CN116659856A - Fault diagnosis method of motor bearing based on WPT-1DCNN - Google Patents
Fault diagnosis method of motor bearing based on WPT-1DCNN Download PDFInfo
- Publication number
- CN116659856A CN116659856A CN202310480764.4A CN202310480764A CN116659856A CN 116659856 A CN116659856 A CN 116659856A CN 202310480764 A CN202310480764 A CN 202310480764A CN 116659856 A CN116659856 A CN 116659856A
- Authority
- CN
- China
- Prior art keywords
- motor bearing
- fault diagnosis
- detected
- vibration signal
- layer
- 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 169
- 238000000034 method Methods 0.000 title claims abstract description 79
- 239000013598 vector Substances 0.000 claims abstract description 83
- 238000006243 chemical reaction Methods 0.000 claims abstract description 13
- 238000012549 training Methods 0.000 claims description 98
- 238000012360 testing method Methods 0.000 claims description 31
- 238000011176 pooling Methods 0.000 claims description 29
- 238000003860 storage Methods 0.000 claims description 23
- 230000007246 mechanism Effects 0.000 claims description 17
- 238000000354 decomposition reaction Methods 0.000 claims description 15
- 230000004927 fusion Effects 0.000 claims description 12
- 239000004973 liquid crystal related substance Substances 0.000 claims description 10
- 230000008569 process Effects 0.000 claims description 10
- 238000000605 extraction Methods 0.000 claims description 8
- 238000005457 optimization Methods 0.000 claims description 6
- 230000009466 transformation Effects 0.000 claims description 3
- 230000009977 dual effect Effects 0.000 claims 1
- 230000006870 function Effects 0.000 description 29
- 238000010586 diagram Methods 0.000 description 8
- 239000011159 matrix material Substances 0.000 description 8
- 238000004590 computer program Methods 0.000 description 6
- 230000008859 change Effects 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 230000007423 decrease Effects 0.000 description 2
- 238000010892 electric spark Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 230000001133 acceleration Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000010219 correlation analysis Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000003111 delayed effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000003754 machining Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000002035 prolonged effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
Classifications
-
- 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
-
- 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
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention relates to the technical field of bearing fault diagnosis, in particular to a fault diagnosis method of a motor bearing based on WPT-1DCNN, and aims to solve the problem that the accuracy of the existing bearing fault diagnosis method on bearing fault diagnosis is low. For this purpose, the fault diagnosis method of the motor bearing based on the WPT-1DCNN comprises the following steps: acquiring a vibration signal of a motor bearing to be detected; performing WPT conversion on a vibration signal of a motor bearing to be detected, determining that a multi-layer wavelet packet is decomposed, and extracting a frequency domain feature vector of the vibration signal of the motor bearing to be detected; and inputting the vibration signals of the motor bearing to be detected and the frequency domain feature vectors of the vibration signals of the motor bearing to be detected into a pre-trained double-branch fault diagnosis model, and outputting the fault classification result of the vibration signals of the motor bearing to be detected by the double-branch fault diagnosis model. The diagnosis method can further improve the accuracy of bearing faults and has good generalization.
Description
Technical Field
The invention relates to the technical field of bearing fault diagnosis, and particularly provides a fault diagnosis method and device, a storage medium and a control device for a motor bearing based on WPT-1 DCNN.
Background
With the increase of the automation level, the performance requirements on mechanical equipment are also increased. Bearings are an integral part of mechanical equipment, and as a result of their prolonged operation in harsh natural environments, are subject to various uncertainties, with the result that bearings are highly likely to fail at a certain time. Once the fault occurs, the construction period is delayed and economic loss is caused to a certain extent, and the casualties are extremely likely to be caused when the fault occurs seriously. The fault diagnosis technology of the mechanical equipment is receiving more and more attention and research. If the judgment can be timely made and fed back to related staff for processing, the loss can be reduced to the greatest extent.
When the bearing breaks down, vibration signals containing a large amount of impact noise are generated, and the generated vibration signals have obvious time-varying characteristics, so that obvious nonlinear behaviors are shown, and if the fault characteristics can be accurately extracted, the fault identification of the bearing can be effectively carried out. The existing fault diagnosis method of the motor bearing comprises the steps of extracting time domain features or extracting frequency domain features, and then carrying out fault diagnosis on the extracted time domain features or frequency domain features by utilizing a convolutional neural network.
Accordingly, there is a need in the art for a new motor bearing failure diagnosis solution to address the above-described problems.
Disclosure of Invention
In order to overcome the defects, the invention provides a fault diagnosis method and device, a storage medium and a control device for a motor bearing based on WPT-1DCNN, which are used for solving or at least partially solving the technical problems of low accuracy, low generalization and the like of the prior bearing fault diagnosis method for bearing fault diagnosis.
In a first aspect, the present invention provides a fault diagnosis method for a WPT-1 DCNN-based motor bearing, the method comprising:
acquiring a vibration signal of a motor bearing to be detected;
performing WPT conversion on the vibration signal of the motor bearing to be detected, determining the decomposition of the multi-layer wavelet packet, and extracting the frequency domain feature vector of the vibration signal of the motor bearing to be detected;
and inputting the vibration signals of the motor bearing to be detected and the frequency domain feature vectors of the vibration signals of the motor bearing to be detected into a pre-trained double-branch fault diagnosis model, and outputting a fault classification result of the vibration signals of the motor bearing to be detected by the double-branch fault diagnosis model.
In one technical scheme of the fault diagnosis method for the motor bearing based on the WPT-1DCNN, the method comprises the steps of performing WPT conversion on the vibration signal of the motor bearing to be detected, determining the decomposition of a multi-layer wavelet packet, and extracting the frequency domain feature vector of the vibration signal of the motor bearing to be detected, wherein the method comprises the following steps:
Decomposing a 3-layer wavelet packet of a vibration signal of a motor bearing to be detected to obtain decomposed 8 sub-frequency bands;
determining a frequency domain characteristic vector H of a vibration signal of the motor bearing to be detected in 8*1 dimensions according to the product of the maximum amplitude of each of the 8 sub-bands and the frequency corresponding to the maximum amplitude of each of the 8 sub-bands k The method comprises the following steps:
wherein A is k Maximum amplitude in each sub-band; f (f) k A frequency corresponding to the maximum amplitude in each sub-band; k=1, 2,3,4,5,6,7,8.
