Singular value decomposition sub-signal selection method for fault diagnosis of rotating machinery

Abstract

The invention discloses a singular value decomposition sub-signal selection method for rotating machine fault diagnosis, and belongs to the technical field of rotating machine state monitoring and fault diagnosis. The method comprises the following steps: receiving an original vibration signal of the rotating machine, and adaptively determining the optimal decomposition times of singular value decomposition through exhaustion search by taking the number of periodic pulse quantized values as an optimization target; constructing a Hankel matrix for the original vibration signal and performing singular value decomposition to obtain a group of sub-signals; the square envelope of each sub-signal is calculated, the periodic pulse performance of each sub-signal is quantized, a clustering guide line is drawn, and key sub-signals are screened out; continuously adjacent key sub-signals are classified into one class for reconstruction mixing, fault components are formed for envelope spectrum analysis, and rotating machine fault diagnosis is achieved. According to the invention, periodic impact components related to faults in the signals can be effectively highlighted, and the accuracy and reliability of composite fault diagnosis of the rotating machinery under complex working conditions are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of rotating machinery condition monitoring and fault diagnosis technology, specifically to a method for selecting singular value decomposition sub-signals for rotating machinery fault diagnosis. Background Technology

[0002] Rotating machinery, such as gearboxes and bearings, are core components in industrial equipment, and their operating status directly affects the safety and efficiency of the entire production system. Analyzing the vibration signals generated during equipment operation allows for monitoring its health and fault diagnosis. However, the vibration signals actually collected are often very complex, containing not only weak fault characteristic information but also strong background noise and interference signals between different components. Especially when equipment experiences compound faults (i.e., multiple faults coexist), the characteristic signals of different faults couple with each other, making fault feature extraction and separation particularly difficult.

[0003] Singular value decomposition (SVD) SingularValueDecomposition , SVD As a powerful matrix factorization tool, it is widely used for signal denoising and feature extraction. Its basic idea is to construct the original signal into a Hankel matrix and then perform... SVD Decomposition involves breaking down a signal into a series of sub-signal components. Theoretically, fault-related characteristic information is concentrated in a few specific sub-signals. However, existing methods based on... SVD The fault diagnosis method has two main problems: First, it cannot adaptively determine the optimal number of decomposition layers (i.e., the number of sub-signals). m Insufficient decomposition layers lead to ineffective separation of fault components from noise; excessive decomposition layers cause "over-decomposition," scattering the originally complete fault features into multiple sub-signals, thus increasing the difficulty of identification. Secondly, there is a lack of effective criteria to filter out the key components that truly contain fault information from the numerous decomposed sub-signals, and to further cluster these key components according to their respective fault types. Existing methods typically simply mix all sub-signals or a few sub-signals selected empirically. This fails to achieve targeted separation of different fault characteristic signals when dealing with complex faults, resulting in poor diagnostic performance.

[0004] Therefore, there is an urgent need for a method that can adaptively determine... SVD A method for decomposing parameters and intelligently filtering and clustering sub-signals can be used to accurately diagnose complex faults in rotating machinery. Summary of the Invention

[0005] The purpose of this invention is to provide a singular value decomposition sub-signal selection method for fault diagnosis of rotating machinery, so as to solve the problems of existing technologies being unable to adaptively determine the optimal decomposition parameters and lacking effective sub-signal screening and clustering criteria.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for selecting singular value decomposition sub-signals for fault diagnosis of rotating machinery includes the following steps: S 1. Receive the raw vibration signal of rotating machinery, and with the optimization objective of maximizing the number of periodic pulse quantization values ​​greater than a preset threshold, adaptively determine the optimal number of singular value decompositions through exhaustive search. m ; S 2: Based on the optimal number of decompositions m The Hankel matrix is ​​constructed from the original vibration signal and singular value decomposition is performed to obtain... m Individual signals; S 3: Calculate the squared envelope of each sub-signal and use... CESMp 2 q 1. Statistical quantification of the periodicity and impulsivity of each sub-signal, plotting a line graph of the quantization value as a function of the sub-signal number as a clustering guideline, and screening out key sub-signals with significant periodicity and impulsivity based on the clustering guideline. S 4: Group the selected consecutive adjacent key sub-signals into one category, and reconstruct and mix the sub-signals within the same category to form fault components corresponding to specific faults. By performing envelope spectrum analysis on each fault component, the rotating machinery fault diagnosis is realized and the diagnosis results are output.

