CN110657985B - Gearbox fault diagnosis method and system based on singular value spectrum manifold analysis - Google Patents

Gearbox fault diagnosis method and system based on singular value spectrum manifold analysis Download PDF

Info

Publication number
CN110657985B
CN110657985B CN201910961446.3A CN201910961446A CN110657985B CN 110657985 B CN110657985 B CN 110657985B CN 201910961446 A CN201910961446 A CN 201910961446A CN 110657985 B CN110657985 B CN 110657985B
Authority
CN
China
Prior art keywords
singular value
fault
gearbox
manifold
fault diagnosis
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201910961446.3A
Other languages
Chinese (zh)
Other versions
CN110657985A (en
Inventor
苏祖强
罗子澳
于洪
萧红
谢海琼
谭峰
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Chongqing University of Post and Telecommunications
Original Assignee
Chongqing University of Post and Telecommunications
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Chongqing University of Post and Telecommunications filed Critical Chongqing University of Post and Telecommunications
Priority to CN201910961446.3A priority Critical patent/CN110657985B/en
Publication of CN110657985A publication Critical patent/CN110657985A/en
Application granted granted Critical
Publication of CN110657985B publication Critical patent/CN110657985B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • GPHYSICS
    • G01MEASURING; TESTING
    • G01MTESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
    • G01M13/00Testing of machine parts
    • G01M13/02Gearings; Transmission mechanisms
    • G01M13/021Gearings
    • GPHYSICS
    • G01MEASURING; TESTING
    • G01MTESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
    • G01M13/00Testing of machine parts
    • G01M13/02Gearings; Transmission mechanisms
    • G01M13/028Acoustic or vibration analysis
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/12Classification; Matching

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Acoustics & Sound (AREA)
  • Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)

Abstract

The invention relates to the technical fields of mechanical fault diagnosis, mode identification and the like, in particular to a method and a system for diagnosing faults of a gearbox based on singular value spectrum manifold analysis. The method comprises the steps of obtaining fault vibration signals of the gearbox and preprocessing the fault vibration signals to form a plurality of one-dimensional original vibration signal data, and performing phase space reconstruction processing to obtain a plurality of two-dimensional matrixes; singular value decomposition is carried out on the reconstructed two-dimensional matrix to obtain a singular value spectrum of the two-dimensional matrix; calculating the slope of the singular value spectrum to obtain singular value spectrum manifold characteristics; training a support vector machine by using the characteristic data to construct a fault diagnosis model; and inputting the vibration signal data of the gear box to be tested into the fault diagnosis model, and outputting the fault diagnosis classification result of the gear box to be tested. The method adopts singular value spectrum manifold analysis to realize the feature extraction of the gearbox fault data, can effectively extract the change trend of signal components, removes the influence of noise, enhances the characterization capability of the features on the fault, and can improve the precision of the gearbox fault diagnosis.

