CN114136619A - Rolling bearing fault diagnosis method under variable working conditions based on convolution self-coding - Google Patents
Rolling bearing fault diagnosis method under variable working conditions based on convolution self-coding Download PDFInfo
- Publication number
- CN114136619A CN114136619A CN202111258171.0A CN202111258171A CN114136619A CN 114136619 A CN114136619 A CN 114136619A CN 202111258171 A CN202111258171 A CN 202111258171A CN 114136619 A CN114136619 A CN 114136619A
- Authority
- CN
- China
- Prior art keywords
- data set
- sample data
- coding
- rolling bearing
- self
- 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
- 238000005096 rolling process Methods 0.000 title claims abstract description 72
- 238000003745 diagnosis Methods 0.000 title claims abstract description 65
- 238000000034 method Methods 0.000 title claims abstract description 41
- 239000011159 matrix material Substances 0.000 claims abstract description 31
- 238000013528 artificial neural network Methods 0.000 claims abstract description 12
- 238000013527 convolutional neural network Methods 0.000 claims description 35
- 238000004422 calculation algorithm Methods 0.000 claims description 4
- 230000011218 segmentation Effects 0.000 claims description 3
- 238000012549 training Methods 0.000 abstract description 16
- 238000000605 extraction Methods 0.000 abstract description 5
- 230000006870 function Effects 0.000 description 29
- 230000008569 process Effects 0.000 description 7
- 230000007423 decrease Effects 0.000 description 6
- 238000010586 diagram Methods 0.000 description 6
- 238000012706 support-vector machine Methods 0.000 description 5
- 238000012360 testing method Methods 0.000 description 4
- 238000013136 deep learning model Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 230000036541 health Effects 0.000 description 2
- 238000001228 spectrum Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 238000012935 Averaging Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000002405 diagnostic procedure Methods 0.000 description 1
- 230000003862 health status Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000007637 random forest analysis Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Images
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
- G01M13/045—Acoustic or vibration analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Computational Linguistics (AREA)
- Computational Mathematics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Optimization (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Analysis (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Operations Research (AREA)
- Probability & Statistics with Applications (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Algebra (AREA)
- Databases & Information Systems (AREA)
- Acoustics & Sound (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention discloses a rolling bearing fault diagnosis method under variable working conditions based on convolution self-coding, which specifically comprises the following steps; obtaining a sample data set: acquiring vibration signals of a rolling bearing under a plurality of working conditions through a public data set to obtain a sample data set of the vibration signals, wherein the sample data set comprises a source domain data set and a target domain data set; respectively inputting the source domain data set and the target domain data set into a one-dimensional convolution self-coding encoder to obtain a source domain characteristic parameter matrix and a target domain characteristic parameter matrix; reconstructing the sample data set, and calculating to obtain a mean square error loss function through the reconstructed sample data set and the sample data set; training a one-dimensional convolution self-coding and convolution neural network; calculating the diagnosis accuracy of the target domain data set through the trained network; by combining the one-dimensional convolution self-coding with the convolution neural network, the self-adaptive feature extraction and the field self-adaptive rolling bearing fault diagnosis under the variable working condition are realized.
Description
Technical Field
The invention relates to the technical field of intelligent diagnosis of bearing faults, in particular to a fault diagnosis method of a rolling bearing under variable working conditions based on one-dimensional convolution self-coding.
Background
Rolling bearings are common parts in rotating machinery, and their health status relates to the reliability and stability of equipment operation. However, when the device runs under the working conditions of high rotating speed and heavy load for a long time, the failure is frequent, and if the device cannot be maintained in time, other parts of the device are damaged, so that greater economic loss is caused. Therefore, the method has important significance for accurately identifying the health state of the rolling bearing. Under the influence of various factors such as requirements and environment, the rolling bearing does not operate under a single working condition, but often operates under the conditions of variable speed and variable load, the complexity of the mapping relation between fault characteristics and fault modes is increased to a certain extent due to the change of the working condition, and great challenges are brought to the monitoring of the health state of mechanical equipment. To solve the problem of fault diagnosis of rolling bearings, a large amount of research work is carried out by domestic and foreign scholars. In recent years, with the continuous and deep study of machine learning, the diagnostic method driven by data, such as support vector machine, random forest, deep belief network, etc., has been developed greatly. The fault diagnosis method usually needs manual feature extraction, has strong dependence on expert experience, and has certain limitation in application under variable working conditions. In order to get rid of the limitation caused by manually extracting the features, the deep learning method is gradually applied to bearing fault diagnosis under variable working conditions, the method can automatically learn feature extraction, but the problems of difficult convergence and low training speed often exist when a deep learning model is trained, so that the wide application of the deep learning model is influenced.