[0007] Furthermore, S The optimization objective function of the exhaustive search described in section 1 is: ; in, For each decomposition, the first j The square envelope of the decomposed signal; As an indicator function, when the quantized value Greater than or equal to the preset threshold T When the value is 1, its value is 1; otherwise, it is 0.

[0008] Furthermore, the aforementioned m The search range is [10, 40], and the preset threshold... T It is 1.6.

[0009] Furthermore, the aforementioned CESMp 2 q The calculation process of a statistic includes: Calculate the squared envelope of each sub-signal SE ; square envelope SE Cut off without overlapS part; Calculate the signal for each segment pq -mean, where p =2, q =1; right S indivual pq - Perform a power-mean operation on the mean, where the power-mean parameter... α The value is taken as -10, and the sub-signal is obtained. CESMp 2 q 1 value; The pq -The formula for calculating the mean is: In the formula, c This represents a truncated square envelope signal. express c The first in n Data points, N This indicates the length of the signal segment.

[0010] Furthermore, S The key sub-signals described in section 3 are selected using a threshold method, specifically: consecutive adjacent clusters with quantization values ​​greater than or equal to a preset threshold are selected. T The sub-signal is determined to be the key sub-signal.

[0011] Furthermore, S The key sub-signals described in section 3 are selected using the local maximum method, which involves traversing the clustering guide lines and identifying sub-signals whose quantization values ​​are simultaneously greater than the quantization values ​​of their left and right adjacent sub-signals as key sub-signals.

[0012] Furthermore, S The sub-signals within the same class described in section 4 are reconstructed and mixed by performing an anti-diagonal averaging of the Hankel matrix corresponding to each sub-signal.

[0013] Furthermore, S If there are multiple sub-signal clusters that meet the conditions on the clustering guideline, then each cluster is reconstructed and mixed to form an independent fault component, thereby realizing the composite fault diagnosis of rotating machinery.

[0014] Another object of the present invention is to provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned method for selecting singular value decomposition sub-signals for fault diagnosis of rotating machinery.

[0015] Another object of the present invention is to provide an electronic device, including a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to cause the electronic device to perform the singular value decomposition sub-signal selection method for fault diagnosis of rotating machinery.

[0016] The present invention provides a singular value decomposition sub-signal selection method for fault diagnosis of rotating machinery, which has the following significant advantages compared with the prior art: Highly adaptive: This invention utilizes quantified fault pulse characteristics... CESMp 2 q Using a statistic as the objective function for an exhaustive search, it is possible to adaptively determine... SVD Number of optimal sub-signals in decomposition m This overcomes the limitations of traditional methods that rely on experience, and achieves optimal decomposition results. Precise screening and clustering: By constructing "clustering guidelines" and using thresholding or local maximum methods, key fault-related components can be objectively and accurately screened from a large number of sub-signals. More importantly, this invention innovatively proposes a clustering criterion of "grouping consecutive adjacent sub-signals into one class," which aligns with... SVD The characteristic that similar fault features are continuously distributed on the component index after decomposition can effectively distinguish and aggregate feature components belonging to different faults, thus perfectly solving the problem of separating composite fault signals. High diagnostic accuracy: This method can clearly separate the independent components corresponding to different faults from complex composite fault signals. On the envelope spectrum of each component, the corresponding fault characteristic frequencies and their harmonics are clearly visible with few interference components, which significantly improves the accuracy and reliability of composite fault diagnosis. Attached Figure Description

[0017] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a schematic diagram of clustering guide lines and decomposition signal singular values ​​of the present invention, wherein (a) is a schematic diagram of clustering guide lines and (b) is a schematic diagram of decomposition signal singular values. Figure 3 This is a schematic diagram of the time domain, frequency domain, and square envelope spectrum of the signal being analyzed according to the present invention, wherein (a) is the time domain diagram, (b) is the frequency domain diagram, and (c) is the square envelope spectrum. Figure 4The diagram shows the first component (bearing fault) and the second component (gear fault) extracted by the method described in this invention. In the diagram, (a) is the time-domain diagram of the bearing fault component, (b) is the Fourier spectrum of the bearing fault component, (c) is the square envelope spectrum of the bearing fault component, (d) is the time-domain diagram of the gear fault component, (e) is the Fourier spectrum of the gear fault component, and (f) is the square envelope spectrum of the gear fault component. Detailed Implementation

[0018] 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.

