A rolling bearing fault diagnosis method based on multi-channel feature fusion image

Abstract

The application discloses a kind of based on multi-channel feature fusion image rolling bearing fault diagnosis method, including fault diagnosis model training step and fault diagnosis step, fault diagnosis model training step includes: the multi-channel vibration signal of collection rolling bearing;The vibration signal of each channel is carried out feature modal decomposition, and select a part of feature modal component as target feature modal component from it;Target feature modal component is converted to the same polar coordinate system and is shown output, obtains multi-channel feature fusion image;Build fault diagnosis model;Fault diagnosis step includes: the multi-channel feature fusion image of rolling bearing is input to fault diagnosis model.The rolling bearing fault diagnosis method of the application can reduce data processing amount.By converting multiple one-dimensional time series into a single two-dimensional feature image, the complex relationship between the overall image structure and key features is learned, and the automation of rolling bearing fault diagnosis is achieved.

Description

Technical Field

[0001] This invention belongs to the field of rolling bearing fault diagnosis technology, specifically relating to a rolling bearing fault diagnosis method based on multi-channel feature fusion images. Background Technology

[0002] Rotating equipment is one of the main pieces of equipment in industrial production; a malfunction can cause production stoppages or even serious accidents. According to incomplete statistics, approximately 30% of rotating machinery failures are caused by rolling bearings. Conducting condition monitoring and fault diagnosis research on rolling bearings can effectively reduce the operating costs of rotating machinery and ensure the long-term safety and stability of its operation.

[0003] Because rolling bearings operate in harsh and variable environments for extended periods, their vibration signals in industrial applications are often subject to interference from other components and noise, making direct fault diagnosis difficult. Signal decomposition methods can break down a signal into multiple sub-signal components with clear physical meanings, containing easily extractable rolling bearing fault features. Variational mode decomposition establishes a constrained optimization problem using narrowband conditions for components, thereby estimating the center frequency of the signal components and their reconstructive effects. Eigenmode decomposition considers the impulsiveness and periodicity of the equipment fault response, adaptively adjusting the filter's center frequency and bandwidth to obtain a series of eigenmode components. However, each eigenmode component only contains features of a specific rolling bearing fault; therefore, fault diagnosis requires checking each eigenmode component individually, which is time-consuming and labor-intensive.

[0004] In addition, due to the influence of operating conditions and noise, one-dimensional signals may exhibit phenomena such as shrinkage, blurring, and displacement of fault information, leading to a decrease in the accuracy of fault diagnosis.

[0005] With the development of computer technology, shallow machine learning algorithms such as Support Vector Machines (SVMs) and Artificial Neural Networks (ANNs) have been increasingly applied to fault diagnosis. However, the extracted features have relatively weak generalization capabilities. Deep learning, with its powerful feature extraction capabilities, can learn efficient feature representations from inputs. While recurrent neural networks (RNNs) are used to extract features from time series data, their accuracy is easily affected by noise. Using the two-dimensional feature image of the original signal as input to a neural network, and extracting fault information from the image through a convolutional neural network (CNN), can extract more complex features. However, CNNs suffer from the problem of not focusing on global information and the inability to dynamically change the weights of the convolutional kernels. Therefore, a fault diagnosis model with strong global information capture and generalization capabilities is needed. Summary of the Invention

[0006] To address the technical problem that existing rolling bearing fault diagnosis methods for rotating equipment require checking each feature modal component one by one, which is time-consuming, labor-intensive, and has low accuracy in single-channel signal fault diagnosis, this invention proposes a rolling bearing fault diagnosis method based on multi-channel feature fusion images, which can solve the above problems.

[0007] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0008] A rolling bearing fault diagnosis method based on multi-channel feature fusion images includes a fault diagnosis model training step and a fault diagnosis step. The fault diagnosis model training step includes:

[0009] Step 1: Collect multi-channel vibration signals of the rolling bearing, preprocess the multi-channel vibration signals to obtain multi-channel vibration signals x(n), where n represents the number of sampling points;

[0010] Step 2: Set the characteristic mode decomposition parameters, perform characteristic mode decomposition on the vibration signal x(n) of each channel respectively, obtain multiple characteristic mode components, and select a portion of the characteristic mode components as the target characteristic mode components;

[0011] Step 3: Convert the target feature modal components to the same polar coordinate system for display output to obtain a multi-channel feature fusion image. Collect the multi-channel feature fusion images of rolling bearings with different fault types together to construct a rolling bearing fault dataset. Preprocess the multi-channel feature fusion images and divide the rolling bearing fault dataset into a training set and a test set according to the proportion.