In one technical scheme of the above fault diagnosis method for a motor bearing based on WPT-1DCNN, the frequency domain feature vectors of the vibration signal of the motor bearing to be detected and the vibration signal of the motor bearing to be detected are input into a pre-trained dual-branch fault diagnosis model, and the dual-branch fault diagnosis model outputs a fault classification result of the vibration signal of the motor bearing to be detected, including:
inputting the vibration signal of the motor bearing to be detected into a 1DCNN convolution layer of a double-branch fault diagnosis model, wherein the 1DCNN convolution layer outputs a feature vector of 128 x t dimensions;
inputting the feature vector of the 128 x t dimension into a multi-head self-attention mechanism layer of a double-branch fault diagnosis model, wherein the multi-head self-attention mechanism layer outputs a time domain feature vector K of a vibration signal of a motor bearing to be detected of the 128 x 1 dimension;
Frequency domain characteristic vector H of vibration signal of 8*1-dimensional motor bearing to be detected k The time domain feature vector K of the vibration signal of the motor bearing to be detected with the dimension of 128 x 1 is input into a double-branch feature fusion layer of a double-branch fault diagnosis model, and the double-branch feature fusion layer outputs the time-frequency domain feature vector of the vibration signal of the motor bearing to be detected with the dimension of 136 x 1;
the method comprises the steps of inputting a time-frequency domain feature vector of a vibration signal of a 136 x 1-dimensional motor bearing to be detected into a full-connection layer of a double-branch fault diagnosis model, and outputting a fault diagnosis signal of the vibration signal of the 4*1-dimensional motor bearing to be detected by the full-connection layer;
and inputting a fault diagnosis signal of the vibration signal of the 4*1-dimensional motor bearing to be detected into a Softmax layer of the double-branch fault diagnosis model, and outputting a fault classification result of the vibration signal of the motor bearing to be detected by the Softmax layer.
In one technical scheme of the fault diagnosis method of the motor bearing based on the WPT-1DCNN, the method further comprises the following steps:
and training the double-branch fault diagnosis model to obtain a pre-trained double-branch fault diagnosis model.
In one technical scheme of the fault diagnosis method for the motor bearing based on the WPT-1DCNN, the training the double-branch fault diagnosis model to obtain a pre-trained double-branch fault diagnosis model comprises the following steps:
Acquiring a data set of an original vibration signal of a motor bearing, and dividing the data set into a training set and a testing set according to a certain proportion;
training the double-branch fault diagnosis model by the training set, wherein the training process comprises the following steps: performing WPT conversion on the original vibration signals of the motor bearings in the training set, determining the decomposition of the multi-layer wavelet packet, and extracting frequency domain feature vectors of the original vibration signals of the motor bearings in the training set; inputting the frequency domain feature vector of the original vibration signal of the motor bearing in the training set and the original vibration signal of the motor bearing in the training set into a double-branch fault diagnosis model under training, and outputting a fault classification result of the original vibration signal of the motor bearing in the training set by the double-branch fault diagnosis model;
comparing the fault classification result of the original vibration signals of the motor bearings in the training set with the fault labels of the original vibration signals of the motor bearings in the training set, and calculating the loss function of the double-branch fault diagnosis model;
when the loss function reaches the requirement of meeting the model optimization, stopping training the double-branch fault diagnosis model;
and testing the trained fault diagnosis model by the test set, and obtaining the trained double-branch fault diagnosis model after the test is completed.
In one technical scheme of the fault diagnosis method of the motor bearing based on the WPT-1DCNN, the loss function is a cross entropy loss function.
In one technical scheme of the fault diagnosis method of the motor bearing based on the WPT-1DCNN, the 1DCNN convolution layer in the trained double-branch fault diagnosis model comprises an input layer, a first convolution layer, a first pooling layer, a second convolution layer and a second pooling layer; wherein, the liquid crystal display device comprises a liquid crystal display device,
the first convolution layer is set to be 1 input channel, 64 output channels, the step length is 1, and the edge filling is set to be 4 layers; the first pooling layer is the maximum time pooling layer; the second convolution layer is set to 64 input channels, 128 output channels, the step size is 1, and the edge filling is set to 2 layers; the second pooling layer is an average time pooling layer.
In a second aspect, the present invention provides a fault diagnosis apparatus for a WPT-1 DCNN-based motor bearing, the apparatus comprising:
the acquisition module is used for acquiring a vibration signal of the motor bearing to be detected;
the frequency domain feature extraction module is used for carrying out WPT conversion on the vibration signal of the motor bearing to be detected, determining the decomposition of the multi-layer wavelet packet and extracting the frequency domain feature vector of the vibration signal of the motor bearing to be detected;
The diagnosis module is used for inputting the vibration signals of the motor bearing to be detected and the frequency domain feature vectors of the vibration signals of the motor bearing to be detected into a pre-trained double-branch fault diagnosis model, and the double-branch fault diagnosis model outputs fault classification results of the vibration signals of the motor bearing to be detected.
In a third aspect, a control device is provided, the control device includes a processor and a storage device, the storage device is adapted to store a plurality of program codes, the program codes are adapted to be loaded and executed by the processor to perform the method according to any one of the above solutions of the fault diagnosis method of the WPT-1 DCNN-based motor bearing.
In a fourth aspect, a computer readable storage medium is provided, in which a plurality of program codes are stored, the program codes being adapted to be loaded and executed by a processor to perform the method according to any one of the above solutions for fault diagnosis methods of WPT-1 DCNN-based motor bearings.
The technical scheme provided by the invention has at least one or more of the following beneficial effects:
in the technical scheme of implementing the invention, a fault diagnosis method of a motor bearing based on WPT-1DCNN is provided, and the method comprises the following steps: acquiring a vibration signal of a motor bearing to be detected; performing WPT conversion on the vibration signal of the motor bearing to be detected, determining the decomposition of the multi-layer wavelet packet, and extracting the frequency domain feature vector of the vibration signal of the motor bearing to be detected; and inputting the vibration signals of the motor bearing to be detected and the frequency domain feature vectors of the vibration signals of the motor bearing to be detected into a pre-trained double-branch fault diagnosis model, and outputting a fault classification result of the vibration signals of the motor bearing to be detected by the double-branch fault diagnosis model. According to the prediction method, the frequency domain feature vector and the time domain feature vector fused double-branch fault diagnosis model is adopted, so that the accuracy of bearing fault diagnosis can be effectively improved, and the prediction method has good generalization.
Drawings
The present disclosure will become more readily understood with reference to the accompanying drawings. As will be readily appreciated by those skilled in the art: the drawings are for illustrative purposes only and are not intended to limit the scope of the present invention. Moreover, like numerals in the figures are used to designate like parts, wherein:
FIG. 1 is a flow chart of the main steps of a fault diagnosis method for a WPT-1 DCNN-based motor bearing according to one embodiment of the invention;
FIG. 2 is a schematic diagram of the main steps of step S102 according to one embodiment of the invention;
FIG. 3 is a schematic diagram of the main steps of step S103 according to one embodiment of the invention;
FIG. 4 is a schematic diagram of the structure of 1 DCNN;
FIG. 5 is a schematic diagram of the self-attention mechanism layer;
FIG. 6 is a schematic view of a CWRU bearing test stand;
FIG. 7 is a schematic diagram of the accuracy change curves of training and test sets according to one embodiment of the invention;
FIG. 8 is a graphical illustration of a change in the loss function of a training set and a test set according to one embodiment of the invention;
fig. 9 is a main structural diagram of a control device according to an embodiment of the present invention.
