CN113221996A - Bearing fault diagnosis method and system based on multi-head attention mechanism - Google Patents
Bearing fault diagnosis method and system based on multi-head attention mechanism Download PDFInfo
- Publication number
- CN113221996A CN113221996A CN202110489147.1A CN202110489147A CN113221996A CN 113221996 A CN113221996 A CN 113221996A CN 202110489147 A CN202110489147 A CN 202110489147A CN 113221996 A CN113221996 A CN 113221996A
- Authority
- CN
- China
- Prior art keywords
- bearing fault
- fault diagnosis
- head attention
- signals
- bearing
- 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
- 238000003745 diagnosis Methods 0.000 title claims abstract description 76
- 238000000034 method Methods 0.000 title claims abstract description 39
- 230000007246 mechanism Effects 0.000 title claims abstract description 25
- 230000001133 acceleration Effects 0.000 claims abstract description 32
- 230000009466 transformation Effects 0.000 claims abstract description 32
- 230000006870 function Effects 0.000 claims description 27
- 238000012549 training Methods 0.000 claims description 20
- 239000011159 matrix material Substances 0.000 claims description 19
- 230000015654 memory Effects 0.000 claims description 9
- 238000003860 storage Methods 0.000 claims description 6
- 238000010586 diagram Methods 0.000 description 17
- 238000010606 normalization Methods 0.000 description 14
- 238000004590 computer program Methods 0.000 description 7
- 238000013528 artificial neural network Methods 0.000 description 6
- 238000013135 deep learning Methods 0.000 description 6
- 239000013598 vector Substances 0.000 description 6
- 238000004422 calculation algorithm Methods 0.000 description 4
- 238000013527 convolutional neural network Methods 0.000 description 4
- 125000004122 cyclic group Chemical group 0.000 description 4
- 238000000354 decomposition reaction Methods 0.000 description 4
- 238000009826 distribution Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 238000013519 translation Methods 0.000 description 3
- 230000008034 disappearance Effects 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000004880 explosion Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000005096 rolling process Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000005284 excitation Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000006698 induction Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 238000007637 random forest analysis Methods 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M13/00—Testing of machine parts
- G01M13/04—Bearings
- G01M13/045—Acoustic or vibration analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
- G06F2218/04—Denoising
- G06F2218/06—Denoising by applying a scale-space analysis, e.g. using wavelet analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Acoustics & Sound (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The bearing fault diagnosis method and system based on the multi-head attention mechanism disclosed by the disclosure comprise the following steps: acquiring a vibration acceleration signal of a bearing; wavelet packet transformation is carried out on the vibration acceleration signals to obtain signals of different frequency bands; grouping the signals of different frequency bands to obtain a plurality of groups of frequency band signals; inputting a plurality of groups of frequency band signals into a trained bearing fault diagnosis model to obtain a bearing fault diagnosis result; the bearing fault diagnosis model comprises a plurality of parallel multi-head attention networks, a plurality of groups of frequency band signals are respectively input into the corresponding multi-head attention networks, and bearing fault judgment is carried out after the output of the multi-head attention networks is connected through a full connection layer. The bearing fault can be accurately diagnosed.
Description
Technical Field
The invention relates to the technical field of fault diagnosis, in particular to a bearing fault diagnosis method and system based on a multi-head attention mechanism.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
The bearing is the most important component in the rotary machine, the main function of the bearing is to support the mechanical rotating body and reduce the friction coefficient in the movement process, but the bearing is one of the vulnerable parts in the rotary machine due to high rotating speed, complex structure and overload operation of the bearing in the working process, and according to various researches, the bearing fault is the main source of the fault of the rotary machine, so that an effective state detection and fault diagnosis method has very important significance, which is one of the most challenging tasks in fault diagnosis.
In the conventional data-driven fault diagnosis method, a large number of statistical parameters need to be extracted and input into a machine learning algorithm, such as a K-means algorithm, a random forest, a naive bayes model, a support vector machine and a multi-layer perceptron. In order to obtain accurate diagnostic performance, the distribution of these high-dimensional statistical parameters is distinguishable for each condition (i.e., category) to be considered in the diagnosis. However, if the distribution of data is not a condition that can be distinguished, it is difficult to perform highly accurate failure diagnosis. Due to the complex structure, long transmission path, variable working conditions and strong background noise, these distributions are easily overlapped, resulting in low diagnostic accuracy.
