Time-varying rotating speed bearing fault diagnosis method and device

By constructing a time-varying speed bearing fault diagnosis model, and comprehensively considering time and space information, the problem of accuracy in rolling bearing fault diagnosis under time-varying speed conditions is solved, and accurate fault classification is achieved in noisy environments.

CN118981708AActive Publication Date: 2024-11-19ZHENGZHOU EXPERT TECH CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202411032859.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-30
Publication Date
2024-11-19
Estimated Expiration
2044-07-30

AI Technical Summary

Technical Problem

Existing technologies for diagnosing rolling bearing faults under time-varying speed conditions cannot accurately identify fault characteristics, especially in environments with external interference and noise, where the diagnostic results are inaccurate.

Method used

A time-varying speed bearing fault diagnosis model is adopted, which includes two channels and a classifier. The first channel obtains time fault information through a gated recurrent unit, and the second channel obtains spatial fault information through a multi-scale wide kernel layer and a small kernel convolution module. The outputs of the two are combined to perform fault diagnosis.

Benefits of technology

It improves the accuracy of fault diagnosis for time-varying speed bearings, can accurately classify fault types in noisy environments, adapts to complex working conditions, and enhances the interdependence between time series data.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118981708A_ABST
    Figure CN118981708A_ABST
Patent Text Reader

Abstract

The invention belongs to the technical field of variable speed bearing faults, and particularly relates to a time-varying rotating speed bearing fault diagnosis method and device. The method comprises the following steps: inputting bearing operation data in a prediction time period into a time-varying rotating speed bearing fault diagnosis model to obtain a fault type result; wherein the operation data comprises vibration data and rotation speed pulse data; the time-varying rotating speed bearing fault diagnosis model comprises two channels and a classifier; the first channel is used for obtaining time fault information of the operation data, the second channel is used for obtaining space fault information of the operation data, and the classifier is used for combining output of the first channel and output of the second channel to complete fault diagnosis. According to the time-varying rotating speed bearing fault diagnosis method, the time information and the space information of the input signals are comprehensively considered, the interdependence between the time sequence data is enhanced, the space fault information and the time fault information of the input signals can be obtained at the same time, and the accuracy of fault diagnosis of the time-varying rotating speed bearing is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of variable speed bearing faults, and in particular relates to a method and a device for diagnosing time-varying speed bearing faults. Background Art

[0002] Rolling bearings play a vital role in various industrial machinery. Due to the complexity of the working conditions, rolling bearings are also components that often fail. Therefore, timely fault diagnosis is extremely important for the normal operation of mechanical equipment. The operating environment of rolling bearings is often complex and constantly changing. Such operating conditions will cause changes in the vibration characteristics of rolling bearings, thereby reducing the diagnostic effectiveness of rolling bearings in practical applications. It is worth noting that speed changes are quite common. Considering torque pulsation, vibration loads and speed fluctuations, rolling bearings often operate at speeds that vary with time in practical applications. Therefore, faults under variable speed conditions are not easy to detect and will cause unpredictable losses.

[0003] For fault diagnosis of time-varying speed, the speed change will cause the vibration signal spectrum to change, which increases the difficulty of feature extraction and selection. At different speeds, the fault characteristics may show different frequencies and amplitudes, so the traditional frequency domain analysis method is used to diagnose time-varying speed bearings.

[0004] The existing Chinese invention patent application with application publication number CN114112398A discloses a rolling bearing fault diagnosis method under variable speed conditions. The method realizes rolling bearing fault diagnosis under variable speed conditions by building a convolutional neural network based on deep learning, and utilizing the horizontal vibration signal and vertical vibration signal of the rolling bearing under different speed conditions. The convolutional neural network extracts fault features by alternating operations of convolution layer and pooling layer. However, the method does not take into account that the equipment may be subject to external interference such as surrounding equipment and environmental noise, electromagnetic interference, etc. during actual operation. These external interferences may cause the original signal to be confused or masked, causing signal distortion, thereby affecting the accuracy and reliability of signal processing. Moreover, the method only takes into account the spatial fault features of the input signal, and may not be able to accurately identify the fault feature frequency, resulting in the inability to accurately extract the fault features of the time-varying speed bearing, increasing the difficulty of fault diagnosis and resulting in inaccurate diagnosis results. Summary of the invention

[0005] The object of the present invention is to provide a time-varying speed bearing fault diagnosis method and device, so as to solve the problem that the bearing fault diagnosis method in the prior art provides inaccurate diagnosis results for time-varying speed bearings.