Description

Gearbox fault diagnosis method and system based on singular value spectrum manifold analysis
Technical Field
The invention relates to the technical fields of mechanical fault diagnosis, mode identification and the like, in particular to a method and a system for diagnosing faults of a gearbox based on singular value spectrum manifold analysis.
Background
Along with the improvement of the integration and automation degree of mechanical equipment, the mechanical equipment is more and more prone to failure, the damage caused by equipment failure is more and more large, and the requirements on the reliability and the safety of the mechanical equipment are also improved. The early failure diagnosis of the mechanical equipment can avoid huge loss caused by accidents, so that enterprises can obtain good economic and social benefits. The gear box is used as a transmission mechanism which is most widely applied, the structure of the gear box is complex, the working environment is severe, and faults are very easy to occur, so that the gear box has great significance in fault diagnosis.
The fault of the gearbox is usually expressed in the form of a vibration signal, and characteristic information related to the fault can be extracted from the vibration signal to realize fault diagnosis of the gearbox. In most cases, the operating environment of the gearbox has strong interfering noise or the signature is weak, so it is necessary to select a suitable fault signature extraction method.
The spectrum analysis method based on Fourier transform is widely applied to feature extraction, but is only suitable for stationary signals, and fault signals are usually non-stationary signals, so that the application of spectrum analysis is greatly limited; wavelet transformation can also be used for feature extraction of fault signals, but the selection of wavelet bases has great influence on the result of feature extraction, and the selection of wavelet bases has no uniform standard which can be referred to; empirical mode decomposition is an effective method for processing non-stationary signals, but a 'boundary effect' occurs in the processing process, thereby causing data pollution. The Singular Value Decomposition (SVD) is used for processing the operating state signal of the mechanical device, and is essentially a matrix decomposition method, which decomposes the signal into a plurality of subspaces, can separate noise while extracting the signal variation trend, and the singular value spectrum obtained after SVD decomposition reflects the distribution of each signal component in the signal, so that the method is a feature extraction method with wide application.
However, the range of the singular value is usually very large, and a few singular values having a large value are actually used when the singular value is directly used for fault diagnosis, so that fault information contained in the singular value spectrum cannot be fully utilized. In addition, although the singular value can reflect the distribution of each signal component in the signal, the actual numerical value of the singular value is affected by various factors such as interference noise and the like, so that the singular value is not sensitive enough to the contained fault information.
Disclosure of Invention
Aiming at the problems, the invention provides a gearbox fault diagnosis method and system based on singular value spectrum manifold analysis, signal component distribution changes under different fault states can be directly reflected as the distribution changes of a singular value spectrum, and a manifold topological structure can directly represent the changes of the singular value spectrum, so that the running state change conditions of equipment are reflected. The method adopts the singular value spectrum slope to extract the manifold topological structure characteristic of the singular value spectrum, and simultaneously adopts a variable scale method to calculate the singular value spectrum slope in order to eliminate the influence of the singular value numerical value on the characteristic extraction, so that the characteristic extraction process has a self-weighting effect, and the characterization capability of the characteristic on the fault can be improved.
A gearbox fault diagnosis method based on singular value spectral manifold analysis comprises the following steps:
step 1, acquiring various fault vibration signals of a gear box, preprocessing the fault vibration signals, correspondingly forming a plurality of one-dimensional original vibration signal data, and respectively performing phase space reconstruction processing to obtain a two-dimensional matrix of the original vibration signal data;
step 2, singular value decomposition is carried out on the reconstructed two-dimensional matrix to obtain a singular value spectrum of the two-dimensional matrix;
step 3, extracting the singular value spectrum manifold topological structure characteristics by calculating the slope of the singular value spectrum, thereby obtaining the singular value spectrum manifold characteristics;
step 4, training a support vector machine by using singular value spectrum manifold characteristic data to complete the construction of a fault diagnosis model;
and 5, inputting the vibration signal data of the gear box to be tested into a fault diagnosis model, and outputting a fault diagnosis classification result of the gear box to be tested.
Further, a gearbox fault diagnosis system based on singular value spectral manifold analysis is also provided, the system comprising:
the acceleration sensor is used for extracting a fault vibration signal of the gearbox;
the filter is used for filtering the extracted fault vibration signal of the gearbox;
the data reconstruction module is used for performing phase space reconstruction processing on the filtered one-dimensional original vibration signal and obtaining a two-dimensional Hankel matrix;
the data decomposition module is used for carrying out singular value decomposition on the data after reconstruction processing to obtain a singular value spectrum of the two-dimensional matrix;
the data characteristic module is used for processing singular value spectrum data to obtain singular value spectrum popular characteristics;
the model generation module is used for carrying out optimization training on the characteristic data output by the data characteristic module on a parameter optimization support vector machine adopting a competition mechanism firework algorithm to generate a fault diagnosis model;
and the fault diagnosis module is used for calling the fault diagnosis model generated by the model generation module, predicting the vibration signal data of the gear box to be tested and outputting the fault diagnosis classification result of the gear box to be tested.
Further, the present invention also provides a computer readable storage medium having stored thereon computer instructions which, when executed by a processor, implement the steps of a method for gearbox fault diagnosis based on singular value spectral manifold analysis.