Disclosure of Invention
The invention provides a rolling bearing fault diagnosis method under variable working conditions based on one-dimensional convolution self-coding, which automatically extracts and identifies fault characteristics of a rolling bearing under variable working conditions by combining the one-dimensional convolution self-coding with a convolution neural network, and realizes self-adaptive characteristic extraction and field self-adaptive rolling bearing fault diagnosis under variable working conditions.
In order to solve the technical problems, the invention adopts the technical scheme that:
a rolling bearing fault diagnosis method under variable working conditions based on convolution self-coding specifically comprises the following steps:
step S1, obtaining a sample data set: acquiring vibration signals of a rolling bearing under a plurality of working conditions through a public data set, and segmenting the acquired vibration signals by using a data overlapping segmentation method to obtain a sample data set of the vibration signals, wherein the sample data set comprises a source domain data set and a target domain data set;
step S2, respectively inputting the source domain data set and the target domain data set into a one-dimensional convolution self-coding encoder to obtain a source domain characteristic parameter matrix and a target domain characteristic parameter matrix;
s3, inputting the source domain characteristic parameter matrix and the target domain characteristic parameter matrix in the step S2 into a one-dimensional convolution self-coding decoder to obtain a reconstructed sample data set, and calculating through the reconstructed sample data set and the sample data set to obtain a mean square error loss function;
step S4, inputting the source domain characteristic parameter matrix into a convolutional neural network, outputting the state type of the source domain and calculating to obtain a cross entropy loss function;
s5, obtaining a total loss function of the sample data set through summation of a mean square error loss function and a cross entropy loss function, updating parameters of the one-dimensional convolutional self-coding and convolutional neural network by using a back propagation algorithm, stopping updating the parameters of the one-dimensional convolutional self-coding and convolutional neural network until the total loss of the sample data set is 0, and entering the step S6;
and step S6, calculating the diagnosis accuracy of the target domain data set.
As a further preferable embodiment of the present invention, the step S1 specifically includes the following steps:
step S11, obtaining a sample data set I: acquiring vibration signals of a rolling bearing under the same rotating speed and different loads through a public data set, wherein the rolling bearing comprises four states and operates under four working conditions, and the four states comprise a rolling element fault, an outer ring fault, an inner ring fault and a normal state; the four working conditions comprise 1797rpm/0Hp, 1772rpm/1Hp, 1750rpm/2Hp and 1730rpm/3 Hp; the method comprises the steps of giving a sample data label according to the rolling bearing state by taking 1024 data points as sample data, forming a plurality of sample data to obtain a sample data set I, wherein the sample data set I comprises four different loads, taking the sample data set I as a variable load data set, taking data of any working condition in the sample data set I as a source domain data set, and taking data of other working conditions in the sample data set I as a target domain data set.
As a further preferable embodiment of the present invention, the step S1 further includes the following steps:
step S12, obtaining a sample data set II: the method comprises the steps of obtaining vibration signals of the rolling bearing under different rotating speeds and the same load through a public data set, wherein the rotating speed range [900rpm, 1620rpm ] of the rolling bearing, at the moment, the state of the rolling bearing comprises an outer ring fault, an inner ring fault and a normal state, at the moment, 1024 data points are used as one sample data, a sample data label is given according to the state of the rolling bearing, a plurality of sample data in the state are combined to obtain a sample data set II, the sample data set II comprises four different rotating speeds, the sample data set II is used as a variable rotating speed data set, data of any working condition in the sample data set II is used as a source domain data set, and data of other working conditions in the sample data set II is used as a target domain data set.
As a further preferable aspect of the present invention, the one-dimensional convolutional self-encoding in step S2 includes an encoder composed of 5 convolutional layers and a decoder composed of 7 deconvolution layers.
As a further preferable embodiment of the present invention, the step S3 specifically includes the following steps:
inputting the source domain characteristic parameter matrix and the target domain characteristic parameter matrix obtained in the step S2 into a decoder of one-dimensional convolution self-coding to obtain reconstructed sample data and calculate a mean square error loss function, inputting the source domain characteristic parameter matrix and the target domain characteristic parameter matrix obtained in the step S2 into the decoder of one-dimensional convolution self-coding to obtain a reconstructed sample data set, and calculating the mean square error loss function according to the reconstructed sample data set and the sample data set:
where x is the input to the encoder for one-dimensional convolutional self-encoding, i is the sample number, xiIs the i-th input sample of the encoder, y is the output of the decoder, yiIs the output of the ith sample decoder, and n is the number of samples.