In recent years, various fault diagnosis methods based on deep learning are popular, and features in an original signal are extracted continuously by a feature extractor which can be used in the deep learning, low-level features are basic details of the original signal, high-level features are more abstract, and the high-level features can be better used for classification or prediction. Deep learning the representation learning of the multi-layer nonlinear transformation can solve the limitation of the traditional shallow machine learning algorithm.
Currently, the widely applied Deep learning algorithm includes a Deep Auto Encoder (DAN), a Convolutional Neural Network (CNN), a Recurrent Neural Network (RNN), and a Long Short-Term Memory Network (LSTM). The deep auto-encoder can perform unsupervised pre-training on the weights, which can reduce the difficulty of subsequent supervised training on the deep network, but when the input is the original vibration signal or the time-frequency representation thereof, the parameters needing training are very many, and a large amount of computing resources are consumed. The convolutional neural network can reduce the number of parameters needing training by using a local receptive field and a weight distribution strategy for optimization, so that the computational burden in the training process is effectively reduced, but the translation invariance and the existence of a pooling layer of the convolutional neural network can cause loss of much valuable spatial information, and meanwhile, the association between the local part and the whole part can be ignored. The cyclic neural network is easy to generate gradient explosion and gradient disappearance in training, so that the gradient cannot be transmitted in a long sequence all the time during training, the cyclic neural network cannot capture the influence of a long distance, although the problems of gradient disappearance and gradient explosion of the cyclic neural network are solved to a certain extent by a long-time memory network, the internal structure of the cyclic neural network is complex, if the time span is large and the network is also deep, the calculated amount is large, a large amount of training time is needed, and the deep neural networks are easy to be influenced by non-sensitive adjustment, so that the accuracy of bearing fault diagnosis is influenced.
Disclosure of Invention
In order to solve the problems, the disclosure provides a bearing fault diagnosis method and system based on a multi-head attention mechanism, and accurate diagnosis of bearing faults is achieved.
In order to achieve the purpose, the following technical scheme is adopted in the disclosure:
in a first aspect, a bearing fault diagnosis method based on a multi-head attention mechanism is provided, including:
acquiring a vibration acceleration signal of a bearing;
wavelet packet transformation is carried out on the vibration acceleration signals to obtain signals of different frequency bands;
grouping the signals of different frequency bands to obtain a plurality of groups of frequency band signals;
inputting a plurality of groups of frequency band signals into a trained bearing fault diagnosis model to obtain a bearing fault diagnosis result;
the bearing fault diagnosis model comprises a plurality of parallel multi-head attention networks, a plurality of groups of frequency band signals are respectively input into the corresponding multi-head attention networks, and bearing fault judgment is carried out after the output of the multi-head attention networks is connected through a full connection layer.
In a second aspect, a bearing fault diagnosis system based on a multi-head attention mechanism is provided, including:
the signal acquisition module is used for acquiring a vibration acceleration signal of the bearing;
the wavelet packet transformation module is used for carrying out wavelet packet transformation on the vibration acceleration signals to obtain signals of different frequency bands;
the signal grouping module is used for grouping the signals of different frequency bands to obtain a plurality of groups of frequency band signals;
the bearing fault diagnosis module is used for inputting a plurality of groups of frequency band signals into a trained bearing fault diagnosis model to obtain a bearing fault diagnosis result;
the bearing fault diagnosis model comprises a plurality of parallel multi-head attention networks, a plurality of groups of frequency band signals are respectively input into the corresponding multi-head attention networks, and bearing fault judgment is carried out after the output of the multi-head attention networks is connected through a full connection layer.
In a third aspect, an electronic device is provided, which includes a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein the computer instructions, when executed by the processor, perform the steps of the multi-attention-based bearing fault diagnosis method.
In a fourth aspect, a computer-readable storage medium is provided for storing computer instructions, which when executed by a processor, perform the steps of a multi-attention-based bearing fault diagnosis method.
Compared with the prior art, the beneficial effect of this disclosure is:
1. the vibration acceleration signals of the bearing are input into a bearing fault diagnosis model formed by a plurality of multi-head attention networks in parallel after wavelet packet conversion for fault diagnosis, and the information of different frequency bands of the signals is subjected to feature extraction by using a plurality of independent multi-head attention networks, so that the relation between the features of different fault types in a high-dimensional space and the information contained in different frequency bands of the original signals can be effectively established, and the accuracy of fault diagnosis is improved.