Furthermore, the invention also provides gearbox fault diagnosis equipment based on singular value spectral manifold analysis, which comprises a memory, a processor and a computer program which is run on the memory and the processor, wherein the processor realizes the steps of the gearbox fault diagnosis method based on singular value spectral manifold analysis when executing the program.
The invention has the beneficial effects that:
1. the method adopts singular value spectrum manifold analysis to realize the feature extraction of the gearbox fault data, and compared with other feature extraction methods, the method can effectively extract the change trend of signal components, remove the influence of noise, enhance the characterization capability of features on faults, and improve the precision of gearbox fault diagnosis.
2. The method adopts an improved self-adaptive Unscented Kalman Filter (UKF) to filter the extracted fault vibration signal of the gearbox, and compared with the traditional Unscented Kalman filter, a suboptimal unbiased MAP constant noise statistical estimator is deduced according to the Maximum A Posteriori (MAP) estimation principle; and then, on the basis, an exponential weighting method is adopted to provide a recursion formula of the time-varying noise statistical estimator, so that the self-adaptive UKF filter with the noise statistical estimator is obtained. The method has the main advantages that under the condition that the noise statistics is unknown and time-varying, the filtering still converges, the filtering precision and the filtering stability are obviously improved, and the method has the self-adaptive capacity of coping with the noise variation.
Drawings
FIG. 1 is a schematic flow chart diagram of a fault diagnosis method based on singular value spectral manifold analysis according to a preferred embodiment of the present invention;
FIG. 2 is a schematic diagram of a vibration signal vector of an outer ring fault in an embodiment of the present invention;
FIG. 3 is a schematic diagram of a vibration signal vector of an inner ring fault in an embodiment of the present invention;
FIG. 4 is a vibration signal vector diagram of a combinational fault in an embodiment of the present invention;
FIG. 5 is a vibration signal vector diagram illustrating a root erosion fault in an embodiment of the present invention;
FIG. 6 is a vibration signal vector diagram of a root break fault in an embodiment of the present invention;
FIG. 7 is a diagram of a gearbox test stand in an embodiment of the present invention;
FIG. 8 is a singular value spectral manifold feature diagram in an embodiment of the invention;
FIG. 9 is a singular value feature map in an embodiment of the invention;
fig. 10 is a block diagram of a fault diagnosis system based on singular value spectral manifold analysis according to the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more clearly and completely apparent, the technical solutions in the embodiments of the present invention are described below with reference to the accompanying drawings, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments.
A preferred embodiment of the fault diagnosis method based on singular value spectrum manifold analysis according to the present invention, as shown in fig. 1, includes the following steps:
s11, acquiring various fault vibration signals of the gearbox, and preprocessing the fault vibration signals, including filtering, smoothing and the like;
s21, correspondingly forming a plurality of one-dimensional original vibration signal data by the preprocessed fault vibration signals, and respectively carrying out phase space reconstruction processing to obtain a two-dimensional matrix of the plurality of original vibration signal data;
s31, performing singular value decomposition on the reconstructed two-dimensional matrix to obtain a singular value spectrum of the two-dimensional matrix;
s41, extracting singular value spectrum manifold topological structure characteristics by calculating the slope of a singular value spectrum, thereby obtaining singular value spectrum manifold characteristics;
s51, training by using singular value spectrum manifold characteristic data, and completing construction of a fault diagnosis model by using a parameter optimization support vector machine of a competition mechanism firework algorithm;
and S61, inputting the vibration signal data of the gear box to be tested into the fault diagnosis model, and outputting the fault diagnosis classification result of the gear box to be tested.
In this embodiment, mainly 5 kinds of fault vibration signals in certain gearbox fault data are analyzed, including: outer ring failure, inner ring failure, combination failure, tooth root corrosion failure, tooth root fracture failure.
Firstly, receiving one-dimensional original vibration signal data of a gearbox, wherein a vibration signal tested each time comprises N sampling points; the time domain waveform of the vibration signal of the inner ring fault is shown in figure 2, the time domain waveform of the vibration signal of the outer ring fault is shown in figure 3, the time domain waveform of the vibration signal of the combined fault is shown in figure 4, the time domain waveform of the vibration signal of the tooth root corrosion fault is shown in figure 5, and the time domain waveform of the vibration signal of the tooth root fracture fault is shown in figure 6; carrying out phase space reconstruction processing on the vibration signal of the gearbox to obtain a two-dimensional matrix A; the phase space reconstruction means reconstructing the vibration signal vector of the gearbox into a high-dimensional matrix Hankel matrix, wherein the Hankel matrix is defined as a given set of vibrationMotion signal
Figure BDA0002229049700000051
Wherein p is more than 1 and less than N, q is more than 1 and less than N, and N is p + q; and N is the number of sampling points of the vibration signal.
In this embodiment, N is 4096, and the vibration signal vector phase space is reconstructed into a Hanke matrix of 25 × 4072, that is, p is 25 and q is 4072. Of course, the specific data size can be selected according to actual conditions.
Carrying out SVD on the two-dimensional matrix A to obtain a singular value spectrum of the two-dimensional matrix A; the SVD decomposition refers to the decomposition of a two-dimensional Hanke matrix A belonging to R25×4072The following transformations are performed:
A=UΣVT
wherein U is [ U ]1,u2,…,u25]∈R25×25And V ═ V1,v2,…,v4072]∈R4072×4072Is two orthogonal matrices, Σ ═ diag (λ)12,…λn),O]∈R25×4072Is a singular value matrix, where n ═ min (p, q), i.e. n ═ min (25,4072) ═ 25, λ1>λ2>…>λ25For the obtained singular values, SVs ═ λ12,…,λ25]The obtained singular value spectrum.
Calculating the slope of the obtained singular value spectrum to obtain the manifold topological structure characteristic of the singular value spectrum; the characteristic information contained in the singular value spectrum needs to be further extracted, and through calculation of the slope of the characteristic information, the method can be expressed as follows:
Figure BDA0002229049700000061