As a further preferable embodiment of the present invention, the convolutional neural network in step S4 is composed of two convolutional layers and two fully-connected layers.
As a further preferred aspect of the present invention, the cross entropy loss function is:
wherein C is the number of bearing state classes, z is the output of the convolutional neural network, j is the class number, j takes the value of 0,1,2 … C-1,
the j-dimension output value of the ith sample after passing through the convolutional neural network is obtained; 1 {. is an exponential function when
1{ · } -, 1; when in use
When the value is 1{ · } -, 0; p is
The probability of occurrence.
As a further preferable embodiment of the present invention, the step S6 specifically includes the following steps:
step S61: inputting the target domain data set in the S2 into a trained one-dimensional convolution self-coding encoder to obtain a target domain characteristic parameter matrix;
step S62: and when the parameters of the one-dimensional convolutional self-coding and convolutional neural network stop updating, obtaining the trained one-dimensional convolutional self-coding and convolutional neural network, inputting the target domain characteristic parameter matrix in the step S61 into the trained one-dimensional convolutional self-coding and convolutional neural network, outputting to obtain the bearing state category of the screened target domain, comparing the bearing state category of the screened target domain with the sample data label of the target domain, and calculating to obtain the diagnosis and identification precision of the rolling bearing fault.
As a further preferable scheme of the invention, before the sample data set is obtained, a rolling bearing fault diagnosis model under variable working conditions of one-dimensional convolution self-coding is established.
As a further preferable scheme of the invention, the rolling bearing fault diagnosis model under the variable working condition of the one-dimensional convolution self-coding comprises the one-dimensional convolution self-coding and a convolution neural network
The invention has the following beneficial effects:
1. the rolling bearing fault diagnosis method under the variable working condition based on the convolution self-coding, disclosed by the invention, has the advantages that the one-dimensional convolution self-coding and the convolution neural network are combined, so that the fault characteristics of the rolling bearing under the variable working condition are automatically extracted and identified without a large amount of test data, the field self-adaption is realized, and the fault diagnosis accuracy of the rolling bearing under the variable working condition is greatly improved;
2. according to the rolling bearing fault diagnosis method under the variable working condition based on the convolution self-coding, the encoder, the decoder and the neural network are all composed of the convolution layer and the deconvolution layer, so that the training parameters are not too much, the training process is more stable, and the rolling bearing fault diagnosis method has the advantages of easiness in convergence and convenience in training.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a diagram of the convolutional self-coding network structure of the present invention;
FIG. 3 is a diagram of the convolutional neural network structure of the present invention;
FIG. 4 is a diagram of a bearing fault diagnosis model of the present invention;
FIG. 5 is a graph of the loss function of the model of the present invention during training;
FIG. 6 is a diagram of the one-dimensional convolutional self-coding training effect of the present invention;
FIG. 7 is a comparison of the diagnostic accuracy of the model of the present invention and the comparison model.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings and specific preferred embodiments.
In the description of the present invention, it is to be understood that the terms "left side", "right side", "upper part", "lower part", etc., indicate orientations or positional relationships based on those shown in the drawings, and are only for convenience of describing the present invention and simplifying the description, but do not indicate or imply that the referred device or element must have a specific orientation, be constructed in a specific orientation, and be operated, and that "first", "second", etc., do not represent an important degree of the component parts, and thus are not to be construed as limiting the present invention. The specific dimensions used in the present example are only for illustrating the technical solution and do not limit the scope of protection of the present invention.
As shown in fig. 1, a rolling bearing fault diagnosis method under variable working conditions based on convolutional self-coding specifically includes the following steps:
firstly, establishing a rolling bearing fault diagnosis model under the variable working condition of one-dimensional convolution self-coding, wherein the rolling bearing fault diagnosis model under the variable working condition of the one-dimensional convolution self-coding comprises one-dimensional convolution self-coding and a convolution neural network.
Step S1, obtaining a sample data set: the method comprises the steps of obtaining vibration signals of the rolling bearing under multiple working conditions through a public data set, and segmenting the collected vibration signals by using a data overlapping segmentation method to obtain a sample data set of the vibration signals, wherein the sample data set comprises a source domain data set and a target domain data set.
