Intermediate bearing fault diagnosis device and method under complex working condition

Abstract

The invention relates to an intermediate bearing fault diagnosis method and device under a complex working condition, and the device comprises a data collection module which collects an intermediate bearing vibration signal through installing an acceleration sensor; the data preprocessing module is used for converting the collected vibration signals into a time-frequency diagram so as to express the characteristics of the signals from two angles of a time domain and a frequency domain; the feature extraction module is used for effectively capturing a complex mode and correlation in a conversion signal by combining an attention mechanism and a deep residual network so as to improve the fault diagnosis accuracy; and the model verification module is used for training the feature extraction module on a constructed data set and evaluating a model effect on a verification set. According to the method and the device for diagnosing the intermediate bearing fault under the complex working condition, the accuracy of fault diagnosis of the bearing under the complex working condition can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of mechanical fault diagnosis, and in particular to a device and method for diagnosing intermediate bearing faults under complex working conditions, which are applied to the detection of bearing faults in rotating mechanical equipment. Background Art

[0002] Fault diagnosis is an important part of the health management of rotating machinery and equipment, and is of great significance to ensure the healthy and safe operation of manufacturing equipment. As the key core component of rotating machinery and equipment, the engine has a complex structure that brings a high cost to the disassembly and assembly of the engine. As the core part inside, the rolling bearing has a harsh working environment. Once a failure occurs, it will cause huge economic losses and even casualties. Therefore, it needs to be inspected regularly. However, frequent disassembly and inspection will result in higher engine maintenance costs. With the continuous development of information technology, the use of computer technology to perform non-disassembly fault diagnosis of rolling bearings in rotating machinery and equipment engines to reduce engine maintenance costs has become a research hotspot in the manufacturing field.

[0003] When actually collecting aircraft engine bearing signals, the signals are collected during a period of time when the speed is almost stable without external force, but it is difficult to maintain a stable speed under this condition. Therefore, the engine works under complex working conditions such as variable load and variable speed, making the vibration signal a more complex non-stationary signal. The change in running speed during signal feature extraction is very similar to the feature change caused by the fault. In addition, in actual environments, the collected bearing signals usually contain a lot of noise from the environment and other components, resulting in nonlinear characteristics in the collected data. However, the existing deep learning-based bearing fault diagnosis solutions are difficult to meet the requirements for bearing health status classification under noise and variable loads. Summary of the invention

[0004] Based on the technical problems raised above, the present invention proposes a device and method for diagnosing intermediate bearing faults under complex working conditions to solve the problem that the signals collected in complex scenarios such as variable speed and variable load are complex non-stationary signals, thereby eliminating the problem that the fault features are difficult to extract.

[0005] The technical solution adopted by the present invention to achieve the above-mentioned purpose is:

[0006] A device for diagnosing intermediate bearing faults under complex working conditions, comprising:

[0007] A signal acquisition module, used to control the acceleration sensor to acquire the vibration signal of the intermediate bearing through an acquisition controller;

[0008] A data preprocessing module, used for converting the intermediate bearing vibration signal into a time-frequency diagram through short-time Fourier transform;

[0009] Feature extraction module, used to extract key features in the time-frequency graph through the hole residual convolution network, the global residual network, and the feature fusion network;

[0010] The model verification module is used to train and verify the key feature extraction module and convert its output features into bearing status for fault diagnosis.

[0011] The acquisition controller is implemented by FPGA.

[0012] The dilated residual convolution network is composed of four sequentially connected convolution basic blocks Block, wherein Block0 and Block2 have the same structure, which are: sequentially connected dilated convolution layer Dconv, batch normalization layer BN and activation function ReLU, and the features output by the activation function ReLU are fused through the feature fusion network; Block1 and Block3 have the same structure, which are based on Block0 or Block2, and the fusion results are corrected through the features output by ReLU.

[0013] The global residual block consists of sequentially connected convolutional layers and batch normalization layers.

[0014] The feature fusion network is composed of a stack of a maximum pooling layer, an average pooling layer, a fully connected layer, and an activation function layer, specifically:

[0015]

[0016]

[0017]

[0018]

[0019] Among them, x represents the input data, MaxPool represents the adaptive maximum pooling layer, FC represents the fully connected layer, δ represents the use of activation function ReLU, σ represents the use of activation function Sigmoid, y1 represents the first layer feature extraction result, y se represents the feature extraction result after SENet, y2 represents the feature extraction result of the first layer after SENet, and y represents the final feature extraction result.