[0019] This embodiment provides a method for selecting singular value decomposition sub-signals for fault diagnosis of rotating machinery, such as... Figure 1 As shown, it includes: receiving the raw vibration signal of rotating machinery, and using the goal of maximizing the number of periodic pulse quantization values ​​greater than a preset threshold, adaptively determining the optimal number of singular value decompositions through exhaustive search. m According to the optimal number of decompositions m The Hankel matrix is ​​constructed from the original vibration signal and singular value decomposition is performed to obtain... m Each sub-signal; calculate the squared envelope of each sub-signal, and use... CESMp 2 q 1. The periodicity and impulsivity of each sub-signal are quantified using statistical methods. A line graph showing the change of quantized values ​​with the sub-signal index is plotted as a clustering guideline. Based on this guideline, key sub-signals with significant periodicity and impulsivity are selected. These selected, consecutively adjacent key sub-signals are grouped into one category, and sub-signals within the same category are reconstructed and mixed to form fault components corresponding to specific faults. Envelope spectrum analysis is performed on each fault component to achieve rotating machinery fault diagnosis and output diagnostic results. This is explained in detail below.

[0020] 1. Receive the raw vibration signal of rotating machinery, and with the optimization objective of maximizing the number of periodic pulse quantization values ​​greater than a preset threshold, adaptively determine the optimal number of singular value decompositions through exhaustive search. m ; First, the raw vibration signal sequence generated by the rotating machinery during operation is acquired using vibration monitoring equipment such as accelerometers. This raw signal typically contains strong background noise, which drowns out the periodic impact components associated with the fault. To effectively extract these weak features, this invention utilizes singular value decomposition (SVD). SVD The optimal number of decompositionsm Exhaustive search, as an adaptive signal decomposition method, yields a large number of sub-signals. m This is the only parameter that needs optimization, and its value directly affects the quality of subsequent sub-signals. Since the computational cost of a one-dimensional discrete exhaustive search is not high, this invention employs an exhaustive search strategy, traversing all possible sub-signals within a reasonable range (e.g., 10 to 40). m Value. For each candidate m Value, execute a complete operation once SVD Decompose to obtain m Each sub-signal. Then, a target function value is calculated, which is defined as: the periodic impulse quantization value of each sub-signal ( CESMp 2 q 1) Exceeding the preset threshold T The sum of the quantized values ​​(typically 1.6) is what we get. The largest m This is the optimal number of decompositions. The principle is that the optimal decomposition should maximize the prominence of the fault-related periodic impulse components in the signal. CESMp 2 q 1 (A type based on) p =2 and q Cyclic energy separation method with parameter =1 Cyclic Energy Separation Method ( ), is a statistical index used to quantify the periodicity and impulsivity of a signal. It works by segmenting the squared envelope of the signal and calculating the periodicity of each segment. pq - Mean (where the parameter) p =2, q =1), then perform a power mean operation on the obtained mean (where the power mean parameter is 1). α =-10) is obtained. The optimization objective function of the above exhaustive search is: and in, For each decomposition, the first j The square envelope of the decomposed signal; As an indicator function, when the quantized value Greater than or equal to the preset threshold T When the time condition is met, its value is 1; otherwise, it is 0. m The exhaustive search range is [10, 40], and its core principle is that the optimal number of decompositions should maximize the periodicity and impulsivity of the typical fault components. After exhaustive search, the optimal number of decompositions can be determined. SVD Optimal number of decompositions m .

[0021] In the above objective function CESMIt is a statistical indicator with excellent characteristics suitable for transient quantification of repetitive faults. Therefore, it has been selected to quantify the sub-signals obtained by singular value decomposition in order to realize the fault diagnosis of rotating machinery. CESM The definition and characteristics are as follows: First, for a discrete signal Its square envelope signal It can be done Calculation, where The imaginary unit, This is the Hilbert transform.

[0022] Then Cut off without overlap S Each length is The segment, marked as , , …, ,in A sparse measure It is a statistical index that can quantify the sparsity or impulsivity of signal components. Random impulse noise typically contains only single-pulse transients, while repetitive fault transients have approximately periodic pulse transients. Therefore, by combining the original sparsity measure with power values, a method using... CESM Distinguishing between repetitive fault transients and random impulse noise: in, It is a parameter of the power mean. When When the power mean is equal to the geometric mean, then... The value is set to .

[0023] Then calculate the signal for each segment. pq -Mean, using parameters p =2 and q =1 pq - The mean, which is the ratio of the squared root mean of the signal to the arithmetic mean, is defined as follows: In the formula, c This represents a truncated square envelope signal. express c The first in n Data points, N This indicates the length of the signal segment.