[0006] In order to solve the above technical problems, the present invention provides a time-varying speed bearing fault diagnosis method, which includes: obtaining the operating data of the bearing in a predicted time period, wherein the operating data includes vibration data and speed pulse data; and inputting the obtained operating data into a trained time-varying speed bearing fault diagnosis model to obtain a fault type result; wherein the time-varying speed bearing fault diagnosis model includes two channels and a classifier; the first channel is used to obtain the time fault information of the operating data, the second channel is used to obtain the spatial fault information of the operating data, and the classifier is used to complete the fault diagnosis in combination with the outputs of the first channel and the second channel.

[0007] Furthermore, the first channel includes a gated recurrent unit for acquiring temporal information of the running data to enhance the interdependence between the data.

[0008] Furthermore, the second channel includes a multi-scale wide kernel layer, and the multi-scale wide kernel layer is used to extract multi-scale features of the operating data through at least two wide kernel layers of different sizes, and splice the extracted multi-scale features.

[0009] Furthermore, the second channel also includes a second wide core layer, which is used to process the input operation data, and the processing includes constructing a new signal to distinguish more fault information and high-frequency noise suppression processing, and then inputting the processed operation data into the multi-scale wide core layer, so that the multi-scale wide core layer performs multi-scale feature extraction on the processed operation data.

[0010] Furthermore, the second channel also includes a small-core convolution module, which is used to reduce the neurons output by the multi-scale wide-core layer. The small-core convolution module includes a small-core convolution unit or at least two small-core convolution units connected in sequence; the small-core convolution unit includes a small-core convolution layer and a pooling layer; the size of the convolution kernel in the small-core convolution layer is smaller than the size of the convolution kernel in the wide-core layer.

[0011] Furthermore, the first channel also includes a first wide core layer, which is used to process the input operating data, and the processing includes constructing a new signal to distinguish more fault information and high-frequency noise suppression processing, and then inputting the processed operating data into the gated loop unit so that the gated loop unit obtains the time information of the processed operating data.

[0012] Furthermore, the number of wide core layers of different sizes is 3.

[0013] Furthermore, the specific calculation formula of the activation function used in the small-core convolution unit is:

[0014]

[0015] Among them, α is a positive hyperparameter, e x is the exponential function, x is the input value, and ELU is the activation function.

[0016] To solve the above technical problems, the present invention also provides a time-varying speed bearing fault diagnosis device, comprising a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the time-varying speed bearing fault diagnosis method as described above.

[0017] The beneficial effects are as follows: the present invention is an improved invention creation, which utilizes the operating data of the bearing under different speed conditions, and completes the fault diagnosis of the bearing by building a time-varying speed bearing fault diagnosis model; the operating data includes vibration data and speed pulse data, and the time-varying speed bearing fault diagnosis model includes two channels and a classifier; the first channel is used to obtain the time fault information of the operating data, and the second channel is used to obtain the spatial fault information of the operating data, and the classifier is used to complete the fault diagnosis by combining the outputs of the first channel and the second channel. The time-varying speed bearing fault diagnosis method of the present invention comprehensively considers the time information and spatial information of the input signal, enhances the interdependence between time series data, can simultaneously obtain the spatial and temporal fault information of the input signal, and improves the accuracy of the fault diagnosis of the time-varying speed bearing. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a flow chart of a time-varying speed bearing fault diagnosis method according to an embodiment of the present invention;