Detailed Description
Some embodiments of the invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are merely for explaining the technical principles of the present invention, and are not intended to limit the scope of the present invention.
In the description of the present invention, a "module," "processor" may include hardware, software, or a combination of both. A module may comprise hardware circuitry, various suitable sensors, communication ports, memory, or software components, such as program code, or a combination of software and hardware. The processor may be a central processor, a microprocessor, an image processor, a digital signal processor, or any other suitable processor. The processor has data and/or signal processing functions. The processor may be implemented in software, hardware, or a combination of both. Non-transitory computer readable storage media include any suitable medium that can store program code, such as magnetic disks, hard disks, optical disks, flash memory, read-only memory, random access memory, and the like. The term "a and/or B" means all possible combinations of a and B, such as a alone, B alone or a and B. The term "at least one A or B" or "at least one of A and B" has a meaning similar to "A and/or B" and may include A alone, B alone or A and B. The singular forms "a", "an" and "the" include plural referents.
Referring to fig. 1, fig. 1 is a schematic flow chart of main steps of a fault diagnosis method of a motor bearing based on WPT-1DCNN according to one embodiment of the present invention. As shown in fig. 1, the fault diagnosis method for the motor bearing based on WPT-1DCNN in the embodiment of the present invention mainly includes the following steps S101 to S103.
Step S101: and obtaining a vibration signal of the motor bearing to be detected.
In this embodiment, under the actual working condition, the vibration sensor is disposed at a detection point of the motor bearing to be detected, and when the vibration sensor receives an instruction for acquiring the vibration signal of the motor bearing, the vibration sensor installed at the driving end of the motor is used for acquiring the one-dimensional vibration signal of the motor bearing to be detected.
Step S102: and carrying out WPT conversion on the vibration signal of the motor bearing to be detected, determining the decomposition of the multi-layer wavelet packet, and extracting the frequency domain feature vector of the vibration signal of the motor bearing to be detected.
In this embodiment, the principle of WPT transformation is to decompose a signal into multiple layers of wavelet packets, each layer of wavelet packet containing different frequency components, so that the characteristics of the signal can be better analyzed.
In order to better represent the frequency domain feature vector of the vibration signal of the motor bearing to be detected, through verification of a large amount of experimental data, the embodiment of the invention selects to decompose the vibration signal of the motor bearing to be detected by a 3-layer wavelet packet, and 8 sub-frequency bands after the vibration signal of the motor bearing to be detected is decomposed are obtained.
In one implementation of the embodiment of the present invention, as shown in fig. 2, the step S102 further includes steps S1021 to S1022:
step S1021: and decomposing the 3-layer wavelet packet of the vibration signal of the motor bearing to be detected to obtain decomposed 8 sub-frequency bands.
Step S1022: determining a frequency domain characteristic vector H of a vibration signal of the motor bearing to be detected in 8*1 dimensions according to the product of the maximum amplitude of each of the 8 sub-bands and the frequency corresponding to the maximum amplitude of each of the 8 sub-bands k The method comprises the following steps:
wherein A is k Maximum amplitude in each sub-band; f (f) k A frequency corresponding to the maximum amplitude in each sub-band; k=1, 2,3,4,5,6,7,8.
In this embodiment, in order to remove redundant information and noise interference, a product of the maximum amplitude value in each sub-band and its corresponding frequency is selected and used to characterize the most significant feature in each sub-band. The feature selection not only greatly saves the computing resources, but also can obtain the feature with stronger robustness, and is expressed by the following formula:
wherein H is k Representing the maximum amplitude A in a certain subband k And a frequency f corresponding thereto k The multiplied features, where the subscript k denotes different frequency bands, in this embodiment k=1, 2,3,4,5,6,7,8, are obtained by calculating H in each frequency band k And combine them together to collectively represent a 8*1-dimensional frequency domain feature vector,and the final frequency domain characteristic vector of the vibration signal of the motor bearing to be detected is obtained.
Step S103: and inputting the vibration signals of the motor bearing to be detected and the frequency domain feature vectors of the vibration signals of the motor bearing to be detected into a pre-trained double-branch fault diagnosis model, and outputting a fault classification result of the vibration signals of the motor bearing to be detected by the double-branch fault diagnosis model.
In one implementation of the embodiment of the present invention, as shown in fig. 3, step S103 may further include steps S1031 to S1035:
step S1031: and inputting the vibration signal of the motor bearing to be detected into a 1DCNN convolution layer of a double-branch fault diagnosis model, and outputting a feature vector of 128 x t dimensions by the 1DCNN convolution layer.
In one specific example, the dual-branch fault diagnosis model includes a 1DCNN convolution layer, a multi-headed self-attention mechanism layer, a dual-branch feature fusion layer, a fully-connected layer, and a Softmax layer. The structure of the 1DCNN convolutional layer is shown in fig. 4, the structure of the 1DCNN convolutional layer is a one-dimensional convolutional neural network, the 1DCNN convolutional layer comprises an input layer and an implicit layer, the implicit layer comprises one or more one-dimensional convolutional Layers (Convolutional Layers) and one or more Pooling Layers (Pooling Layers), a full connection layer and an output layer, one-dimensional vibration signals of a motor bearing to be detected are input into the input layer, after the one-dimensional convolutional Layers and the Pooling Layers are processed, time domain feature extraction of the one-dimensional vibration signals of the motor bearing to be detected is output, the 1DCNN convolutional layer can automatically perform feature extraction through training, and the specific number of the one-dimensional convolutional Layers and the Pooling Layers included in the implicit layer is determined, so that the time domain feature extraction effect is the best.
Thus, after the one-dimensional vibration signal of the motor bearing to be detected is input to the time domain branch formed by the 1DCNN convolution layer, the one-dimensional vibration signal X of the motor bearing to be detected in Extracting time domain features, and processing the features X of the extracted data by a one-dimensional convolution layer and a pooling layer 1DCNN The method comprises the following steps:
wherein, the liquid crystal display device comprises a liquid crystal display device,indicate->The layer corresponds to the weight of the node; b is the bias.
Step S1032: and inputting the feature vector of the 128 x t dimension into a multi-head self-attention mechanism layer of the double-branch fault diagnosis model, wherein the multi-head self-attention mechanism layer outputs a time domain feature vector K of a vibration signal of the motor bearing to be detected of the 128 x 1 dimension.
