A gear box fault diagnosis method and system based on scatter recursive analysis
By combining scatter recursive analysis and SVM, the problem of fault diagnosis of nonlinear vibration signals in wind turbine gearboxes is solved, achieving high-precision and highly adaptable fault identification, which is applicable to fault diagnosis of wind turbine gearboxes.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- NINGBO LIDOU INTELLIGENT TECH CO LTD
- Filing Date
- 2022-10-25
- Publication Date
- 2026-06-26
AI Technical Summary
Existing technologies struggle to effectively extract fault information from the nonlinear and non-stationary vibration signals of wind turbine gearboxes, resulting in insufficient accuracy and adaptability in fault diagnosis.
We employ scatter recursive analysis combined with support vector machine (SVM) to extract features and diagnose gearbox vibration signals. We obtain multidimensional nonlinear dynamic features through scatter recursive analysis and use SVM for efficient fault diagnosis.
It achieves high-precision and highly adaptable fault diagnosis of wind turbine gearboxes, improving the accuracy and adaptability of fault identification, and is suitable for small sample problems.
Smart Images
Figure CN115791152B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of wind turbine generator fault diagnosis technology, and more specifically, relates to a gearbox fault diagnosis method and system based on scatter recursive analysis, which specifically integrates scatter recursive analysis and the machine learning algorithm SVM. Background Technology
[0002] Wind power generation has been recognized as one of the most important means of alleviating urgent electricity demand in a low-carbon way, and has achieved widespread adoption and significant development. However, with the large-scale construction, operation, and production of wind turbine units, a series of new problems have emerged. Unexpected gearbox failures will lead to wind turbine unit shutdowns, significantly increasing maintenance costs. To address this issue, effective fault diagnosis of wind turbine gearboxes is urgently needed.
[0003] In recent years, extracting fault information from vibration signals to reflect the performance degradation of wind turbine gearboxes and combining this with the powerful learning capabilities of machine learning for fault diagnosis has become a research focus. However, extracting appropriate fault information remains a challenging problem. Most wind turbine fault diagnosis methods generally employ time-domain and frequency-domain methods to obtain fault information. Furthermore, some signal processing methods have been developed, such as wavelet decomposition, Hilbert transform, and variational mode decomposition. However, due to the influence of various nonlinear factors (such as friction, clearance, and stiffness), the actual vibration signals of wind turbine gearboxes typically exhibit nonlinear and non-stationary characteristics. The aforementioned methods have inherent limitations in processing these vibration signals and struggle to reveal the inherent nonlinear dynamic characteristics of the system. Moreover, these methods rely on the experience and knowledge of technicians, meaning that ordinary users find it difficult to make accurate decisions. Therefore, developing a novel and effective fault extraction and diagnosis technique based on nonlinear dynamics is a pressing technical problem that needs to be solved. Summary of the Invention
[0004] To address the aforementioned deficiencies or improvement needs of existing technologies, this invention proposes a gearbox fault diagnosis method and system based on distributed recursive analysis. By mining the underlying dynamic characteristics of the system from the perspective of nonlinear dynamics and combining it with machine learning, high-efficiency and high-precision fault diagnosis can be achieved, providing valuable guidance for wind farm operation and maintenance.
[0005] To achieve the above objectives, according to one aspect of the present invention, a gearbox fault diagnosis method based on scatter recursive analysis is provided, comprising:
[0006] Collect vibration signals from the gearbox of the wind turbine unit to be diagnosed;
[0007] Dispersive recursive analysis was used to analyze the vibration signal of the gearbox of the wind turbine under diagnosis and obtain multidimensional nonlinear dynamic characteristics.
[0008] By inputting the multidimensional nonlinear dynamic features into a trained intelligent fault diagnosis model based on support vector machine (SVM), the fault diagnosis results of the gearbox are obtained.
[0009] In some optional implementations, the analysis of the gearbox vibration signal of the wind turbine under diagnosis using scatter recursive analysis to obtain multidimensional nonlinear dynamic characteristics includes:
[0010] Collect vibration signal x of the gearbox of the wind turbine to be diagnosed with a length of N, and sort the amplitude of the data points in the vibration signal x of the gearbox of the wind turbine to be diagnosed in ascending order to obtain the amplitude range.
[0011] The entire amplitude range is divided into c segments to obtain the boundary point of each segment. Based on the boundary point, the vibration signal x of the gearbox of the wind turbine to be diagnosed is converted into a dispersed integer symbol representation to obtain a symbol sequence.