[0019] Figure 2-1 is a schematic diagram of a multi-scale module convolution process according to an embodiment of the present invention;

[0020] Figure 2-2 is a schematic diagram of the multi-scale module maximum pooling process of an embodiment of the present invention;

[0021] Figure 2-3 is a schematic diagram of the multi-scale module average pooling process of an embodiment of the present invention;

[0022] Figure 3 It is a schematic diagram of the GRU part of an embodiment of the present invention;

[0023] Figure 4 t-SNE classification effect diagram of three fault types of time-varying speed data in the absence of noise in an embodiment of the present invention;

[0024] Figure 5 This is a t-SNE classification effect diagram of three fault types of time-varying speed data under strong noise in an embodiment of the present invention;

[0025] Figure 6-1This is the t-SNE classification effect diagram of the CNN model of the existing technology on three types of faults under strong noise of time-varying speed data;

[0026] Figure 6-2 This is a t-SNE classification effect diagram of the WDCNN model of the prior art for three types of faults under strong noise in time-varying speed data;

[0027] Figure 6-3 This is the t-SNE classification effect diagram of the GRU model of the existing technology for three types of faults under strong noise of time-varying speed data;

[0028] Figure 6-4 This is a t-SNE classification effect diagram of the GRU-WDCNN model in the existing technology for three types of fault types of time-varying speed data under strong noise. DETAILED DESCRIPTION

[0029] The present invention obtains the operating data of the bearing at the prediction time, and inputs the obtained operating data into the trained time-varying speed bearing fault diagnosis model to obtain the fault type result; wherein the time-varying speed bearing fault diagnosis model includes two channels and a classifier; the first channel is used to obtain the time fault information of the operating data, and the second channel is used to obtain the spatial fault information of the operating data, and the classifier is used to complete the fault diagnosis by combining the outputs of the first channel and the second channel. The time-varying speed bearing fault diagnosis method of the present invention comprehensively considers the time information and spatial information of the input signal, enhances the mutual dependence between time series data, can simultaneously obtain the spatial and temporal fault information of the input signal, and solves the problem that the bearing fault diagnosis method in the prior art has inaccurate diagnosis results for time-varying speed bearings.

[0030] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention is further described in detail below with reference to the accompanying drawings and embodiments.

[0031] Method Example:

[0032] A time-varying speed bearing fault diagnosis method according to an embodiment of the present invention mainly comprises the following steps:

[0033] Step 1: Construct a time-varying speed bearing fault diagnosis model.

[0034] like Figure 1As shown, the model mainly includes two channels and a classifier. Among them, the first channel includes: a gated recurrent unit GRU for obtaining the temporal fault information of the input data, and a first wide kernel layer for suppressing high-frequency noise is set in front of the GRU. The second channel includes: a multi-scale wide kernel layer for obtaining the spatial fault information of the input data; a second wide kernel layer for suppressing high-frequency noise is set in front of the multi-scale wide kernel layer, and a small kernel convolution module is set behind the multi-scale wide kernel layer; the small kernel convolution module includes three small kernel convolution units connected in sequence, and the small kernel convolution unit includes a small kernel convolution layer and a pooling layer. The classifier completes the fault diagnosis by combining the outputs of the first channel and the second channel. Among them, the first channel and the second channel are set in parallel, and the structure, function and connection relationship between each module are specifically introduced as follows:

[0035] (1) The first wide core layer and gated recurrent unit GRU:

[0036] The operation data obtained by the fault acquisition module is processed through the first wide core layer of 64*1*16. The processing includes constructing new signals to distinguish more fault information and high-frequency noise suppression. The processed operation data is then input into the gated recurrent unit GRU. The wide core layer can obtain a larger receptive field, that is, it can see more input data, so that it can better capture the overall characteristics of the input signal and help construct a more global new signal. The gated recurrent unit GRU is used to obtain the time information of the operation data, so as to comprehensively consider the time information of the input signal to enhance the interdependence between time series data. The structure diagram of GRU is shown in the figure below. Figure 3 As shown, the specific implementation method is as follows:

[0037] X t =Wconv(input) (1)

[0038] Z t =σ(W z X t +V z H t-1 +b z ) (2)

[0039] R t =σ(W r X t +V r H t-1 +b r ) (3)

[0040]

[0041] Among them, X t Represents the input sequence of data, Zt and R t Represent the two gates of the GRU module: update gate and reset gate, H t is the current implicit state of the input sequence, is the hidden state of the candidate input sequence, H t-1 is the implicit state of the input sequence at the previous moment; W z represents the weight matrix of the update gate, W r Represents the weight matrix of the reset gate, W c Represents the weight matrix under the hidden state; V z represents the recurrent connection weight matrix of the update gate, V r represents the recurrent connection weight matrix of the update gate, V c represents the weight matrix of the hidden layer of the recurrent connection; b z , b r represents the bias term, σ and tanh represent the activation function, ⊙ represents the dot product, and W represents the weight matrix.

[0042] (2) The second wide core layer and multi-scale wide core layer:

[0043] First, a 64*1*16 second wide kernel layer is constructed to process the input operation data. The processing includes constructing new signals to distinguish more fault information and high-frequency noise suppression processing; the wide kernel layer can obtain a larger receptive field, that is, it can see more input data, so that it can better capture the overall characteristics of the input signal, which helps to construct a more global new signal. Then, a multi-layer multi-scale wide kernel layer is constructed under the second wide kernel layer to extract the multi-scale features of the operation data processed by the second wide kernel layer, and the extracted multi-scale features are spliced; the multi-scale wide kernel layer of the embodiment of the present invention is 3 layers, and the kernel size (convolution kernel size) is [100, 200, 300] respectively. As other implementation methods, the number and size of the multi-scale wide kernel layer can be set according to actual needs.

[0044] (3) Small kernel convolution module:

[0045] A small-core convolution module is set behind the multi-scale wide-core layer, and the small-core convolution module includes three small-core convolution units connected in sequence; wherein the small-core convolution unit includes a small-core convolution layer and a pooling layer, which is used to reduce a large number of neurons output by the multi-scale wide-core layer. As other implementations, the number of small-core convolution units can be set according to actual needs.

[0046] The scales of the small-core convolution units are 8*1*8, 32*1*8, and 16*1*8, respectively. This not only reduces a large number of neurons, but also allows for the acquisition of more potential sensitive features. The specific implementation of the constructed multi-scale module is as follows:

[0047] Convolutional layer: Figure 2-1 The schematic diagram of the multi-scale module convolution process shown in Figure 1 is that the input signal is assumed to be x∈T n and filter w∈T m , then the convolution process is described as follows:

[0048]

[0049] Among them, x l is the output feature of the lth layer, b l is the bias term, z l is the linear activation vector of layer l, Y l Refers to the feature vector after convolution processing, w l refers to the kernel size of the lth layer, wl(i) where the symbol represents the kernel of the lth layer; σ is a nonlinear activation function, conv represents convolution operation, same represents all zero padding, and x l-1 is the output feature of layer l-1; (yl(t),…,yl(n-m+1))∈Tn-m+ 1 represents the convolution process, Represents the convolution process y l The calculation process of (t), x l-1 (t+i-1) is the signal of the l-1th layer.

[0050] Pooling layer: usually located between two consecutive convolutional layers, its main function is to reduce the dimension of the output value of the upper layer, to maintain the consistency of the feature scale of the output value to a certain extent, and significantly reduce the amount of calculated data. In the embodiment of the present invention, the pooling layer is used as maximum pooling or average pooling, such as Figure 2-2 , 2-3 As shown. It can be expressed as:

[0051] x l =f(down(x l-1 )+b l ) (7)

[0052] Among them, x l-1 refers to the feature vector of the previous layer, down refers to downsampling, and f represents the pooling operation.