2. Wavelet packet transformation used by the method better analyzes high-frequency and non-stationary vibration acceleration signals; the multi-head attention mechanism can realize weighted expression of different characteristics, so that the classified characteristics have higher expression capability, and meanwhile, the parameters of the network to be trained are much less than those of other deep learning methods, so that the training speed is increased, and the speed of fault diagnosis is increased on the basis of increasing the accuracy of fault diagnosis.
3. The method for extracting the features of the information of the signals in different frequency bands by using the multiple independent multi-head attention networks not only improves the fault diagnosis precision, but also can well analyze the relationship between the features of different fault types in a high-dimensional space and the information contained in the original signals in different frequency bands. Meanwhile, a plurality of networks are processed in parallel, so that a large amount of computing time is saved.
Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this application, illustrate embodiments of the application and, together with the description, serve to explain the application and are not intended to limit the application.
FIG. 1 is a bearing fault diagnosis model disclosed in embodiment 1 of the present disclosure;
fig. 2 is a structure diagram of a multi-head attention network disclosed in embodiment 1 of the present disclosure;
FIG. 3 is a structure diagram of a multi-head attention layer disclosed in embodiment 1 of the present disclosure;
FIG. 4 is a block diagram of a test stand according to embodiment 1 of the present disclosure;
fig. 5 is a diagram of an original waveform of a vibration acceleration signal involved in embodiment 1 of the present disclosure;
fig. 6 is a waveform diagram of each node of the fifth stage after wavelet packet transformation of the vibration acceleration signal according to embodiment 1 of the present disclosure.
The specific implementation mode is as follows:
the present disclosure is further described with reference to the following drawings and examples.
It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the disclosure. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments according to the present application. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.
In the present disclosure, terms such as "upper", "lower", "left", "right", "front", "rear", "vertical", "horizontal", "side", "bottom", and the like indicate orientations or positional relationships based on those shown in the drawings, and are only relational terms determined for convenience in describing structural relationships of the parts or elements of the present disclosure, and do not refer to any parts or elements of the present disclosure, and are not to be construed as limiting the present disclosure.
In the present disclosure, terms such as "fixedly connected", "connected", and the like are to be understood in a broad sense, and mean either a fixed connection or an integrally connected or detachable connection; may be directly connected or indirectly connected through an intermediate. The specific meanings of the above terms in the present disclosure can be determined on a case-by-case basis by persons skilled in the relevant art or technicians, and are not to be construed as limitations of the present disclosure.
Example 1
In order to realize accurate diagnosis of the bearing fault, the embodiment discloses a bearing fault diagnosis method based on a multi-head attention mechanism, which comprises the following steps:
acquiring a vibration acceleration signal of a bearing;
wavelet packet transformation is carried out on the vibration acceleration signals to obtain signals of different frequency bands;
grouping the signals of different frequency bands to obtain a plurality of groups of frequency band signals;
inputting a plurality of groups of frequency band signals into a trained bearing fault diagnosis model to obtain a bearing fault diagnosis result;
the bearing fault diagnosis model comprises a plurality of parallel multi-head attention networks, a plurality of groups of frequency band signals are respectively input into the corresponding multi-head attention networks, and bearing fault judgment is carried out after the output of the multi-head attention networks is connected through a full connection layer.
Further, the full connection layer is connected with a Softmax function, and the output of the full connection layer is calculated through the Softmax function to obtain the type of the bearing fault.
Further, the obtained signals of different frequency bands are arranged according to the frequency size, and the arranged signals are grouped to obtain a plurality of groups of frequency band signals.
Further, the multi-head attention network includes a plurality of identical layers consisting of a multi-head attention layer and a feedforward connection layer.
Furthermore, the multi-head attention layer distributes different weights to the input, carries out matrix transformation on the data obtained after the different weights are distributed, inputs the data after the matrix transformation into a plurality of parallel attention layers, and carries out matrix transformation on the output of each parallel attention layer again to obtain the output of the multi-head attention layer.
Further, the feedforward connecting layer comprises two connected Linear layers.
Further, acquiring an existing bearing vibration acceleration signal;
classifying the obtained existing bearing vibration acceleration signals according to load and fault types;
segmenting each type of data, and extracting a plurality of data points from each segment of data;
constructing a label corresponding to the fault type for each data point to obtain a training sample;
and training the bearing fault diagnosis model through the training sample to obtain the trained bearing fault diagnosis model.
A detailed description will be given of a bearing fault diagnosis method based on a multi-head attention mechanism disclosed in this embodiment.
