A rolling bearing fault diagnosis method under variable working conditions based on convolutional autoencoder
By combining one-dimensional convolutional autoencoders and convolutional neural networks, the fault features of rolling bearings under varying operating conditions are automatically extracted, solving the problems of strong dependence on manual feature extraction and training difficulties in existing technologies, and achieving high-accuracy fault diagnosis.
Patent Information
- Application Number
- CN202111258171.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-10-27
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2041-10-27
Smart Images

Figure CN114136619B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of intelligent diagnosis of bearing faults, and in particular to a rolling bearing fault diagnosis method under variable working conditions based on one-dimensional convolutional autoencoding. Background Art
[0002] Rolling bearings are commonly used components in rotating machinery and equipment, and their health status is related to the reliability and stability of equipment operation. However, they frequently fail under high speed and heavy load conditions for a long time. If they are not maintained in time, other parts of the equipment will be damaged and cause greater economic losses. Therefore, it is of great significance to accurately identify the health status of rolling bearings. Affected by various factors such as demand and environment, rolling bearings do not operate under a single working condition, but often operate under variable speed and variable load conditions. The change of working conditions increases the complexity of the mapping relationship between fault characteristics and fault modes to a certain extent, which brings great challenges to the health status monitoring of mechanical equipment. Domestic and foreign scholars have carried out a lot of research on the fault diagnosis of rolling bearings. The diagnosis method based on digital signal processing was studied earlier. In recent years, with the continuous deepening of machine learning research, great progress has been made in the use of data-driven diagnosis methods, such as support vector machines, random forests, and deep belief networks. This fault diagnosis method often requires manual feature extraction, has a strong dependence on expert experience, and has certain limitations in application under variable working conditions. In order to get rid of the limitations brought by manual feature extraction, deep learning methods have gradually been applied in bearing fault diagnosis under variable working conditions. This type of method can automatically learn feature extraction, but there are often problems such as convergence difficulties and slow training speed when training deep learning models, which affects its widespread application. Summary of the invention
[0003] The present invention proposes a rolling bearing fault diagnosis method under variable working conditions based on one-dimensional convolutional autoencoder. By combining the one-dimensional convolutional autoencoder with a convolutional neural network, the fault characteristics of the rolling bearing under variable working conditions are automatically extracted and identified, thereby realizing adaptive feature extraction and domain-adaptive rolling bearing fault diagnosis under variable working conditions.
[0004] In order to solve the above technical problems, the technical solution adopted by the present invention is:
[0005] A rolling bearing fault diagnosis method under variable working conditions based on convolutional autoencoder specifically comprises the following steps:
[0006] Step S1, obtaining a sample data set: obtaining vibration signals of rolling bearings under multiple working conditions through public data sets, and segmenting the collected vibration signals using a data overlapping segmentation method to obtain a sample data set of vibration signals, wherein the sample data set includes a source domain data set and a target domain data set;
[0007] Step S2: input the source domain data set and the target domain data set into the encoder of the one-dimensional convolutional autoencoder respectively to obtain the source domain feature parameter matrix and the target domain feature parameter matrix;
[0008] Step S3, inputting the source domain feature parameter matrix and the target domain feature parameter matrix in step S2 into the decoder of the one-dimensional convolutional autoencoder to obtain a reconstructed sample data set, and calculating the mean square error loss function through the reconstructed sample data set and the sample data set;
[0009] Step S4: input the source domain feature parameter matrix into the convolutional neural network, output the state category of the source domain and calculate the cross entropy loss function;
[0010] Step S5, obtain the total loss function of the sample data set by summing the mean square error loss function and the cross entropy loss function, and then use the back propagation algorithm to update the parameters of the one-dimensional convolutional autoencoder and the convolutional neural network until the total loss of the sample data set is 0, stop updating the parameters of the one-dimensional convolutional autoencoder and the convolutional neural network, and enter step S6;
[0011] Step S6: Calculate the diagnostic accuracy of the target domain data set.
[0012] As a further preferred embodiment of the present invention, step S1 specifically comprises the following steps:
[0013] Step S11, obtain sample data set I: obtain vibration signals of rolling bearings at the same speed and different loads through public data sets, the rolling bearings include four states and operate under four working conditions, the four states include rolling element fault, outer ring fault, inner ring fault and normal state; the four working conditions include 1797rpm / 0Hp, 1772rpm / 1Hp, 1750rpm / 2Hp and 1730rpm / 3Hp; take 1024 data points as one sample data, and assign sample data labels according to the rolling bearing state, and combine multiple sample data to obtain sample data set I, the sample data set I includes four different loads, take sample data set I as a variable load data set, take data of any working condition in sample data set I as a source domain data set, and the data of other working conditions in sample data set I are the target domain data set.