[0053] In addition, in order to better extract effective features in a noisy environment, a new activation function is introduced after each small convolution layer and pooling layer. The establishment of the activation function allows the input data to be processed by the activation function, making the model more adaptable to changes in negative noise. The specific implementation method is as follows:

[0054]

[0055] Among them, α is a positive hyperparameter, usually set to 1, ex is the exponential function, x is the input value, and ELU is the activation function.

[0056] (4) SoftMax classifier:

[0057] The SoftMax classifier completes fault diagnosis based on the output of the channel composed of the second wide kernel layer and the gated recurrent unit GRU and the output of the multi-scale module channel, thereby realizing the diagnosis and classification of bearing faults with time-varying speed. The specific implementation method is as follows:

[0058]

[0059] Among them, z i is the output value of the ith node, C is the number of output nodes, that is, the number of classification categories. e is an exponential function.

[0060] Step 2: Obtain the operating data of time-varying speed bearings with different fault types, which include vibration data and speed pulse data; randomly divide these data into data sets for training and testing, and train and test the time-varying speed bearing fault diagnosis model constructed in step 1.

[0061] Among them, for different fault types and different speeds, by replacing bearings with different fault types, using acceleration sensors and speed encoders to collect vibration data and speed pulse data respectively, the multiple signals collected from bearings with different fault types at different speeds constitute the required data set. Different speeds include: the speed increases all the time (for example, the speed range is 0-1500r / min), the speed decreases all the time (for example, the speed range is 1500-0r / min), the speed increases first and then decreases (for example, the speed range is 0-1500-0r / min), and the speed decreases first and then increases (for example, the speed range is 1500-0-1500r / min). The different fault types used in this embodiment for training and testing the time-varying speed bearing fault diagnosis model include three fault types: normal state, inner ring fault and outer ring fault.

[0062] The time-varying speed bearing fault diagnosis method of the present invention is specially designed for time-varying speed. The speed signal that varies with time is more complex, which significantly increases the difficulty of fault diagnosis. However, it is precisely because of this complexity that the method is more suitable for common working conditions in actual work, and improves the practicality and accuracy of fault diagnosis.

[0063] The vibration data in the acquired operating data is named Channel_1, and the speed pulse data in the acquired operating data is named Channel_2. The acquired vibration data and speed pulse data are used to create training data with a window length of 2048 data points and a step length of 64 to train the time-varying speed bearing fault diagnosis model.

[0064] In order to make full use of the data source, this embodiment introduces dual signal input, namely Channel_1 and Channel_2. This dual signal input design enables the network to independently extract fault features from vibration and speed pulse signals, effectively improving the characterization ability of the model.

[0065] Step 3: Obtain the operating data of the bearing within the prediction time period, and use the time-varying speed bearing fault diagnosis model obtained in step 2 to perform bearing fault diagnosis.

[0066] like Figure 4 As shown in FIG. 1 , the t-SNE classification effect diagram of the method of the present invention for the three types of faults in the absence of noise for the time-varying speed data provided. Each sample is visualized as a point, and the same fault type has the same color. Figure 5 As shown, it is a t-SNE classification effect diagram of the method of the present invention for the three fault types under the strong noise interference level of 6dB for the provided time-varying speed data.

[0067] like Figure 6-1 , 6-2 , 6-3, and 6-4 are the t-SNE classification results of three fault types using conventional models for time-varying speed data under 6dB strong noise interference. The model used in 6-1 is the CNN model, the model used in 6-2 is the WDCNN model, the model used in 6-3 is the GRU model, and the model used in 6-4 is the GRU-WDCNN model.

[0068] Compare Figure 4 , 5 , 6-1, 6-2, 6-3, and 6-4 show that the speed change data targeted by the method of the present invention can be accurately separated into various types of faults, achieving accurate classification, regardless of whether the speed change bearing is subject to noise interference or not.