In one specific example, the self-attention mechanism layer is an attention mechanism that associates different positions of a single sequence, and calculates the response of each position in the sequence by focusing on all positions in the same sequence. The essence of the method is that important information of a target object is highlighted through a series of attention weight coefficients, and irrelevant information is restrained at the same time, so that global information can be associated, and the relation between the global information and local information can be captured efficiently. The schematic diagram of the self-attention mechanism is shown in fig. 5, and as can be seen from fig. 5, the basic steps of the self-attention mechanism are as follows:
N sequences of single samplesPerforming adaptive linear mapping, and converting into 3 vectors with length d>；
VectorParallel connection is carried out, and a query matrix Q, an index matrix K and a content matrix V are respectively synthesized; calculating the product of Q and K to obtain the attention weight +.>The calculation formula is as follows, S is a weight set among N sequences;
normalizing S to obtainThe formula is as follows:
converting weights into probability forms using a Softmax function. The formula is as follows:
obtaining a weighting matrix Z:
the entire self-attention process can be summarized as the formula:
in this embodiment, after extracting the time domain feature of the one-dimensional vibration signal of the motor bearing to be detected, the feature X output by the 1DCNN convolution layer 1DCNN Inputting the vibration signal to a multi-head self-attention mechanism layer, carrying out correlation analysis of characteristics, and outputting a final time domain characteristic vector K of the vibration signal of the motor bearing to be detected in 128 x 1 dimensions, wherein the time domain characteristic vector K is as follows:
wherein, the liquid crystal display device comprises a liquid crystal display device,n is the length of the input sequence; />，/>，/>Features X respectively representing outputs of 1DCNN convolution layers 1DCNN A corresponding query matrix, an index matrix, and a content matrix.
Step S1033: frequency domain characteristic vector H of vibration signal of 8*1-dimensional motor bearing to be detected k And the time domain feature vector K of the vibration signal of the motor bearing to be detected with the dimension of 128 x 1 is input into a double-branch feature fusion layer of the double-branch fault diagnosis model, and the double-branch feature fusion layer outputs the time-frequency domain feature vector of the vibration signal of the motor bearing to be detected with the dimension of 136 x 1.
In a specific example, the frequency domain feature vector and the time domain feature vector are fused to obtain a 136X 1-dimensional time-frequency domain feature vector X out Is that
Wherein a is m The mth vector of 136 x 1-dimensional time-frequency domain feature vectors, m=1, 2,..136.
Step S1034: and (3) inputting the time-frequency domain feature vector of the vibration signal of the 136 x 1-dimensional motor bearing to be detected into a full-connection layer of the double-branch fault diagnosis model, and outputting a fault diagnosis signal of the vibration signal of the 4*1-dimensional motor bearing to be detected by the full-connection layer.
Step S1035: and inputting a fault diagnosis signal of the vibration signal of the 4*1-dimensional motor bearing to be detected into a Softmax layer of the double-branch fault diagnosis model, and outputting a fault classification result of the vibration signal of the motor bearing to be detected by the Softmax layer.
In this embodiment, the time-frequency domain feature vector of the vibration signal of the motor bearing to be detected in 136 x 1 dimensions is input to the full connection layer of the dual-branch fault diagnosis model, the fault diagnosis signal of the vibration signal of the motor bearing to be detected in 4*1 dimensions is obtained through the full connection layer, then the weight under each corresponding channel in the fault diagnosis signal of the vibration signal of the motor bearing to be detected in 4*1 dimensions is obtained through the Softmax layer, the weight is mapped to the (0, 1) numerical space, the sum of the weights of the features of each channel is 1, each channel represents a bearing state, the bearing state comprises four cases of an inner ring fault, an outer ring fault, a retainer fault and a normal state, and in one specific example, if the result of the outer ring fault p=0.94, the inner ring fault p=0.05, and the retainer fault p=0.01 and the normal state p=0.00 is obtained, the bearing state to be detected is represented as the outer ring fault state.
In the embodiment, two branches are input into a pre-trained double-branch fault diagnosis model, wherein a first branch inputs 8*1-dimensional frequency domain feature vectors of vibration signals of a motor bearing to be detected; the second branch inputs the vibration signal of the motor bearing to be detected, and in the second branch, the vibration signal of the motor bearing to be detected is extracted to 128 x 1-dimensional time domain feature vectors of the vibration signal of the motor bearing to be detected through a 1DCNN convolution layer and a multi-head self-attention mechanism layer in the double-branch fault diagnosis model; then, inputting a frequency domain feature vector of a 8*1-dimensional vibration signal of the motor bearing to be detected, which is input by a first branch, and a 128 x 1-dimensional time domain feature vector extracted by a second branch into a double-branch feature fusion layer in a double-branch fault diagnosis model to perform feature fusion, wherein the double-branch feature fusion layer outputs a 136 x 1-dimensional time-frequency domain feature vector of the vibration signal of the motor bearing to be detected; then, inputting 136 x 1-dimensional time-frequency domain feature vectors into a full-connection layer in a double-branch fault diagnosis model, and outputting 4*1-dimensional fault diagnosis signals of vibration signals of the motor bearing to be detected by the full-connection layer; and finally, inputting a 4*1-dimensional fault diagnosis signal into a Softmax layer of the double-branch fault diagnosis model, and outputting a fault classification result of a vibration signal of the motor bearing to be detected by the Softmax layer.
In one implementation of the embodiment of the present invention, before using the pre-trained dual-branch fault diagnosis model, the method further includes:
step S100: and training the double-branch fault diagnosis model to obtain a pre-trained double-branch fault diagnosis model.
In one implementation of the embodiment of the present invention, the step S100 further includes step S1001 to step S1005:
step S1001: and acquiring a data set of an original vibration signal of the motor bearing, and dividing the data set into a training set and a testing set according to a certain proportion.
In this embodiment, the data set selects CWRU (Case Western Reserve University) bearing vibration data of the bearing data center, the CWRU data set adopts SKF6205 bearings, faults are introduced by electric spark machining, and the fault diameters are 0.18mm, 0.36mm and 0.53mm, respectively. As shown in fig. 6, the CWRU bearing test stand is comprised of a 2 horsepower motor (left side of fig. 6), torque sensor/encoder (connection in fig. 6), power tester (right side of fig. 6), and control electronics (not shown in fig. 6). The rated output power of the motor is 2.2kW, the sampling frequency is 12kHz and 48kHz respectively, and in order to obtain bearing data of different faults, 4 different types of vibration signals of a normal state, an inner ring fault, an outer ring fault and a retainer fault are totally contained by adopting electric spark counting to manually process on the inner ring, the outer ring and the retainer respectively.
In one specific example, the experimental motor speed is 1797r/min and the dataset is acquired at a sampling frequency of 12 kHz. Vibration data in the experiment were collected using an accelerometer mounted on the drive end of the motor housing, and in this example, SKF6205 type bearing data from the drive end were used, and the relevant bearing parameters are shown in table 1.