[0012] Step 4: Construct a fault diagnosis model and train the model using the training set. Once the loss function converges, the fault diagnosis model is obtained.

[0013] The fault diagnosis steps include:

[0014] The multi-channel feature fusion image of the rolling bearing is input into the fault diagnosis model, and the rolling bearing fault diagnosis result is output.

[0015] In some embodiments, the method for preprocessing the multi-channel vibration signal in step one includes:

[0016] The multi-channel vibration signal is sampled at high frequency, and the sampling frequency of the high-frequency sampling is at least twice the highest frequency of the rolling bearing vibration signal;

[0017] The multi-channel vibration signals after high-frequency sampling are aligned on the time axis, and the signals of a set duration are extracted to obtain the multi-channel vibration signal x(n).

[0018] In some embodiments, the characteristic mode decomposition parameters set in step two include the number of modes N, the number of filters K, the maximum number of iterations M, and the filter length L.

[0019] In some embodiments, step two, the method for obtaining the target feature modal components includes:

[0020] The vibration signal of each channel is decomposed into characteristic modes, and the vibration signal of each channel is decomposed into multiple characteristic mode components.

[0021] Calculate the correlation kurtosis of each feature modal component in each channel, sort the correlation kurtosis in descending order, and select the feature modal components corresponding to the first A correlation kurtosis as the target feature modal components.

[0022] In some embodiments, the relevant kurtosis is calculated as follows:

[0023] ;

[0024] in, Represents the modal component y Point vibration value, Represents the modal component y Point vibration value, T represents the fault period, and N represents the length of the modal component y.

[0025] In some embodiments, the method for obtaining the fault period T includes:

[0026] Establish the autocorrelation function of the modal component y :

[0027] ;

[0028] Where τ represents the time delay value for acquiring the multi-channel vibration signal of the rolling bearing. Represents the modal component y Point vibration value;

[0029] calculate The local maxima are taken as the fault period T.

[0030] In some embodiments,

[0031] Polar radius of multi-channel feature fusion image With polar angle , , The calculation method is as follows: ; ;

[0032] ; ;

[0033] In the formula, Indicates the first Each characteristic modal component; Indicates the first each modal component Vibration value of the point; This represents the maximum value of the vibration modal components; This represents the minimum value of the vibration modal components; , , Indicates the offset coefficient; Indicates the magnification factor; Indicates the number of target feature modal components; Indicates the first each modal component The vibration value of the point.

[0034] In some embodiments, step three, the method for preprocessing the multi-channel feature fusion image, includes:

[0035] Set the resolution of the multi-channel feature fusion image to q. 2 ×q 2 ;

[0036] Convert the multi-channel feature fusion image into a grayscale image.

[0037] In some embodiments, in step four, during the construction of the fault diagnosis model, the grayscale image of the multi-channel feature fusion image is divided into q×q segments to obtain q 2 There are 1 image patch, and each image patch is mapped to q. 2 The vector.

[0038] In some embodiments, the fault diagnosis model is constructed based on a self-attention mechanism.

[0039] Compared with existing technologies, the advantages and positive effects of this invention are as follows: The rolling bearing fault diagnosis method based on multi-channel feature fusion images of this invention can reduce the amount of data processing by selecting a portion of the feature modal components as target feature modal components. By converting the target feature modal components to the same polar coordinate system for display output, a multi-channel feature fusion image is obtained, realizing the transformation of multiple one-dimensional time series into a single two-dimensional feature image for presentation. The global information of the rolling bearing fault state is reflected from the visual information of the feature image, the complex relationship between the overall image structure and key features is learned, and the fault diagnosis model is trained based on the pattern presented by the multi-channel feature fusion image and the corresponding fault, thus realizing the automation of rolling bearing fault diagnosis.