Step S11, obtaining a sample data set I: acquiring vibration signals of a rolling bearing under the same rotating speed and different loads through a public data set, wherein the rolling bearing comprises four states and operates under four working conditions, and the four states comprise a rolling element fault, an outer ring fault, an inner ring fault and a normal state; the four working conditions comprise 1797rpm/0Hp, 1772rpm/1Hp, 1750rpm/2Hp and 1730rpm/3 Hp; the method comprises the following steps of taking 1024 data points as sample data, giving a sample data label according to the state of a rolling bearing, forming a plurality of sample data to obtain a sample data set I, wherein the sample data set I comprises four different loads, the sample data set is taken as a variable load data set, and the specific conditions of the sample data set I are shown in a table 1:
TABLE 1
Step S12, obtaining a sample data set II: acquiring vibration signals of a rolling bearing under different rotating speeds and the same load through a public data set, wherein the rotating speed range [900rpm, 1620rpm ] of the rolling bearing, at the moment, the state of the rolling bearing comprises an outer ring fault, an inner ring fault and a normal state, at the moment, 1024 data points are taken as one sample data, a sample data label is given according to the state of the rolling bearing, a plurality of sample data in the state are combined to obtain a sample data set II, the sample data set II comprises four different rotating speeds, the sample data set II is taken as a variable rotating speed data set, and the concrete conditions of the sample data set II are shown in a table 2:
TABLE 2
And step S13, taking the sample data set I and the sample data set II as two variable working condition data sets, taking data of any working condition in the sample data set I or the sample data set II as a source domain data set, and taking other working condition data in the sample data set I or the sample data set II as a target domain data set.
Step S2, inputting the source domain data set and the target domain data set into a one-dimensional convolution self-coding encoder respectively to obtain a source domain characteristic parameter matrix and a target domain characteristic parameter matrix, where the one-dimensional convolution self-coding comprises an encoder and a decoder, the one-dimensional convolution self-coding structure is shown in fig. 2, where the encoder comprises 5 layers of convolution layers, and the decoder comprises 7 layers of deconvolution layers.
Step S3, inputting the source domain characteristic parameter matrix and the target domain characteristic parameter matrix in the step S2 into a decoder of one-dimensional convolution self-coding to obtain a reconstructed sample data set, and calculating through the reconstructed sample data set and the sample data set to obtain a mean square error loss function:
where x is the input to the encoder for one-dimensional convolutional self-encoding, i is the sample number, xiFor an encoderIs the output of the decoder, y is the ith input sample of (1)iIs the output of the ith sample decoder, and n is the number of samples.
Step S4, inputting the source domain characteristic parameter matrix into a convolutional neural network, outputting the state type of the source domain and calculating to obtain a cross entropy loss function;
wherein C is the number of bearing state classes, z is the output of the convolutional neural network, j is the class number, j takes the value of 0,1,2 … C-1,
the j-dimension output value of the ith sample after passing through the convolutional neural network is obtained; 1 {. is an exponential function when
1{ · } -, 1; when in use
When the value is 1{ · } -, 0; p is
The probability of occurrence.
The convolutional neural network consists of two convolutional layers and two full-connection layers.
The structure of the bearing fault diagnosis model under the variable working condition based on the one-dimensional convolution self-coding is shown in fig. 4, and it can be seen that the model can be divided into three parts: decoder, encoder and convolutional neural network, the specific network parameters are shown in table 3:
TABLE 3
Step S5, obtaining a total loss function of the sample data set through summation of the mean square error loss function and the cross entropy loss function, updating parameters of the one-dimensional convolutional self-coding and convolutional neural network by using a back propagation algorithm, and stopping updating the parameters of the one-dimensional convolutional self-coding and convolutional neural network until the total loss of the sample data set is 0:
specifically, a mean square error loss function and a cross entropy loss function are summed to obtain a total loss function of the sample data set, an Adam optimizer is adopted, the learning rate is 0.0003, the number of training rounds is 600, the size of a small batch mini-batch is 32, parameters of the one-dimensional convolutional self-coding network and the convolutional neural network are iteratively updated by using a back propagation algorithm until the total loss function result of the sample data set tends to 0, the updated parameters of the one-dimensional convolutional self-coding network and the convolutional neural network tend to be stable, the rolling bearing fault diagnosis model training under the variable working condition of the one-dimensional convolutional self-coding is finished, the iterative updating process is stopped, and the step S6 is entered.
The total loss function of the sample data set is shown in fig. 5, and it can be seen that the whole training process is good, the total loss function rapidly decreases at the initial stage of training, after 30 rounds, the total loss function has reached 0.01, the diagnosis accuracy of the verification set has reached 100%, and the process is a process in which the total loss functions of the convolutional neural network and the one-dimensional convolutional self-coding neural network both rapidly decrease; the total loss function then starts to decline smoothly and slowly, the process is mainly the decline process of the total loss function of the one-dimensional convolution self-coding, and the loss function is 0.0009 in 600 rounds, and the decline is not shown because the decline is slow.