TABLE 1
In this example, the failure conditions of the motor bearings are categorized into an outer ring failure, an inner ring failure, a cage failure, and a normal condition. The diameters of the three faults are respectively set to be 0.007, 0.014 and 0.021inch (1 inch=25.4 mm), then the faults are reinstalled on the motor, the load of the motor can be set to be 0, 1, 2 and 3hp (1 hp= 735.50 w) so as to correspond to different motor speeds, and vibration acceleration signals of the bearing can be acquired through the sensor, so that a sample data set of the bearing working under various working conditions is obtained. The fault information of the relevant bearings is shown in table 2.
TABLE 2
The collected data set is divided into a training set and a testing set according to 8:2, namely 800 samples in the data set in a normal state are training sets, and 200 samples are testing sets; 800 samples in the data set under the fault of the inner ring are training sets, and 200 samples are test sets; 800 samples in the data set under the outer ring fault are training sets, and 200 samples are test sets; 800 samples in the data set under the cage fault are training sets and 200 samples are test sets.
Step S1002: training the double-branch fault diagnosis model by the training set, wherein the training process comprises the following steps: step S10021: performing WPT conversion on the original vibration signals of the motor bearings in the training set, determining the decomposition of the multi-layer wavelet packet, and extracting frequency domain feature vectors of the original vibration signals of the motor bearings in the training set; step S10022: and inputting the frequency domain feature vector of the original vibration signal of the motor bearing in the training set and the original vibration signal of the motor bearing in the training set into a double-branch fault diagnosis model under training, and outputting a fault classification result of the original vibration signal of the motor bearing in the training set by the double-branch fault diagnosis model.
In this embodiment, the number of layers of wavelet packet decomposition is determined to be 3, and 8*1-dimensional frequency domain feature vectors corresponding to original vibration signals of the motor bearings in the training set are extractedWherein A is K training Maximum amplitude in each sub-band of the original vibration signal for the motor bearing in the training set; f (f) K training The frequency corresponding to the maximum amplitude value in each sub-band of the original vibration signal of the motor bearing in the training set is obtained; k=1, 2,3,4,5,6,7,8.
In a specific example, the step S10022 further includes:
Inputting vibration signals of a motor bearing in a training set to a 1DCNN convolution layer of a double-branch fault diagnosis model which is being trained, wherein the 1DCNN convolution layer outputs feature vectors of 128 x t dimensions;
inputting the feature vector in the 128 x t dimension to a multi-head self-attention mechanism layer of a training double-branch fault diagnosis model, wherein the multi-head self-attention mechanism layer outputs a time domain feature vector K of a vibration signal of a 128 x 1 dimension training concentrated motor bearing Training device ；
Frequency domain characteristic vector H of original vibration signal of 8*1-dimensional training set motor bearing K training Time domain feature vector K of original vibration signal of motor bearing in training set with dimension of 128 x 1 Training device The method comprises the steps that a double-branch feature fusion layer is input to a double-branch fault diagnosis model under training, and the double-branch feature fusion layer outputs 136 x 1-dimensional time-frequency domain feature vectors of original vibration signals of a motor bearing in a training set;
inputting the time-frequency domain feature vector of the original vibration signal of the 136 x 1-dimensional training set motor bearing into a full-connection layer of the double-branch fault diagnosis model under training, and outputting a fault diagnosis signal of the original vibration signal of the 4*1-dimensional training set motor bearing by the full-connection layer;
The fault diagnosis signal of the original vibration signal of the 4*1-dimensional motor bearing in the training set is input to a Softmax layer of the double-branch fault diagnosis model which is being trained, and the Softmax layer outputs the fault classification result of the original vibration signal of the motor bearing in the training set.
Step S1003: and comparing the fault classification result of the original vibration signals of the motor bearings in the training set with the fault labels of the original vibration signals of the motor bearings in the training set, and calculating the loss function of the double-branch fault diagnosis model.
In a specific example, the type of fault is set to a label, for example, a normal state is set to "0", an outer ring fault is set to "1", an inner ring fault is set to "2", a cage fault is set to "3", a fault label corresponding to an original vibration signal in a training set is known, in the process of training a double-branch fault diagnosis model, the fault classification result of the original vibration signal of a motor bearing in the training set and the fault label of the original vibration signal of the motor bearing in the training set are output by the double-branch fault diagnosis model are compared, and a loss function of the double-branch fault diagnosis model is calculated, and when the loss function does not meet the condition of model optimization, the double-branch fault diagnosis model is continuously trained.
In one implementation of the embodiment of the present invention, the loss function is a cross entropy loss function.
Step S1004: and stopping training the double-branch fault diagnosis model when the loss function meets the requirement of model optimization.
When training is performed iteratively all the time, the parameters of the double-branch fault diagnosis model are updated all the time, the model is more and more optimized, and when the loss function reaches the requirement of meeting the model optimization, the training process is ended, and at the moment, the training of the double-branch fault diagnosis model is stopped.
Step S1005: and testing the trained fault diagnosis model by the test set, and obtaining the trained double-branch fault diagnosis model after the test is completed.
In a specific example, the design of the dual-branch fault diagnosis model is adjusted through the performance on the training set to obtain a final dual-branch fault diagnosis model, and the trained dual-branch fault diagnosis model is tested on the testing set to obtain the accuracy of the model and is compared with other methods. Experimental results show that the double-branch fault diagnosis model provided by the embodiment can learn the internal corresponding relation between the data and the classification labels through the training set, and deduce bearing fault positions corresponding to the test data according to the internal corresponding relation.
In one implementation manner of the embodiment of the present invention, the 1DCNN convolution layers in the trained dual-branch fault diagnosis model include an input layer, a first convolution layer, a first pooling layer, a second convolution layer and a second pooling layer; wherein, the liquid crystal display device comprises a liquid crystal display device,
the first convolution layer is set to be 1 input channel, 64 output channels, the step length is 1, and the edge filling is set to be 4 layers; the first pooling layer is the maximum time pooling layer; the second convolution layer is set to 64 input channels, 128 output channels, the step size is 1, and the edge filling is set to 2 layers; the second pooling layer is an average time pooling layer.
In this embodiment, the training set is used to train the dual-branch fault diagnosis model, and after verification by the test set, parameters of each layer of the trained dual-branch fault diagnosis model are set as follows:
performing three-layer wavelet packet transformation, and setting the parameter of the layer number corresponding to the wavelet packet decomposition to be 3;
the method comprises the steps of setting two 1D convolution layers and two pooling layers, wherein the first convolution layer is set to be 1 input channel, 64 output channels, the step length is 1, and the edge filling is set to be 4 layers; the first pooling layer is the maximum time pooling layer; the second convolution layer is set to 64 input channels, 128 output channels, the step size is 1, and the edge filling is set to 2 layers; the second pooling layer is an average time pooling layer.
And dividing the data of the vibration signal passing through the 1DCNN convolution layer into three parts, inputting the three parts into a multi-head attention mechanism layer to form a Q, K, V matrix, and finally outputting 128 x 1-dimensional characteristics after a series of calculation processing.