[0069] In summary, no matter in the case of variable speed bearing with or without noise interference, it can be seen that various types of faults diagnosed by the method of the present invention are accurately separated, which shows that the method of the present invention can still meet the actual application requirements in a noisy environment and can effectively handle the fault diagnosis of bearings with time-varying speeds. Therefore, the present invention takes into account the working conditions of noise interference and time-varying speed at the same time, which is more in line with the actual working conditions, making the diagnosis results of time-varying speed bearings more accurate.

[0070] Device Example:

[0071] A time-varying speed bearing fault diagnosis device of the present invention includes a memory, a processor and an internal bus, and the processor and the memory communicate and exchange data with each other through the internal bus. The memory includes at least one software function module stored in the memory, and the processor executes various functional applications and data processing by running the computer program and module stored in the memory, thereby realizing a time-varying speed bearing fault diagnosis method introduced in the method embodiment of the present invention. The principle, implementation process, and achievable effect of the method have been fully introduced in the method embodiment and will not be repeated here.

Claims

1. A time-varying speed bearing fault diagnosis method, characterized in that: The method includes: obtaining the operating data of the bearing in the predicted time period, wherein the operating data includes vibration data and speed pulse data; and inputting the obtained operating data into the trained time-varying speed bearing fault diagnosis model to obtain the fault type result; wherein the time-varying speed bearing fault diagnosis model includes two channels and a classifier; the first channel is used to obtain the time fault information of the operating data, the second channel is used to obtain the space fault information of the operating data, and the classifier is used to complete the fault diagnosis by combining the outputs of the first channel and the second channel.

2. The time-varying speed bearing fault diagnosis method according to claim 1 is characterized in that: The first channel includes a gated recurrent unit to obtain the temporal information of the running data to enhance the interdependence between the data.

3. The time-varying speed bearing fault diagnosis method according to claim 1 is characterized in that: The second channel includes a multi-scale wide kernel layer, which is used to extract multi-scale features from the operating data through at least two wide kernel layers of different sizes, and to splice the extracted multi-scale features.

4. The time-varying speed bearing fault diagnosis method according to claim 3 is characterized in that: The second channel also includes a second wide core layer, which is used to process the input operation data, and the processing includes constructing a new signal to distinguish more fault information and high-frequency noise suppression processing, and then inputting the processed operation data into the multi-scale wide core layer, so that the multi-scale wide core layer performs multi-scale feature extraction on the processed operation data.

5. The time-varying speed bearing fault diagnosis method according to claim 3 is characterized in that: The second channel also includes a small-core convolution module, which is used to reduce the neurons output by the multi-scale wide-core layer. The small-core convolution module includes a small-core convolution unit or at least two small-core convolution units connected in sequence; the small-core convolution unit includes a small-core convolution layer and a pooling layer; the size of the convolution kernel in the small-core convolution layer is smaller than the size of the convolution kernel in the wide-core layer.

6. The time-varying speed bearing fault diagnosis method according to claim 2 is characterized in that: The first channel also includes a first wide core layer, which is used to process the input operation data, and the processing includes constructing a new signal to distinguish more fault information and high-frequency noise suppression processing, and then inputting the processed operation data into the gated loop unit so that the gated loop unit obtains the time information of the processed operation data.

7. The time-varying speed bearing fault diagnosis method according to claim 3 is characterized in that: The number of the wide core layers of different sizes is 3.

8. The time-varying speed bearing fault diagnosis method according to claim 5 is characterized in that: The specific calculation formula of the activation function used in the small kernel convolution unit is: Among them, α is a positive hyperparameter, e x is the exponential function, x is the input value, and ELU is the activation function.

9. A time-varying speed bearing fault diagnosis device, comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the time-varying speed bearing fault diagnosis method according to any one of claims 1 to 8.

Citation Information

Patent Citations

  • Fault diagnosis method based on improved unsupervised domain adaptive network

    CN116522118A

  • Cross-domain mechanical fault diagnosis method based on multi-channel feature fusion of CBAM and use thereof

    US20240142342A1