Fig. 6 shows a comparison graph of original data and reconstructed data after passing through a one-dimensional convolutional self-coding network under the condition of 0 load, wherein the left four oscillograms are the original data, the right four oscillograms are the reconstructed data, and the types of states from top to bottom are respectively a normal state, an inner ring fault, a rolling element fault and an outer ring fault. It can be seen from fig. 6 that the waveform diagrams of the original data and the reconstructed data are substantially consistent, and since the one-dimensional convolutional self-coding has a denoising function, the two diagrams have slight differences in some places, and in addition, after 600 rounds of training, the loss function reaches 0.0009, which also indicates that the training effect of the one-dimensional convolutional self-coding network is better.
And step S6, calculating the diagnosis accuracy of the target domain data set.
Step S61: and (4) inputting the target domain data set in the S2 into a trained encoder of the rolling bearing fault diagnosis model based on the one-dimensional convolution self-coding under the variable working condition to obtain a target domain characteristic parameter matrix.
Step S62: and when the parameters of the one-dimensional convolutional self-coding and the convolutional neural network stop updating, obtaining a trained fault diagnosis model of the rolling bearing under the variable working condition based on the one-dimensional convolutional self-coding, inputting the target domain characteristic parameter matrix in the step S61 into the trained fault diagnosis model of the rolling bearing under the variable working condition based on the one-dimensional convolutional self-coding, outputting to obtain the state class of the bearing of the screened target domain, comparing the screened working condition class of the target domain with the sample data label of the target domain, and calculating to obtain the diagnosis and identification accuracy of the fault of the rolling bearing.
In order to analyze the diagnosis performance of the rolling bearing fault diagnosis model under the variable working condition of the one-dimensional convolution self-coding more specifically, three models are selected for comparison, wherein the three models comprise a convolution neural network, a residual convolution neural network and a support vector machine. The convolutional neural network is a model obtained by removing a decoder from the model in the invention; the residual convolution neural network adopts a classical network structure; the support vector machine extracts characteristic parameters including 6 time domain characteristics (root mean square, kurtosis, skewness, form factor, peak factor and standard deviation), 5 frequency domain characteristics (center frequency, mean square frequency, root mean square frequency, frequency variance, spectrum divergence) and 4 time frequency characteristics (amplitude at rotation order and fault characteristic order in envelope order spectrum).
Firstly, a data set I (variable load data set) is used for verification, a data set of a certain working condition is used as a training set, and data sets of other working conditions are used as a test set, so that the diagnosis accuracy of each model is shown in a table 4, wherein each diagnosis task in each model is a result of three times of training and prediction averaging. It can be seen that for the A diagnosis task, namely, the diagnosis accuracy of each model of the test set under the working condition of 0hp load is higher, the three models based on the convolutional neural network can reach 100%, the support vector machine reaches 97.6%, but for the data under the variable working condition, the diagnosis accuracy of each model is reduced; comparing B, C, D the accuracy of the diagnosis task and the average value of the three, it can be seen that the fault diagnosis accuracy of the support vector machine based on the manual feature extraction is low under three variable working conditions, and the reason may be that the extracted features do not have good expression capability and the limitation of the model itself under the variable working conditions; the average diagnosis accuracy of the three models based on the convolutional neural network is more than 92%, the diagnosis capability of the residual convolutional neural network under variable working conditions is stable and good in performance, and the diagnosis accuracy is higher on three diagnosis tasks compared with that of a simple convolutional neural network model; the model provided by the invention is added with the one-dimensional convolution self-coding network on the basis of the convolution neural network model, so that the diagnosis accuracy is greatly improved, the diagnosis tasks B and C reach more than 99.5%, the diagnosis task D has lower accuracy of 93.5%, and the model still has advantages compared with other models.
TABLE 4
And then, verifying again by using a data set II (variable rotation speed data set), using the data set with the rotation speed of 900-1080rpm as a training set, using the data sets with the rest of the rotation speeds as a test set, wherein the diagnosis accuracy of each model under each working condition is shown in FIG. 7, and each model has four diagnosis tasks.
Although the preferred embodiments of the present invention have been described in detail, the present invention is not limited to the details of the embodiments, and various equivalent modifications can be made within the technical spirit of the present invention, and the scope of the present invention is also within the scope of the present invention.