[0012] The symbol sequence is reconstructed from the phase space and mapped to a high-dimensional space. Based on whether the patterns formed by the components of the symbol elements in each embedding vector in the high-dimensional space are the same, a scatter recursive matrix is constructed.
[0013] The recursion rate is obtained by calculating the percentage of recursive points in the scatter recursion matrix. The degree of certainty is used to measure the ratio of recursive points forming diagonal lines in the scatter recursion matrix to all recursive points. The degree of layering is used to represent the ratio of recursive points forming vertical lines in the scatter recursion matrix to all recursive points.
[0014] Faults are characterized by recursion rate, determinism, and layering as multidimensional nonlinear dynamic features.
[0015] In some alternative implementations, the symbol sequence is S = {s} i ,i=1,2,...,N}, p k k = 1, 2, ..., c-1 represents the dividing point. y kL x represents the amplitude of the data point ranked kL in the amplitude domain. i denoted as the i-th data point in the vibration signal x of the gearbox of the wind turbine to be diagnosed, and c represents the number of segments into which the amplitude domain is divided.
[0016] In some alternative implementations, DR = {DR i,j The scattered recursion matrix is obtained. Among them, the point with element 1 is called the recursion point. and For each embedding vector in a high-dimensional space, ... The corresponding pattern is formed according to the components of the symbolic elements. like These two vectors have the same pattern, where m is the embedding dimension and d is the time delay.
[0017] In some alternative implementations, by The recursion rate RR is obtained, n = N - (m-1)d.
[0018] In some alternative implementations, the determination measure of the ratio of the recursive points forming the diagonal line in the scatter recursion matrix to all recursive points includes:
[0019] Starting from point (i, j) in DR, through and DR i+l,j+l =DR i-1,j-1 =0 forms a diagonal line of length l parallel to the main diagonal;
[0020] Depend on The degree of certainty, DET, is obtained to measure the ratio of the recursive point that forms the diagonal to all recursive points, where l min and l max For the minimum and maximum lengths, N d (l) indicates the number of diagonal lines with length l.
[0021] In some optional implementations, the use of layer degree to represent the ratio of recursive points forming vertical lines to all recursive points in the scatter recursion matrix includes:
[0022] Starting from point (i, j) in DR, through and DR i,j-1 =DR i,j+l =0 forms a vertical line of length l;
[0023] Depend on Determine the layer degree LAM to represent the ratio of the recursive point forming the vertical line to all recursive points, where N v (l) indicates the number of vertical lines with length l.
[0024] In some optional implementations, the training method for the intelligent fault diagnosis model based on support vector machine (SVM) is as follows:
[0025] Vibration signals of wind turbine gearboxes under different health conditions were collected;
[0026] Dispersive recursive analysis was used to analyze the vibration signals of wind turbine gearboxes under different health conditions in order to obtain multidimensional nonlinear dynamic characteristics from the complex gearbox vibration signals.
[0027] The multidimensional nonlinear dynamic features are divided into training and testing sets according to a certain ratio;
[0028] The intelligent fault diagnosis model based on support vector machine (SVM) is trained using the input training set.
[0029] The test set data is input into the trained SVM, and the output of the SVM is compared with the actual label category of the gearbox to obtain the diagnostic accuracy, thus completing the gearbox fault diagnosis.
[0030] Among them, vibration signals of wind turbine gearboxes under different health conditions are collected, including vibration signals of common health conditions such as normal, broken teeth, base crack, and tooth damage.
[0031] According to another aspect of the present invention, a gearbox fault diagnosis system based on distributed recursive analysis is provided, comprising:
[0032] The data acquisition module is used to collect vibration signals from the gearbox of the wind turbine to be diagnosed.
[0033] The feature extraction module is used to analyze the vibration signal of the gearbox of the wind turbine under diagnosis using scatter recursive analysis to obtain multidimensional nonlinear dynamic characteristics.
[0034] The diagnostic module is used to input multidimensional nonlinear dynamic features into a trained intelligent fault diagnosis model based on support vector machine (SVM) to obtain gearbox fault diagnosis results.
[0035] According to another aspect of the present invention, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the methods described above.
