CN117232846A - Marine turbocharger fault diagnosis method, device and equipment based on coarse granularity - Google Patents
Marine turbocharger fault diagnosis method, device and equipment based on coarse granularity Download PDFInfo
- Publication number
- CN117232846A CN117232846A CN202311193524.2A CN202311193524A CN117232846A CN 117232846 A CN117232846 A CN 117232846A CN 202311193524 A CN202311193524 A CN 202311193524A CN 117232846 A CN117232846 A CN 117232846A
- Authority
- CN
- China
- Prior art keywords
- fault
- coarse
- frequency domain
- neural network
- signal
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 48
- 238000003745 diagnosis Methods 0.000 title claims abstract description 24
- 238000001228 spectrum Methods 0.000 claims abstract description 70
- 238000012549 training Methods 0.000 claims abstract description 57
- 238000013528 artificial neural network Methods 0.000 claims abstract description 52
- 238000003062 neural network model Methods 0.000 claims abstract description 32
- 230000003595 spectral effect Effects 0.000 claims abstract description 32
- 238000010586 diagram Methods 0.000 claims abstract description 31
- 238000012545 processing Methods 0.000 claims abstract description 31
- 238000005070 sampling Methods 0.000 claims abstract description 25
- 238000000605 extraction Methods 0.000 claims abstract description 14
- 238000012795 verification Methods 0.000 claims description 12
- 230000006870 function Effects 0.000 claims description 11
- 238000004590 computer program Methods 0.000 claims description 10
- 238000012360 testing method Methods 0.000 claims description 10
- 230000008569 process Effects 0.000 claims description 8
- 238000005520 cutting process Methods 0.000 claims description 4
- 238000000638 solvent extraction Methods 0.000 claims description 4
- 238000003860 storage Methods 0.000 claims description 4
- 238000005096 rolling process Methods 0.000 abstract description 10
- 230000002159 abnormal effect Effects 0.000 abstract description 5
- 230000004927 fusion Effects 0.000 abstract description 4
- 238000005192 partition Methods 0.000 description 5
- 230000008901 benefit Effects 0.000 description 4
- 238000012512 characterization method Methods 0.000 description 4
- 239000011159 matrix material Substances 0.000 description 4
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 230000008878 coupling Effects 0.000 description 2
- 238000010168 coupling process Methods 0.000 description 2
- 238000005859 coupling reaction Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 230000003213 activating effect Effects 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000002485 combustion reaction Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000001816 cooling Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 239000000446 fuel Substances 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000005461 lubrication Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
Abstract
The invention provides a coarse-grained based marine turbocharger bearing fault diagnosis method, which comprises the following steps: coarsening the frequency domain signals of each vibration sample signal of the bearing based on coarse granularity sampling coefficients to obtain a second frequency spectrum characteristic value; generating a coarse-grained lattice feature map based on the second spectral feature values; training the fault identification neural network by using the coarse-granularity lattice feature diagram, and outputting a fault identification neural network model; identifying the current coarse-grain lattice feature diagram of the bearing to be diagnosed by utilizing a fault identification neural network model, and outputting a corresponding fault type; after the turbocharger bearing fault signals are collected, coarse granularity processing is carried out on the vibration signals, so that the characteristics of the fault signals are obviously enhanced; after the first frequency spectrum characteristic value is processed, the influence of abnormal fault data is reduced; the fault recognition neural network has good feature fusion and feature extraction capability, accurately completes fault classification recognition tasks, and improves the fault diagnosis accuracy of the rolling bearing.
Description
Technical Field
The application relates to the technical field of marine turbocharger fault diagnosis, in particular to a marine turbocharger fault diagnosis method, device and equipment based on coarse granularity.
Background
The marine turbocharger plays an important role in improving the power density and the fuel efficiency of a marine diesel engine, improving the in-cylinder combustion process, reducing the emission index of the diesel engine and the like, and the safe, stable and efficient operation of the marine turbocharger is an important guarantee for the operation safety and the economy of the marine diesel engine.
The inside of the high-supercharging diesel engine turbocharger generally adopts a compact structure, the lubrication and cooling conditions of a rotor bearing are poor, and the working condition of the turbocharger can fluctuate in a large range along with the change of a diesel engine control instruction and external load, so that the bearing failure rate is high under the action of the unstable alternating load formed by the high-supercharging diesel engine turbocharger, and the power output stability and the continuity of a power system are seriously influenced.
The sensor elements capable of effectively sensing the state of the bearing cannot be distributed in the supercharger, the turbocharging system can only collect the integral coupling vibration signals of the rotor, weak vibration signals of bearing faults are easy to annihilate by strong background noise in the integral coupling vibration signals, so that fault characteristics of the weak faults of the bearing are difficult to discover early, and the accuracy of fault diagnosis of the marine turbocharging bearing is further affected.
Disclosure of Invention
Aiming at the problems existing in the prior art, the embodiment of the invention provides a method, a device and equipment for realizing fault diagnosis of a marine turbocharger based on coarse granularity, which are used for solving or partially solving the technical problems that the fault signal of the marine turbocharger bearing cannot be accurately extracted in the prior art, so that the accuracy of the fault diagnosis of the marine turbocharger bearing cannot be ensured.
In a first aspect of the present invention, there is provided a coarse-grained based marine turbocharger fault diagnosis method, the method comprising:
obtaining a vibration sample signal of a marine turbocharger bearing; the vibration sample signal comprises a first vibration signal of a normal bearing and a second vibration signal of a fault bearing; the first vibration signal and the second vibration signal are preset with corresponding state labels;
converting each vibration sample signal into a corresponding frequency domain signal, and performing coarse granularity processing on each frequency domain signal based on a preset coarse granularity sampling coefficient to obtain a corresponding first frequency spectrum characteristic value;
processing the first spectrum characteristic value to obtain a corresponding second spectrum characteristic value; generating a corresponding coarse-granularity lattice feature map based on the second spectrum feature value;
Training a pre-constructed fault recognition neural network by using the coarse-granularity lattice feature diagram, and outputting a trained fault recognition neural network model; the fault identification neural network is a Swin Transformer network;
and determining a current coarse-grain lattice characteristic diagram corresponding to the bearing to be diagnosed, identifying the current coarse-grain lattice characteristic diagram by using the fault identification neural network model, and outputting a corresponding fault type.
In the above scheme, the performing coarse granularity processing on each frequency domain signal based on a preset coarse granularity sampling coefficient to obtain a corresponding first spectrum characteristic value includes:
for any one of the frequency domain signals, dividing the frequency domain signal into a plurality of segments of sub-frequency domain signals using the formula j=fix (r/d) based on the coarse-granularity sampling coefficients;
according to formula F CN (j)＝sum(F c ((j-1) d+1, j d)) determining a first spectral feature value F of said each segment of sub-frequency domain signal CN (j) The method comprises the steps of carrying out a first treatment on the surface of the Wherein,
the j is the number of segments of the sub-frequency domain signal, d is the coarse-granularity sampling coefficient, fix is a rounding function, r is the length of the frequency domain signal, F C () Is the amplitude of the spectral signal.