Claims (10)
1. A rolling bearing fault diagnosis method under variable working conditions based on convolution self-coding is characterized in that: the method specifically comprises the following steps:
step S1, obtaining a sample data set: acquiring vibration signals of a rolling bearing under a plurality of working conditions through a public data set, and segmenting the acquired vibration signals by using a data overlapping segmentation method to obtain a sample data set of the vibration signals, wherein the sample data set comprises a source domain data set and a target domain data set;
step S2, respectively inputting the source domain data set and the target domain data set into a one-dimensional convolution self-coding encoder to obtain a source domain characteristic parameter matrix and a target domain characteristic parameter matrix;
s3, inputting the source domain characteristic parameter matrix and the target domain characteristic parameter matrix in the step S2 into a one-dimensional convolution self-coding decoder to obtain a reconstructed sample data set, and calculating through the reconstructed sample data set and the sample data set to obtain a mean square error loss function;
step S4, inputting the source domain characteristic parameter matrix into a convolutional neural network, outputting the state type of the source domain and calculating to obtain a cross entropy loss function;
s5, obtaining a total loss function of the sample data set through summation of a mean square error loss function and a cross entropy loss function, updating parameters of the one-dimensional convolutional self-coding and convolutional neural network by using a back propagation algorithm, stopping updating the parameters of the one-dimensional convolutional self-coding and convolutional neural network until the total loss of the sample data set is 0, and entering the step S6;
and step S6, calculating the diagnosis accuracy of the target domain data set.
2. The rolling bearing fault diagnosis method under the variable working condition based on the convolution self-coding as claimed in claim 1, characterized in that: the step S1 specifically includes the following steps:
step S11, obtaining a sample data set I: acquiring vibration signals of a rolling bearing under the same rotating speed and different loads through a public data set, wherein the rolling bearing comprises four states and operates under four working conditions, and the four states comprise a rolling element fault, an outer ring fault, an inner ring fault and a normal state; the four working conditions comprise 1797rpm/0Hp, 1772rpm/1Hp, 1750rpm/2Hp and 1730rpm/3 Hp; the method comprises the steps of giving a sample data label according to the rolling bearing state by taking 1024 data points as sample data, forming a plurality of sample data to obtain a sample data set I, wherein the sample data set I comprises four different loads, taking the sample data set I as a variable load data set, taking data of any working condition in the sample data set I as a source domain data set, and taking data of other working conditions in the sample data set I as a target domain data set.
3. The rolling bearing fault diagnosis method under the variable working condition based on the convolution self-coding as claimed in claim 2, characterized in that: the step S1 specifically includes the following steps:
step S12, obtaining a sample data set II: the method comprises the steps of obtaining vibration signals of the rolling bearing under different rotating speeds and the same load through a public data set, wherein the rotating speed range [900rpm, 1620rpm ] of the rolling bearing, at the moment, the state of the rolling bearing comprises an outer ring fault, an inner ring fault and a normal state, at the moment, 1024 data points are used as one sample data, a sample data label is given according to the state of the rolling bearing, a plurality of sample data in the state are combined to obtain a sample data set II, the sample data set II comprises four different rotating speeds, the sample data set II is used as a variable rotating speed data set, data of any working condition in the sample data set II is used as a source domain data set, and data of other working conditions in the sample data set II is used as a target domain data set.
4. The rolling bearing fault diagnosis method under the variable working condition based on the convolution self-coding as claimed in claim 1, characterized in that: the one-dimensional convolutional self-encoding in step S2 includes an encoder composed of 5 convolutional layers and a decoder composed of 7 deconvolution layers.
5. The rolling bearing fault diagnosis method under the variable working condition based on the convolution self-coding as claimed in claim 1, characterized in that: the step S3 is specifically as follows:
inputting the source domain characteristic parameter matrix and the target domain characteristic parameter matrix obtained in the step S2 into a decoder of one-dimensional convolution self-coding to obtain reconstructed sample data and calculate a mean square error loss function, inputting the source domain characteristic parameter matrix and the target domain characteristic parameter matrix obtained in the step S2 into the decoder of one-dimensional convolution self-coding to obtain a reconstructed sample data set, and calculating the mean square error loss function according to the reconstructed sample data set and the sample data set:
where x is the input to the encoder for one-dimensional convolutional self-encoding, i is the sample number, xiIs the i-th input sample of the encoder, y is the output of the decoder, yiIs the output of the ith sample decoder, and n is the number of samples.