[0036] In summary, compared with the prior art, the above-described technical solutions conceived by this invention can achieve the following beneficial effects:
[0037] This invention specifically integrates scatter recursive analysis and SVM. It innovatively proposes scatter recursive analysis, which, based on nonlinear dynamics, reveals the complex nonlinear dynamic evolution characteristics of signals. This analysis can effectively characterize the health status of wind turbine gearboxes, capture the essential differences between different states, and provide high-quality information for downstream fault diagnosis tasks. Considering the limited number of actual fault samples, SVM is introduced to address this issue. Combined with the multidimensional features of scatter recursive analysis, it enables the effective identification of various health states. The adaptability and diagnostic accuracy of the wind turbine gearbox fault diagnosis method provided by this invention are significantly improved. Therefore, the wind turbine fault diagnosis method integrating scatter recursive analysis and SVM can achieve accurate and reliable fault diagnosis. Attached Figure Description
[0038] Figure 1 This is a flowchart illustrating the implementation of a gearbox fault diagnosis method based on scatter recursive analysis provided in an embodiment of the present invention.
[0039] Figure 2 This is a flowchart illustrating the implementation of scatter recursive analysis provided in this embodiment of the invention. Detailed Implementation
[0040] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.
[0041] The purpose of this invention is to provide a new method for fault diagnosis of wind turbine gearboxes, which offers higher detection accuracy and adaptability. The overall fault diagnosis process can be divided into two stages: the first stage uses scatter recursive analysis to extract key fault dynamic features; the second stage uses SVM to achieve fault diagnosis based on these dynamic features.
[0042] Example 1
[0043] like Figure 1 As shown in the figure, a flowchart of a gearbox fault diagnosis method based on scatter recursive analysis provided by an embodiment of the present invention includes the following steps:
[0044] Step 1: Collect vibration signals of the wind turbine gearbox under different health conditions;
[0045] In this embodiment of the invention, vibration signals were collected under five healthy states: normal, missing tooth, fractured abutment, cracked tooth, and broken tooth. These were labeled {L1, L2, L3, L4, L5}. Each sample contained 3600 data points. The specific data settings are shown in Table 1.
[0046] Table 1 Data Settings
[0047]
[0048] Step 2: Analyze all samples using scatter recursive analysis to construct the feature space;
[0049] In this embodiment of the invention, the parameter settings for the scatter recursive analysis are shown in Table 2.
[0050] Table 2 Parameter settings for scatter recursion analysis
[0051]
[0052] In this embodiment of the invention, scatter recursive analysis is used to analyze and process the vibration signal to obtain multidimensional nonlinear dynamic characteristics, such as... Figure 2As shown, step 2 can be implemented in the following way:
[0053] Given a vibration signal sample x = {x} of length N. i Given the data points x in the sample, i = 1, 2, ..., N, ... i The amplitude range y is obtained by arranging the amplitude values in ascending order, that is:
[0054]
[0055] Among them, y i i = 1, 2, ..., N represents the magnitude of the data points.
[0056] Divide the entire amplitude domain y into c equal segments, obtain the boundary points of each segment, and form a vector P:
[0057]
[0058] in, p represents the floor operation. k This is the dividing point.
[0059] Based on the dividing point, the vibration signal x is converted into a dispersed integer symbol representation, that is:
[0060]
[0061] In this way, the vibration signal x is converted into a new symbol sequence S = {s} i Let ,i=1,2,...,N}, where each sample point s i Represent it using an integer from 1 to c.
[0062] Reconstructing the symbol sequence S from phase space to a higher-dimensional space:
[0063]
[0064] Where m is the embedding dimension and d is the time delay. Each embedding vector The corresponding pattern is formed according to the components of the symbolic elements. like These two vectors have the same pattern.
[0065] Construct the scatter recursion matrix DR = {DR i,j}:
[0066]
[0067] Among them, the point with an element of 1 is called the recursion point.
[0068] Calculate the percentage of recursive points in matrix DR, i.e., the recursion rate RR:
[0069]
[0070] Where n = N - (m - 1)d.
[0071] Starting from point (i, j) in DR, through and DR i+l,j+l =DR i-1,j-1 =0 forms a diagonal line of length l parallel to the main diagonal. Determinism DET measures the ratio of the recursive point forming the diagonal line to all recursive points; the formula is:
[0072]
[0073] Among them, l min and l max The minimum and maximum lengths are given, where l is an integer variable, and the minimum value l is taken. min up to the maximum value l max All integers between N d (l) indicates the number of diagonal lines of length l; l min The usual value is 2.