Dropout layer. In order to effectively prevent the model from over fitting, the training speed is increased, and the Dropout coefficient parameter is set to be 0.5.
Batch Norm layer. Input data is normalized, training speed can be effectively improved, and parameters set for each convolution layer are as follows: the batch_size is set to 64; momentum used by dynamic means and dynamic variance = 0.1; to ensure numerical stability (denominator cannot approach or take 0), the value eps=0.00001 is added to the denominator.
A loss function. The loss function used in this embodiment is a cross entropy loss function, and the description of the cross entropy loss function in the prior art is already very detailed, and will not be repeated here.
Adam optimizer. Adam, as an adaptive algorithm, can further increase the training speed of the model. The model proposed in this embodiment sets Adam parameters as=0.99；/>=0.999。
In order to more clearly demonstrate the layer parameters of the 1DCNN convolution layer used in this embodiment, table 3 details the parameter settings of the dual-branch fault diagnosis model proposed in this embodiment, including layer type, convolution kernel size, activation function, and dimensions of input and output data.
TABLE 3 Table 3
| Layer type | Convolution kernel size | Activation function | Input dimension | Output dimension |
| Input layer | - | - | - | (1,2048) |
| One-dimensional convolution | 9 | ReLU | (1,2048) | (64,2048) |
| Batch Norm layer 1 | - | - | (64,2048) | (64,2048) |
| Dropout layer 1 | - | - | (64,2048) | (64,2048) |
| One-dimensional convolution | 5 | ReLU | (64,2048) | (128,2048) |
| Batch Norm layer 2 | - | - | (128,2048) | (128,2048) |
| Dropout layer 2 | - | - | (128,2048) | (128,2048) |
| Self-attention mechanismLayer(s) | - | - | (128,2048) | 128 |
| Time pooling layer | - | - | (128,2048) | 128 |
| Full connection layer | - | Softmax | 4 | 4 |
Training the model after setting the double-branch fault diagnosis model based on the parameters, setting the iteration times to be 20 times, setting the batch_Size Size to be 64, namely putting 64 samples at one time for training, recording the optimization process of the model, and respectively recording the change of the accuracy and the loss function along with the iteration times (Epochs) in the model training process in the figures 7 and 8. In fig. 7, curve 1 represents the Accuracy (Accuracy) of the training set, and curve 2 represents the Accuracy of the test set; in fig. 8 curve 3 represents the Loss function (Loss) of the training set and curve 4 represents the Loss function of the test set. As can be seen from a combination of fig. 7 and 8, the accuracy curve of the test set shows a spiral-up trend and the loss function curve shows a spiral-down trend under the influence of Dropout. As is evident from fig. 7, the training set is stabilized with small fluctuation and variation, but the accuracy of the test set is quite obvious in the fluctuation of the first 10 times, the accuracy is increased from 0.9 to more than 0.98, and from 11 th time, the accuracy is fluctuated, but the fluctuation is still kept within the range of 0.98-0.99, and the final result is higher than 0.99; as can be seen from an observation of fig. 8, the loss function overall shows a decreasing trend, although peaks exist at the 7 th, 12 th, 15 th and 18 th times of the test set due to Dropout, the loss function is always lower than 0.07, and for the training set, from the second iteration, the loss function is always kept around 0, which means that the network has already tended to be stable at the second iteration; as can be seen from a comprehensive observation of the curve change conditions of fig. 7 and fig. 8, the model proposed in this embodiment has good stability.
It should be noted that, the judging of the quality of the model is not focused on the accuracy of the training set, but focuses on whether the performance of the test set can reach the level close to the training set. As can be seen from fig. 7 and 8, the accuracy of the test set and the final value of the loss function reach a high level close to that of the training set, which fully illustrates that the model provided by the embodiment has good generalization, namely, has good classification capability on strange data. In addition, it can be seen that our dual-branch model achieves extremely high accuracy after a very short number of iterations. The training curve then gradually goes to a plateau, with a consequent decrease in training yield, so we stopped training at 20 epochs.
Based on the steps S101-S103, a fault diagnosis method of a motor bearing based on WPT-1DCNN is provided, and the method comprises the following steps: acquiring a vibration signal of a motor bearing to be detected; performing WPT conversion on the vibration signal of the motor bearing to be detected, determining the decomposition of the multi-layer wavelet packet, and extracting the frequency domain feature vector of the vibration signal of the motor bearing to be detected; and inputting the vibration signals of the motor bearing to be detected and the frequency domain feature vectors of the vibration signals of the motor bearing to be detected into a pre-trained double-branch fault diagnosis model, and outputting a fault classification result of the vibration signals of the motor bearing to be detected by the double-branch fault diagnosis model. According to the prediction method, the frequency domain feature vector and the time domain feature vector fused double-branch fault diagnosis model is adopted, so that the accuracy of bearing fault diagnosis can be effectively improved, and the prediction method has good generalization.
It should be noted that, although the foregoing embodiments describe the steps in a specific order, it will be understood by those skilled in the art that, in order to achieve the effects of the present invention, the steps are not necessarily performed in such an order, and may be performed simultaneously (in parallel) or in other orders, and these variations are within the scope of the present invention.
Furthermore, the invention also provides a fault diagnosis device of the motor bearing based on the WPT-1 DCNN.
Referring to fig. 9, fig. 9 is a main structural block diagram of a WPT-1 DCNN-based motor bearing fault diagnosis apparatus according to an embodiment of the present invention. As shown in fig. 9, the fault diagnosis device for the motor bearing based on WPT-1DCNN in the embodiment of the present invention mainly includes an acquisition module 11, a frequency domain feature extraction module 12, and a diagnosis module 13. In some embodiments, one or more of the acquisition module 11, the frequency domain feature extraction module 12, and the diagnostic module 13 may be combined together into one module. The acquisition module 11 may in some embodiments be configured to acquire a vibration signal of the motor bearing to be detected. The frequency domain feature extraction module 12 may be configured to perform WPT conversion on the vibration signal of the motor bearing to be detected, determine decomposition of the multi-layer wavelet packet, and extract a frequency domain feature vector of the vibration signal of the motor bearing to be detected. The diagnosis module 13 may be configured to input the vibration signal of the motor bearing to be detected and the frequency domain feature vector of the vibration signal of the motor bearing to be detected to a pre-trained dual-branch fault diagnosis model that outputs a fault classification result of the vibration signal of the motor bearing to be detected.
In one embodiment, the description of the specific implementation functions may be described with reference to step S101 to step S103.
The above-mentioned fault diagnosis device for WPT-1 DCNN-based motor bearing is used to implement the embodiment of the fault diagnosis method for WPT-1 DCNN-based motor bearing shown in fig. 1, and the technical principles, the technical problems to be solved and the technical effects to be produced are similar, and those skilled in the art can clearly understand that, for convenience and brevity of description, the specific working process and related description of the fault diagnosis device for WPT-1 DCNN-based motor bearing may refer to the description of the embodiment of the fault diagnosis method for WPT-1 DCNN-based motor bearing.