6. The rolling bearing fault diagnosis method under the variable working condition based on the convolution self-coding as claimed in claim 1, characterized in that: the convolutional neural network in step S4 is composed of two convolutional layers and two fully-connected layers.
7. The rolling bearing fault diagnosis method under the variable working condition based on the convolution self-coding as claimed in claim 4, characterized in that: the cross entropy loss function is:
wherein C is the number of bearing state classes, z is the output of the convolutional neural network, j is the class number, j takes the value of 0,1,2 … C-1,
the j-dimension output value of the ith sample after passing through the convolutional neural network is obtained; 1 {. is an exponential function when
1{ · } -, 1; when in use
When the value is 1{ · } -, 0; p is
The probability of occurrence.
8. The rolling bearing fault diagnosis method under the variable working condition based on the convolution self-coding as claimed in claim 1, characterized in that: the step S6 specifically includes the following steps:
step S61: inputting the target domain data set in the S2 into a trained one-dimensional convolution self-coding encoder to obtain a target domain characteristic parameter matrix;
step S62: and when the parameters of the one-dimensional convolutional self-coding and convolutional neural network stop updating, obtaining the trained one-dimensional convolutional self-coding and convolutional neural network, inputting the target domain characteristic parameter matrix in the step S61 into the trained one-dimensional convolutional self-coding and convolutional neural network, outputting to obtain the bearing state category of the screened target domain, comparing the bearing state category of the screened target domain with the sample data label of the target domain, and calculating to obtain the diagnosis and identification precision of the rolling bearing fault.
9. The rolling bearing fault diagnosis method under the variable working condition based on the convolution self-coding as claimed in claim 1, characterized in that:
before the sample data set is obtained, a rolling bearing fault diagnosis model under the variable working condition of one-dimensional convolution self-coding is established.
10. The rolling bearing fault diagnosis method under the variable working condition based on the convolution self-coding as claimed in claim 8, characterized in that: the rolling bearing fault diagnosis model under the variable working condition of the one-dimensional convolution self-coding comprises the one-dimensional convolution self-coding and a convolution neural network.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111258171.0A CN114136619A (en) | 2021-10-27 | 2021-10-27 | Rolling bearing fault diagnosis method under variable working conditions based on convolution self-coding |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111258171.0A CN114136619A (en) | 2021-10-27 | 2021-10-27 | Rolling bearing fault diagnosis method under variable working conditions based on convolution self-coding |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114136619A true CN114136619A (en) | 2022-03-04 |
Family
ID=80394632
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111258171.0A Pending CN114136619A (en) | 2021-10-27 | 2021-10-27 | Rolling bearing fault diagnosis method under variable working conditions based on convolution self-coding |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114136619A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115310498A (en) * | 2022-10-12 | 2022-11-08 | 青岛明思为科技有限公司 | Neural network-based rotating machine fault classification method under variable rotating speed working condition |
| CN115389247A (en) * | 2022-11-01 | 2022-11-25 | 青岛睿发工程咨询服务合伙企业(有限合伙) | Rotating machinery fault monitoring method based on speed self-adaptive encoder |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110210381A (en) * | 2019-05-30 | 2019-09-06 | 盐城工学院 | A kind of adaptive one-dimensional convolutional neural networks intelligent failure diagnosis method of domain separation |
| CN110361176A (en) * | 2019-06-05 | 2019-10-22 | 华南理工大学 | A kind of intelligent failure diagnosis method for sharing neural network based on multitask feature |
| CN111860677A (en) * | 2020-07-29 | 2020-10-30 | 湖南科技大学 | Rolling bearing transfer learning fault diagnosis method based on partial domain confrontation |
| CN112434602A (en) * | 2020-11-23 | 2021-03-02 | 西安交通大学 | Fault diagnosis method based on migratable common feature space mining |
| CN112665852A (en) * | 2020-11-30 | 2021-04-16 | 南京航空航天大学 | Variable working condition planetary gearbox fault diagnosis method and device based on deep learning |
| CN112729835A (en) * | 2021-01-30 | 2021-04-30 | 温州大学 | Multi-domain semi-supervised fault diagnosis method and device for axial plunger pump bearing |
| CN113033309A (en) * | 2021-02-25 | 2021-06-25 | 北京化工大学 | Fault diagnosis method based on signal downsampling and one-dimensional convolution neural network |