[0014] As a further preferred embodiment of the present invention, step S1 specifically further includes the following steps:
[0015] Step S12, obtain sample data set II: obtain the vibration signal of the rolling bearing under different speeds and the same load through the public data set, the speed range of the rolling bearing is [900rpm, 1620rpm]. At this time, the state of the rolling bearing includes outer ring fault, inner ring fault and normal state. At this time, 1024 data points are taken as a sample data, and the sample data label is assigned according to the rolling bearing state. Multiple sample data under this state are combined to obtain sample data set II. The sample data set II includes four different speeds. Sample data set II is used as a variable speed data set, and the data of any working condition in sample data set II is used as the source domain data set, and the other working condition data in sample data set II are the target domain data set.
[0016] As a further preferred embodiment of the present invention, the one-dimensional convolutional autoencoder in step S2 includes an encoder and a decoder, the encoder consists of 5 convolutional layers, and the decoder consists of 7 deconvolutional layers.
[0017] As a further preferred embodiment of the present invention, the step S3 is as follows:
[0018] The source domain feature parameter matrix and the target domain feature parameter matrix obtained in step S2 are input into the decoder of the one-dimensional convolutional autoencoder to obtain the reconstructed sample data set and calculate the mean square error loss function. The source domain feature parameter matrix and the target domain feature parameter matrix in step S2 are input into the decoder of the one-dimensional convolutional autoencoder to obtain the reconstructed sample data set. The mean square error loss function is calculated by the reconstructed sample data set and the sample data set:
[0019]
[0020] Among them, x is the input of the encoder of the one-dimensional convolutional self-encoder, i is the sample number, x i is the i-th input sample of the encoder, y is the output of the decoder, y i is the output of the i-th sample decoder, and n is the number of samples.
[0021] As a further preferred embodiment of the present invention, the convolutional neural network in step S4 consists of two convolutional layers and two fully connected layers.
[0022] As a further preferred embodiment of the present invention, the cross entropy loss function is:
[0023]
[0024] Where c is the number of bearing status categories, z is the output of the convolutional neural network, j is the category number, and j takes values of 0, 1, 2...C-1. is the output value of the jth dimension after the i-th sample passes through the convolutional neural network; 1{·} is an exponential function. When , 1{·}=1; when When 1{·}=0; P is Probability of occurrence.
[0025] As a further preferred embodiment of the present invention, step S6 specifically comprises the following steps:
[0026] Step S61: input the target domain data set described in S2 into the encoder of the trained one-dimensional convolutional autoencoder to obtain a target domain feature parameter matrix;
[0027] Step S62: When the parameters of the one-dimensional convolutional autoencoder and the convolutional neural network stop updating, the trained one-dimensional convolutional autoencoder and the convolutional neural network are obtained, the target domain feature parameter matrix in step S61 is input into the trained one-dimensional convolutional autoencoder and the convolutional neural network, the identified bearing state category of the target domain is output, the identified bearing state category of the target domain is compared with the sample data label of the target domain, and the diagnostic recognition accuracy of the rolling bearing fault is calculated.
[0028] As a further preferred solution of the present invention, before obtaining the sample data set, a rolling bearing fault diagnosis model under variable working conditions of a one-dimensional convolutional autoencoder is established.
[0029] As a further preferred embodiment of the present invention, the rolling bearing fault diagnosis model under variable working conditions of the one-dimensional convolutional autoencoder includes a one-dimensional convolutional autoencoder and a convolutional neural network.