wherein i belongs to {1, 2.., n-1 }; the singular value spectrum SVs can be calculated to obtain 24 singular value spectrum slopes SLs ═ β12,…,βn-1]SLs are singular value spectrum manifold features, represent manifold topological structure of singular value spectrum, and include fault information of vibration signal vectors.
The method comprises the following steps of training a support vector machine by using singular value spectrum manifold features (SLs) to complete the construction of a fault diagnosis model, inputting vibration signal data of a gear box to be tested into the fault diagnosis model to obtain a final classification result, and training the support vector machine by using singular value spectrum manifold feature data to optimize parameters of a competition mechanism firework algorithm, wherein the support vector machine comprises:
the method for training the support vector machine by using the singular value spectral manifold feature data comprises the following steps of training the support vector machine by using the singular value spectral manifold feature data, and optimizing parameters of the support vector machine by using a competition mechanism firework algorithm, and specifically comprises the following steps:
s401, randomly selecting two groups of firework parameters by using a firework algorithm, wherein the two groups of firework parameters respectively correspond to a group of kernel function parameters and a group of punishment function parameters of a support vector machine;
s402, constructing a support vector machine from each kernel function parameter and each penalty function parameter; inputting singular value spectrum manifold feature data into each support vector machine, and outputting the identification rate of fault diagnosis;
s403, sorting according to the recognition rate, and respectively selecting a kernel function parameter and a penalty function parameter corresponding to the support vector machine with a larger recognition rate;
s404, optimizing the kernel function parameters and the penalty function parameters selected in the step S403 by adopting a competition mechanism firework algorithm, taking the optimized result as a new group of kernel function parameters and a new group of penalty function parameters, and returning to the step S402 until the kernel function parameters and the penalty function parameters corresponding to the support vector machine with the highest recognition rate, namely the optimal kernel function parameters and penalty function parameters, are selected.
In a still further aspect of the present invention,
1) competition mechanism firework algorithm:
randomly initializing two groups of mu fireworks in a search space; evaluating the fitness of each firework; the following process is repeated until the termination condition is satisfied: respectively calculating lambda of the two groups of fireworks from 1 to mujAnd AjEvaluating the fitness of all fireworks in each group respectively, and selecting better fireworks individuals (including the generation, fireworks, explosion fireworks and guide fireworks) as the next generation of smoke of the j generation in the groupInputting the fireworks into a support vector machine for training, selecting part of fireworks corresponding to a higher recognition rate as the next generation of fireworks, and executing a competition mechanism; finally, returning the position and the fitness of the optimal firework individual; the optimal firework individuals in the two groups respectively correspond to a kernel function parameter and a penalty function parameter of the support vector machine;
Figure BDA0002229049700000071
Figure BDA0002229049700000072
wherein λ isrRepresenting the number of explosion sparks of the r-th firework;
Figure BDA0002229049700000073
represents a constant control parameter; r represents the fitness value ranking of the fireworks, mu is the total number of the fireworks, and alpha is the shape distribution controlled by one parameter;
Figure BDA0002229049700000074
the explosion amplitude of the jth fireworks in g generation
Figure BDA0002229049700000075
Presetting as a constant;
Figure BDA0002229049700000076
g generation j firework; ca represents an amplification factor; represents a reduction coefficient;
Figure BDA0002229049700000077
representing the fitness value of the jth firework in the g generation;
wherein the contention mechanism includes defining a maximum generation algebra gmaxThe number of fireworks is from 1 to mu, and finding out the fireworks with better quality in g generations is reasonable
Figure BDA0002229049700000081
Time, calculate
Figure BDA0002229049700000082
If it is
Figure BDA0002229049700000083
G is substituted for j fireworks
Figure BDA0002229049700000084
The firework is used as a better firework of the generation;
wherein the content of the first and second substances,
Figure BDA0002229049700000085
representing the improvement degree of the jth firework in the g generation;
Figure BDA0002229049700000086
representing the minimum value in the fitness value of g-generation fireworks;
2) a support vector machine:
Figure BDA0002229049700000087
Figure BDA0002229049700000088
Figure BDA0002229049700000089
wherein | · | purple sweet2Expressing two norms, w is a normal vector of the hyperplane, C is a punishment parameter calculated by a competition mechanism firework algorithm, epsilonlAs a relaxation variable, ylA label for classifying the l hyperplane; b is the displacement, i.e. the distance to the hyperplane;
Figure BDA00022290497000000810
is the corresponding first vibration signal x after the kernel function processinglAnd M denotes the dimension of the support vector machine.
In one embodiment, the kernel function parameters and the penalty function parameters are respectively trained, that is, each kernel function parameter corresponds to μ penalty function parameters, that is, μ × μ support vector machines are generated based on μ kernel function parameters and μ penalty function parameters, the support vector machines are respectively trained, singular value spectral manifold feature data is input, the recognition rate of fault diagnosis is output, and the kernel function parameters and the penalty function parameters corresponding to the support vector machine with a higher recognition rate are selected, for example, before selection
Figure BDA00022290497000000811
And continuously optimizing the selected kernel function parameters and penalty function parameters by adopting a competition mechanism firework algorithm, and repeatedly selecting and optimizing until the optimal kernel function parameters and penalty function parameters are selected.
The purpose of fault diagnosis of mechanical equipment is to classify fault types, which is essentially a pattern recognition problem. The main steps of fault diagnosis of mechanical equipment comprise signal preprocessing, feature extraction, fault identification and the like, and the feature extraction plays an important role in fault diagnosis precision.