A bearing fault diagnosis method based on a multi-head attention mechanism comprises the following steps:
s1: and acquiring a vibration acceleration signal of the bearing.
S2: and carrying out wavelet packet transformation on the vibration acceleration signals to obtain signals of different frequency bands.
In specific implementation, the vibration acceleration signal is subjected to wavelet packet transformation, and the higher the wavelet packet decomposition level, the higher the resolution of the signal in the time domain and the frequency domain, and the more information is contained. In the embodiment, the Meyer wavelet function is used for carrying out wavelet packet transformation on the vibration acceleration signal obtained in the step S1, and the signal decomposed to the k layer is reconstructed to obtain 2kData of different frequency bands, 2kThe vector of data of different frequency bands is divided into 2 from low frequency to high frequencykGroup/n, give 2kN × dmodelThe matrix of (c) is selected from k 5, n 4, dmodel=1024。
In the binary wavelet packet transformation, the scale function of the adjacent levels of the wavelet packets at each level and the wavelet function have a recurrence relation, and the father wavelet in the wavelet packet transformation is recorded asMother wavelet isWherein the superscript represents the waveletThe decomposition level of the packet, and the subscript indicates the position of the wavelet packet in the level, the recurrence relation can be expressed as:
namely:
In the discrete wavelet transform, when the discrete wavelet transform is decomposed stage by stage, the space formed by the scale function is as follows:
wherein ViSpace, s, spanned by scale function of i-th orderiIs a scale variable, k is a translation variable,is a normalization factor, phi (x) is the generationParent wavelets of a family of scale functions.
The space spanned by the wavelet function is:
wherein WiDeveloped for wavelet function of i-th levelSpace, siIs a scale variable, k is a translation variable,is a normalization factor, psi (x) is generatedA mother wavelet of a family of wavelet functions.
WiIs ViAbout Vi+1The two have a relationship:from this relationship, recursive expansion can be achieved, resulting in:
as can be seen from the above equation, the higher the number of decomposition stages, the higher the resolution of the signal in the time domain and the frequency domain, and the more information it contains. In this embodiment, a Meyer wavelet function is used to sequentially perform wavelet packet transformation on the obtained vibration acceleration signal, and reconstruct the signal decomposed to the fifth layer to obtain 32 data of different frequency bands.
S3: grouping different frequency band signals to obtain a plurality of groups of frequency band signals, specifically: and arranging the obtained signals of different frequency bands according to the frequency size, and grouping the arranged signals to obtain a plurality of groups of frequency band signals.
In specific implementation, a vector composed of 32 data of different frequency bands obtained after S2 wavelet packet transformation is divided into 8 groups from low frequency to high frequency, and 8 4 × 1024 matrices are obtained.
S4: and inputting the multiple groups of frequency band signals into the trained bearing fault diagnosis model to obtain a bearing fault diagnosis result.
The bearing fault diagnosis model comprises a plurality of parallel multi-head attention networks, a plurality of groups of frequency band signals are respectively input into the corresponding multi-head attention networks, data output by the multi-head attention networks are flattened, flattened row vectors are input into a full connection layer, the output of the full connection layer is calculated by using a Softmax function to obtain the prediction probability of each fault type, and the maximum value of the probability is taken as the predicted fault type of the bearing.
The expression of the Softmax function is:
wherein: p (y ═ j) denotes the probability of the j-th output value, xjRepresenting the jth input value.
The multi-headed attention network includes a plurality of identical layers of a multi-headed attention layer and a feedforward connection layer.
The specific structure of the multi-head attention network is shown in fig. 2, and comprises a first normalization layer and a plurality of identical layers consisting of a multi-head attention layer and a feedforward connection layer.
S41: inputting the vibration acceleration signal into a first normalization layer for normalization, wherein the function expression of normalization is as follows:
wherein: σ is the standard deviation of the input matrix X,average value of input matrix X, Xi,jIs an element of the ith row and the jth column of the input matrix, Yi,jIs the element of the ith row and the jth column of the output matrix.
S42: and inputting the data output by the first normalization layer into the multi-head attention layer, adding the data output by the multi-head attention layer and the data output by the first normalization layer to form a first residual connecting block, and inputting the data after residual connection into the second normalization layer for normalization.