In the above scheme, the processing the first spectrum feature value to obtain a corresponding second spectrum feature value includes:
Using the formulaProcessing the first spectrum characteristic value to obtain the second spectrum characteristic value F cn ' (j); wherein,
the F is cn (j) For the first spectral feature value, j is the number of segments dividing the frequency domain signal into sub-frequency domain signals, F cnmin For the minimum spectral feature value in each segment of the sub-frequency domain signal, the F cnmax And for the maximum spectrum characteristic value in each segment of sub-frequency domain signal, S is the preset number of spectrum characteristic values, and the value of S is consistent with the value of j.
In the above solution, the generating a corresponding coarse-grained lattice feature map based on the second spectral feature value includes:
acquiring the number of segments for dividing the frequency domain signal into sub-frequency domain signals;
and drawing the coarse-granularity lattice characteristic diagram by taking the number of segments of the sub-frequency domain signal as an abscissa and the corresponding second frequency spectrum characteristic value as an ordinate.
In the above scheme, the training the pre-constructed fault identification neural network by using the coarse-granularity lattice feature map, and outputting a trained fault identification neural network model, includes:
adjusting the coarse-grain lattice feature map to a preset size, and dividing the size-adjusted coarse-grain lattice feature map into a training set, a verification set and a data set according to a preset proportion;
Acquiring preset training parameters; the training parameters include: learning rate, weight decay, iteration number, optimizer and loss function;
training the fault recognition neural network by using the training set and the training parameters, and adjusting the training parameters by using a verification set in the training process;
and testing the trained fault recognition neural network by using the test set, and outputting a trained fault recognition neural network model if the accuracy of the trained fault recognition neural network reaches a preset accuracy threshold.
In the above scheme, the identifying the current coarse-granularity lattice feature map by using the fault identification neural network model, and outputting the corresponding fault type includes:
cutting the current coarse-granularity lattice feature map by utilizing an image block partitioning layer of the fault identification network model to obtain a plurality of non-overlapping image blocks;
performing feature extraction on a plurality of non-overlapping image blocks by using a fault feature value extraction layer of the fault identification network model to obtain local fault feature values;
fitting the local fault characteristic values by using a full connection layer of the fault identification network model to obtain a plurality of global fault characteristic values;
And determining the probability of each global fault characteristic value by using a fault characteristic classifier of the fault identification network model, and outputting the fault type according to the probability of the global fault characteristic value.
In a second aspect of the present invention, there is provided a coarse-grained based marine turbocharger fault diagnosis apparatus, the apparatus comprising:
the acquisition unit is used for acquiring a vibration sample signal of the marine turbocharger bearing; the vibration sample signal comprises a first vibration signal of a normal bearing and a second vibration signal of a fault bearing; the first vibration signal and the second vibration signal are preset with corresponding state labels;
the processing unit is used for converting each vibration sample signal into a corresponding frequency domain signal, and performing coarse granularity processing on each frequency domain signal based on a preset coarse granularity sampling coefficient to obtain a corresponding first frequency spectrum characteristic value;
the generating unit is used for processing the first frequency spectrum characteristic value to obtain a corresponding second frequency spectrum characteristic value; generating a corresponding coarse-granularity lattice feature map based on the second spectrum feature value;
the training unit is used for training the pre-constructed fault recognition neural network by utilizing the coarse-granularity lattice feature diagram and outputting a trained fault recognition neural network model;
And the identification unit is used for determining a current coarse grain lattice characteristic diagram corresponding to the bearing to be diagnosed, identifying the current coarse grain lattice characteristic diagram by utilizing the fault identification neural network model and outputting a corresponding fault type.
In the above scheme, the processing unit is specifically configured to:
for any frequency domain signal, dividing the frequency domain signal into a plurality of sections of sub-frequency domain signals by utilizing the coarse granularity sampling coefficient;
acquiring spectrum characteristic values of a plurality of frequency domain signals contained in each segment of sub-frequency domain signal;
a sum of spectral eigenvalues of a plurality of frequency domain signals is determined as the first spectral eigenvalue.
In a third aspect of the invention, there is provided a computer readable storage medium having stored thereon a computer program which when executed by a processor performs the steps of the method of any of the first aspects.
In a fourth aspect the present invention provides a computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the steps of the method of any one of the first aspects when the program is executed.
The invention provides a coarse-grained based marine turbocharger fault diagnosis method, device and equipment, wherein the method comprises the following steps: obtaining a vibration sample signal of a marine turbocharger bearing; the vibration sample signal comprises a first vibration signal of a normal bearing and a second vibration signal of a fault bearing; the first vibration signal and the second vibration signal are preset with corresponding state labels; converting each vibration sample signal into a corresponding frequency domain signal, and performing coarse granularity processing on each frequency domain signal based on a preset coarse granularity sampling coefficient to obtain a corresponding first frequency spectrum characteristic value; processing the first spectrum characteristic value to obtain a corresponding second spectrum characteristic value; generating a corresponding coarse-granularity lattice feature map based on the second spectrum feature value; training a pre-constructed fault recognition neural network by using the coarse-granularity lattice feature diagram, and outputting a trained fault recognition neural network model; the fault identification neural network is a Swin Transformer network; determining a current coarse-grain lattice feature map corresponding to a bearing to be diagnosed, identifying the current coarse-grain lattice feature map by using the fault identification neural network model, and outputting a corresponding fault type; therefore, even if the bearing fault signal of the marine turbocharger is weak, the vibration signal of the bearing can be subjected to coarse granularity treatment, so that the characteristics of the fault signal are obviously enhanced; after the first frequency spectrum characteristic value is processed, the accuracy of sample data can be further improved, and the influence of abnormal fault data is reduced; the Swin Transformer network also has good feature fusion and feature extraction capability, and can accurately complete fault classification and identification tasks, so that the fault diagnosis accuracy of the rolling bearing is improved.
Drawings
Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The drawings are only for purposes of illustrating the preferred embodiments and are not to be construed as limiting the invention. Also, like reference numerals are used to designate like parts throughout the figures.