[0040] Other features and advantages of the present invention will become clearer after reading the detailed description of the embodiments of the present invention in conjunction with the accompanying drawings. Attached Figure Description

[0041] Figure 1 This is a flowchart of the fault diagnosis model training steps in one embodiment of the rolling bearing fault diagnosis method based on multi-channel feature fusion image proposed in this invention.

[0042] Figure 2 This is a schematic diagram of the mechanical performance and fault comprehensive simulation test bench structure in one embodiment of the rolling bearing fault diagnosis method based on multi-channel feature fusion image proposed in this invention;

[0043] Figure 3a This is a vibration signal in the x-direction collected in one embodiment of the rolling bearing fault diagnosis method based on multi-channel feature fusion image proposed in this invention;

[0044] Figure 3b This is a vibration signal in the y-direction collected in one embodiment of the rolling bearing fault diagnosis method based on multi-channel feature fusion image proposed in this invention;

[0045] Figure 3c This is a vibration signal in the z-direction collected in one embodiment of the rolling bearing fault diagnosis method based on multi-channel feature fusion image proposed in this invention;

[0046] Figure 4a This is a decomposed feature mode component image of one embodiment of the rolling bearing fault diagnosis method based on multi-channel feature fusion image proposed in this invention;

[0047] Figure 4b This is another feature mode component image after decomposition in one embodiment of the rolling bearing fault diagnosis method based on multi-channel feature fusion image proposed in this invention;

[0048] Figure 4c This is another feature mode component image after decomposition in one embodiment of the rolling bearing fault diagnosis method based on multi-channel feature fusion image proposed in this invention;

[0049] Figure 5 This is a grayscale image of a multi-channel feature fusion image in one embodiment of the rolling bearing fault diagnosis method based on multi-channel feature fusion image proposed in this invention;

[0050] Figure 6 This is a comparison image of multi-channel feature fusion images of different rolling bearing fault types and multi-channel feature fusion images of fault-free rolling bearings in one embodiment of the rolling bearing fault diagnosis method based on multi-channel feature fusion images proposed in this invention.

[0051] Figure 7 This is a schematic diagram of the fault diagnosis model in one embodiment of the rolling bearing fault diagnosis method based on multi-channel feature fusion image proposed in this invention;

[0052] Figure 8 This is an iterative-loss graph of the fault diagnosis model in one embodiment of the rolling bearing fault diagnosis method based on multi-channel feature fusion image proposed in this invention.

[0053] Figure 9 This is an iteration-accuracy diagram of the fault diagnosis model in one embodiment of the rolling bearing fault diagnosis method based on multi-channel feature fusion image proposed in this invention. Detailed Implementation

[0054] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0055] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0056] It should be noted that in the description of this invention, terms such as "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," indicating directional or positional relationships, are based on the directional or positional relationships shown in the accompanying drawings. These are merely for ease of description and do not indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation; therefore, they should not be construed as limitations on this invention. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0057] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0058] Example 1, see Figure 1As shown, this embodiment proposes a rolling bearing fault diagnosis method based on multi-channel feature fusion images, including a fault diagnosis model training step and a fault diagnosis step. The fault diagnosis model training step includes:

[0059] Step 1: Collect multi-channel vibration signals of rolling bearings with different fault types using sensors, preprocess the multi-channel vibration signals to obtain multi-channel vibration signals x(n), where n represents the number of sampling points.

[0060] Step 2: Set the characteristic mode decomposition parameters, perform characteristic mode decomposition on the vibration signal x(n) of each channel respectively, obtain multiple characteristic mode components, and select a portion of the characteristic mode components as the target characteristic mode components.

[0061] Step 3: Transform the target feature modal components to the same polar coordinate system for display output to obtain a multi-channel feature fusion image. Combine the multi-channel feature fusion images of rolling bearings with different fault types to construct a rolling bearing fault dataset. Preprocess the multi-channel feature fusion images and divide the rolling bearing fault dataset into training and test sets according to the proportion.

[0062] Step 4: Build a fault diagnosis model and train the model using the training set. Once the loss function converges, the fault diagnosis model is obtained.

[0063] The fault diagnosis steps include:

[0064] The multi-channel feature fusion image of the rolling bearing is input into the fault diagnosis model, and the rolling bearing fault diagnosis result is output.