S43: inputting the second normalization layer into the feedforward connection layer, adding the processed data of the feedforward connection layer and the data output by the second normalization layer to formA second residual connecting block for inputting the data output by the second residual connecting block into a third normalization layer for normalization, wherein the feedforward connecting layer comprises two Linear layers, and the input dimension of the first layer is dmodelOutput dimension dff2048, input dimension d of the second layerff2048, output dimension dmodelAnd a ReLU excitation function is used between the two Linear layers, and the function expression of the feedforward connecting layer is as follows:
FFN(x)=max(0,xW1+b1)W2+b2 (9)
wherein: w1Weight matrix of the first Linear layer, b1Is the offset vector of the first Linear layer, W2Weight matrix for the first Linear layer, b2Offset vector of the first Linear layer.
And repeating the steps of S42 and S43 for three times, and outputting the data, wherein the output data is output by a multi-head attention network.
The multi-head attention layer assigns different weights to the input, performs matrix transformation on the data obtained after assigning different weights, inputs the data after matrix transformation into a plurality of parallel attention layers, and performs matrix transformation again on the output of each parallel attention layer to obtain the output of the multi-head attention layer, as shown in fig. 3, specifically:
weight matrix Q that differentiates the input of multi-head attention layer from three initializationsW、KWAnd VWObtaining three matrixes of Q (Query), K (Key) and V (value) by respective multiplication, wherein: qW、KWAnd VWAll dimensions are dmodel×dmodelQ, K and V dimensions are both n x dmodel. Dividing the obtained Q, K and V matrixes into n h x d matrixeskIn this embodiment, n is 4, dmodel=1024,h=8,dk=dmodelH 128, the n h x dkCombined to be one nxh x dkAnd the first and second dimensions are swapped to obtain hxnxdkIs divided into h n × d tensorskMatrices, respectively Qi、KiAnd ViWherein i is 1,2,…,h。
Then the h Q are puti、KiAnd ViThe input is input into h parallel attention layers, and the output of each parallel layer is calculated through an attention function. The attention function used was:
n x d of the output of the h parallel attention layerskThe matrix combination is h x n x dkAnd interchanging the first and second dimensions to obtain an nxhxdkAnd combining the second and third dimensions to obtain an nxdmodelAnd then the matrix passes through a Linear layer to obtain the output of the multi-head attention layer. The mathematical expression of the multi-head attention layer is as follows:
MultiHead(Q,K,V)=Contat(head1,…,headi,…,headh)WO (7)
Training the constructed bearing fault diagnosis model, specifically comprising the following steps:
the experiment table shown in fig. 4 is adopted, and the bearing experiment table mainly comprises a motor, a torque sensor, a fan end bearing, a motor driving end bearing and a dynamometer. Vibration data of rolling bearings at the motor driving end and the fan end are obtained by an acceleration sensor arranged on an induction motor shell, and sampling frequencies are 12kHz and 48 kHz. The method comprises the steps of classifying data of a bearing fault vibration data set of the university of Keiss Xizhi according to different fault types and loads, specifically classifying the data into inner ring faults, outer ring faults, rolling body faults and normal states, and taking motor drive end bearing data with the sampling frequency of 12kHz as a training set. Dividing the _ DE _ time data in the mat file in the data set into 100 sections, taking 1024 data points in each section, discarding redundant data, and constructing fault labels (normal data-0, inner ring fault-1, outer ring fault-2 and roller body fault-3) corresponding to each section of data to obtain training samples for model training.
And sequentially carrying out wavelet packet transformation on each segment by using a Meyer wavelet function, reconstructing the signals decomposed to the fifth layer to obtain 32 data of different frequency segments, and dividing vectors formed by the 32 data of different frequency segments into 8 groups from low frequency to high frequency to obtain 8 matrixes of 4 multiplied by 1024. Fig. 5 is a waveform diagram of an original vibration acceleration signal. Fig. 6 is a waveform diagram reconstructed by nodes of a fifth stage after wavelet packet transformation is performed on an original vibration acceleration signal, 32 waveform diagrams obtained by decomposition are divided into 8 groups from low frequency to high frequency, wherein the 8 groups are A, B, C, D, E, F, G and H respectively, and each group comprises 4 waveform diagrams, namely (a), (b), (c) and (d) respectively.
And inputting the obtained 8 4 multiplied by 1024 matrixes into the constructed bearing fault diagnosis model to train the model, so as to obtain the trained bearing fault diagnosis model.
And inputting the multiple groups of frequency band signals into the trained bearing fault diagnosis model to obtain a bearing fault diagnosis result.