In order to verify the application effect of the invention in mechanical fault diagnosis, the effectiveness of the method is illustrated by certain gearbox fault data, and a gearbox test bed is shown in FIG. 7. The present invention may be used for mechanical fault diagnosis including, but not limited to, gearboxes.
The experiment selects and analyzes 5 kinds of fault vibration signals of the variable-speed gearbox of the intermediate shaft with the rotating speed of 20Hz-40Hz, the load of 0HP and the sampling frequency of 25600Hz, and the fault vibration signals comprise: outer ring failure, inner ring failure, combination failure, root erosion failure, and root fracture failure. 320 groups of samples are respectively measured under various fault states, and the sampling point number of each group of samples is 4096 points. 160 of the data sets were used as training samples, and the remaining 160 were used as test samples. The feature of the invention and the other five feature extraction methods are respectively extracted from the vibration signal of each sample: singular value characteristics, an AR model (AR model for short), time-frequency characteristics, wavelet characteristics and entropy characteristics, and fault diagnosis is performed by adopting a support vector machine for parameter optimization of a competition mechanism firework algorithm.
TABLE 1 Pattern recognition accuracy of the present invention corresponding to singular values, AR models, time-frequency, wavelets, entropy
Figure BDA0002229049700000091
Table 1 shows the final pattern recognition accuracy for 6 different feature extraction methods. The characteristic diagram of the method provided by the invention is shown in figure 8, and the singular value characteristic diagram is shown in figure 9. As can be seen from fig. 9, the singular value features in various fault states are mixed together, and the range of the numerical variation of the singular value is very large, so that the singular value features cannot identify the fault type, and the identification rate is 20%. As can be seen from the table, the pattern recognition accuracy of other feature extraction methods does not reach the classification accuracy of the invention. Therefore, the invention is a good application background in mechanical fault diagnosis.
In addition, a gearbox fault diagnosis system based on singular value spectral manifold analysis according to the present invention is shown in fig. 10, and the system includes:
the acceleration sensor is used for extracting a fault vibration signal of the gearbox;
the filter is used for filtering the extracted fault vibration signal of the gearbox;
the data reconstruction module is used for performing phase space reconstruction processing on the filtered one-dimensional original vibration signal and obtaining a two-dimensional Hankel matrix;
the data decomposition module is used for carrying out singular value decomposition on the data after reconstruction processing to obtain a singular value spectrum of the two-dimensional matrix;
the data characteristic module is used for processing singular value spectrum data to obtain singular value spectrum popular characteristics;
the model generation module is used for carrying out optimization training on the characteristic data output by the data characteristic module on a parameter optimization support vector machine adopting a competition mechanism firework algorithm to generate a fault diagnosis model;
and the fault diagnosis module is used for calling the fault diagnosis model generated by the model generation module, predicting the vibration signal data of the gear box to be tested and outputting the fault diagnosis classification result of the gear box to be tested.
As a preferred embodiment, the filter is a modified adaptive Unscented kalman filter. For example, an adaptive Unscented Kalman filter algorithm based on Maximum a posteriori estimation and exponential weighting proposed by zhao lin et al in "automated science and newspapers" 2010, which is designed to solve the problem of nonlinear filtering precision reduction and even divergence of a conventional Unscented Kalman filter (Unscented Kalman filter ) under the condition of unknown time-varying noise prior statistics, is a suboptimal unbiased MAP constant noise statistical estimator according to the principle of MAP estimation based on Maximum A Posteriori (MAP); and then, on the basis, an exponential weighting method is adopted to provide a recursion formula of the time-varying noise statistical estimator, so that the self-adaptive UKF filter with the noise statistical estimator is obtained. The method has the main advantages that under the condition that the noise statistics is unknown and time-varying, the filtering still converges, the filtering precision and the filtering stability are obviously improved, and the method has the self-adaptive capacity of coping with the noise variation. In the gearbox fault diagnosis model in the embodiment, various fault vibration signals can be acquired through the acceleration sensor additionally arranged on the gearbox.
Further, the present invention also provides a computer readable storage medium having stored thereon computer instructions which, when executed by a processor, implement the steps of a method for gearbox fault diagnosis based on singular value spectral manifold analysis.
Those skilled in the art will appreciate that all or part of the steps in the methods of the above embodiments may be implemented by associated hardware instructed by a program, which may be stored in a computer-readable storage medium, and the storage medium may include: ROM, RAM, magnetic or optical disks, and the like.
Furthermore, the invention also provides gearbox fault diagnosis equipment based on singular value spectral manifold analysis, which comprises a memory, a processor and a computer program which is run on the memory and the processor, wherein the processor realizes the steps of the gearbox fault diagnosis method based on singular value spectral manifold analysis when executing the program.
The fault diagnosis device comprises one or more processors and a memory, wherein one processor is taken as an example;
the device for executing list item operation may further comprise a collecting device and an output device.
The processor, memory, acquisition device and output device may be connected by a CAN bus or other means.
The processor may be a central processing unit CPU, but may also be other general purpose processors, digital signal processors, etc.
It is to be understood that some features of the fault diagnosis method, system, computer readable storage medium and fault diagnosis apparatus of the present invention may be mutually cited, and the present invention is not to be construed as illustrative for the sake of brevity.
The above-mentioned embodiments, which further illustrate the objects, technical solutions and advantages of the present invention, should be understood that the above-mentioned embodiments are only preferred embodiments of the present invention, and should not be construed as limiting the present invention, and any modifications, equivalents, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims (9)