[0065] In the fault diagnosis process, the real-time acquired multi-channel vibration signals of the rolling bearing fault can be preprocessed, followed by eigenmode decomposition in step two, and a multi-channel feature fusion image obtained in step three. This multi-channel feature fusion image is then input into the fault diagnosis model for diagnosis, and the diagnostic results are output. Alternatively, historical acquisition data of the rolling bearing fault multi-channel vibration signals can be processed using the above steps to obtain a multi-channel feature fusion image, which is then input into the fault diagnosis model for diagnosis, and the diagnostic results are output.

[0066] This embodiment of the rolling bearing fault diagnosis method based on multi-channel feature fusion images reduces data processing volume by selecting a subset of feature modal components as target feature modal components. By converting the target feature modal components to the same polar coordinate system for display output, a multi-channel feature fusion image is obtained. This transforms multiple one-dimensional time series into a single two-dimensional feature image for presentation. The global information of the rolling bearing fault state is reflected from the visual information of the feature image. The complex relationship between the overall image structure and key features is learned, and a fault diagnosis model is trained based on the pattern presented by the multi-channel feature fusion image and the corresponding fault, thus achieving automated rolling bearing fault diagnosis.

[0067] By constructing polar coordinate mapping, multi-channel one-dimensional vibration signals are transformed into two-dimensional feature fusion images. While preserving the independent features of each channel, cross-channel feature collaborative analysis is established. At the same time, the differences in the images can accurately reflect the state of the rolling bearing, enhancing the overall expression of different fault states of the rolling bearing.

[0068] In some embodiments, the method for preprocessing the multi-channel vibration signal in step one includes:

[0069] High-frequency sampling is performed on the multi-channel vibration signals, with the sampling frequency being at least twice the highest frequency of the rolling bearing vibration signal. High-frequency sampling can completely record the high-frequency transient impact signals generated by rolling bearing faults, avoiding signal aliasing or loss due to insufficient sampling rate. In this embodiment, by setting the high-frequency sampling rate to at least twice the highest frequency of the rolling bearing vibration signal, the multi-channel vibration signals are aligned and truncated based on the sampling frequency, ensuring that the vibration signals of different channels maintain the consistency of fault-related characteristics in the time dimension.

[0070] The multi-channel vibration signals after high-frequency sampling are aligned on the time axis, and the signals of a set duration are extracted to obtain the multi-channel vibration signal x(n).

[0071] In some embodiments, the method for preprocessing multi-channel vibration signals includes, but is not limited to, noise reduction processing.

[0072] In some embodiments, the characteristic mode decomposition parameters set in step two include the number of modes N, the number of filters K, the maximum number of iterations M, and the filter length L.

[0073] In some embodiments, step two, the method for obtaining the target feature modal components includes:

[0074] The vibration signals of each channel are decomposed into characteristic modes, and multiple characteristic mode components are obtained from the decomposition of the vibration signal of each channel.

[0075] Calculate the correlation kurtosis of each feature mode component in each channel, sort the correlation kurtosis in descending order, and select the feature mode components corresponding to the top A correlation kurtosis as the target feature mode components. It is understood that A should be a positive integer, and A is not greater than the total number of all feature mode components.

[0076] Correlation kurtosis can reflect the periodicity and impulsivity of vibration signals, and rolling bearing faults also exhibit periodicity and impulsivity. This scheme selects the characteristic mode components with larger correlation kurtosis as target characteristic mode components, thereby filtering out characteristic mode components that are only related to the fault for processing. This can greatly reduce the amount of computation, improve the fault diagnosis speed, and realize real-time fault diagnosis.

[0077] In some embodiments, the relevant kurtosis The calculation method is as follows:

[0078] ;

[0079] in, Represents the modal component y Point vibration value, Represents the modal component y Point vibration value, T represents the fault period, and N represents the length of the modal component y.

[0080] In some embodiments, the method for obtaining the fault period T includes:

[0081] Establish the autocorrelation function of the modal component y :

[0082] ;

[0083] Where τ represents the time delay value for acquiring the multi-channel vibration signal of the rolling bearing. Represents the modal component y Point vibration value.