In the drawings:
FIG. 1 shows a flow diagram of a coarse-grained based marine turbocharger fault diagnosis method according to one embodiment of the invention;
FIG. 2A illustrates a spectral signature of a first vibration signal according to one embodiment of the invention;
FIG. 2B illustrates a coarse-grain lattice characterization of a first vibration signal according to one embodiment of the present invention;
FIG. 3A illustrates a spectral signature of a second vibration signal when the type of fault is an outer ring fault, according to one embodiment of the present invention;
FIG. 3B illustrates a coarse-grained lattice characterization of a second vibration signal when the failure type is an outer ring failure, according to an embodiment of the invention;
FIG. 4A illustrates a spectral signature of a second vibration signal when the type of fault is an inner ring fault, according to one embodiment of the present invention;
FIG. 4B illustrates a coarse-grain lattice characterization of a second vibration signal when the fault type is an inner ring fault, according to one embodiment of the present invention;
FIG. 5A illustrates a spectral signature of a second vibration signal when the type of fault is a rolling element fault, according to one embodiment of the invention;
FIG. 5B illustrates a coarse-grain lattice characterization of a second vibration signal when the type of failure is a rolling element failure, according to one embodiment of the invention;
FIG. 6 illustrates a schematic diagram of recognition accuracy of a failure recognition neural network according to one embodiment of the present invention;
FIG. 7 illustrates a schematic diagram of loss values for a failure recognition neural network, according to one embodiment of the invention;
FIG. 8 illustrates a confusion matrix schematic of a validation set according to one embodiment of the invention;
FIG. 9 shows a schematic diagram of a failure recognition neural network according to one embodiment of the invention;
fig. 10 shows a schematic structural diagram of Swin Transformer Block according to an embodiment of the invention;
FIG. 11 is a schematic diagram of fault diagnosis logic of a fault-recognizing neural network when the CGLF graph has a size of 224×224×3, according to an embodiment of the invention;
FIG. 12 shows a schematic structural view of a coarse-grained based marine turbocharger fault diagnosis device according to an embodiment of the invention;
FIG. 13 shows a schematic diagram of a computer device architecture according to one embodiment of the invention;
FIG. 14 shows a schematic diagram of a computer-readable storage medium structure according to one embodiment of the invention.
Detailed Description
Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.
The invention provides a fault diagnosis method of a marine turbocharger based on coarse granularity, which is shown in fig. 1 and comprises the following steps:
s110, obtaining a vibration sample signal of a marine turbocharger bearing; the vibration sample signal comprises a first vibration signal of a normal bearing and a second vibration signal of a fault bearing; the first vibration signal and the second vibration signal are preset with corresponding state labels;
the invention can sample the vibration signals of the marine turbocharger bearing in different states to obtain the original vibration signals. The states include a normal state and a fault state.
The invention can intercept original vibration signals of the rolling bearing in different states by utilizing the sliding window as vibration sample signals of the bearing. The length of the sliding window may be preset, for example, 8192, and then the length of each vibration sample signal acquired is 8192. The number of vibration sample signals may be determined according to the need, for example, 10000, which is not limited herein.
Since the bearings include normal bearings and bearings having various fault types, the vibration sample signal includes a first vibration signal of the normal bearing and a second vibration signal of the faulty bearing. The fault types may include: failure types such as outer ring failure, inner ring failure, rolling element failure, etc. Since the fault size may be different when there is a fault in the bearing, each fault type may also be divided into 3 fault states for each fault type, the fault size of each fault state being different.
For example, for each fault type, the fault states may include: failure diameters were 7 mil, 14mil, and 21mil failures.
In order to improve the training accuracy of the fault identification neural network, the first vibration signal and the second vibration signal are further provided with corresponding state labels.
For example, for a first vibration signal to be a fault-free signal, the corresponding status tag may be 0; and setting a corresponding state label according to the fault state, wherein the state label corresponding to the fault type with the fault diameter of 7mils is 1, the state label corresponding to the fault type with the fault diameter of 14mils is 2, and the state label corresponding to the fault type with the fault diameter of 21mils is 3.
For the second vibration signal with the fault type of the inner ring fault, the state label corresponding to the fault type with the fault diameter of 7mils is 4, the state label corresponding to the fault type with the fault diameter of 14mils is 5, and the state label corresponding to the fault type with the fault diameter of 21mils is 6.
For the second vibration signal of which the failure type is a rolling element failure, the state label corresponding to the failure type of which the failure diameter is 7mils is 7, the state label corresponding to the failure type of which the failure diameter is 14mils is 8, and the state label corresponding to the failure type of which the failure diameter is 21mils is 9.
S111, converting each vibration sample signal into a corresponding frequency domain signal, and performing coarse granularity processing on each frequency domain signal based on a preset coarse granularity sampling coefficient to obtain a corresponding first frequency spectrum characteristic value;
After each vibration sample signal is obtained, each vibration sample signal can be converted into a corresponding frequency domain signal by using a fast fourier transform method. Because of the influence of some environmental noise, the corresponding fault characteristics in the frequency domain signals are weak, and in order to enhance the significance of the fault characteristic values, the invention carries out coarse granularity processing on each frequency domain signal based on a preset coarse granularity sampling coefficient to obtain the corresponding first frequency spectrum characteristic values.
In one embodiment, performing coarse granularity processing on each frequency domain signal based on a preset coarse granularity sampling coefficient to obtain a corresponding first spectrum characteristic value, including:
for any frequency domain signal, dividing the frequency domain signal into a plurality of sections of sub-frequency domain signals by using coarse granularity sampling coefficients;
acquiring spectrum characteristic values of a plurality of frequency domain signals contained in each segment of sub-frequency domain signal;
a sum of spectral eigenvalues of the plurality of frequency domain signals is determined as a first spectral eigenvalue.
Specifically, the coarse-granularity sampling coefficient needs to be determined according to the length of the sliding window, and in general, the quotient of the sliding window length and the coarse-granularity sampling coefficient is ensured to be an integer as much as possible.
After the coarse-granularity sampling coefficients are set, the division of the frequency domain signal into j segments of sub-frequency domain signals can be determined using the formula j=fix (r/d). Where r is the length of the frequency domain signal, d is the coarse-granularity sampling coefficient, and fix is a rounding function.
For example, if the length of the vibration sample signal is 8192, after the vibration sample signal is converted into the frequency domain signal, the length of the frequency domain signal is 8192/2=4096; assuming that the coarse-granularity sampling coefficient is 32, j=fix (4096/32) =128.
That is, each frequency domain signal may be divided into 128 segments of sub-frequency domain signals, each segment of sub-frequency domain signals containing 32 frequency domain signals.
Then for any segment of the sub-frequency domain signal, the method can be based on the formula F CN (j)＝sum(F c ((j-1) d+1, j d)) determines a first spectral feature value for each segment of the sub-frequency domain signal.