1. A gearbox fault diagnosis method based on singular value spectral manifold analysis is characterized by comprising the following steps:
s1, acquiring various fault vibration signals of the gearbox, preprocessing the fault vibration signals, correspondingly forming a plurality of one-dimensional original vibration signal data, and respectively performing phase space reconstruction processing to obtain a two-dimensional matrix of the original vibration signal data;
s2, performing singular value decomposition on the reconstructed two-dimensional matrix to obtain a singular value spectrum of the two-dimensional matrix;
step S3, extracting singular value spectrum manifold topological structure characteristics by calculating the slope of a singular value spectrum, thereby obtaining singular value spectrum manifold characteristics; i.e. calculating each phaseTwo adjacent singular values λiAnd λi+1The slope of (1), a vector SLs formed by the obtained n-1 singular value spectrum slopes is [ beta ]12,…,βn-1]As a singular value spectral manifold feature; wherein the content of the first and second substances,
Figure FDA0003065730370000011
n is min (p, q), and the two-dimensional matrix is a p × q two-dimensional matrix; 1<p<N,1<q<N, wherein N is p + q, and N is the number of sampling points of the vibration signal of the gearbox;
s4, training a support vector machine by using singular value spectrum manifold characteristic data to complete the construction of a fault diagnosis model;
and step S5, inputting the vibration signal data of the gear box to be tested into the fault diagnosis model, and outputting the fault diagnosis classification result of the gear box to be tested.
2. The method of claim 1, wherein the fault vibration signal data comprises an outer ring fault, an inner ring fault, a combination fault, a root corrosion fault, and a root fracture fault.
3. The method for diagnosing the fault of the gearbox based on the singular value spectral manifold analysis as claimed in claim 1, wherein the phase space reconstruction process in the step S1 includes obtaining a one-dimensional vibration signal vector X ═ X of the gearbox1,x2,…,xN]Reconstructing the matrix into a two-dimensional Hankel matrix
Figure FDA0003065730370000012
Wherein x ispRepresenting the vibration signal corresponding to the p sampling point; x is the number ofqRepresenting the vibration signal corresponding to the q sampling point; 1<p<N,1<q<N; and N ═ p + q; and N is the number of sampling points of the vibration signal of the gearbox.
4. Singular value based spectral manifold according to claim 3Method for diagnosing faults of an analytical gearbox, characterized in that said step S2 comprises dividing a two-dimensional matrix A e R of size p qp×qThe following transformations are performed: a ═ U ∑ VT(ii) a Wherein U and V are two orthogonal matrices, and T represents transposition; u ═ U1,u2,...,up]∈Rp×p;V=[v1,v2,...,vq]∈Rq×q;∑=[diag(λ12,...,λn),O]∈Rp×qThe matrix is a singular value matrix, is a diagonal matrix, and the other elements except the diagonal elements are 0; n ═ min (p, q), λ12>...>λnFor the obtained singular values, SVs ═ λ12,...,λn]The obtained singular value spectrum.
5. The method of claim 3, wherein the training of the SVM using the singular value spectral manifold feature data comprises: the method adopts singular value spectrum manifold characteristic data to train a support vector machine, and adopts a competition mechanism firework algorithm to optimize parameters of the support vector machine, and specifically comprises the following steps:
s401, randomly selecting two groups of firework parameters by using a firework algorithm, wherein the two groups of firework parameters respectively correspond to a group of kernel function parameters and a group of punishment function parameters of a support vector machine;
s402, constructing a support vector machine from each kernel function parameter and each penalty function parameter; inputting singular value spectrum manifold feature data into each support vector machine, and outputting the identification rate of fault diagnosis;
s403, sorting according to the recognition rate, and respectively selecting a kernel function parameter and a penalty function parameter corresponding to the support vector machine with a larger recognition rate;
s404, optimizing the kernel function parameters and the penalty function parameters selected in the step S403 by adopting a competition mechanism firework algorithm, taking the optimized result as a new group of kernel function parameters and a new group of penalty function parameters, and returning to the step S402 until the kernel function parameters and the penalty function parameters corresponding to the support vector machine with the highest recognition rate, namely the optimal kernel function parameters and penalty function parameters, are selected.
6. A system for implementing a method for gearbox fault diagnosis based on singular value spectral manifold analysis, the system comprising:
the acceleration sensor is used for extracting a fault vibration signal of the gearbox;
the filter is used for filtering the extracted fault vibration signal of the gearbox;
the data reconstruction module is used for performing phase space reconstruction processing on the filtered one-dimensional original vibration signal and obtaining a two-dimensional Hankel matrix;
the data decomposition module is used for carrying out singular value decomposition on the data after reconstruction processing to obtain a singular value spectrum of the two-dimensional matrix;
the data characteristic module is used for extracting singular value spectrum manifold topological structure characteristics by calculating the slope of a singular value spectrum to obtain singular value spectrum manifold characteristics; i.e. calculating every two adjacent singular values lambdaiAnd λi+1The slope of (1), a vector SLs formed by the obtained n-1 singular value spectrum slopes is [ beta ]12,…,βn-1]As a singular value spectral manifold feature; wherein the content of the first and second substances,
Figure FDA0003065730370000031
n is min (p, q), and the two-dimensional matrix is a p × q two-dimensional matrix; 1<p<N,1<q<N, wherein N is p + q, and N is the number of sampling points of the vibration signal of the gearbox;
the model generation module is used for carrying out optimization training on the support vector machine by using the characteristic data output by the data characteristic module to generate a fault diagnosis model;
and the fault diagnosis module is used for calling the fault diagnosis model generated by the model generation module, predicting the vibration signal data of the gear box to be tested and outputting the fault diagnosis classification result of the gear box to be tested.
7. The system for implementing the gearbox fault diagnosis method based on the singular value spectral manifold analysis according to claim 6, wherein the filter is a modified adaptive Unscented Kalman filter.
8. A computer readable storage medium having stored thereon computer instructions, wherein the instructions when executed by a processor implement the steps of a method for gearbox fault diagnosis based on singular value spectral manifold analysis according to any of claims 1 to 5.
9. A gearbox fault diagnosis device based on singular value spectral manifold analysis, comprising a memory, a processor and a computer program running on the memory and on the processor, characterized in that the processor implements the steps of a gearbox fault diagnosis method based on singular value spectral manifold analysis according to any one of claims 1 to 5 when executing the program.
CN201910961446.3A 2019-10-11 2019-10-11 Gearbox fault diagnosis method and system based on singular value spectrum manifold analysis Active CN110657985B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910961446.3A CN110657985B (en) 2019-10-11 2019-10-11 Gearbox fault diagnosis method and system based on singular value spectrum manifold analysis