[0084] calculate The local maxima are taken as the fault period T.

[0085] In some embodiments, before calculating the correlation kurtosis of each characteristic mode component in each channel, a step of processing the characteristic mode components is included, including:

[0086] All feature mode components are time-aligned.

[0087] Based on the extraction order of the multi-channel vibration signal x(n), the characteristic modal components of each channel are spliced ​​together.

[0088] In some embodiments, step three, the method for obtaining the multi-channel feature fusion image, includes:

[0089] The time series coordinates of the characteristic modal components are transformed to polar coordinates. The center of the plane is taken as the pole, and a ray is drawn from the pole as the polar axis. The polar radius coordinate unit is 1, and the polar angle coordinate unit is 1°, with counterclockwise as the positive direction of the angle. The polar radius is calculated. polar angle , , ,Will( , ), ( , ), ( , The image is formed by polar coordinates to create a multi-channel feature fusion image. The image can effectively characterize the state differences caused by different fault types through radial texture features and angular pole distribution features, and realize the mapping between fault modes and image visual features. , , , The details are as follows:

[0090] ;

[0091] ;

[0092] ; ;

[0093] In the formula, Indicates the first Each characteristic modal component; Indicates the first each modal component Vibration value of the point; This represents the maximum value of the vibration modal components; This represents the minimum value of the vibration modal components; , , Indicates the offset coefficient; Indicates the magnification factor; Indicates the number of target feature modal components; Indicates the first each modal component The vibration value of the point.

[0094] In the process of calculating the extreme radius, data normalization was performed to reduce the influence of operating conditions and prevent certain features from having an excessive impact on the results while ignoring the contributions of other features.

[0095] In some embodiments, step three, the method for preprocessing the multi-channel feature fusion image, includes:

[0096] Set the resolution of the multi-channel feature fusion image to q. 2 ×q 2 .

[0097] Converting multi-channel feature fusion images into grayscale images reduces redundant color information, effectively reducing the amount of input data and lowering the computational cost and memory consumption of training. This allows subsequent diagnosis to focus more on the structural and textural features of the image, improving the model's operating efficiency.

[0098] In some embodiments, in step four, during the construction of the fault diagnosis model, the grayscale image of the multi-channel feature fusion image is divided into q×q segments to obtain q 2 There are 1 image patch, and each image patch is mapped to q. 2 The vector. In this embodiment, the resolution of the multi-channel feature fusion image is set to q. 2 ×q 2 This facilitates image segmentation.

[0099] To facilitate the operation of subsequent methods, the image is preprocessed by dividing the grayscale image of the multi-channel feature fusion image into several small image blocks. Each image block has a smaller resolution and is more compact than the original image, which can reduce the computational complexity and improve the computational speed in the training of the fault diagnosis model.

[0100] In some embodiments, the fault diagnosis model is built based on a self-attention mechanism. The fault diagnosis model comprises multiple encoders, which capture global information from the image through a self-attention mechanism. This is achieved by defining three learnable weight matrices: Q, key K, and value V. The output of the self-attention layer is calculated as shown in the following formula:

[0101] ;

[0102] In the formula, This represents the output of the self-attention layer; softmax represents... A function that transforms into a probability distribution; QKT represents the dot product between the query and all keys; It is the normalization factor.

[0103] A fault diagnosis model based on a self-attention mechanism is used to extract features from multi-channel feature fusion images for rolling bearing fault diagnosis. The self-attention mechanism captures global features in the image and learns the complex relationship between the overall image structure and key features, thus automating the rolling bearing fault diagnosis. In industrial applications, due to the existence of massive amounts of monitoring data, the model exhibits good generalization ability and accuracy after pre-training and fine-tuning.