[0030] The present invention has the following beneficial effects:
[0031] 1. A rolling bearing fault diagnosis method under variable working conditions based on convolutional autoencoder disclosed in the present invention combines one-dimensional convolutional autoencoder with convolutional neural network to automatically extract and identify the fault characteristics of rolling bearings under variable working conditions without requiring a large amount of test data, thereby achieving domain adaptation and greatly improving the accuracy of rolling bearing fault diagnosis under variable working conditions;
[0032] 2. The present invention discloses a rolling bearing fault diagnosis method under variable working conditions based on convolutional autoencoding. Since the encoder, decoder and neural network are all composed of convolutional layers and deconvolutional layers, the training parameters are not too many, the training process is more stable, and has the advantages of easy convergence and convenient training. BRIEF DESCRIPTION OF THE DRAWINGS
[0033] Figure 1 is a flow chart of the present invention;
[0034] Figure 2 It is a convolutional autoencoder network structure diagram of the present invention;
[0035] Figure 3 It is a convolutional neural network structure diagram of the present invention;
[0036] Figure 4 It is a structural diagram of a bearing fault diagnosis model of the present invention;
[0037] Figure 5 is a graph of the loss function of the model of the present invention during the training process;
[0038] Figure 6 This is a one-dimensional convolutional autoencoder training effect diagram of the present invention;
[0039] Figure 7 It is a comparison chart of the diagnostic accuracy of the model of the present invention and the comparison model. DETAILED DESCRIPTION
[0040] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific preferred embodiments.
[0041] In the description of the present invention, it should be understood that the terms "left side", "right side", "upper part", "lower part" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, which is only for the convenience of describing the present invention and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. "First", "second" and the like do not indicate the importance of the components, and therefore cannot be understood as limiting the present invention. The specific dimensions used in this embodiment are only for illustrating the technical solution, and do not limit the scope of protection of the present invention.
[0042] like Figure 1 As shown, a rolling bearing fault diagnosis method under variable working conditions based on convolutional autoencoder specifically includes the following steps:
[0043] Firstly, a rolling bearing fault diagnosis model under variable working conditions of a one-dimensional convolutional autoencoder is established. The rolling bearing fault diagnosis model under variable working conditions of the one-dimensional convolutional autoencoder includes a one-dimensional convolutional autoencoder and a convolutional neural network.
[0044] Step S1, obtaining a sample data set: obtaining vibration signals of rolling bearings under multiple working conditions through public data sets, and segmenting the collected vibration signals using a data overlapping segmentation method to obtain a sample data set of vibration signals, wherein the sample data set includes a source domain data set and a target domain data set.
[0045] Step S11, obtaining sample data set I: obtaining vibration signals of rolling bearings at the same speed and different loads through public data sets, the rolling bearings include four states and operate under four working conditions, the four states include rolling element fault, outer ring fault, inner ring fault and normal state; the four working conditions include 1797rpm / 0Hp, 1772rpm / 1Hp, 1750rpm / 2Hp and 1730rpm / 3Hp; taking 1024 data points as one sample data, and assigning sample data labels according to the rolling bearing state, combining multiple sample data to obtain sample data set I, the sample data set I includes four different loads, and the sample data set is used as a variable load data set. The specific situation of sample data set I is shown in Table 1:
[0046] Table 1
[0047]
[0048]
[0049] Step S12, obtain sample data set II: obtain the vibration signal of the rolling bearing at different speeds and the same load through the public data set, the speed range of the rolling bearing is [900rpm, 1620rpm]. At this time, the state of the rolling bearing includes outer ring fault, inner ring fault and normal state. At this time, 1024 data points are used as a sample data, and the sample data label is assigned according to the rolling bearing state. Multiple sample data under this state are combined to obtain sample data set II. The sample data set II includes four different speeds. The sample data set II is used as a variable speed data set. The specific situation of the sample data set II is shown in Table 2:
[0050] Table 2
[0051]
[0052] Step S13: Take sample data set I and sample data set II as two variable working condition data sets, take the data of any working condition in sample data set I or sample data set II as the source domain data set, and the other working condition data in sample data set I or sample data set II as the target domain data set.
[0053] Step S2: Input the source domain dataset and the target domain dataset into the encoder of the one-dimensional convolutional autoencoder respectively to obtain the source domain feature parameter matrix and the target domain feature parameter matrix, wherein the one-dimensional convolutional autoencoder is composed of an encoder and a decoder, and the one-dimensional convolutional autoencoder structure is as follows: Figure 2 As shown in the figure, the encoder consists of 5 convolutional layers and the decoder consists of 7 deconvolutional layers.
[0054] Step S3: Input the source domain feature parameter matrix and the target domain feature parameter matrix in step S2 into the decoder of the one-dimensional convolutional autoencoder to obtain a reconstructed sample data set, and calculate the mean square error loss function through the reconstructed sample data set and the sample data set:
[0055]
[0056] Among them, x is the input of the encoder of the one-dimensional convolutional self-encoder, i is the sample number, x i is the i-th input sample of the encoder, y is the output of the decoder, y i is the output of the i-th sample decoder, and n is the number of samples.