It will be appreciated that the number of components,after the vibration signal is converted into frequency domain signals, each frequency domain signal has a corresponding amplitude (frequency spectrum characteristic value) F C () Then for each segment of the sub-frequency-domain signal, the sum of the magnitudes of all the frequency-domain signals contained in the segment of the sub-frequency-domain signal can be determined as the first spectral feature value F of the segment of the sub-frequency-domain signal CN (j)。
For example, assuming j=1, then F CN (1)＝sum(F c ((1-1)*32+1,1*32))＝sum(F c (1,32)). That is, the 1 st segment of the sub-frequency domain signal contains 1 st to 32 nd frequency domain signals, the spectral eigenvalues of the 1 st to 32 nd frequency domain signals are obtained, and then the sum of the 32 spectral eigenvalues is determined as the first spectral eigenvalue of the 1 st segment of the sub-frequency domain signal.
S112, processing the first spectrum characteristic value to obtain a corresponding second spectrum characteristic value; generating a corresponding coarse-granularity lattice feature map based on the second spectrum feature value;
After the first spectrum characteristic value is determined, a larger or smaller abnormal first spectrum characteristic value may exist, and in order to avoid the influence of the larger or smaller value on the training precision of the subsequent fault identification neural network, the invention continues to process the first spectrum characteristic value to obtain a corresponding second spectrum characteristic value.
In one embodiment, processing the first spectral feature value to obtain a corresponding second spectral feature value includes:
using the formulaProcessing the first spectrum characteristic value to obtain a second spectrum characteristic value F cn ' j) to eliminate the influence of the odd samples on the subsequent prediction precision; wherein,
the F is cn (j) For the first spectral feature value, j is the number of segments dividing the frequency domain signal into sub-frequency domain signals, F cnmin For the minimum frequency spectrum characteristic value in each segment of sub-frequency domain signal, F cnmax And for the maximum spectrum characteristic value in each segment of sub-frequency domain signal, S is the preset spectrum characteristic value quantity, and the value of S is consistent with the value of j.
The method is equivalent to removing the influence of the training process of the maximum first frequency spectrum characteristic value and the minimum first frequency spectrum characteristic value, and improving the training precision of the follow-up fault identification neural network.
After the second spectrum characteristic value is determined, a corresponding coarse-granularity lattice characteristic diagram is generated based on the second spectrum characteristic value, and the method specifically comprises the following steps:
Acquiring the number of segments for dividing the frequency domain signal into sub-frequency domain signals;
and drawing the coarse-granularity lattice feature map (CGLF, coarse Grained Lattice Feature) by taking the number of segments of the sub-frequency domain signal as an abscissa and the corresponding second frequency spectrum feature value as an ordinate.
Specifically, the spectrogram corresponding to the first vibration signal may be shown in fig. 2A, and the coarse-grain lattice feature map corresponding to the first vibration signal may be shown in fig. 2B.
When the bearing is in a fault type and is in an outer ring fault, a spectrogram corresponding to the second vibration signal is shown in fig. 3A, and a coarse grain lattice characteristic diagram corresponding to the second vibration signal is shown in fig. 3B.
When the bearing is in a fault type and is in an inner ring fault, a spectrogram corresponding to the second vibration signal is shown in fig. 4A, and a coarse grain lattice characteristic diagram corresponding to the second vibration signal is shown in fig. 4B.
When the bearing is in a fault type and is in a rolling body ring fault, a spectrogram corresponding to the second vibration signal is shown in fig. 5A, and a coarse-grain lattice characteristic diagram corresponding to the second vibration signal is shown in fig. 5B.
In this way, the one-dimensional vibration signal is converted into a two-dimensional coarse-grain lattice characteristic diagram, and as apparent from reference to fig. 2A to 5B, the fault characteristic can be remarkably enhanced, and meanwhile, the fault frequency information is reserved.
S113, training a pre-constructed fault identification neural network by using the coarse-granularity lattice feature map, and outputting a trained fault identification neural network model; the fault identification neural network is a Swin Transformer network;
after the coarse-granularity lattice feature patterns corresponding to the vibration sample signals are determined, training a pre-constructed fault recognition neural network by using the coarse-granularity lattice feature patterns, and outputting a trained fault recognition neural network model; the fault-recognition neural network is a Swin transducer network.
In one embodiment, training a pre-built failure recognition neural network using a coarse-grained lattice feature map, outputting a trained failure recognition neural network model, comprising:
adjusting the coarse-grain lattice feature map to a preset size, and dividing the coarse-grain lattice feature map after the size adjustment into a training set, a verification set and a data set according to a preset proportion;
acquiring preset training parameters; the training parameters include: learning rate, weight decay, iteration number, optimizer and loss function;
training the fault recognition neural network by using the training set and the training parameters;
and adjusting training parameters of the fault identification neural network by using the verification set, testing the trained fault identification neural network by using the test set, and outputting a trained fault identification neural network model if the accuracy of the trained fault identification neural network is determined to reach a preset accuracy threshold.
The verification set is mainly used for adjusting training parameters of the fault identification neural network, so that the phenomenon of over-fitting of the network is avoided, and the generalization capability of the fault identification neural network is improved. The test set is used to evaluate the predictive performance of the model.
Specifically, after each coarse-grain lattice feature map is determined, in order to improve the calculation efficiency, the coarse-grain lattice feature map may be adjusted to a preset size (generally 224×224), and then the lattice feature map after the size adjustment is divided into a training set, a verification set and a test set according to a ratio of 3:1:1.
Then setting training parameters, setting a learning rate to be 1e-3, setting weight attenuation to be the training parameters 1e-5, setting the iteration number to be 50, setting an optimizer to random gradient descent, and setting a loss function to be an angle difference entropy loss function.
Training the pre-constructed fault recognition neural network based on the training parameters and the training set, adjusting the training parameters of the fault recognition neural network by using the verification set in the training process, testing the trained fault recognition neural network by using the testing set when the preset iteration times are reached, and outputting a corresponding fault recognition neural network model if the recognition accuracy meets the requirements.
Wherein, as shown in fig. 6, after training the trained failure recognition neural network by using the training set, the recognition accuracy of the failure recognition neural network is shown as a mark 61 in fig. 6. After the training parameters of the trained failure recognition neural network are adjusted by using the verification set, the recognition accuracy of the failure recognition neural network is shown as a reference numeral 62 in fig. 6.
Similarly, after training the trained failure recognition neural network by using the training set, a loss value is output at the same time, as shown by a reference numeral 71 in fig. 7; after the training parameters of the trained failure recognition neural network are adjusted using the verification set, a loss value is also output, as indicated by reference numeral 72 in fig. 7.
As can be seen from fig. 6 and fig. 7, the recognition accuracy is higher, the loss value is lower, and the accuracy of the trained fault recognition neural network is higher.