[0104] Example 2, in a specific embodiment, such as Figure 2 As shown, taking a comprehensive simulation test bench for fluid machinery performance and faults as an example, the test bench includes a motor 1, a photoelectric sensor 2, a coupling 3, a vibration acceleration sensor 4, a bearing housing 5, and a gearbox 6. The photoelectric sensor 2 is installed on the front side of the motor 1 to measure the movement of the inner ring of the bearing. The motor 1 is connected to the rotor via the coupling 3 to drive the movement of the inner ring of the bearing. The outer ring of the bearing is fixed on the bearing housing 5. The vibration acceleration sensor 4 is installed at the bearing housing 5 to measure the vibration signal of the bearing. The gearbox 6 is installed at the tail of the rotor and connected via a coupling. The experiment mainly includes various conditions such as rolling bearing outer ring fault, inner ring fault, and healthy condition. In a specific example, the collected fault vibration signal is subjected to high-frequency sampling, truncation, and noise reduction processing. The sampling frequency is 51200Hz, and data with 51200 sampling points in the entire signal is extracted, such as... Figures 3a-3c The figures show vibration signals collected in the x, y, and z directions in space, respectively.

[0105] The original fault signal is a multi-channel one-dimensional time-domain signal. Due to noise and interference from other components, the various fault characteristics of the rolling bearing are hidden in the time-domain signal, making direct diagnosis of rolling bearing faults difficult. Combining the impulsive and periodic nature of rolling bearing faults, the multi-channel one-dimensional time-domain signal is decomposed into multi-channel characteristic mode components using the eigenmode decomposition method. This separates the various fault characteristics into multiple characteristic mode components, thereby determining the bearing fault type.

[0106] In step S2, the parameters set include the number of modes N, the number of filters K, the filter length L, and the maximum number of iterations. In a specific example, the number of modes is 1, the number of filters is 10, the filter length is 30, and the maximum number of iterations is 30.

[0107] According to the method described in steps two to three of Embodiment 1, the multi-channel vibration signal is decomposed into multi-channel characteristic mode components through characteristic mode decomposition, resulting in... Figures 4a-4c The multi-channel characteristic mode components shown will Figures 4a-4c The multi-channel feature modal components are converted into a multi-channel feature fusion image, and the multi-channel feature fusion image is converted into a grayscale image, such as... Figure 5As shown, the image effectively characterizes the state differences caused by different fault types through radial texture features and angular pole distribution features, realizing the mapping between fault modes and image visual features. The multi-channel feature fusion image fully contains the global features and overall information of the vibration signal, enhancing the visualization of fault features.

[0108] like Figure 6 As shown in the figure, the multi-channel feature fusion images formed by different fault types have significant differences, which can accurately reflect the state of different rolling bearing fault types, facilitating subsequent fault diagnosis and demonstrating the effectiveness of multi-channel feature fusion images for rolling bearing fault diagnosis.

[0109] Self-attention mechanisms can focus on and capture global features of an image without convolutional layers. Furthermore, in industrial applications, due to the complexity of equipment environments and the massive amounts of monitoring data, deep learning fault diagnosis models exhibit better generalization ability and accuracy. The structure of the fault diagnosis model is as follows: Figure 7 As shown, the image is segmented into a series of image patches, and each image patch is converted into a low-dimensional vector to meet the requirements of the input data.

[0110] The hyperparameters of the fault diagnosis model are set as shown in Table 1.

[0111] Table 1 Model Hyperparameter Settings

[0112]

[0113] The initial model is trained using a rolling bearing training set. If the set number of iterations is reached, the model is output. In a specific embodiment of this invention, after processing the rolling bearing vibration data, it is input into the rolling bearing fault diagnosis model to classify the rolling bearings and provide corresponding diagnostic results. The iteration loss and accuracy of the model are as follows: Figure 8 and Figure 9 As shown, the proposed method achieves an accuracy rate of 97% for rolling bearings. The results demonstrate that the proposed method can efficiently and accurately perform rolling bearing fault diagnosis, effectively ensuring the efficient operation of rolling bearings.

[0114] This invention proposes a rolling bearing fault diagnosis method based on multi-channel feature fusion images. It employs Eigenmode Decomposition (EMD) to decompose the rolling bearing vibration signal, obtaining multi-channel feature mode components containing different rolling bearing fault information. All mode components are fused using a multi-channel feature fusion image. Simultaneously, normalization reduces the influence of operating conditions, better reflecting the impact and periodicity of rolling bearing faults. A model is used to extract global and local features from the multi-channel feature fusion image, realizing the rolling bearing fault diagnosis function.

[0115] Of course, the above description is not intended to limit the present invention, and the present invention is not limited to the examples given above. Any changes, modifications, additions or substitutions made by those skilled in the art within the scope of the present invention should also fall within the protection scope of the present invention.