| CN113095413A (en) * | 2021-04-14 | 2021-07-09 | 山东建筑大学 | Variable working condition fault diagnosis method, system, storage medium and equipment |
| CN113378778A (en) * | 2021-06-30 | 2021-09-10 | 东南大学 | On-load tap-changer fault diagnosis method based on self-encoder |
-
2021
- 2021-10-27 CN CN202111258171.0A patent/CN114136619A/en active Pending
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110210381A (en) * | 2019-05-30 | 2019-09-06 | 盐城工学院 | A kind of adaptive one-dimensional convolutional neural networks intelligent failure diagnosis method of domain separation |
| CN110361176A (en) * | 2019-06-05 | 2019-10-22 | 华南理工大学 | A kind of intelligent failure diagnosis method for sharing neural network based on multitask feature |
| CN111860677A (en) * | 2020-07-29 | 2020-10-30 | 湖南科技大学 | Rolling bearing transfer learning fault diagnosis method based on partial domain confrontation |
| CN112434602A (en) * | 2020-11-23 | 2021-03-02 | 西安交通大学 | Fault diagnosis method based on migratable common feature space mining |
| CN112665852A (en) * | 2020-11-30 | 2021-04-16 | 南京航空航天大学 | Variable working condition planetary gearbox fault diagnosis method and device based on deep learning |
| CN112729835A (en) * | 2021-01-30 | 2021-04-30 | 温州大学 | Multi-domain semi-supervised fault diagnosis method and device for axial plunger pump bearing |
| CN113033309A (en) * | 2021-02-25 | 2021-06-25 | 北京化工大学 | Fault diagnosis method based on signal downsampling and one-dimensional convolution neural network |
| CN113095413A (en) * | 2021-04-14 | 2021-07-09 | 山东建筑大学 | Variable working condition fault diagnosis method, system, storage medium and equipment |
| CN113378778A (en) * | 2021-06-30 | 2021-09-10 | 东南大学 | On-load tap-changer fault diagnosis method based on self-encoder |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115310498A (en) * | 2022-10-12 | 2022-11-08 | 青岛明思为科技有限公司 | Neural network-based rotating machine fault classification method under variable rotating speed working condition |
| CN115389247A (en) * | 2022-11-01 | 2022-11-25 | 青岛睿发工程咨询服务合伙企业(有限合伙) | Rotating machinery fault monitoring method based on speed self-adaptive encoder |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109218114B (en) | Decision tree-based server fault automatic detection system and detection method | |
| CN110163261B (en) | Unbalanced data classification model training method, device, equipment and storage medium | |
| CN114136619A (en) | Rolling bearing fault diagnosis method under variable working conditions based on convolution self-coding | |
| CN108241873B (en) | A kind of intelligent failure diagnosis method towards pumping plant main equipment | |
| CN110703057B (en) | Power equipment partial discharge diagnosis method based on data enhancement and neural network | |
| CN111680820B (en) | Distributed photovoltaic power station fault diagnosis method and device | |
| CN111046916A (en) | Motor fault diagnosis method and system based on void convolution capsule network | |
| CN110647830B (en) | Bearing fault diagnosis method based on convolutional neural network and Gaussian mixture model | |
| CN113177357B (en) | Transient stability assessment method for power system | |
| CN113111820B (en) | Rotary part fault diagnosis method and device based on improved CNN and relation module | |
| CN111238815B (en) | Bearing fault identification method based on data enhancement under sample imbalance | |
| CN111582396A (en) | Fault diagnosis method based on improved convolutional neural network | |
| CN107194415A (en) | Peak clustering method based on Laplace centrality | |
| CN110458189A (en) | Compressed sensing and depth convolutional neural networks Power Quality Disturbance Classification Method | |
| CN117150359A (en) | Small sample fault diagnosis method, system, device and medium based on model independent element learning | |
| CN115452362A (en) | Fault diagnosis method for gear box | |
| CN115186564A (en) | Analog circuit fault diagnosis method based on feature fusion and improved particle swarm algorithm | |
| CN110307981B (en) | Bearing fault diagnosis method based on PNN-IFA | |
| CN114648060A (en) | Fault signal standardization processing and classification method based on machine learning | |
| CN112380932B (en) | Vibration signal characteristic value selection method and elevator health state evaluation or fault diagnosis method | |
| CN117170980A (en) | Early warning method, device, equipment and storage medium for server hardware abnormality | |
| CN107016440A (en) | The multiresolution deep neural network intelligent diagnosing method of machine driving failure | |
| CN116222753A (en) | Rotor system fault sensitivity feature extraction method and system | |
| CN115906959A (en) | Parameter training method of neural network model based on DE-BP algorithm | |
| CN115291091A (en) | Analog circuit fault diagnosis method based on graph neural network |
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 |