Wavelet packet transformation used by the method better analyzes high-frequency and non-stationary vibration acceleration signals; the multi-head attention mechanism can realize the weighted expression of different characteristics, so that the classified characteristics have higher expression capability, and meanwhile, the parameters of the network to be trained are much less compared with other deep learning methods, so that the training speed is improved; the method for extracting the characteristics of the information of different frequency bands of the signal by using the multiple independent multi-head attention networks not only improves the fault diagnosis precision, but also can well analyze the relationship between the characteristics of different fault types in a high-dimensional space and the information contained in different frequency bands of the original signal. Meanwhile, a plurality of networks are processed in parallel, so that a large amount of computing time is saved. Compared with the prior art, the method not only can achieve higher accuracy, but also can better analyze the relationship between different fault types and different frequency band signals.
Example 2
In this embodiment, a bearing fault diagnosis system based on a multi-head attention mechanism is disclosed, including:
the signal acquisition module is used for acquiring a vibration acceleration signal of the bearing;
the wavelet packet transformation module is used for carrying out wavelet packet transformation on the vibration acceleration signals to obtain signals of different frequency bands;
the signal grouping module is used for grouping the signals of different frequency bands to obtain a plurality of groups of frequency band signals;
the bearing fault diagnosis module is used for inputting a plurality of groups of frequency band signals into a trained bearing fault diagnosis model to obtain a bearing fault diagnosis result;
the bearing fault diagnosis model comprises a plurality of parallel multi-head attention networks, a plurality of groups of frequency band signals are respectively input into the corresponding multi-head attention networks, and bearing fault judgment is carried out after the output of the multi-head attention networks is connected through a full connection layer.
Example 3
In this embodiment, an electronic device is disclosed, which comprises a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein the computer instructions, when executed by the processor, perform the steps of the multi-attention mechanism-based bearing fault diagnosis method disclosed in embodiment 1.
Example 4
In this embodiment, a computer readable storage medium is disclosed for storing computer instructions which, when executed by a processor, perform the steps described in the method for diagnosing a bearing fault based on a multi-attention mechanism disclosed in embodiment 1.
The above description is only a preferred embodiment of the present application and is not intended to limit the present application, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, improvement and the like made within the spirit and principle of the present application shall be included in the protection scope of the present application.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solutions of the present invention and not for limiting the same, and although the present invention is described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: modifications and equivalents may be made to the embodiments of the invention without departing from the spirit and scope of the invention, which is to be covered by the claims.
Claims (10)
1. A bearing fault diagnosis method based on a multi-head attention mechanism is characterized by comprising the following steps:
acquiring a vibration acceleration signal of a bearing;
wavelet packet transformation is carried out on the vibration acceleration signals to obtain signals of different frequency bands;
grouping the signals of different frequency bands to obtain a plurality of groups of frequency band signals;
inputting a plurality of groups of frequency band signals into a trained bearing fault diagnosis model to obtain a bearing fault diagnosis result;
the bearing fault diagnosis model comprises a plurality of parallel multi-head attention networks, a plurality of groups of frequency band signals are respectively input into the corresponding multi-head attention networks, and bearing fault judgment is carried out after the output of the multi-head attention networks is connected through a full connection layer.
2. The multi-attention mechanism-based bearing fault diagnosis method according to claim 1, wherein the fully-connected layer is connected with a Softmax function, and the type of the bearing fault is obtained by calculating the output of the fully-connected layer through the Softmax function.
3. The multi-head attention mechanism-based bearing fault diagnosis method according to claim 1, wherein the obtained signals of different frequency bands are arranged according to frequency magnitude, and the arranged signals are grouped to obtain a plurality of groups of frequency band signals.
4. The multi-head attention mechanism-based bearing fault diagnosis method according to claim 1, wherein the multi-head attention network comprises a plurality of identical layers consisting of a multi-head attention layer and a feedforward connection layer.
5. The method for diagnosing the bearing fault based on the multi-head attention mechanism as claimed in claim 4, wherein the multi-head attention layer assigns different weights to the input, performs matrix transformation on the data obtained after assigning different weights, inputs the data obtained after the matrix transformation into a plurality of parallel attention layers, and performs matrix transformation again on the output of each parallel attention layer to obtain the output of the multi-head attention layer.
6. The multi-head attention mechanism-based bearing fault diagnosis method according to claim 4, wherein the feed-forward connection layer comprises two connected Linear layers.