In addition, the prediction accuracy of the trained fault recognition neural network can be visually displayed based on the confusion matrix. When the trained neural network is subjected to fault recognition by using the verification set, a confusion matrix corresponding to the recognition result is shown in fig. 8. Each row of the confusion matrix represents the actual category, each column represents the predicted category, and the values on the diagonal represent the duty cycle of the number of correct classifications. For example, with a status tag of 0 in fig. 8, the corresponding correct classification number has a duty ratio of 1, which illustrates that the prediction accuracy for the failure category with a status tag of 0 is 100%.
The fault recognition neural network in the present invention is a Swin Transformer network, as shown in fig. 9, which mainly includes: the system comprises an image block partitioning layer, a fault characteristic value extraction layer, a full connection layer and a fault characteristic classifier;
the Partition layer Patch Partition of the image block is mainly used for performing a cutting operation on the input two-dimensional coarse-granularity lattice feature map Images (h×w×3) to obtain 4×4 non-overlapping image blocks patches, wherein the feature dimension of each non-overlapping image block is 4×4×3=48, that is, H/4*W/4×48.
The fault characteristic value extraction layer comprises: including Stage1, stage2, stage3, and Stage4.
Stage1 includes Linear embedding and Swin Transformer Block; the linear embedded layer may be represented by C through a manually preset dimension.
Stage2, stage3 and Stage4 each comprise: tie layers patch bonding and Swin Transformer Block; the link layer is used for linking the features of adjacent patches of each group 2x 2, and is also used for reducing the resolution of the image block (the resolution is reduced by 0.5 times of the original resolution), and improving the number of feature channels (the number of feature channels is equal to 2) so as to realize a hierarchical structure. Therefore, after one stage, the width of the image block is reduced to 1/2 of the original width, the height is also reduced to 1/2 of the original width, and the dimension is changed to 2 times of the original dimension. For example, after one Stage, H/4 becomes H/8,W/4 becomes W/8 and the dimension changes from C to 2C.
Swin Transformer Block can refer to fig. 10, including:
window Multi-Head Self-Attention module (W-MSA, windows Multi-Head Self-Attention), one offset window Multi-Head Self-Attention module (SW-MSA, shifted Windows Multi-Head Self-Attention), two Multi-layer perceptrons (MLP, multilayer Perceptron), between which the non-linear capability of the whole network is enhanced by activating function GELU. A normalization layer (LN, layer Normalization) is provided before each of the W-MSA, SW-MSA and MLP, and a residual may be attached after each module to improve the fit of the overall network structure.
The window multi-head self-attention module W-MSA is mainly used for taking attention between every two patches in each window, and the SW-MSA module is mainly used for taking attention among image blocks of different window windows into consideration, so that local fault features are combined into global fault features according to the attention.
The full-connection layer is used for fitting the local fault characteristic value to the global fault characteristic value, and an output formula of the full-connection layer is as follows:
y s (x)＝f(w*x+b)
wherein x represents a local fault characteristic value, w is a weight, b is a bias, f () is an activation function, y j (x) And the global fault characteristic value of the s fault class.
After passing through the full connection layer, a plurality of global fault characteristic values y can be output s (x) To obtain each global fault characteristic value y s (x) The fault feature classifier needs to first use the following formula to determine y s (x) The mapping is carried out to (0, ++ infinity a) of the above-mentioned components, then normalized to (0, 1), the formula is as follows:
where s is the failure category and K is the total number of failure categories.
It can be seen that after the global fault feature value is converted by softmax, the characteristics of the probability expression are satisfied, that is, for any type of fault class, a corresponding fault probability is output finally. Therefore, the probability equivalent to different global fault characteristic values is expressed, and the specific fault type can be determined according to the fault probability.
Generally, the fault class corresponding to the greatest probability of fault is determined as the final fault type.
S114, determining a current coarse-grain lattice feature map corresponding to the bearing to be diagnosed, identifying the current coarse-grain lattice feature map by using the fault identification neural network model, and outputting a corresponding fault type.
After the fault identification neural network model is output, the fault type of the bearing to be diagnosed can be identified by using the model.
Specifically, the current coarse-grain lattice feature map corresponding to the bearing to be diagnosed needs to be determined, where the current coarse-grain lattice feature map may refer to the implementation manner of the coarse-grain lattice feature map of the vibration signal sample described above, and thus will not be described herein.
And then identifying the current coarse-grain lattice feature map by using the fault identification neural network model, wherein the fault identification neural network model comprises the following steps:
cutting the current coarse-granularity lattice feature map by utilizing an image block partitioning layer of the fault identification network model to obtain a plurality of non-overlapping image blocks;
performing dimension reduction processing on the plurality of non-overlapping image blocks by using the linear embedded layer to obtain corresponding characteristic image blocks;
performing feature extraction on the feature image blocks by using a fault feature value extraction layer of the fault identification network model to obtain local fault feature values;
fitting the local fault characteristic values by using a full connection layer of the fault identification network model to obtain a plurality of global fault characteristic values;
and determining the probability of each global fault characteristic value by using a fault characteristic classifier of the fault identification network model, and outputting the fault type according to the probability of the global fault characteristic value.
The fault-identifying neural network model has the same structure as the Swin Transformer network, and includes an image block partition layer, a linear embedding layer, a fault feature value extraction layer, a full connection layer, and a fault feature classifier, and specific execution logic of the image block partition layer, the linear embedding layer, the fault feature value extraction layer, the full connection layer, and the fault feature classifier may refer to corresponding descriptions in the step S113, so that details are not repeated herein.
For example, in practical applications, such as 224×224×3 of the input CGLF graph, as shown in fig. 11, the fault diagnosis logic is as follows:
the CGLF graph is divided into patches by Patch Partition, the size of each Patch is 4x4 (the pixel size of the image, i.e. height and width), the size of the image becomes 56 x 48, 48 is 4x 3, and 3 is the rgb channel of the image.
Then, the vector passes through Linear encoding in Stage1, so that the dimension of the vector is changed into a preset value C, the size of C is 96, and the obtained network output value is 56×56×96. Swin Transformer block in Stage1 is used to calculate a feature value, and the output is 56×56×96.
For Stage2, the Patch merge layer in Stage2 is used to reduce the resolution of the segmented image, and adjust and increase the number of feature channels. Therefore, the downsampling window size of the Patch Merging layer is set to be 2×2, after passing through the first Patch Merging in Stage2, the output image becomes 28×28×192, and after calculating the feature value through a Swin Transformer block module in Stage2, the output is still 28×28×192.
Stage3, stage4 repeats the same operation, and the image passes Stage3 to become 14×14×384; after Stage4, the image becomes 7×7×768.
The 7×7×768 image is then input to the fully connected layer, the number of neurons of the fully connected layer is 37632, and the output is 10 classifications.
Finally, fault classification is achieved through a softmax classifier.
