σ) (8)
[0091] like Figure 5 shown.
[0092] S4.3: Data fusion of Teager energy superharmonic dynamics models at each scale is performed as follows:
[0093] At the kth scale, the crack appears at position x i The probability at (i=1,2…,n) is expressed as:
[0094]
[0095] Therefore, at the kth scale, the crack appears between x1 and x n The probability vector
[0096] By using the DS evidence fusion rule, the basic probability distribution functions at each scale are fused to obtain the probability of cracks at each location. The data source corresponding to the Teager energy superharmonic dynamics mode at each scale can be obtained from S1 to S M According to the fusion rule, the result of the fusion of the first two scales is expressed as:
[0097]
[0098] Where K1 represents the conflict weight between evidence 1 and evidence 2:
[0099]
[0100] When K1=1, Does not exist, indicating that the two pieces of evidence are completely contradictory, which is also a defect of the classical DS evidence theory, which cannot handle completely contradictory evidence. However, in the present invention, due to the similarity of each scale, the problem of completely contradictory evidence will not occur.
[0101] According to this rule, the remaining scales are fused in order, and finally we get:
[0102]
[0103] In the formula, Final income is the damage factor. Figure 6 shown.
[0104] S5: Determine the damage index distribution of each measuring point. If the distribution is relatively uniform, there is no crack.
[0105] S6: Otherwise, there is a crack in the shaft, and the crack position corresponds to the peak position of the damage index.
[0106] Depend on Figure 6 It can be seen that the damage factor shows an obvious double peak distribution, so it can be judged that there are two cracks in the shaft. Specifically, the damage factor at measuring points 9 and 16 shows obvious peaks, which is consistent with the location of the real crack, and excludes the interference caused by the decrease in the stiffness of the stepped shaft. Therefore, the accuracy of crack location of this method can be explained, and by comparing the results with and without noise, it can be seen that the method is more robust to noise.
[0107] The above is only a preferred embodiment of the present invention. It should be noted that, for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered as the protection scope of the present invention. All components not specified in this embodiment can be implemented by existing technologies.
Claims
1. A method for locating shaft fatigue cracks driven by superharmonic dynamics mode, characterized in that: The method comprises the following steps: S1: non-contact test of the lateral vibration response of the rotating shaft at multiple measuring points under stable working conditions; S2: construct the response matrix; S3: Extract superharmonic dynamical modes φ(x) by dynamical mode decomposition; S4: Fusion of structural damage indicators using multi-scale superharmonic pattern evidence; S5: Determine the damage index distribution of each measuring point. If the distribution is relatively uniform, there is no crack. S6: Otherwise, there is a crack in the shaft, and the crack position corresponds to the peak position of the damage index; Step S2 is specifically as follows: The vibration time domain responses of n measuring points are combined into a response matrix Y where y i is the response measured at measuring point i, and m is the number of points in the response data; Step S3 is specifically as follows: S3.1: Take the first l data points in the response matrix to form the response matrix Y1 and its time shift matrix Y2; S3.2: Assume that there is a linear transformation between Y1 and Y2 such that: Y2=AY1 Then the transformation matrix in represents the pseudo-inverse of Y1, and the eigenvectors and eigenvalues of the transformation matrix A are the dynamic decomposition modes and eigenvalues corresponding to the system response matrix; S3.3: Perform singular value decomposition on Y1, and we have: Y1=UΣV * In the formula, * represents the transpose of the matrix, then we have therefore, A=Y2VΣ -1 U * S3.4: Use eigenmode decomposition to reduce the dimension of A and perform r-order rank truncation. The approximate matrix after dimensionality reduction is Right now: S3.5: Solution The eigenvalues and eigenvectors of can be used to obtain the eigenvalues and dynamic modes of the system, that is, φ=Uw In the formula, μ is the eigenvalue of the dynamic mode decomposition, and φ is the corresponding dynamic mode; S3.6: According to the rotation speed of the shaft, extract the superharmonic dynamic mode corresponding to the high-order component of the rotation frequency, i.e., φ(x); Step S4 is specifically as follows: S4.1: The extracted superharmonic dynamics mode φ(x) is decomposed into multiple scales using the Gaussian multiscale decomposition method: in is the Gaussian kernel function S4.2: Calculate the Teager energy of the superharmonic dynamics mode at each scale: T(x,σ)=L 2 (x,σ)-L(x-1,σ)L(x+1,σ) S4.3: Data fusion of Teager energy superharmonic dynamics models at each scale is performed as follows: At the kth scale, the crack appears at position x i The probability at (i=1,2…,n) is expressed as: Therefore, at the kth scale, the crack appears between x1 and x n The probability vector of the position By using the DS evidence fusion rule, the basic probability distribution functions at each scale are fused to obtain the probability of cracks at each position. The Teager energy superharmonic dynamics mode at each scale corresponds to the data source, and the data from S1 to S M According to the fusion rule, the result of the fusion of the first two scales is expressed as: Where K1 represents the conflict weight between evidence 1 and evidence 2: According to this rule, the remaining scales are fused in order, and finally we get: In the formula, Final income That is the damage factor.
Citation Information
Patent Citations
Rotor crack fault detection method based on interval control strategy
CN110134110A
Dynamic modeling and analyzing method for cracked rotor system
CN110569516A
Structural fatigue crack monitoring system and method
US20150330950A1
Cited By
Rotating shaft crack positioning method based on time-frequency spatial domain feature fusion under unstable working condition
CN122409839A