[0024] Finally, S indivual pq - Perform a power-mean operation on the mean, where the power-mean parameter... αThe value is taken as -10, and the sub-signal is obtained. CESMp 2 q 1. When quantizing the squared envelope signal, its value ranges from 1 to 1. The stronger the sparsity and impulsiveness of the quantized signal, the larger its quantization value. When quantizing the squared envelope of white Gaussian noise, pq -mean ( p =2 and q The theoretical quantization value of =1) is .based on pq -mean ( p =2 and q =1) CESM for CESMp 2 q 1, that is, able to calculate CESMp 2 q The value of 1 is given by the following formula: II. Based on the optimal number of decompositions m The Hankel matrix is ​​constructed from the original vibration signal and singular value decomposition is performed to obtain... m Sub-signals.

[0025] The optimal solution determined based on the above process m Value, for the original vibration signal x Constructing the Hankel matrix A Then, singular value decomposition (SVD) is performed on the matrix. SVD is a matrix factorization method whose core principle is to decompose any matrix into the product of three matrices: in, and The sum is an orthogonal matrix. It is a diagonal matrix, and the diagonal elements are singular values ​​(arranged in descending order).

[0026] After constructing the Hankel matrix from the original signal, singular value decomposition is performed. and All are real unitary matrices and satisfy , The matrix is and Hankel matrix The order of the singular values ​​satisfies ,in ( ) is a matrix The rank of the matrix. and Also known as a matrix The left and right singular matrices. Therefore, the Hankel matrix. It can be expressed by the following formula: Decomposed matrix Indicates the first i A decomposed signal, with m The reason for the decomposition of the matrix is ​​that the constructed Hankel matrix is ​​usually of full row rank. Theoretically, the decomposition of the matrix... It should be a Hankel matrix, where the elements on the inverse diagonal should be the same, but the elements on the antidiagonal should have some differences. Therefore, the formula for reconstructing the sub-signal is as follows: In summary, the original signal It can be decomposed into The sub-signals are respectively , , , and .

[0027] 3. Calculate the squared envelope of each sub-signal and use... CESMp 2 q 1. Statistical quantification of the periodicity and impulsivity of each sub-signal, plotting a line graph of the quantization value as a function of the sub-signal number as a clustering guideline, and screening out key sub-signals with significant periodicity and impulsivity based on the clustering guideline. This step is the core of achieving intelligent filtering and clustering of sub-signals. First, calculate each sub-signal. square envelope Then use CESMp 2 q 1. Statistic for each Quantify it. CESMp 2 q 1. It can effectively quantify the periodic pulse intensity of repetitive fault transients; the larger the value, the higher the probability that the sub-signal contains fault impulse characteristics. This involves applying the values ​​of all sub-signals... CESMp 2 q The values ​​are connected in order of sub-signal index and plotted as a broken line, which is the "clustering guide line", such as... Figure 2 As shown.

[0028] Based on clustering guidelines, there are two strategies for filtering key sub-signals: Thresholding method: Clustering guide lines with consecutive adjacent quantized values ​​greater than or equal to a preset threshold are selected. T The sub-signals are determined to be key sub-signals. Consecutive adjacent sub-signals with quantization values ​​greater than a certain quantization threshold are identified. T The decomposed signal is preserved (set in this embodiment). T =1.6), and further mixed into an extracted fault component. As shown in Figure 2 ( aAs shown in the figure, the quantized values ​​are plotted as a line graph, i.e. SVD Component clustering guideline. Singular values ​​of the decomposed signal are also present. Figure 2 ( b Draw it in ) . Figure 2 ( a )and Figure 2 ( b By comparing them, we can see that some decomposed signals have weaker energy but... CESMp 2 q 1. The quantization value is relatively large. By observing this guide line, two conditions can be clearly identified (i.e., the quantization value is continuously greater than the threshold). T Sub-signal clustering (=1.6).

[0029] Local maximum method: Traverse the clustering guide lines, identifying sub-signals whose quantization values ​​are simultaneously greater than those of their left and right adjacent sub-signals as key sub-signals. Traverse all quantization values; if the quantization value of a decomposed signal is simultaneously greater than the quantization values ​​of its left and right adjacent decomposed signals, this signal is identified as a target signal with significant fault characteristics and retained. Finally, all retained target signals are mixed to generate an extracted fault component, which is then plotted as a polyline using the quantization values. SVD Component clustering guide lines. This scheme uses local maximum identification to accurately locate the signal with the most prominent fault characteristics in the clustering guide lines, reducing interference from irrelevant signals.