The invention firstly carries out coarse granularity treatment on the vibration signal to obviously enhance the characteristics of the fault signal; after the first frequency spectrum characteristic value is processed, the accuracy of sample data can be further improved, and the influence of abnormal fault data is reduced; the Swin Transformer network also has good feature fusion and feature extraction capability, and can accurately complete fault classification and identification tasks, so that the fault diagnosis accuracy of the rolling bearing is improved.
Based on the same inventive concept as the previous embodiments, the present embodiments also provide a coarse-grained based marine turbocharger fault diagnosis apparatus, as shown in fig. 12, including:
an acquisition unit 121 for acquiring a vibration sample signal of a marine turbocharger bearing; the vibration sample signal comprises a first vibration signal of a normal bearing and a second vibration signal of a fault bearing; the first vibration signal and the second vibration signal are preset with corresponding state labels;
The processing unit 122 is configured to convert each of the vibration sample signals into a corresponding frequency domain signal, and perform coarse granularity processing on each of the frequency domain signals based on a preset coarse granularity sampling coefficient, so as to obtain a corresponding first spectrum characteristic value;
a generating unit 123, configured to process the first spectrum feature value to obtain a corresponding second spectrum feature value; generating a corresponding coarse-granularity lattice feature map based on the second spectrum feature value;
the training unit 124 is configured to train the pre-constructed fault identification neural network by using the coarse-granularity lattice feature map, and output a trained fault identification neural network model;
and the identifying unit 125 is configured to determine a current coarse-grain lattice feature map corresponding to the bearing to be diagnosed, identify the current coarse-grain lattice feature map by using the fault identification neural network model, and output a corresponding fault type.
Since the device described in the embodiment of the present invention is a device used for implementing the method for diagnosing the fault of the marine turbocharger based on coarse granularity in the embodiment of the present invention, based on the method described in the embodiment of the present invention, a person skilled in the art can know the specific structure and deformation of the device, and therefore, the detailed description thereof is omitted herein. All devices used in the method of the embodiment of the invention are within the scope of the invention.
Based on the same inventive concept, this embodiment provides a computer device 1000, as shown in fig. 13, including a memory 1010, a processor 1020, and a computer program 1011 stored on the memory 1010 and capable of running on the processor 1020, wherein the processor 1020 implements any one of the steps of the method described above when executing the computer program 1011.
Based on the same inventive concept, this embodiment provides a computer-readable storage medium 1100, as shown in fig. 14, on which a computer program 1111 is stored, which computer program 1111 implements the steps of any of the methods described above when being executed by a processor.
Through one or more embodiments of the present invention, the present invention has the following benefits or advantages:
the invention provides a coarse-grained based marine turbocharger fault diagnosis method, device, medium and equipment, wherein the method comprises the following steps: acquiring a vibration sample signal of a bearing; the vibration sample signal comprises a first vibration signal of a normal bearing and a second vibration signal of a fault bearing; the first vibration signal and the second vibration signal are preset with corresponding state labels; converting each vibration sample signal into a corresponding frequency domain signal, and performing coarse granularity processing on each frequency domain signal based on a preset coarse granularity sampling coefficient to obtain a corresponding first frequency spectrum characteristic value; processing the first spectrum characteristic value to obtain a corresponding second spectrum characteristic value; generating a corresponding coarse-granularity lattice feature map based on the second spectrum feature value; training a pre-constructed fault recognition neural network by using the coarse-granularity lattice feature diagram, and outputting a trained fault recognition neural network model; the fault identification neural network is a Swin Transformer network; determining a current coarse-grain lattice feature map corresponding to a bearing to be diagnosed, identifying the current coarse-grain lattice feature map by using the fault identification neural network model, and outputting a corresponding fault type; therefore, even if the bearing fault signal of the marine turbocharger is weak, the vibration signal of the bearing can be subjected to coarse granularity treatment, so that the characteristics of the fault signal are obviously enhanced; after the first frequency spectrum characteristic value is processed, the accuracy of sample data can be further improved, and the influence of abnormal fault data is reduced; the Swin Transformer network also has good feature fusion and feature extraction capability, and can accurately complete fault classification and identification tasks, so that the fault diagnosis accuracy of the rolling bearing is improved.
The algorithms and displays presented herein are not inherently related to any particular computer, virtual system, or other apparatus. Various general-purpose systems may also be used with the teachings herein. The required structure for a construction of such a system is apparent from the description above. In addition, the present invention is not directed to any particular programming language. It will be appreciated that the teachings of the present invention described herein may be implemented in a variety of programming languages, and the above description of specific languages is provided for disclosure of enablement and best mode of the present invention.
In the description provided herein, numerous specific details are set forth. However, it is understood that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures and techniques have not been shown in detail in order not to obscure an understanding of this description.
Similarly, it should be appreciated that in the foregoing description of exemplary embodiments of the invention, various features of the invention are sometimes grouped together in a single embodiment, figure, or description thereof for the purpose of streamlining the disclosure and aiding in the understanding of one or more of the various inventive aspects. However, the disclosed method should not be construed as reflecting the intention that: i.e., the claimed invention requires more features than are expressly recited in each claim. Rather, as the following claims reflect, inventive aspects lie in less than all features of a single foregoing disclosed embodiment. Thus, the claims following the detailed description are hereby expressly incorporated into this detailed description, with each claim standing on its own as a separate embodiment of this invention.
Those skilled in the art will appreciate that the modules in the apparatus of the embodiments may be adaptively changed and disposed in one or more apparatuses different from the embodiments. The modules or units or components of the embodiments may be combined into one module or unit or component and, furthermore, they may be divided into a plurality of sub-modules or sub-units or sub-components. Any combination of all features disclosed in this specification (including any accompanying claims, abstract and drawings), and all of the processes or units of any method or apparatus so disclosed, may be used in combination, except insofar as at least some of such features and/or processes or units are mutually exclusive. Each feature disclosed in this specification (including any accompanying claims, abstract and drawings), may be replaced by alternative features serving the same, equivalent or similar purpose, unless expressly stated otherwise.
Furthermore, those skilled in the art will appreciate that while some embodiments herein include some features but not others included in other embodiments, combinations of features of different embodiments are meant to be within the scope of the invention and form different embodiments. For example, in the following claims, any of the claimed embodiments can be used in any combination.
Various component embodiments of the invention may be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. Those skilled in the art will appreciate that some or all of the functions of some or all of the components in a gateway, proxy server, system according to embodiments of the present invention may be implemented in practice using a microprocessor or Digital Signal Processor (DSP). The present invention can also be implemented as an apparatus or device program (e.g., a computer program and a computer program product) for performing a portion or all of the methods described herein. Such a program embodying the present invention may be stored on a computer readable medium, or may have the form of one or more signals. Such signals may be downloaded from an internet website, provided on a carrier signal, or provided in any other form.