[0074] Starting from point (i, j) in DR, through and DR i,j-1 =DR i,j+l =0 forms a vertical line of length l. The layering degree (LAM) represents the ratio of the recursive point forming the vertical line to all recursive points, and is calculated using the formula:
[0075]
[0076] Where, N v (l) represents the number of vertical lines of length l;
[0077] Faults are characterized by the values of RR, DET, and LAM.
[0078] Step 3: Divide the multidimensional features into training and test sets according to the specified proportions;
[0079] In this embodiment of the invention, 60% of the samples are randomly selected as the training set for the Support Vector Machine (SVM), and the remaining 40% are used as the test set to test the performance of the trained SVM model.
[0080] Step 4: Input the training set to train the SVM-based intelligent fault diagnosis model;
[0081] In this embodiment of the invention, the kernel function of the SVM is selected as the radial basis kernel function, and the parameters of the SVM are selected using a sine-cosine algorithm.
[0082] The radial basis functions are: x and x′ are two feature vectors of the input training sample.
[0083] In this embodiment of the invention, other algorithms may also be used for the parameter selection method of SVM, and this embodiment of the invention does not limit the uniqueness of the algorithm.
[0084] Step 5: Use the constructed SVM model to test on the test set, calculate the diagnostic accuracy, and give the fault diagnosis results.
[0085] In this embodiment of the invention, the performance of the diagnostic accuracy evaluation method can be employed:
[0086]
[0087] Wherein, TP refers to the number of positive classes predicted as positive, TN refers to the number of positive classes predicted as negative, FP refers to the number of negative classes predicted as positive, and FN refers to the number of negative classes predicted as negative.
[0088] The results of the fault diagnosis are shown in Table 3. Compared with the fault diagnosis results of the mainstream feature extraction algorithms permutation entropy and scattering entropy, the fault diagnosis method provided by this invention is advanced.
[0089] Table 3 Comparison of Gearbox Fault Diagnosis Results
[0090]
[0091] Example 2
[0092] In another embodiment of the present invention, a gearbox fault diagnosis system based on scatter recursive analysis is also provided, comprising:
[0093] The data acquisition module is used to collect vibration signals from the gearbox of the wind turbine to be diagnosed.
[0094] The feature extraction module is used to analyze the vibration signal of the gearbox of the wind turbine under diagnosis using scatter recursive analysis to obtain multidimensional nonlinear dynamic characteristics.
[0095] The diagnostic module is used to input multidimensional nonlinear dynamic features into a trained intelligent fault diagnosis model based on support vector machine (SVM) to obtain gearbox fault diagnosis results.
[0096] The specific implementation methods of each module can be referred to the description of the above method embodiments, and will not be repeated in this embodiment.
[0097] It should be noted that, depending on the implementation needs, the various steps / components described in this application can be broken down into more steps / components, or two or more steps / components or parts of the operation of steps / components can be combined into new steps / components to achieve the purpose of this invention.
[0098] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A gearbox fault diagnosis method based on scatter recursive analysis, characterized in that, include: Collect vibration signals from the gearbox of the wind turbine unit to be diagnosed; Dispersive recursive analysis was used to analyze the vibration signal of the gearbox of the wind turbine under diagnosis and obtain multidimensional nonlinear dynamic characteristics. By inputting the multidimensional nonlinear dynamic features into the trained intelligent fault diagnosis model based on support vector machine (SVM), the gearbox fault diagnosis results are obtained. The method utilizes scatter recursive analysis to analyze the vibration signal of the gearbox of the wind turbine under diagnosis, obtaining multidimensional nonlinear dynamic characteristics, including: The collection length is N Vibration signal of the gearbox of the wind turbine to be diagnosed x The vibration signal of the gearbox of the wind turbine to be diagnosed x The amplitude range is obtained by arranging the amplitude values of the data points in ascending order. The entire amplitude range is divided into c segments, and the boundary point of each segment is obtained. Based on the boundary point, the vibration signal of the gearbox of the wind turbine to be diagnosed is... x The symbol sequence is obtained by converting the integers into scattered symbolic representations, where the symbol sequence is... , , Indicates the dividing point. , Indicates the number ranked in the amplitude range The magnitude of the data points in bits. This indicates the vibration signal of the gearbox of the wind turbine unit to be diagnosed. x The first in i Data points, This indicates the number of segments into which the amplitude range is divided; The symbol sequence is reconstructed from the phase space and mapped to a high-dimensional space. Based on whether the patterns formed by the components of the symbol elements in each embedding vector in the high-dimensional space are the same, a scatter recursive matrix is constructed. The recursion rate is obtained by calculating the percentage of recursive points in the scatter recursion matrix. The degree of certainty is used to measure the ratio of recursive points forming diagonal lines in the scatter recursion matrix to all recursive points. The degree of layering is used to represent the ratio of recursive points forming vertical lines in the scatter recursion matrix to all recursive points. Faults are characterized by recursion rate, determinism, and layering as multidimensional nonlinear dynamic features.