7. The multi-attention mechanism-based bearing fault diagnosis method according to claim 1, wherein an existing bearing vibration acceleration signal is acquired;
classifying the obtained existing bearing vibration acceleration signals according to load and fault types;
segmenting each type of data, and extracting a plurality of data points from each segment of data;
constructing a label corresponding to the fault type for each data point to obtain a training sample;
and training the bearing fault diagnosis model through the training sample to obtain the trained bearing fault diagnosis model.
8. A bearing fault diagnostic system based on a multi-head attention mechanism, comprising:
the signal acquisition module is used for acquiring a vibration acceleration signal of the bearing;
the wavelet packet transformation module is used for carrying out wavelet packet transformation on the vibration acceleration signals to obtain signals of different frequency bands;
the signal grouping module is used for grouping the signals of different frequency bands to obtain a plurality of groups of frequency band signals;
the bearing fault diagnosis module is used for inputting a plurality of groups of frequency band signals into a trained bearing fault diagnosis model to obtain a bearing fault diagnosis result;
the bearing fault diagnosis model comprises a plurality of parallel multi-head attention networks, a plurality of groups of frequency band signals are respectively input into the corresponding multi-head attention networks, and bearing fault judgment is carried out after the output of the multi-head attention networks is connected through a full connection layer.
9. An electronic device comprising a memory and a processor and computer instructions stored on the memory and executed on the processor, wherein the computer instructions, when executed by the processor, perform the steps of the method for bearing fault diagnosis based on the multi-attention mechanism as claimed in any one of claims 1 to 7.
10. A computer readable storage medium storing computer instructions which, when executed by a processor, perform the steps of the method for multi-attention mechanism based bearing fault diagnosis according to any one of claims 1 to 7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110489147.1A CN113221996A (en) | 2021-04-30 | 2021-04-30 | Bearing fault diagnosis method and system based on multi-head attention mechanism |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110489147.1A CN113221996A (en) | 2021-04-30 | 2021-04-30 | Bearing fault diagnosis method and system based on multi-head attention mechanism |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113221996A true CN113221996A (en) | 2021-08-06 |
Family
ID=77090934
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110489147.1A Pending CN113221996A (en) | 2021-04-30 | 2021-04-30 | Bearing fault diagnosis method and system based on multi-head attention mechanism |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113221996A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114818835A (en) * | 2022-06-27 | 2022-07-29 | 山东大学 | Rotary machine fault diagnosis method and system based on difference feature extraction and fusion |
CN114993677A (en) * | 2022-05-11 | 2022-09-02 | 山东大学 | Rolling bearing fault diagnosis method and system based on unbalanced small sample data |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
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 |
WO2019179340A1 (en) * | 2018-03-19 | 2019-09-26 | 河北工业大学 | Eemd- and msb-based failure feature extraction method for rolling-element bearing |
CN110672343A (en) * | 2019-09-29 | 2020-01-10 | 电子科技大学 | Rotary machine fault diagnosis method based on multi-attention convolutional neural network |
CN110795843A (en) * | 2019-10-24 | 2020-02-14 | 北京建筑大学 | Method and device for identifying faults of rolling bearing |
CN111275007A (en) * | 2020-02-24 | 2020-06-12 | 山东大学 | Bearing fault diagnosis method and system based on multi-scale information fusion |
CN111860410A (en) * | 2020-07-29 | 2020-10-30 | 南京邮电大学 | Myoelectric gesture recognition method based on multi-feature fusion CNN |
CN112244873A (en) * | 2020-09-29 | 2021-01-22 | 陕西科技大学 | Electroencephalogram time-space feature learning and emotion classification method based on hybrid neural network |
CN112268615A (en) * | 2020-11-27 | 2021-01-26 | 华侨大学 | Electromechanical equipment vibration signal feature extraction method |
CN112629854A (en) * | 2020-11-25 | 2021-04-09 | 西安交通大学 | Bearing fault classification method based on neural network attention mechanism |
-
2021
- 2021-04-30 CN CN202110489147.1A patent/CN113221996A/en active Pending
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2019179340A1 (en) * | 2018-03-19 | 2019-09-26 | 河北工业大学 | Eemd- and msb-based failure feature extraction method for rolling-element bearing |