[0030] Fourth, the selected consecutive adjacent key sub-signals are grouped into one category, and the sub-signals within the same category are reconstructed and mixed to form fault components corresponding to specific faults. By performing envelope spectrum analysis on each fault component, the fault diagnosis of rotating machinery is realized and the diagnosis results are output.

[0031] The identified clusters (thresholding method) or all retained target signals (local maximum method) are reconstructed and mixed. The reconstructing and mixing of sub-signals within the same cluster is achieved by averaging the anti-diagonal matrix corresponding to each sub-signal. The arithmetic mean of all elements with the same index on the anti-diagonal line in each matrix is ​​taken to reconstruct a one-dimensional sub-signal. Then, all reconstructed one-dimensional sub-signals belonging to the same cluster are directly added together to form a final fault component. Thus, the number of clusters corresponds to the number of reconstructed fault components. These components theoretically correspond to different fault sources in the original composite signal. Finally, envelope spectrum analysis is performed on each reconstructed fault component. The characteristic frequencies and harmonics of the corresponding fault can be clearly observed on the envelope spectrum, thereby achieving accurate fault type identification and diagnosis, and outputting diagnostic results.

[0032] Figure 2Due to the existence of the two clusters mentioned above, two independent fault components are generated. The first cluster (containing 2 sub-signals) is reconstructed and mixed to form "fault component 1"; the second cluster (containing 3 sub-signals) is reconstructed and mixed to form "fault component 2". Envelope spectrum analysis is performed on these two fault components respectively, and the results are as follows: Figure 4 As shown. Therefore, the method of the present invention successfully achieves feature separation and diagnosis of different fault sources (bearing faults and gear faults) in composite fault signals by identifying multiple clusters on the clustering guide line.

[0033] The implementation process of the above method is explained in detail below using a gearbox complex fault diagnosis experiment. This embodiment uses an experimental platform as follows: MCC 5- THU The dataset for a gearbox, which is a two-stage parallel gearbox, has a signal sampling frequency of 12.8 GHz. kHz The vibration fault signal with a central length of 1 second was analyzed. This signal was acquired from experiments involving broken teeth and bearing outer ring faults. The input shaft speed was 3000 rpm. RPM Based on the structural parameters of the gearbox, the characteristic frequencies of gear and bearing outer ring failures can be calculated to be 15.29. Hz and 54.57 Hz The time-domain, frequency-domain, and squared envelope spectra of the analyzed signal are as follows: Figure 3 As shown.

[0034] The optimal decomposition number can be obtained exhaustively based on formula (4). m =33, after finding the optimal decomposition number, the singular value decomposition method is used based on the optimal decomposition number. m Signal decomposition is performed. After signal decomposition, steps three and four are used to cluster and diagnose faults in the decomposed signal. The results are as follows: Figure 4 As shown. Figure 4 In the square envelope spectrum of the extracted signal components, the fault characteristic frequencies of bearings and gears are clearly visible, and there are few other interfering fault characteristic components, which effectively realizes the diagnosis of complex faults.

[0035] Therefore, experiments show that by introducing clustering guidelines and quantization screening criteria, this invention overcomes the limitations of existing singular value decomposition methods in signal component selection and clustering, and significantly improves the accuracy and reliability of complex fault diagnosis.

[0036] This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the singular value decomposition sub-signal selection method for fault diagnosis of rotating machinery.

[0037] This embodiment provides an electronic device, including a memory and a processor. The memory stores a computer program, and the processor runs the computer program to enable the electronic device to perform the singular value decomposition sub-signal selection method for fault diagnosis of rotating machinery.

[0038] In summary, by combining adaptive signal decomposition with periodic impulse index optimization, this invention can effectively highlight the periodic impulse components in the signal that are related to the fault, and significantly improve the accuracy and reliability of complex fault diagnosis for rotating machinery such as gearboxes under complex working conditions.

[0039] Embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of one or more computer-usable storage media (including, but not limited to, disk storage, etc.) containing computer-usable program code. CD - ROM It takes the form of a computer program product implemented on (such as optical memory, etc.).

[0040] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0041] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0042] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0043] Contents not described in detail in this specification are prior art known to those skilled in the art. It is hereby indicated that the above description is intended to help those skilled in the art understand this invention, but does not limit the scope of protection of this invention. Any equivalent substitutions, modifications, improvements, or simplifications of the above descriptions that do not depart from the essential content of this invention fall within the scope of protection of this invention.

Claims

1. A singular value decomposition sub-signal selection method for rotating machinery fault diagnosis, characterized in that, Comprising the steps of: S 1: receiving original vibration signal of rotating machinery, and adaptively determining optimal decomposition number of singular value decomposition through exhaustive search with the optimization target of maximizing the number of periodic pulse quantization values greater than a preset threshold m ; S 2: according to the optimal decomposition number m , a Hankel matrix is constructed from the original vibration signal and singular value decomposition is performed to obtain m sub-signals; S 3: Calculate the square envelope of each sub-signal, and adopt CESMp 2 q 1: Statistically quantify the periodic pulse nature of each sub-signal, draw a broken line graph of the quantized value changing with the sub-signal serial number as a clustering guide line, and based on the clustering guide line, screen out key sub-signals with significant periodic pulse nature; S 4: The selected continuous adjacent key sub-signals are classified into one category, and the sub-signals in the same category are reconstructed and mixed to form a fault component corresponding to a specific fault. Through envelope spectrum analysis of each fault component, the rotating machinery fault diagnosis is realized and the diagnosis result is output.

2. The singular value decomposition sub-signal selection method for fault diagnosis of rotating machinery according to claim 1, characterized in that, S The optimization objective function for the exhaustive search described in 1 is: ； in, For each decomposition, the first j The square envelope of the decomposed signal; As an indicator function, when the quantized value Greater than or equal to the preset threshold T When the value is 1, its value is 1; otherwise, it is 0.

3. The singular value decomposition sub-signal selection method for rotating machinery fault diagnosis according to claim 2, characterized in that: The m The search range is [10, 40], and the preset threshold... T It is 1.

6.

4. The singular value decomposition sub-signal selection method for rotating machinery fault diagnosis according to claim 3, characterized in that, The CESMp 2 q 1The calculation process of the statistical quantity comprises: computing a squared envelope of each sub-signal SE ; square envelope SE without overlap truncated to S segments; the mean value of each segment of the signal pq - the mean value, wherein p = 2, q = 1; On S one pq - the mean is power averaged with a power average parameter α of -10, resulting in a CESMp 2 q 1 value for the sub-signal; The pq - The mean calculation formula is: ； wherein c denotes a truncated square envelope signal, denotes c the n data point in N denotes the length of the signal segment.

5. The singular value decomposition sub-signal selection method for fault diagnosis of rotating machinery according to claim 1, characterized in that, S The screening of the key sub-signal in step 3 adopts a threshold method, specifically: the sub-signals in the clustering guide line that are continuous and adjacent and have quantized values greater than or equal to a preset threshold are determined as key sub-signals. T The screening of the key sub-signal in step 3 adopts a threshold method, specifically: the sub-signals in the clustering guide line that are continuous and adjacent and have quantized values greater than or equal to a preset threshold are determined as key sub-signals.

6. The singular value decomposition sub-signal selection method for rotating machinery fault diagnosis according to claim 1, characterized in that, S The screening of the key sub-signals in step 3 adopts a local maximum method, specifically: traversing the clustering guide line, and determining a sub-signal as a key sub-signal if the quantization value of the sub-signal is greater than the quantization values of the left and right adjacent sub-signals.

7. The singular value decomposition sub-signal selection method for fault diagnosis of rotating machinery according to claim 1, characterized in that: S 4 The same class of sub-signals are reconstructed and mixed by inverse diagonal averaging of the Hankel matrix corresponding to each sub-signal.

8. The singular value decomposition sub-signal selection method for fault diagnosis of rotating machinery according to claim 1, characterized in that: S 4If there are multiple sub-signal clusters meeting the conditions on the cluster guide line, each cluster is reconstructed separately to form an independent fault component after mixing, realizing compound fault diagnosis of rotating machinery.

9. A computer-readable storage medium, characterized in that: A computer program product having stored therein a computer program that, when executed by a processor, implements a singular value decomposition sub-signal selection method for rotating machinery fault diagnosis according to any one of claims 1-8.

10. An electronic device, comprising: An electronic device comprising a memory and a processor, the memory configured to store a computer program, and the processor configured to execute the computer program to cause the electronic device to perform a singular value decomposition sub-signal selection method for rotating machinery fault diagnosis according to any one of claims 1-8.

Images