[0057] Step S4: input the source domain feature parameter matrix into the convolutional neural network, output the state category of the source domain and calculate the cross entropy loss function;
[0058]
[0059] Where c is the number of bearing status categories, z is the output of the convolutional neural network, j is the category number, and j takes values of 0, 1, 2...C-1. is the output value of the jth dimension after the i-th sample passes through the convolutional neural network; 1{·} is an exponential function. When , 1{·}=1; when When 1{·}=0; P is Probability of occurrence.
[0060] Among them, the above convolutional neural network consists of two convolutional layers and two fully connected layers.
[0061] The structure of the bearing fault diagnosis model under variable working conditions based on one-dimensional convolutional autoencoder is as follows: Figure 4 As shown, we can see that the model can be divided into three parts: decoder, encoder and convolutional neural network. The specific network parameters are shown in Table 3:
[0062] Table 3
[0063]
[0064] Step S5, the total loss function of the sample data set is obtained by summing the mean square error loss function and the cross entropy loss function, and then the back propagation algorithm is used to update the parameters of the one-dimensional convolutional autoencoder and the convolutional neural network until the total loss of the sample data set is 0, and then the updating of the parameters of the one-dimensional convolutional autoencoder and the convolutional neural network is stopped:
[0065] Specifically, the mean square error loss function and the cross entropy loss function are summed to obtain the total loss function of the sample data set. The Adam optimizer is used, the learning rate is 0.0003, the number of training rounds is 600, the mini-batch size is 32, and the back-propagation algorithm is used to iteratively update the parameters of the one-dimensional convolutional autoencoder network and the convolutional neural network until the total loss function result of the sample data set tends to 0, indicating that the parameters of the updated one-dimensional convolutional autoencoder network and the convolutional neural network tend to be stable, indicating that the training of the rolling bearing fault diagnosis model under variable working conditions of the one-dimensional convolutional autoencoder is completed, the iterative update process is stopped, and step S6 is entered.
[0066] The total loss function of the sample data set is as follows: Figure 5 As shown, it can be seen that the overall training process is good. In the early stage of training, the total loss function drops rapidly. After 30 rounds, the total loss function has reached 0.01, and the diagnostic accuracy of the verification set has reached 100%. This process is a process in which the total loss functions of both the convolutional neural network and the one-dimensional convolutional autoencoder neural network drop rapidly; after that, the total loss function begins to drop steadily and slowly. This process is mainly the drop process of the total loss function of the one-dimensional convolutional autoencoder. The loss function is 0.0009 at 600 rounds, which is not shown because of its slow drop.
[0067] Figure 6 The figure shows the comparison between the original data under zero load and the reconstructed data after passing through the one-dimensional convolutional autoencoder network. The four waveforms on the left are the original data, and the four waveforms on the right are the reconstructed data. From top to bottom, the state types are normal state, inner race fault, rolling element fault, and outer race fault. Figure 6 It can be seen that the waveforms of the original data and the reconstructed data are basically the same. Since the one-dimensional convolutional autoencoder has a denoising function, there are slight differences between the two graphs in some places. In addition, after 600 rounds of training, the loss function reaches 0.0009, which also shows that the one-dimensional convolutional autoencoder network has a good training effect.
[0068] Step S6: Calculate the diagnostic accuracy of the target domain data set.
[0069] Step S61: input the target domain data set described in S2 into the encoder of the rolling bearing fault diagnosis model under variable working conditions that has been trained based on one-dimensional convolutional autoencoder to obtain the target domain feature parameter matrix.
[0070] Step S62: When the parameters of the one-dimensional convolutional autoencoder and the convolutional neural network stop updating, the rolling bearing fault diagnosis model under variable working conditions based on the one-dimensional convolutional autoencoder after training is obtained, and the target domain feature parameter matrix in step S61 is input into the rolling bearing fault diagnosis model under variable working conditions based on the one-dimensional convolutional autoencoder after training. The state category of the bearing in the identified target domain is output, and the identified working condition category of the target domain is compared with the sample data label of the target domain to calculate the diagnostic recognition accuracy of the rolling bearing fault.