2. The method according to claim 1, characterized in that, Depend on The scattered recursion matrix is obtained. Among them, the point with element 1 is called the recursion point. and For each embedding vector in a high-dimensional space, ... The corresponding pattern is formed according to the components of the symbolic elements. ,like These two vectors have the same pattern. m For the embedding dimension, d This is a time delay.
3. The method according to claim 2, characterized in that, Depend on The recursion rate RR is obtained. n = N -( m -1) d .
4. The method according to claim 3, characterized in that, The method of using determination to measure the ratio of the recursive points forming the diagonal line in the scatter recursion matrix to all recursive points includes: From the point in DR ( i , j (Start, through) and Form a line of length l A diagonal line parallel to the main diagonal; Depend on The degree of certainty, DET, is obtained to measure the ratio of the recursive point that forms the diagonal to all recursive points, where, l min and l max For minimum and maximum length, Indicates the length of the slash is l The number of.
5. The method according to claim 4, characterized in that, The method of using layered degree to represent the ratio of recursive points forming vertical lines to all recursive points in the scatter recursion matrix includes: From the point in DR ( i , j (Start, through) and Form a line of length l Vertical lines; Depend on Determine the layer degree LAM to represent the ratio of the recursive point forming the vertical line to all recursive points, where, Indicates the length of the vertical line is l The number of.
6. The method according to any one of claims 1 to 5, characterized in that, The training method for the intelligent fault diagnosis model based on Support Vector Machine (SVM) is as follows: Vibration signals of wind turbine gearboxes under different health conditions were collected; Dispersive recursive analysis was used to analyze the vibration signals of wind turbine gearboxes under different health conditions in order to obtain multidimensional nonlinear dynamic characteristics from the complex gearbox vibration signals. The multidimensional nonlinear dynamic features are divided into training and testing sets according to a certain ratio; The intelligent fault diagnosis model based on support vector machine (SVM) is trained using the input training set. The test set data is input into the trained SVM, and the output of the SVM is compared with the actual label category of the gearbox to obtain the diagnostic accuracy, thus completing the gearbox fault diagnosis.
7. A gearbox fault diagnosis system based on scatter recursive analysis, characterized in that, include: The data acquisition module is used to collect vibration signals from the gearbox of the wind turbine to be diagnosed. The feature extraction module is used to analyze the vibration signal of the gearbox of the wind turbine under diagnosis using scatter recursive analysis to obtain multidimensional nonlinear dynamic characteristics. Specifically used to perform the following steps: The collection length is N Vibration signal of the gearbox of the wind turbine to be diagnosed x The vibration signal of the gearbox of the wind turbine to be diagnosed x The amplitude range is obtained by arranging the amplitude values of the data points in ascending order. The entire amplitude range is divided into c segments, and the boundary point of each segment is obtained. Based on the boundary point, the vibration signal of the gearbox of the wind turbine to be diagnosed is... x The symbol sequence is obtained by converting the integers into scattered symbolic representations, where the symbol sequence is... , , Indicates the dividing point. , Indicates the number ranked in the amplitude range The magnitude of the data points in bits. This indicates the vibration signal of the gearbox of the wind turbine unit to be diagnosed. x The first in i Data points, This indicates the number of segments into which the amplitude range is divided; The symbol sequence is reconstructed from the phase space and mapped to a high-dimensional space. Based on whether the patterns formed by the components of the symbol elements in each embedding vector in the high-dimensional space are the same, a scatter recursive matrix is constructed. The recursion rate is obtained by calculating the percentage of recursive points in the scatter recursion matrix. The degree of certainty is used to measure the ratio of recursive points forming diagonal lines in the scatter recursion matrix to all recursive points. The degree of layering is used to represent the ratio of recursive points forming vertical lines in the scatter recursion matrix to all recursive points. Faults are characterized by recursion rate, determinism, and layering as multidimensional nonlinear dynamic features; The diagnostic module is used to input multidimensional nonlinear dynamic features into a trained intelligent fault diagnosis model based on support vector machine (SVM) to obtain gearbox fault diagnosis results.
8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.