[0020] A method for diagnosing intermediate bearing faults under complex working conditions comprises the following steps:

[0021] The signal acquisition module controls the acceleration sensor to acquire the vibration signal of the intermediate bearing through the acquisition controller;

[0022] The data preprocessing module converts the intermediate bearing vibration signal into a time-frequency diagram through short-time Fourier transform;

[0023] The feature extraction module extracts key features from the time-frequency graph through the hole residual convolution network, the global residual network, and the feature fusion network;

[0024] The model verification module trains and verifies the key feature extraction module and converts its output features into bearing status for fault diagnosis.

[0025] A system for diagnosing intermediate bearing faults under complex working conditions comprises a memory and a processor; the memory is used to store a computer program; the processor is used to implement a method for diagnosing intermediate bearing faults under complex working conditions when executing the computer program.

[0026] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, a method for diagnosing intermediate bearing faults under complex working conditions is implemented.

[0027] The present invention has the following beneficial effects and advantages:

[0028] 1. This paper constructs a new intermediate bearing fault diagnosis model, which combines the advantages of attention mechanism and void convolution.

[0029] 2. The network structure proposed in the present invention achieves a good balance between speed and accuracy, has strong robustness, is suitable for the field of aircraft engine intermediate bearing fault diagnosis, and can also be extended to various signal processing problems, with good practicality.

[0030] 3. The present invention establishes an intermediate bearing fault diagnosis method under complex scenarios, which can perform accurate fault diagnosis at a faster reasoning speed and has strong stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 It is an overall schematic diagram of the intermediate bearing fault diagnosis device of the present invention;

[0032] Figure 2 It is an overall architecture diagram of the intermediate bearing fault diagnosis method of the present invention;

[0033] Figure 3 It is a schematic diagram of the installation position of the sensor of the present invention;

[0034] Figure 4 It is a structural diagram of the hollow residual convolutional network and the global residual network of the present invention;

[0035] Figure 5 It is a structural diagram of the feature fusion network of the present invention. DETAILED DESCRIPTION

[0036] The present invention is further described in detail below in conjunction with the accompanying drawings and embodiments.

[0037] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation method of the present invention is described in detail below in conjunction with the accompanying drawings. In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without violating the connotation of the invention, so the present invention is not limited by the specific implementation disclosed below.

[0038] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in the specification of the invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0039] A device for diagnosing intermediate bearing faults under complex working conditions, such as Figure 1 As shown in Figure 1, it includes a signal acquisition module, a data preprocessing module, a feature extraction module and a model verification module. The signal acquisition module controls the acceleration sensor to collect signals, sends the collected vibration signal to the data preprocessing module to convert it into a time-frequency graph, and then sends the time-frequency graph to the feature extraction module for fault feature extraction. Finally, the extracted results are sent to the model verification module for bearing fault diagnosis. The overall architecture of the intermediate bearing fault diagnosis method is shown in Figure 1. Figure 2 shown.

[0040] In signal acquisition, in order to reduce the interference of noise on the signal, the sensor is installed at a position close to the intermediate bearing, and the vibration signal related to the intermediate bearing fault is captured as much as possible, so as to perform fault diagnosis and monitoring more accurately. Figure 3 shown.

[0041] The structure diagram of the void residual convolutional network is as follows Figure 4 As shown in (a) and 4(b), its basic block consists of a dilated convolution layer (Dconv), a batch normalization layer (BN), and an activation function (ReLU). These three layers work together on the data, enabling the model to better extract features, normalize data distribution, and enhance the model's nonlinear expression capabilities, thereby improving the model's performance. Among them, the batch normalization layer can accelerate the convergence of the network, suppress the vanishing or exploding gradients, and increase the generalization ability of the network. The dilated convolution layer is a convolution layer with an adjustable receptive field, which can perform a comprehensive convolution on the image and extract richer information from the input image. Block 0 and Block 2 have the same structure, as shown in Figure 4 (a); Block 1 and Block 3 have the same structure, as shown in Figure 4 (b) as shown.

[0042] The structure of the global residual block (GRB) is shown in Figure 4 As shown in (c), it is used to solve the problem of relevant information loss caused by operations such as convolution. The GRB module consists of a convolution layer and a batch normalization layer. Here, the convolution layer uses ordinary convolution, that is, the hole rate is set to 1. This is because, in this case, compared with the hole convolution, the ordinary convolution has fewer calculations and parameters, so it can improve the calculation efficiency and improve the accuracy of the model to a certain extent without significantly affecting the model performance.