It will be appreciated by those skilled in the art that the present invention may implement all or part of the above-described methods according to the above-described embodiments, or may be implemented by means of a computer program for instructing relevant hardware, where the computer program may be stored in a computer readable storage medium, and where the computer program may implement the steps of the above-described embodiments of the method when executed by a processor. Wherein the computer program comprises computer program code which may be in source code form, object code form, executable file or some intermediate form etc. The computer readable storage medium may include: any entity or device, medium, usb disk, removable hard disk, magnetic disk, optical disk, computer memory, read-only memory, random access memory, electrical carrier wave signals, telecommunications signals, software distribution media, and the like capable of carrying the computer program code. It should be noted that the computer readable storage medium may include content that is subject to appropriate increases and decreases as required by jurisdictions and by jurisdictions in which such computer readable storage medium does not include electrical carrier signals and telecommunications signals.
Further, the invention also provides a control device. In one control device embodiment according to the present invention, the control device includes a processor and a storage device, the storage device may be configured to store a program for executing the WPT-1 DCNN-based motor bearing fault diagnosis method of the above-described method embodiment, and the processor may be configured to execute the program in the storage device, including, but not limited to, the program for executing the WPT-1 DCNN-based motor bearing fault diagnosis method of the above-described method embodiment. For convenience of explanation, only those portions of the embodiments of the present invention that are relevant to the embodiments of the present invention are shown, and specific technical details are not disclosed, please refer to the method portions of the embodiments of the present invention. The control device may be a control device formed of various electronic devices.
Further, the invention also provides a computer readable storage medium. In one embodiment of the computer readable storage medium according to the present invention, the computer readable storage medium may be configured to store a program for performing the WPT-1 DCNN-based motor bearing fault diagnosis method of the above-described method embodiment, which may be loaded and executed by a processor to implement the WPT-1 DCNN-based motor bearing fault diagnosis method described above. For convenience of explanation, only those portions of the embodiments of the present invention that are relevant to the embodiments of the present invention are shown, and specific technical details are not disclosed, please refer to the method portions of the embodiments of the present invention. The computer readable storage medium may be a storage device including various electronic devices, and optionally, the computer readable storage medium in the embodiments of the present invention is a non-transitory computer readable storage medium.
Further, it should be understood that, since the respective modules are merely set to illustrate the functional units of the apparatus of the present invention, the physical devices corresponding to the modules may be the processor itself, or a part of software in the processor, a part of hardware, or a part of a combination of software and hardware. Accordingly, the number of individual modules in the figures is merely illustrative.
Those skilled in the art will appreciate that the various modules in the apparatus may be adaptively split or combined. Such splitting or combining of specific modules does not cause the technical solution to deviate from the principle of the present invention, and therefore, the technical solution after splitting or combining falls within the protection scope of the present invention.
Thus far, the technical solution of the present invention has been described in connection with the preferred embodiments shown in the drawings, but it is easily understood by those skilled in the art that the scope of protection of the present invention is not limited to these specific embodiments. Equivalent modifications and substitutions for related technical features may be made by those skilled in the art without departing from the principles of the present invention, and such modifications and substitutions will fall within the scope of the present invention.
Claims (10)
1. A fault diagnosis method for a motor bearing based on WPT-1DCNN, comprising:
acquiring a vibration signal of a motor bearing to be detected;
performing WPT conversion on the vibration signal of the motor bearing to be detected, determining the decomposition of the multi-layer wavelet packet, and extracting the frequency domain feature vector of the vibration signal of the motor bearing to be detected;
and inputting the vibration signals of the motor bearing to be detected and the frequency domain feature vectors of the vibration signals of the motor bearing to be detected into a pre-trained double-branch fault diagnosis model, and outputting a fault classification result of the vibration signals of the motor bearing to be detected by the double-branch fault diagnosis model.
2. The fault diagnosis method for WPT-1 DCNN-based motor bearings according to claim 1, wherein the performing WPT transformation on the vibration signal of the motor bearing to be detected, and determining the decomposition of the multi-layer wavelet packet, extracting the frequency domain feature vector of the vibration signal of the motor bearing to be detected, includes:
decomposing a 3-layer wavelet packet of a vibration signal of a motor bearing to be detected to obtain decomposed 8 sub-frequency bands;
determining a frequency domain characteristic vector H of a vibration signal of the motor bearing to be detected in 8*1 dimensions according to the product of the maximum amplitude of each of the 8 sub-bands and the frequency corresponding to the maximum amplitude of each of the 8 sub-bands k The method comprises the following steps:wherein A is k Maximum amplitude in each sub-band; f (f) k A frequency corresponding to the maximum amplitude in each sub-band; k=1, 2,3,4,5,6,7,8.
3. The WPT-1DCNN based motor bearing fault diagnosis method of claim 2, wherein the frequency domain feature vectors of the vibration signal of the motor bearing to be detected and the vibration signal of the motor bearing to be detected are input to a pre-trained dual-branch fault diagnosis model, which outputs a fault classification result of the vibration signal of the motor bearing to be detected, comprising:
inputting the vibration signal of the motor bearing to be detected into a 1DCNN convolution layer of a double-branch fault diagnosis model, wherein the 1DCNN convolution layer outputs a feature vector of 128 x t dimensions;
inputting the feature vector of the 128 x t dimension into a multi-head self-attention mechanism layer of a double-branch fault diagnosis model, wherein the multi-head self-attention mechanism layer outputs a time domain feature vector K of a vibration signal of a motor bearing to be detected of the 128 x 1 dimension;
frequency domain characteristic vector H of vibration signal of 8*1-dimensional motor bearing to be detected k The time domain feature vector K of the vibration signal of the motor bearing to be detected with the dimension of 128 x 1 is input into a double-branch feature fusion layer of a double-branch fault diagnosis model, and the double-branch feature fusion layer outputs the time-frequency domain feature vector of the vibration signal of the motor bearing to be detected with the dimension of 136 x 1;
The method comprises the steps of inputting a time-frequency domain feature vector of a vibration signal of a 136 x 1-dimensional motor bearing to be detected into a full-connection layer of a double-branch fault diagnosis model, and outputting a fault diagnosis signal of the vibration signal of the 4*1-dimensional motor bearing to be detected by the full-connection layer;
and inputting a fault diagnosis signal of the vibration signal of the 4*1-dimensional motor bearing to be detected into a Softmax layer of the double-branch fault diagnosis model, and outputting a fault classification result of the vibration signal of the motor bearing to be detected by the Softmax layer.
4. The WPT-1DCNN based motor bearing fault diagnosis method of claim 1, further comprising:
and training the double-branch fault diagnosis model to obtain a pre-trained double-branch fault diagnosis model.