[0071] In order to more specifically analyze the diagnostic performance of the rolling bearing fault diagnosis model under variable working conditions of the one-dimensional convolutional autoencoder, three models were selected for comparison, including convolutional neural network, residual convolutional neural network and support vector machine. The convolutional neural network is the model of the present invention after removing the decoder; the residual convolutional neural network adopts the classic network structure; the support vector machine extracts feature parameters including 6 time domain features (root mean square, kurtosis, skewness, waveform factor, peak factor and standard deviation), 5 frequency domain features (center frequency, mean square frequency, root mean square frequency, frequency variance, spectrum divergence) and 4 time-frequency features (the amplitude at the rotation order and fault feature order in the envelope order spectrum).
[0072] First, Dataset I (variable load dataset) is used for verification. The dataset of a certain working condition is used as the training set, and the datasets of other working conditions are used as the test set. The diagnostic accuracy of each model is shown in Table 4, where each diagnostic task in each model is the result of three trainings and predictions and the average value is taken. It can be seen that for diagnostic task A, that is, the test set under the load of 0hp, the diagnostic accuracy of each model is relatively high. The three models based on convolutional neural networks can reach 100%, and the support vector machine can reach 97.6%. However, for data under variable working conditions, the diagnostic accuracy of each model has decreased. By comparing the accuracy of diagnostic tasks B, C, and D and the average of the three, it can be seen that the fault diagnosis accuracy of the support vector machine based on artificially extracted features under the three variable working conditions is relatively low. The reason may be that the extracted features do not have good expression ability and the model itself has limitations under variable working conditions. The average diagnostic accuracy of the three models based on convolutional neural networks is above 92%. The diagnostic ability of the residual convolutional neural network under variable working conditions is relatively stable and performs well. The diagnostic accuracy of the simple convolutional neural network model is higher in the three diagnostic tasks. The model proposed in the present invention adds a one-dimensional convolutional autoencoder network on the basis of the convolutional neural network model. It can be seen that the diagnostic accuracy has been greatly improved. The diagnostic tasks B and C have reached more than 99.5%, and the diagnostic task D has a lower accuracy of 93.5%, but it still has advantages over other models.
[0073] Table 4
[0074]
[0075] Then, the dataset II (variable speed dataset) was used for verification again. The dataset with a speed of 900-1080 rpm was used as the training set, and the dataset with other speeds was used as the test set. The diagnostic accuracy of each model under each working condition is shown in the following figure. Figure 7 As shown, each model has four diagnostic tasks. It can be seen that the results are basically consistent with those on data set I. The support vector machine, a machine learning model based on artificially extracted features, has the worst diagnostic effect. The model proposed in the present invention adds a one-dimensional convolutional autoencoder network to the convolutional neural network, and the diagnostic accuracy is greatly improved, which illustrates the effectiveness of the model proposed in the present invention.
[0076] The preferred embodiments of the present invention are described in detail above; however, the present invention is not limited to the specific details in the above embodiments. Within the technical concept of the present invention, various equivalent transformations can be made to the technical solutions of the present invention, and these equivalent transformations all belong to the protection scope of the present invention.
Claims
1. A rolling bearing fault diagnosis method under variable working conditions based on convolutional autoencoder, characterized in that: The specific steps include: Step S1, obtaining a sample data set: obtaining vibration signals of rolling bearings under multiple working conditions through public data sets, and segmenting the collected vibration signals using a data overlapping segmentation method to obtain a sample data set of vibration signals, wherein the sample data set includes a source domain data set and a target domain data set; Step S2: input the source domain data set and the target domain data set into the encoder of the one-dimensional convolutional autoencoder respectively to obtain the source domain feature parameter matrix and the target domain feature parameter matrix; Step S3, inputting the source domain feature parameter matrix and the target domain feature parameter matrix in step S2 into the decoder of the one-dimensional convolutional autoencoder to obtain a reconstructed sample data set, and calculating the mean square error loss function through the reconstructed sample data set and the sample data set; Step S4: input the source domain feature parameter matrix into the convolutional neural network, output the state category of the source domain and calculate the cross entropy loss function; Step S5, obtain the total loss function of the sample data set by summing the mean square error loss function and the cross entropy loss function, and then use the back propagation algorithm to update the parameters of the one-dimensional convolutional autoencoder and the convolutional neural