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 |
CN110672343A (en) * | 2019-09-29 | 2020-01-10 | 电子科技大学 | Rotary machine fault diagnosis method based on multi-attention convolutional neural network |
CN110795843A (en) * | 2019-10-24 | 2020-02-14 | 北京建筑大学 | Method and device for identifying faults of rolling bearing |
CN111275007A (en) * | 2020-02-24 | 2020-06-12 | 山东大学 | Bearing fault diagnosis method and system based on multi-scale information fusion |
CN111860410A (en) * | 2020-07-29 | 2020-10-30 | 南京邮电大学 | Myoelectric gesture recognition method based on multi-feature fusion CNN |
CN112244873A (en) * | 2020-09-29 | 2021-01-22 | 陕西科技大学 | Electroencephalogram time-space feature learning and emotion classification method based on hybrid neural network |
CN112629854A (en) * | 2020-11-25 | 2021-04-09 | 西安交通大学 | Bearing fault classification method based on neural network attention mechanism |
CN112268615A (en) * | 2020-11-27 | 2021-01-26 | 华侨大学 | Electromechanical equipment vibration signal feature extraction method |
Non-Patent Citations (3)
Title |
---|
ASHISH VASWANI,NOAM SHAZEER,NIKI PARMAR等: "Attention Is All You Need", 《NEURAL INFORMATION PROCESSING SYSTEMS (NIPS 2017)》 * |
刘立,朱健成,韩光洁,远国: "基于1D.CNN联合特征提取的轴承健康监测与故障诊断", 《软件学报》 * |
黄二二: "小波包变换的学习笔记", 《HTTPS://ZHUANLAN.ZHIHU.COM/P/58596902》 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114993677A (en) * | 2022-05-11 | 2022-09-02 | 山东大学 | Rolling bearing fault diagnosis method and system based on unbalanced small sample data |
CN114818835A (en) * | 2022-06-27 | 2022-07-29 | 山东大学 | Rotary machine fault diagnosis method and system based on difference feature extraction and fusion |
CN114818835B (en) * | 2022-06-27 | 2022-09-23 | 山东大学 | Rotary machine fault diagnosis method and system based on difference feature extraction and fusion |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Shao et al. | Highly accurate machine fault diagnosis using deep transfer learning | |
Chen et al. | Intelligent fault diagnosis for rotary machinery using transferable convolutional neural network | |
Pan et al. | A novel deep learning network via multiscale inner product with locally connected feature extraction for intelligent fault detection | |
Tao et al. | Bearing fault diagnosis method based on stacked autoencoder and softmax regression | |
Li et al. | Unsupervised rotating machinery fault diagnosis method based on integrated SAE–DBN and a binary processor | |
Dixit et al. | Intelligent condition-based monitoring of rotary machines with few samples | |
Li et al. | WavCapsNet: An interpretable intelligent compound fault diagnosis method by backward tracking | |
Wu et al. | An expert system for fault diagnosis in internal combustion engines using probability neural network | |
Li et al. | A domain adaptation model for early gear pitting fault diagnosis based on deep transfer learning network | |
Kaya et al. | Classification of bearing vibration speeds under 1D-LBP based on eight local directional filters | |
Kaya et al. | A new automatic bearing fault size diagnosis using time-frequency images of CWT and deep transfer learning methods | |
CN111523081B (en) | Aeroengine fault diagnosis method based on enhanced gate control circulating neural network | |
Gong et al. | Implementation of machine learning for fault classification on vehicle power transmission system | |
He et al. | Wind turbine planetary gearbox feature extraction and fault diagnosis using a deep-learning-based approach | |
CN113221996A (en) | Bearing fault diagnosis method and system based on multi-head attention mechanism | |
Huang et al. | Deep residual networks-based intelligent fault diagnosis method of planetary gearboxes in cloud environments | |
Zhang et al. | Intelligent machine fault diagnosis using convolutional neural networks and transfer learning | |
Han et al. | Data-enhanced stacked autoencoders for insufficient fault classification of machinery and its understanding via visualization | |
CN115290326A (en) | Rolling bearing fault intelligent diagnosis method | |
Cao et al. | Fault diagnosis of rolling bearing based on multiscale one-dimensional hybrid binary pattern | |
Zhiwei | Bearing Fault Diagnosis of End‐to‐End Model Design Based on 1DCNN‐GRU Network | |
Saufi et al. | Machinery fault diagnosis based on a modified hybrid deep sparse autoencoder using a raw vibration time-series signal | |
Kibrete et al. | Applications of artificial intelligence for fault diagnosis of rotating machines: A review | |
CN117113169A (en) | Industrial flow type data on-line fault diagnosis method based on deep knowledge distillation network | |
Ma et al. | Prediction of remaining useful life of rolling bearings based on multiscale efficient channel attention CNN and bidirectional GRU |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210806 |