It should be noted that the above-mentioned embodiments illustrate rather than limit the invention, and that those skilled in the art will be able to design alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps not listed in a claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention may be implemented by means of hardware comprising several distinct elements, and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by one and the same item of hardware. The use of the words first, second, third, etc. do not denote any order. These words may be interpreted as names.
While preferred embodiments of the present application have been described, additional variations and modifications in those embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. It is therefore intended that the following claims be interpreted as including the preferred embodiments and all such alterations and modifications as fall within the scope of the application.
The above description is not intended to limit the scope of the application, but is intended to cover any modifications, equivalents, and improvements within the spirit and principles of the application.
Claims (10)
1. A coarse-grained based marine turbocharger fault diagnosis method, characterized in that the method comprises:
obtaining a vibration sample signal of a marine turbocharger bearing; the vibration sample signal comprises a first vibration signal of a normal bearing and a second vibration signal of a fault bearing; the first vibration signal and the second vibration signal are preset with corresponding state labels;
converting each vibration sample signal into a corresponding frequency domain signal, and performing coarse granularity processing on each frequency domain signal based on a preset coarse granularity sampling coefficient to obtain a corresponding first frequency spectrum characteristic value;
Processing the first spectrum characteristic value to obtain a corresponding second spectrum characteristic value; generating a corresponding coarse-granularity lattice feature map based on the second spectrum feature value;
training a pre-constructed fault recognition neural network by using the coarse-granularity lattice feature diagram, and outputting a trained fault recognition neural network model; the fault identification neural network is a Swin Transformer network;
and determining a current coarse-grain lattice characteristic diagram corresponding to the bearing to be diagnosed, identifying the current coarse-grain lattice characteristic diagram by using the fault identification neural network model, and outputting a corresponding fault type.
2. The method of claim 1, wherein the coarsening each of the frequency domain signals based on a preset coarse-granularity sampling coefficient to obtain a corresponding first spectral feature value comprises:
for any one of the frequency domain signals, dividing the frequency domain signal into a plurality of segments of sub-frequency domain signals using the formula j=fix (r/d) based on the coarse-granularity sampling coefficients;
according to formula F CN (j)＝sum(F c ((j-1) d+1, j d)) determining a first spectral feature value F of said each segment of sub-frequency domain signal CN (j) The method comprises the steps of carrying out a first treatment on the surface of the Wherein,
the j is the number of segments of the sub-frequency domain signal, d is the coarse-granularity sampling coefficient, fix is a rounding function, r is the length of the frequency domain signal, F C () Is the amplitude of the spectral signal.
3. The method of claim 1, wherein said processing the first spectral feature values to obtain corresponding second spectral feature values comprises:
using the formulaProcessing the first spectrum characteristic value to obtain the second spectrum characteristic value F cn ' (j); wherein,
the F is cn (j) For the first spectral feature value, j is the number of segments dividing the frequency domain signal into sub-frequency domain signals, F cnmin For the minimum spectral feature value in each segment of the sub-frequency domain signal, the F cnmax And for the maximum spectrum characteristic value in each segment of sub-frequency domain signal, S is the preset number of spectrum characteristic values, and the value of S is consistent with the value of j.
4. The method of claim 1, wherein the generating a corresponding coarse-grained lattice feature map based on the second spectral feature values comprises:
acquiring the number of segments for dividing the frequency domain signal into sub-frequency domain signals;
and drawing the coarse-granularity lattice characteristic diagram by taking the number of segments of the sub-frequency domain signal as an abscissa and the corresponding second frequency spectrum characteristic value as an ordinate.
5. The method of claim 1, wherein training a pre-built fault-recognition neural network using the coarse-grained lattice feature map, outputting a trained fault-recognition neural network model, comprises:
Adjusting the coarse-grain lattice feature map to a preset size, and dividing the size-adjusted coarse-grain lattice feature map into a training set, a verification set and a data set according to a preset proportion;
acquiring preset training parameters; the training parameters include: learning rate, weight decay, iteration number, optimizer and loss function;
training the fault recognition neural network by using the training set and the training parameters, and adjusting the training parameters by using a verification set in the training process;
and testing the trained fault recognition neural network by using the test set, and outputting a trained fault recognition neural network model if the accuracy of the trained fault recognition neural network reaches a preset accuracy threshold.
6. The method of claim 1, wherein said identifying the current coarse-grained lattice feature map using the fault-recognition neural network model, outputting a corresponding fault type, comprises:
cutting the current coarse-granularity lattice feature map by utilizing an image block partitioning layer of the fault identification network model to obtain a plurality of non-overlapping image blocks;
performing feature extraction on a plurality of non-overlapping image blocks by using a fault feature value extraction layer of the fault identification network model to obtain local fault feature values;
Fitting the local fault characteristic values by using a full connection layer of the fault identification network model to obtain a plurality of global fault characteristic values;
and determining the probability of each global fault characteristic value by using a fault characteristic classifier of the fault identification network model, and outputting the fault type according to the probability of the global fault characteristic value.
7. A coarse-grained-based marine turbocharger fault diagnosis device, characterized in that the device comprises:
the acquisition unit is used for acquiring a vibration sample signal of the marine turbocharger bearing; the vibration sample signal comprises a first vibration signal of a normal bearing and a second vibration signal of a fault bearing; the first vibration signal and the second vibration signal are preset with corresponding state labels;
the processing unit is used for converting each vibration sample signal into a corresponding frequency domain signal, and performing coarse granularity processing on each frequency domain signal based on a preset coarse granularity sampling coefficient to obtain a corresponding first frequency spectrum characteristic value;
the generating unit is used for processing the first frequency spectrum characteristic value to obtain a corresponding second frequency spectrum characteristic value; generating a corresponding coarse-granularity lattice feature map based on the second spectrum feature value;
The training unit is used for training the pre-constructed fault recognition neural network by utilizing the coarse-granularity lattice feature diagram and outputting a trained fault recognition neural network model;
and the identification unit is used for determining a current coarse grain lattice characteristic diagram corresponding to the bearing to be diagnosed, identifying the current coarse grain lattice characteristic diagram by utilizing the fault identification neural network model and outputting a corresponding fault type.
8. The apparatus of claim 7, wherein the processing unit is specifically configured to:
for any frequency domain signal, dividing the frequency domain signal into a plurality of sections of sub-frequency domain signals by utilizing the coarse granularity sampling coefficient;
acquiring spectrum characteristic values of a plurality of frequency domain signals contained in each segment of sub-frequency domain signal;
a sum of spectral eigenvalues of a plurality of frequency domain signals is determined as the first spectral eigenvalue.
9. A computer readable storage medium, on which a computer program is stored, characterized in that the program, when being executed by a processor, implements the steps of the method according to any one of claims 1-7.
10. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the steps of the method according to any of claims 1-7 when the program is executed by the processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311193524.2A CN117232846A (en) | 2023-09-15 | 2023-09-15 | Marine turbocharger fault diagnosis method, device and equipment based on coarse granularity |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311193524.2A CN117232846A (en) | 2023-09-15 | 2023-09-15 | Marine turbocharger fault diagnosis method, device and equipment based on coarse granularity |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117232846A true CN117232846A (en) | 2023-12-15 |
Family
ID=89094236
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311193524.2A Pending CN117232846A (en) | 2023-09-15 | 2023-09-15 | Marine turbocharger fault diagnosis method, device and equipment based on coarse granularity |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117232846A (en) |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108647604A (en) * | 2018-04-28 | 2018-10-12 | 宋浏阳 | Object state identification method based on the extraction of frequency domain histogram feature |
| CN110031227A (en) * | 2019-05-23 | 2019-07-19 | 桂林电子科技大学 | A kind of Rolling Bearing Status diagnostic method based on binary channels convolutional neural networks |
| CN112084974A (en) * | 2020-09-14 | 2020-12-15 | 北京工商大学 | Multi-label rolling bearing fault diagnosis method based on meta-learning |
| CN112883905A (en) * | 2021-03-15 | 2021-06-01 | 福州大学 | Human behavior recognition method based on coarse-grained time-frequency features and multi-level fusion learning |
| CN113191240A (en) * | 2021-04-23 | 2021-07-30 | 石家庄铁道大学 | Multi-task deep neural network method and device for bearing fault diagnosis |
| CN114169377A (en) * | 2021-12-17 | 2022-03-11 | 郑州滕瑟电子科技有限公司 | G-MSCNN-based fault diagnosis method for rolling bearing in noisy environment |
| CN115112372A (en) * | 2022-06-22 | 2022-09-27 | 哈尔滨工业大学（深圳） | Bearing fault diagnosis method and device, electronic equipment and storage medium |
| CN116610993A (en) * | 2023-04-26 | 2023-08-18 | 山东师范大学 | Bearing fault diagnosis method, system, equipment and storage medium |
| CN116625654A (en) * | 2023-04-17 | 2023-08-22 | 中国兵器装备集团自动化研究所有限公司 | Vibration signal fault diagnosis method, device, equipment and storage medium |
-
2023
- 2023-09-15 CN CN202311193524.2A patent/CN117232846A/en active Pending
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108647604A (en) * | 2018-04-28 | 2018-10-12 | 宋浏阳 | Object state identification method based on the extraction of frequency domain histogram feature |
| CN110031227A (en) * | 2019-05-23 | 2019-07-19 | 桂林电子科技大学 | A kind of Rolling Bearing Status diagnostic method based on binary channels convolutional neural networks |
| CN112084974A (en) * | 2020-09-14 | 2020-12-15 | 北京工商大学 | Multi-label rolling bearing fault diagnosis method based on meta-learning |
| CN112883905A (en) * | 2021-03-15 | 2021-06-01 | 福州大学 | Human behavior recognition method based on coarse-grained time-frequency features and multi-level fusion learning |
| CN113191240A (en) * | 2021-04-23 | 2021-07-30 | 石家庄铁道大学 | Multi-task deep neural network method and device for bearing fault diagnosis |
| CN114169377A (en) * | 2021-12-17 | 2022-03-11 | 郑州滕瑟电子科技有限公司 | G-MSCNN-based fault diagnosis method for rolling bearing in noisy environment |
| CN115112372A (en) * | 2022-06-22 | 2022-09-27 | 哈尔滨工业大学（深圳） | Bearing fault diagnosis method and device, electronic equipment and storage medium |
| CN116625654A (en) * | 2023-04-17 | 2023-08-22 | 中国兵器装备集团自动化研究所有限公司 | Vibration signal fault diagnosis method, device, equipment and storage medium |
| CN116610993A (en) * | 2023-04-26 | 2023-08-18 | 山东师范大学 | Bearing fault diagnosis method, system, equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wang et al. | Virtualization and deep recognition for system fault classification | |
| Shao et al. | A deep learning approach for fault diagnosis of induction motors in manufacturing | |
| Pan et al. | A novel deep learning network via multiscale inner product with locally connected feature extraction for intelligent fault detection | |
| Wang et al. | Understanding and learning discriminant features based on multiattention 1DCNN for wheelset bearing fault diagnosis | |
| Souza et al. | Deep learning for diagnosis and classification of faults in industrial rotating machinery | |
| Hasan et al. | A fault diagnosis framework for centrifugal pumps by scalogram-based imaging and deep learning | |
| Gao et al. | ASM1D-GAN: An intelligent fault diagnosis method based on assembled 1D convolutional neural network and generative adversarial networks | |
| Wang et al. | A combination of residual and long–short-term memory networks for bearing fault diagnosis based on time-series model analysis | |
| US20230351177A1 (en) | Enhanced discriminate feature learning deep residual cnn for multi-task rotating machinery fault diagnosis with information fusion | |
| Wei et al. | A method of underwater acoustic signal classification based on deep neural network | |
| Qian et al. | A comparative study of deep learning-based fault diagnosis methods for rotating machines in nuclear power plants | |
| CN115114965B (en) | Wind turbine generator gearbox fault diagnosis method, device, equipment and storage medium | |
| CN115392301A (en) | Converter transformer state identification method, converter transformer state identification device, converter transformer state identification equipment, converter transformer state identification medium and program product | |
| Tyagi et al. | An SVM—ANN hybrid classifier for diagnosis of gear fault | |
| Chen et al. | Intelligent fault diagnosis of rolling bearings with low-quality data: A feature significance and diversity learning method | |
| CN113530850B (en) | Centrifugal pump fault diagnosis method based on ESA and stacked capsule self-encoder | |
| Chao et al. | Extraction of local and global features by a convolutional neural network–long short-term memory network for diagnosing bearing faults | |
| Xu et al. | An improved hybrid CNN-SVM based method for bearing fault diagnosis under noisy environment | |
| CN113758709A (en) | Rolling bearing fault diagnosis method and system combining edge calculation and deep learning | |
| CN115290326A (en) | Rolling bearing fault intelligent diagnosis method | |
| CN112541524B (en) | BP-Adaboost multisource information motor fault diagnosis method based on attention mechanism improvement | |
| CN117232846A (en) | Marine turbocharger fault diagnosis method, device and equipment based on coarse granularity | |
| CN116227556A (en) | Method, device, computer equipment and storage medium for acquiring target network model | |
| CN116842379A (en) | Mechanical bearing residual service life prediction method based on DRSN-CS and BiGRU+MLP models | |
| CN115601649A (en) | TransUnnet-based ocean internal wave stripe segmentation method and equipment and storage medium |
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 |