network until the total loss of the sample data set is 0, stop updating the parameters of the one-dimensional convolutional autoencoder and the convolutional neural network, and enter step S6; Step S6, calculating the diagnostic accuracy of the target domain data set; The specific steps of step S3 are as follows: The source domain feature parameter matrix and the target domain feature parameter matrix obtained in step S2 are input into the decoder of the one-dimensional convolutional autoencoder to obtain the reconstructed sample data set and calculate the mean square error loss function: Among them, x is the input of the encoder of the one-dimensional convolutional self-encoder, i is the sample number, x i is the i-th input sample of the encoder, y is the output of the decoder, y i is the output of the i-th sample decoder, and n is the number of samples; The cross entropy loss function is: Where c is the number of bearing status categories, z is the output of the convolutional neural network, j is the category number, and j takes values of 0, 1, 2...C-1. is the output value of the jth dimension after the i-th sample passes through the convolutional neural network; 1{·} is an exponential function. When , 1{·}=1; when When 1{·}=0; P is Probability of occurrence; The step S6 specifically comprises the following steps: Step S61: input the target domain data set described in S2 into the encoder of the trained one-dimensional convolutional autoencoder to obtain a target domain feature parameter matrix; Step S62: When the parameters of the one-dimensional convolutional autoencoder and the convolutional neural network stop updating, the trained one-dimensional convolutional autoencoder and the convolutional neural network are obtained, the target domain feature parameter matrix in step S61 is input into the trained one-dimensional convolutional autoencoder and the convolutional neural network, the identified bearing state category of the target domain is output, the identified bearing state category of the target domain is compared with the sample data label of the target domain, and the diagnostic accuracy of the rolling bearing fault is calculated.
2. The rolling bearing fault diagnosis method under variable working conditions based on convolutional autoencoder according to claim 1 is characterized in that: The step S1 specifically includes the following steps: Step S11, obtain sample data set I: obtain vibration signals of rolling bearings at the same speed and different loads through public data sets, the rolling bearings include four states and operate under four working conditions, the four states include rolling element fault, outer ring fault, inner ring fault and normal state; the four working conditions include 1797rpm / 0Hp, 1772rpm / 1Hp, 1750rpm / 2Hp and 1730rpm / 3Hp; take 1024 data points as one sample data, and assign sample data labels according to the rolling bearing state, and combine multiple sample data to obtain sample data set I, the sample data set I includes four different loads, take sample data set I as a variable load data set, take data of any working condition in sample data set I as a source domain data set, and the data of other working conditions in sample data set I are the target domain data set.
3. The rolling bearing fault diagnosis method under variable working conditions based on convolutional autoencoder according to claim 2 is characterized in that: The step S1 specifically further comprises the following steps: Step S12, obtain sample data set II: obtain the vibration signal of the rolling bearing under different speeds and the same load through the public data set, the speed range of the rolling bearing is [900rpm, 1620rpm]. At this time, the state of the rolling bearing includes outer ring fault, inner ring fault and normal state. At this time, 1024 data points are taken as a sample data, and the sample data label is assigned according to the rolling bearing state. Multiple sample data under this state are combined to obtain sample data set II. The sample data set II includes four different speeds. Sample data set II is used as a variable speed data set, and the data of any working condition in sample data set II is used as the source domain data set, and the other working condition data in sample data set II are the target domain data set.
4. The rolling bearing fault diagnosis method under variable working conditions based on convolutional autoencoder according to claim 1 is characterized in that: The one-dimensional convolutional autoencoder in step S2 includes an encoder and a decoder, the encoder consists of 5 convolutional layers, and the decoder consists of 7 deconvolutional layers.
5. The rolling bearing fault diagnosis method under variable working conditions based on convolutional autoencoder according to claim 1 is characterized in that: The convolutional neural network in step S4 consists of two convolutional layers and two fully connected layers.
6. The rolling bearing fault diagnosis method under variable working conditions based on convolutional autoencoder according to claim 1 is characterized in that: Before obtaining the sample data set, a rolling bearing fault diagnosis model under variable working conditions based on a one-dimensional convolutional autoencoder is established.
7. The rolling bearing fault diagnosis method under variable working conditions based on convolutional autoencoder according to claim 6 is characterized in that: The rolling bearing fault diagnosis model under variable working conditions of the one-dimensional convolutional autoencoder includes a one-dimensional convolutional autoencoder and a convolutional neural network.
Citation Information
Cited By
Variable working condition domain adaptive fault diagnosis method for intelligent operation and maintenance of rolling bearing
CN122671158A