Claims

1. A method for diagnosing rolling bearing faults based on multi-channel feature fusion images, characterized in that, It includes a fault diagnosis model training step and a fault diagnosis step, wherein the fault diagnosis model training step includes: Step 1: Collect multi-channel vibration signals of the rolling bearing, preprocess the multi-channel vibration signals to obtain multi-channel vibration signals x(n), where n represents the number of sampling points; Step 2: Set the characteristic mode decomposition parameters, perform characteristic mode decomposition on the vibration signal x(n) of each channel respectively, obtain multiple characteristic mode components, and select a portion of the characteristic mode components as the target characteristic mode components; Step 3: Convert the target feature modal components to the same polar coordinate system for display output to obtain a multi-channel feature fusion image. Collect the multi-channel feature fusion images of rolling bearings with different fault types together to construct a rolling bearing fault dataset. Preprocess the multi-channel feature fusion images and divide the rolling bearing fault dataset into a training set and a test set according to the proportion. Step 4: Construct a fault diagnosis model and train the model using the training set. Once the loss function converges, the fault diagnosis model is obtained. The fault diagnosis steps include: The multi-channel feature fusion image of the rolling bearing is input into the fault diagnosis model, and the rolling bearing fault diagnosis result is output. In step two, the methods for obtaining the target feature modal components include: The vibration signal of each channel is decomposed into characteristic modes, and the vibration signal of each channel is decomposed into multiple characteristic mode components. Calculate the correlation kurtosis of each feature mode component in each channel, sort the correlation kurtosis in descending order, and select the feature mode components corresponding to the top A correlation kurtosis as the target feature mode components; In step three, the polar radius of the multi-channel feature fusion image is... With polar angle , , The calculation method is as follows: ； ； ； ； In the formula, Indicates the first Each characteristic modal component; Indicates the first each modal component Vibration value of the point; This represents the maximum value of the vibration modal components; This represents the minimum value of the vibration modal components; , , Indicates the offset coefficient; Indicates the magnification factor; Indicates the number of target feature modal components; Indicates the first each modal component The vibration value of the point.

2. The rolling bearing fault diagnosis method according to claim 1, characterized in that, Step one involves preprocessing the multi-channel vibration signals, including: The multi-channel vibration signal is sampled at high frequency, and the sampling frequency of the high-frequency sampling is at least twice the highest frequency of the rolling bearing vibration signal; The multi-channel vibration signals after high-frequency sampling are aligned on the time axis, and the signals of a set duration are extracted to obtain the multi-channel vibration signal x(n).

3. The rolling bearing fault diagnosis method according to claim 1, characterized in that, The eigenmode decomposition parameters set in step two include the number of modes N, the number of filters K, the maximum number of iterations M, and the filter length L.

4. The rolling bearing fault diagnosis method according to claim 1, characterized in that, The method for calculating the relevant kurtosis is as follows: ； in, Represents the modal component y Point vibration value, Represents the modal component y Point vibration value, T represents the fault period, and N represents the length of the modal component y.

5. The rolling bearing fault diagnosis method according to claim 4, characterized in that, Methods for obtaining the fault period T include: Establish the autocorrelation function of the modal component y : ； Where τ represents the time delay value for acquiring the multi-channel vibration signal of the rolling bearing. Represents the modal component y Point vibration value; calculate The local maxima are taken as the fault period T.

6. The rolling bearing fault diagnosis method according to claim 1, characterized in that, Step three, the method for preprocessing the multi-channel feature fusion image includes: Set the resolution of the multi-channel feature fusion image to q. 2 ×q 2 q is the image size baseline parameter; Convert the multi-channel feature fusion image into a grayscale image.

7. The rolling bearing fault diagnosis method according to claim 6, characterized in that, In step four, during the construction of the fault diagnosis model, the grayscale image of the multi-channel feature fusion image is divided into q×q segments to obtain q. 2 There are 1 image patch, and each image patch is mapped to q. 2 The vector.

8. The rolling bearing fault diagnosis method according to any one of claims 1-7, characterized in that, The fault diagnosis model is built based on a self-attention mechanism.

Images