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910961446.3A CN110657985B (en) 2019-10-11 2019-10-11 Gearbox fault diagnosis method and system based on singular value spectrum manifold analysis

Publications (2)

Publication Number Publication Date
CN110657985A CN110657985A (en) 2020-01-07
CN110657985B true CN110657985B (en) 2021-07-06

Family

ID=69040508

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910961446.3A Active CN110657985B (en) 2019-10-11 2019-10-11 Gearbox fault diagnosis method and system based on singular value spectrum manifold analysis

Country Status (1)

Country Link
CN (1) CN110657985B (en)

Families Citing this family (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111444963B (en) * 2020-03-27 2023-08-25 中南大学 SSA-SVR model-based blast furnace molten iron silicon content prediction method
CN111459838B (en) * 2020-04-20 2021-09-03 武汉大学 Software defect prediction method and system based on manifold alignment
CN111611870B (en) * 2020-04-26 2023-05-09 广东机电职业技术学院 Fault feature identification method, system, device and medium for cold tandem mill
CN112161807B (en) * 2020-09-22 2022-11-01 上海海事大学 Fault diagnosis method and device for speed change gear box and storage medium
CN113295413B (en) * 2021-06-24 2022-10-25 北京交通大学 Traction motor bearing fault diagnosis method based on indirect signals
CN113792397A (en) * 2021-09-02 2021-12-14 江苏新道格自控科技有限公司 Rotary machine gear box fault diagnosis method based on deep learning
CN114279704B (en) * 2021-12-28 2022-11-22 德力佳传动科技(江苏)有限公司 Method and system for monitoring planet-level running state of wind power gear box
CN115100285B (en) * 2022-08-25 2022-12-09 深圳市信润富联数字科技有限公司 Wind power sensor installation method, device, equipment and readable storage medium
CN115169423B (en) * 2022-09-08 2023-05-02 深圳市信润富联数字科技有限公司 Stamping signal processing method, device, equipment and readable storage medium
CN116662793B (en) * 2023-07-31 2024-03-08 昆仑数智科技有限责任公司 Rotor crack fault diagnosis and model construction method and device