[0043] The structure diagram of the feature fusion network (MAB) is as follows Figure 5 As shown in the figure, the main structure includes a pooling layer, a fully connected layer, and an activation function layer. Specifically, in the MAB block, we first use the maximum pooling layer to compress the information of each channel to reduce the computational burden and improve the computational efficiency of the model. Next, the pooled feature map is input into the fully connected layer for feature transformation, and then the ReLU activation function is used for nonlinear transformation to map the input value to a non-negative range, thereby increasing the nonlinear ability of the model and helping to improve the performance of the model. In addition, the module passes through a fully connected layer and a Sigmoid activation function to map the input value to an interval between 0 and 1, indicating the importance of each feature channel and reducing the risk of overfitting, thereby increasing generalization ability. In order to avoid the gradient vanishing problem and speed up training, the module introduces a residual connection, which adds the output of the previous layer to the output of the current layer. The above process is expressed in the following formula.

[0044]

[0045] Among them, MaxPool represents the adaptive maximum pooling layer, FC represents the fully connected layer, δ represents the use of activation function ReLU, and σ represents the use of activation function Sigmoid.

[0046]

[0047]

[0048]

[0049] The obtained output is then input into the SE module to obtain the importance of each feature channel, and a weight value is assigned to each feature channel according to the importance, and then directly multiplied with the output of the previous layer, as shown in equations (2) and (3). Finally, these weight values ​​are weighted and summed with the output of each channel of the previous layer of the feature map to obtain the final output, as shown in equation (4).

[0050] Through the above steps, the MAB module can adaptively adjust the weight of each channel according to the importance of each channel to improve the network's ability to identify key features, thereby improving the performance and accuracy of the entire model.

[0051] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should be regarded as within the scope of protection of the present invention.

Claims

1. A device for diagnosing intermediate bearing faults under complex working conditions, characterized in that: include: A signal acquisition module, used to control the acceleration sensor to acquire the vibration signal of the intermediate bearing through an acquisition controller; A data preprocessing module, used for converting the intermediate bearing vibration signal into a time-frequency diagram through short-time Fourier transform; Feature extraction module, used to extract key features in the time-frequency graph through the hole residual convolution network, the global residual network, and the feature fusion network; The model verification module is used to train and verify the key feature extraction module and convert its output features into bearing status for fault diagnosis.

2. The device for diagnosing intermediate bearing faults under complex working conditions according to claim 1 is characterized in that: The acquisition controller is implemented by FPGA.

3. The device for diagnosing intermediate bearing faults under complex working conditions according to claim 1 is characterized in that: The dilated residual convolutional network is composed of four sequentially connected convolutional basic blocks Block, where Block0 and Block2 have the same structure, which are: sequentially connected dilated convolutional layer Dconv, batch normalization layer BN and activation function ReLU, and the features output by the activation function ReLU are fused through the feature fusion network; Block1 and Block3 have the same structure, which are based on Block0 or Block2, and the fusion results are corrected through the features output by ReLU.

4. The device for diagnosing intermediate bearing faults under complex working conditions according to claim 1 is characterized in that: The global residual block consists of sequentially connected convolutional layers and batch normalization layers.

5. The device for diagnosing intermediate bearing faults under complex working conditions according to claim 1 is characterized in that: The feature fusion network is composed of a stack of a maximum pooling layer, an average pooling layer, a fully connected layer, and an activation function layer, specifically: Among them, x represents the input data, MaxPool represents the adaptive maximum pooling layer, FC represents the fully connected layer, δ represents the use of activation function ReLU, σ represents the use of activation function Sigmoid, y1 represents the first layer feature extraction result, y se represents the feature extraction result after SENet, y2 represents the feature extraction result of the first layer after SENet, and y represents the final feature extraction result.

6. A method for diagnosing intermediate bearing faults under complex working conditions, characterized in that: The following steps are involved: The signal acquisition module controls the acceleration sensor to acquire the vibration signal of the intermediate bearing through the acquisition controller; The data preprocessing module converts the intermediate bearing vibration signal into a time-frequency diagram through short-time Fourier transform; The feature extraction module extracts key features from the time-frequency graph through the hole residual convolution network, the global residual network, and the feature fusion network; The model verification module trains and verifies the key feature extraction module and converts its output features into bearing status for fault diagnosis.

7. A fault diagnosis system for intermediate bearings under complex working conditions, characterized in that: It comprises a memory and a processor; the memory is used to store a computer program; the processor is used to implement a method for diagnosing intermediate bearing faults under complex working conditions as described in claim 6 when executing the computer program.

8. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by the processor, the method for diagnosing intermediate bearing faults under complex working conditions as described in claim 6 is implemented.

Images