5. The WPT-1DCNN based motor bearing fault diagnosis method of claim 4, wherein the training the dual-branch fault diagnosis model to obtain a pre-trained dual-branch fault diagnosis model includes:
acquiring a data set of an original vibration signal of a motor bearing, and dividing the data set into a training set and a testing set according to a certain proportion;
training the double-branch fault diagnosis model by the training set, wherein the training process comprises the following steps: performing WPT conversion on the original vibration signals of the motor bearings in the training set, determining the decomposition of the multi-layer wavelet packet, and extracting frequency domain feature vectors of the original vibration signals of the motor bearings in the training set; inputting the frequency domain feature vector of the original vibration signal of the motor bearing in the training set and the original vibration signal of the motor bearing in the training set into a double-branch fault diagnosis model under training, and outputting a fault classification result of the original vibration signal of the motor bearing in the training set by the double-branch fault diagnosis model;
Comparing the fault classification result of the original vibration signals of the motor bearings in the training set with the fault labels of the original vibration signals of the motor bearings in the training set, and calculating the loss function of the double-branch fault diagnosis model;
when the loss function reaches the requirement of meeting the model optimization, stopping training the double-branch fault diagnosis model;
and testing the trained fault diagnosis model by the test set, and obtaining the trained double-branch fault diagnosis model after the test is completed.
6. The WPT-1DCNN based motor bearing fault diagnosis method of claim 5, wherein the loss function is a cross entropy loss function.
7. The WPT-1DCNN based motor bearing fault diagnosis method of claim 5, wherein the 1DCNN convolution layers in the trained dual branch fault diagnosis model include an input layer, a first convolution layer, a first pooling layer, a second convolution layer, and a second pooling layer; wherein, the liquid crystal display device comprises a liquid crystal display device,
the first convolution layer is set to be 1 input channel, 64 output channels, the step length is 1, and the edge filling is set to be 4 layers; the first pooling layer is the maximum time pooling layer; the second convolution layer is set to 64 input channels, 128 output channels, the step size is 1, and the edge filling is set to 2 layers; the second pooling layer is an average time pooling layer.
8. A WPT-1 DCNN-based motor bearing failure diagnosis apparatus, comprising:
the acquisition module is used for acquiring a vibration signal of the motor bearing to be detected;
the frequency domain feature extraction module is used for carrying out WPT conversion on the vibration signal of the motor bearing to be detected, determining the decomposition of the multi-layer wavelet packet and extracting the frequency domain feature vector of the vibration signal of the motor bearing to be detected;
the diagnosis module is used for inputting the vibration signals of the motor bearing to be detected and the frequency domain feature vectors of the vibration signals of the motor bearing to be detected into a pre-trained double-branch fault diagnosis model, and the double-branch fault diagnosis model outputs fault classification results of the vibration signals of the motor bearing to be detected.
9. A control device comprising a processor and a storage device, the storage device being adapted to store a plurality of program codes, characterized in that the program codes are adapted to be loaded and executed by the processor to perform the WPT-1DCNN based motor bearing fault diagnosis method of any one of claims 1 to 7.
10. A computer readable storage medium, in which a plurality of program codes are stored, characterized in that the program codes are adapted to be loaded and executed by a processor to perform the WPT-1DCNN based motor bearing fault diagnosis method of any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310480764.4A CN116659856A (en) | 2023-04-28 | 2023-04-28 | Fault diagnosis method of motor bearing based on WPT-1DCNN |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310480764.4A CN116659856A (en) | 2023-04-28 | 2023-04-28 | Fault diagnosis method of motor bearing based on WPT-1DCNN |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116659856A true CN116659856A (en) | 2023-08-29 |
Family
ID=87725139
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310480764.4A Pending CN116659856A (en) | 2023-04-28 | 2023-04-28 | Fault diagnosis method of motor bearing based on WPT-1DCNN |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116659856A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117171547A (en) * | 2023-11-02 | 2023-12-05 | 深圳市信润富联数字科技有限公司 | Fault diagnosis method, device, equipment and storage medium based on large model |
-
2023
- 2023-04-28 CN CN202310480764.4A patent/CN116659856A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117171547A (en) * | 2023-11-02 | 2023-12-05 | 深圳市信润富联数字科技有限公司 | Fault diagnosis method, device, equipment and storage medium based on large model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Udmale et al. | Application of spectral kurtosis and improved extreme learning machine for bearing fault classification | |
| CN110516305B (en) | Intelligent fault diagnosis method under small sample based on attention mechanism meta-learning model | |
| CN112906644B (en) | Mechanical fault intelligent diagnosis method based on deep migration learning | |
| Michau et al. | Fully learnable deep wavelet transform for unsupervised monitoring of high-frequency time series | |
| CN112785016A (en) | New energy automobile maintenance and fault monitoring and diagnosis method based on machine learning | |
| Udmale et al. | A bearing vibration data analysis based on spectral kurtosis and ConvNet | |
| CN111160167A (en) | Spindle fault classification and identification method based on S-transform deep convolutional neural network | |
| CN116659856A (en) | Fault diagnosis method of motor bearing based on WPT-1DCNN | |
| CN112329914B (en) | Fault diagnosis method and device for buried transformer substation and electronic equipment | |
| Esfahanian et al. | On contour-based classification of dolphin whistles by type | |
| Benkedjouh et al. | Gearbox fault diagnosis based on mel-frequency cepstral coefficients and support vector machine | |
| CN115830436A (en) | Marine organism intelligent detection method based on deep learning | |
| Li et al. | Transient feature self-enhancement via shift-invariant manifold sparse learning for rolling bearing health diagnosis | |
| CN116150901A (en) | Rolling bearing residual life prediction method based on attention-enhanced time-frequency converter | |
| CN115753101A (en) | Bearing fault diagnosis method based on weight adaptive feature fusion | |
| Behzad et al. | Inaccessible rolling bearing diagnosis using a novel criterion for Morlet wavelet optimization | |
| Wang et al. | A novel rolling bearing fault diagnosis method based on adaptive denoising convolutional neural network under noise background | |
| CN113792597A (en) | Mechanical equipment abnormal sound detection method based on self-supervision feature extraction | |
| CN114639391A (en) | Mechanical failure prompting method and device, electronic equipment and storage medium | |
| Tran et al. | Denoising induction motor sounds using an autoencoder | |
| Pan et al. | Intelligent fault diagnosis of rolling bearing via deep-layerwise feature extraction using deep belief network | |
| Zhou et al. | An attention-based multi-scale convolution network for intelligent underwater acoustic signal recognition | |
| Cui et al. | Aero-engine fault diagnosis using improved local discriminant bases and support vector machine | |
| CN115758237A (en) | Bearing fault classification method and system based on intelligent inspection robot | |
| Chuchra et al. | A deep learning approach for splicing detection in digital audios |
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 |