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH0462451A (en) * 1990-06-29 1992-02-27 Toyota Motor Corp Gear box testing device
CN103810394A (en) * 2014-02-28 2014-05-21 东北电力大学 Design method for singular value decomposition and de-noising of rotating device fault signal
CN104748961A (en) * 2015-03-30 2015-07-01 中国矿业大学 Gear fault diagnosis method based on SVD decomposition and noise reduction and correlation EEMD entropy features
CN105760839A (en) * 2016-02-22 2016-07-13 重庆大学 Bearing fault diagnosis method based on multi-feature manifold learning and support vector machine
CN106197996A (en) * 2016-06-24 2016-12-07 南京理工大学 Offshore crane Fault Diagnosis of Gear Case device and method based on multivariate data
CN106446829A (en) * 2016-09-22 2017-02-22 三峡大学 Hydroelectric generating set vibration signal noise reduction method based on mode autocorrelation analysis of SVD and VMD
CN107192553A (en) * 2017-06-28 2017-09-22 石家庄铁道大学 Gear-box combined failure diagnostic method based on blind source separating

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH0462451A (en) * 1990-06-29 1992-02-27 Toyota Motor Corp Gear box testing device
CN103810394A (en) * 2014-02-28 2014-05-21 东北电力大学 Design method for singular value decomposition and de-noising of rotating device fault signal
CN104748961A (en) * 2015-03-30 2015-07-01 中国矿业大学 Gear fault diagnosis method based on SVD decomposition and noise reduction and correlation EEMD entropy features
CN105760839A (en) * 2016-02-22 2016-07-13 重庆大学 Bearing fault diagnosis method based on multi-feature manifold learning and support vector machine
CN106197996A (en) * 2016-06-24 2016-12-07 南京理工大学 Offshore crane Fault Diagnosis of Gear Case device and method based on multivariate data
CN106446829A (en) * 2016-09-22 2017-02-22 三峡大学 Hydroelectric generating set vibration signal noise reduction method based on mode autocorrelation analysis of SVD and VMD
CN107192553A (en) * 2017-06-28 2017-09-22 石家庄铁道大学 Gear-box combined failure diagnostic method based on blind source separating

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
基于流形-奇异值熵的滚动轴承故障特征提取;王奉涛等;《振动.测试与诊断》;20160430;第36卷(第2期);第288-293页 *
基于相空间奇异值分解与AFSA-SVM的齿轮故障诊断方法;秦波等;《组合机床与自动化加工技术》;20160831(第8期);第67-70页 *

Also Published As

Publication number Publication date
CN110657985A (en) 2020-01-07

Similar Documents

Publication Publication Date Title
CN110657985B (en) Gearbox fault diagnosis method and system based on singular value spectrum manifold analysis
CN111797567B (en) Bearing fault classification method and system based on deep learning network
CN104807534B (en) Equipment eigentone self study recognition methods based on on-line vibration data
CN110647867A (en) Bearing fault diagnosis method and system based on self-adaptive anti-noise neural network
CN105807241B (en) A kind of exponential signal denoising method using prior information
CN111914703A (en) Mechanical rotating part fault diagnosis method based on wavelet transformation and transfer learning GoogLeNet
Jiang et al. A novel method for self-adaptive feature extraction using scaling crossover characteristics of signals and combining with LS-SVM for multi-fault diagnosis of gearbox
CN114091504A (en) Rotary machine small sample fault diagnosis method based on generation countermeasure network
CN114237046A (en) Partial discharge pattern recognition method based on SIFT data feature extraction algorithm and BP neural network model
CN110782041B (en) Structural modal parameter identification method based on machine learning
Han et al. Fault diagnosis of rolling bearing with uneven data distribution based on continuous wavelet transform and deep convolution generated adversarial network
CN112945556B (en) Bearing fault diagnosis method based on wavelet packet decomposition and optimal rejection classification strategy
CN114707534A (en) Rolling bearing fault diagnosis method under small sample data
CN113657244A (en) Fan gearbox fault diagnosis method and system based on improved EEMD and speech spectrum analysis
CN113627539A (en) Rolling bearing composite fault diagnosis method and system, storage medium and computing equipment
CN117473414A (en) Bearing fault position identification method based on low-noise time-frequency image
CN116610990A (en) Method and device for identifying hidden danger of breaker based on characteristic space differentiation
CN110044619B (en) Multi-fault feature identification method based on sparse multi-cycle group lasso
CN111025100A (en) Transformer ultrahigh frequency partial discharge signal mode identification method and device
Chen et al. Gear fault diagnosis model combined with MED-LMD-Hypersphere multiclass SVM
CN115982566A (en) Multi-channel fault diagnosis method for hydroelectric generating set
CN115563480A (en) Gear fault identification method for screening octave geometric modal decomposition based on kurtosis ratio coefficient
CN113869358A (en) Bearing fault diagnosis method based on cyclic correlation entropy and one-dimensional shallow convolution neural network
CN114925724A (en) Mechanical equipment fault diagnosis method and device and storage medium
CN115326396A (en) Bearing fault diagnosis method and device

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant