Rolling bearing fault diagnosis method and system based on improved SK-MSCNN model

Abstract

The embodiment of the invention provides a rolling bearing fault diagnosis method and system based on an improved SK-MSCNN model, and belongs to the technical field of rolling bearing fault diagnosis. The method comprises the following steps: acquiring vibration data of a rolling bearing to be diagnosed; the vibration data is input into a preset improved SK-MSCNN model to obtain the fault type of the rolling bearing, and the improved SK-MSCNN model comprises a feature extraction module which comprises a first convolution layer used for carrying out convolution operation on the vibration data to obtain a first fault signal feature; the second convolution layer comprises three convolution kernels with different scales, and each convolution kernel is used for obtaining a corresponding second fault signal feature according to the first fault signal feature; the first batch normalization layer is used for performing batch standardization operation and index moving average operation according to the second fault signal features to obtain third fault signal features; a selective fault attention module; and a classification module.

Description

Technical Field

[0001] The present invention relates to the technical field of rolling bearing fault diagnosis, and in particular to a rolling bearing fault diagnosis method and system based on an improved SK-MSCNN model. Background Art

[0002] As the core of modern industry, rotating machinery is widely used in many key fields, including manufacturing, chemical industry, energy industry, aerospace, and marine engineering. In the field of fault diagnosis of these large-scale mechanical equipment, the health status monitoring and fault identification of rotating machinery have always been the focus of research. Especially for the core components of rotating machinery, rolling bearings, their health status is directly related to the safe and stable operation of the entire equipment. In the actual production and operation process, rolling bearings are subjected to different loads from multiple directions, and are prone to various forms of defects and faults. These faults may cause unstable operation of mechanical equipment, which in turn causes huge economic losses to industrial production. Therefore, the research on rolling bearing fault diagnosis is of great significance to ensure the safe and stable operation of industrial machinery and equipment.

[0003] With the development of neural networks, the technology of using neural networks to identify equipment faults has gradually emerged. However, rolling bearings are affected by various loads in actual working environments, resulting in significant differences in their fault types and fault degrees. Since the signal frequency caused by vibration presents diverse characteristics in different signal segments, the fault characteristics of rolling bearings are distributed at different scales. Although most existing studies have considered multi-scale issues, the effectiveness of fault features extracted by convolution kernels of different scales is different, ignoring the impact of different importance of scale features learned in different branches on the final diagnosis results. In addition, vibration signals usually contain information from different sources, and sensors usually capture not only valid signals from target fault components, but also invalid signals from other interfering devices. Standard diagnostic methods based on convolutional neural network models are difficult to enhance discriminant information and suppress useless noise and other redundant information, and it is difficult to accurately extract complex fault feature information in a noisy environment. In summary, the existing rolling bearing fault diagnosis technology still has certain limitations in terms of the comprehensiveness of feature extraction and the ability to resist noise interference. Summary of the invention

[0004] The purpose of the embodiments of the present invention is to provide a rolling bearing fault diagnosis method and system based on an improved SK-MSCNN model, wherein the method and system can realize automatic diagnosis of rolling bearings.

[0005] In order to achieve the above object, an embodiment of the present invention provides a rolling bearing fault diagnosis method based on an improved SK-MSCNN model, comprising:

[0006] Acquiring vibration data of a rolling bearing to be diagnosed;

[0007] The vibration data is input into a preset improved SK-MSCNN model to obtain the fault type of the rolling bearing, wherein the improved SK-MSCNN model includes:

[0008] Feature extraction module, including:

[0009] A first convolution layer, used for performing a convolution operation on the vibration data to obtain a first fault signal feature;

[0010] A second convolution layer, including three convolution kernels of different scales, each of which is used to obtain a corresponding second fault signal feature according to the first fault signal feature;

[0011] a first batch normalization layer, used to perform a batch normalization operation and an exponential sliding average operation according to the second fault signal feature, so as to obtain a third fault signal feature;

[0012] a selective fault attention module, configured to determine a fault feature to be identified according to the third fault signal feature;

[0013] A classification module is used to determine the fault type according to the fault characteristics.

[0014] Optionally, obtaining vibration data of a rolling bearing to be diagnosed includes:

[0015] Acquiring original vibration data of the rolling bearing;

[0016] Performing a z-score normalization operation on the original vibration data;

[0017] Using overlapping sliding windows to perform sliding sampling on the original vibration data after the z-score normalization operation to obtain sub-signal samples;

[0018] One-Hot encoding is performed on the sub-signal samples to obtain the vibration data.

[0019] Optionally, performing a z-score normalization operation on the original vibration data includes:

[0020] Formula (1) is used to perform the z-score standardization operation.

[0021]

[0022] Among them, y i The z-score is the original vibration data after the standardization operation, x i is the original vibration data before standardization operation, is the original vibration data before standardization operation x i The average value of S(X) is the original vibration data x before standardization. i The standard deviation of X is the original vibration data before standardization. i A collection of .

[0023] Optionally, the size of the convolution kernel of the first convolution layer is 64×1, the number of channels is 16, and the sliding step is 1.

[0024] Optionally, the sizes of the convolution kernels of the second convolutional layer are 16×1, 64×1 and 128×1 respectively.

[0025] Optionally, the first batch normalization layer is used to:

[0026] Determine the third fault signal characteristic according to formula (2) and formula (3):

[0027]

[0028] in, is the exponential sliding average at time step t, is the current batch mean at time step t, α is the smoothing factor of EMA, is the exponential sliding average at time step t-1, is the exponential moving average of the variance at time step t, is the current batch variance at time step t, is the exponential moving average of the variance at time step t-1.

[0029] Optionally, the selective fault attention module includes:

[0030] A concatenation layer, used for combining the three third fault signal features to obtain a fusion feature;

[0031] A global average pooling layer, used for processing the fused features to obtain a global feature vector;

[0032] An adaptive maximum pooling layer, used for processing the fused features to obtain an adaptive feature vector;

[0033] A fusion layer, used to fuse the global feature vector and the adaptive feature vector using a Sigmoid function to obtain a combined feature;

[0034] The selection layer is used to select a feature vector from the combined features by adopting a soft attention mechanism to obtain the fault feature.

[0035] Optionally, the classification module includes a third convolutional layer, a second batch normalization layer, a maximum pooling layer and a fully connected layer connected in sequence, and the fully connected layer is used to perform a classification operation using a Softmax function to obtain the fault type.

[0036] On the other hand, the present invention also provides a rolling bearing fault diagnosis system based on an improved SK-MSCNN model, the system comprising a processor, and the processor is configured to execute any of the methods described above.

[0037] Through the above technical solution, the embodiment of the present invention provides a rolling bearing fault diagnosis method and system based on the improved SK-MSCNN model. The diagnosis method and system recognize the vibration data of the rolling bearing through the preset improved SK-MSCNN model. During the recognition, the feature extraction module and the selective fault intention model are used to complete the feature extraction, and then the classification module is combined for classification and recognition, so as to realize the automatic recognition of rolling bearing faults.

[0038] Other features and advantages of the embodiments of the present invention will be described in detail in the subsequent detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the following specific implementations, they are used to explain the embodiments of the present invention, but do not constitute a limitation on the embodiments of the present invention. In the accompanying drawings:

[0040] Figure 1 is a flow chart of a rolling bearing fault diagnosis method based on an improved SK-MSCNN model according to an embodiment of the present invention;

[0041] Figure 2 is a schematic diagram of obtaining vibration data of a rolling bearing to be diagnosed according to an embodiment of the present invention;

[0042] Figure 3 is a flow chart of obtaining vibration data of a rolling bearing to be diagnosed according to one embodiment of the present invention;

[0043] Figure 4 is a structural block diagram of an improved SK-MSCNN model according to an embodiment of the present invention;

[0044] Figure 5 is a schematic diagram of an improved SK-MSCNN model according to one embodiment of the present invention;

[0045] Figure 6 is a schematic diagram of a feature extraction module according to one embodiment of the present invention;

[0046] Figure 7 is a block diagram of a selective fault attention module according to an embodiment of the present invention;

[0047] Figure 8 is a schematic diagram of a selective fault attention module according to one embodiment of the present invention. DETAILED DESCRIPTION

[0048] The specific implementation of the embodiment of the present invention is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described here is only used to illustrate and explain the embodiment of the present invention, and is not used to limit the embodiment of the present invention.

[0049] like Figure 1 FIG. 1 is a flow chart of a rolling bearing fault diagnosis method based on an improved SK-MSCNN model according to an embodiment of the present invention. Figure 1 In the method, the diagnosis may include the following steps:

[0050] In step S10, vibration data of the rolling bearing to be diagnosed is obtained;

[0051] In step S11, the vibration data is input into the preset improved SK-MSCNN model to obtain the fault type of the rolling bearing.

[0052] In this Figure 1 In the method shown, step S10 can be used to obtain vibration data of the rolling bearing to be diagnosed. The method for obtaining the vibration data of the rolling bearing can be, for example, installing a vibration sensor device on the equipment where the rolling bearing is located, and then directly accessing the vibration sensor device to obtain the vibration data.

[0053] Since the vibration data is obtained by accessing a vibration sensor device, and the vibration sensor device generally includes a variety of sensors, and different types of sensor signals have different dimensions and amplitude ranges, directly using the original vibration data will result in excessive dimensional differences between features, thereby reducing the effect of feature fusion and model training. Therefore, in this embodiment, step S10 may further include the following: Figure 3 The steps shown, the corresponding schematic diagram is as follows Figure 2 As shown. Figure 3 In the method, the method may include the following steps:

[0054] In step S20, original vibration data of the rolling bearing is obtained;

[0055] In step S21, a z-score normalization operation is performed on the original vibration data. In this embodiment, the z-score normalization operation can also be expressed by the following formula (1):

[0056]

[0057] Among them, y i The z-score is the original vibration data after the standardization operation, x i is the original vibration data before standardization operation, is the original vibration data before standardization operation x i The average value of S(X) is the original vibration data x before standardization. i The standard deviation of X is the original vibration data before standardization. i A collection of .

[0058] In step S22, the original vibration data after the z-score normalization operation is subjected to sliding sampling using an overlapping sliding window to obtain a sub-signal sample. Specifically, the sub-signal sample can also be expressed as X i ={x i+s ,x i+1+s ,x i+2+s ,…,x w+s ,};

[0059] In step S23, one-hot encoding is performed on the sub-signal samples to obtain vibration data.

[0060] Step S11 can be used to input the vibration data into a preset improved SK-MSCNN model to obtain the fault type of the rolling bearing. The improved SK-MSCNN model can be obtained by pre-training.

[0061] In this embodiment, if Figure 4 and Figure 5 As shown, the improved SK-MSCNN model can further include a feature extraction module 10, a selective fault attention module 20 and a classification module 30.

[0062] like Figures 4 to 6 As shown, the feature extraction module may include a first convolution layer 11, a second convolution layer 12 and a first batch normalization layer 13. Among them, the first convolution layer 11 can be used to perform a convolution operation on the vibration data to obtain a first fault signal feature. The second convolution layer 12 may include three convolution kernels of different scales, each of which is used to obtain a corresponding second fault signal feature according to the first fault signal feature. The first batch normalization layer 13 can be used to perform a batch normalization operation and an exponential sliding average operation according to the second fault signal feature, respectively, to obtain a third fault signal feature. The selective fault attention module 20 can be used to determine the fault feature to be identified according to the third fault signal feature. The classification module 30 can be used to determine the fault type according to the fault feature.

[0063] In an implementation, the specific sizes of the convolution kernels of the first convolution layer 11 and the second convolution layer 12 can be in various forms known to those skilled in the art. In one example of the present invention, the size of the convolution kernel of the first convolution layer 11 can be 64×1, the number of channels is 16, and the sliding step is 1. The sizes of the three convolution kernels of the second convolution layer 12 can be 16×1, 64×1, and 128×1, respectively, so as to extract low-frequency, medium-frequency, and high-frequency features, respectively. Through the design of the three convolution kernels, the extracted features can simultaneously reflect the information of the input vibration data at multiple scales, thereby improving the comprehensiveness of the extracted features.

[0064] The first batch normalization layer 13 is used to perform batch normalization operation and exponential sliding average operation according to the second fault signal feature to obtain the third fault signal feature. Due to the smooth update characteristics of the exponential sliding average operation, the fluctuation of the batch normalization operation is significantly reduced in an environment with high noise and unstable data distribution, thereby improving the convergence speed and generalization ability of the improved SK-MSCNN model. In this embodiment, the batch normalization operation and the exponential sliding average operation can also be expressed by the following formula (2) and formula (3),

[0065]

[0066] in, is the exponential sliding average at time step t, is the current batch mean at time step t, α is the smoothing factor of EMA, is the exponential sliding average at time step t-1, is the exponential moving average of the variance at time step t, is the current batch variance at time step t, is the exponential moving average of the variance at time step t-1.

[0067] The selective fault attention module 20 is used to determine the fault feature to be identified according to the third fault signal feature. Figure 7 and Figure 8 As shown, the selective fault attention module 20 may further include a concatenation layer 21, a global average pooling layer 22, an adaptive maximum pooling layer 23, a fusion layer 24, and a selection layer 25. The concatenation layer 21 may be used to combine three third fault signal features to obtain a fusion feature. The global average pooling layer 22 may be used to process the fusion feature to obtain a global feature vector. The adaptive maximum pooling layer 23 may be used to process the fusion feature to obtain an adaptive feature vector. The specific processing process of the adaptive maximum pooling layer 23 may also be expressed by the following formula (4):

[0068] f AMP (Y c )=Max m,n {Y c (p H +s H ·i+m,p W +s W ·j+n)}, (4)

[0069] Among them, f AMP (Y c ) is the feature vector output by the adaptive maximum pooling layer 23, Y c is the input feature map, Y c is the input feature map, m and n represent the internal index of the pooling window, i and j represent the external index of the pooling window, and p H and p W are the starting positions in the height direction and width direction respectively, s H and W is the step size in the height and width directions.

[0070] The fusion layer 24 is used to fuse the global feature vector and the adaptive feature vector using the Sigmoid function to obtain the combined feature, and the selection layer 25 can be used to select the feature vector from the combined feature using the soft attention mechanism to obtain the fault feature.

[0071] The classification module 30 can be used to determine the fault type according to the fault characteristics. The specific structure of the classification module 30 can be various forms known to those skilled in the art. In one example of the present invention, the classification module 30 can include a third convolution layer, a second batch normalization layer, a maximum pooling layer and a fully connected layer connected in sequence. Among them, the third convolution layer can be used to perform a convolution operation, the second batch normalization layer can be used to perform a batch normalization operation, the maximum pooling layer can be used to perform a maximum pooling operation, and the fully connected layer can be used to perform a classification operation using a Softmax function to obtain the fault type.

[0072] In addition, the specific method for training the improved SK-MSCNN model can be in various forms known to those skilled in the art. In one example of the present invention, the method for training the SK-MSCNN model can include the following steps in sequence:

[0073] Step 1: Collect vibration data, perform data preprocessing, and use overlapping sliding windows to randomly divide the vibration data set into training data set, validation data set, and test data set;

[0074] Step 2: Use the training data set to train the improved SK-MSCNN model, and verify and test the improved SK-MSCNN model using the validation data set and the test data set respectively until the classification accuracy of the improved SK-MSCNN model meets the requirements.

[0075] In one example of the present invention, the parameters of the improved SK-MSCNN model obtained by the above training method may be as shown in Table 1 below:

[0076] Table 1

[0077]

[0078] On the other hand, the present invention also provides a rolling bearing fault diagnosis system based on an improved SK-MSCNN model, the system comprising a processor, and the processor is configured to execute any of the methods described above.

[0079] Through the above technical solution, the embodiment of the present invention provides a rolling bearing fault diagnosis method and system based on the improved SK-MSCNN model. The diagnosis method and system recognize the vibration data of the rolling bearing through the preset improved SK-MSCNN model. During the recognition, the feature extraction module and the selective fault intention model are used to complete the feature extraction, and then the classification module is combined for classification and recognition, so as to realize the automatic recognition of rolling bearing faults.

[0080] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0081] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0082] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0083] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0084] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0085] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0086] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0087] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0088] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A rolling bearing fault diagnosis method based on an improved SK-MSCNN model, characterized in that: The method comprises: Acquiring vibration data of a rolling bearing to be diagnosed; The vibration data is input into a preset improved SK-MSCNN model to obtain the fault type of the rolling bearing, wherein the improved SK-MSCNN model includes: Feature extraction module, including: A first convolution layer, used for performing a convolution operation on the vibration data to obtain a first fault signal feature; A second convolution layer, including three convolution kernels of different scales, each of which is used to obtain a corresponding second fault signal feature according to the first fault signal feature; a first batch normalization layer, used to perform a batch normalization operation and an exponential sliding average operation according to the second fault signal feature, so as to obtain a third fault signal feature; a selective fault attention module, configured to determine a fault feature to be identified according to the third fault signal feature; A classification module is used to determine the fault type according to the fault characteristics.

2. The method according to claim 1, characterized in that Obtain vibration data of the rolling bearing to be diagnosed, including: Acquiring original vibration data of the rolling bearing; Performing a z-score normalization operation on the original vibration data; Using overlapping sliding windows to perform sliding sampling on the original vibration data after the z-score normalization operation to obtain sub-signal samples; One-Hot encoding is performed on the sub-signal samples to obtain the vibration data.

3. The method according to claim 2, characterized in that Performing a z-score normalization operation on the raw vibration data includes: Formula (1) is used to perform the z-score standardization operation. Among them, y i The z-score is the original vibration data after the standardization operation, x i is the original vibration data before standardization operation, is the original vibration data before standardization operation x i The average value of S(X) is the original vibration data x before standardization. i The standard deviation of X is the original vibration data before standardization. i A collection of .

4. The method according to claim 1, characterized in that The size of the convolution kernel of the first convolutional layer is 64×1, the number of channels is 16, and the sliding step is 1.

5. The method according to claim 1, characterized in that The sizes of the convolution kernels of the second convolutional layer are 16×1, 64×1, and 128×1 respectively.

6. The method according to claim 1, characterized in that The first batch normalization layer is used to: Determine the third fault signal characteristic according to formula (2) and formula (3): in, is the exponential sliding average at time step t, is the current batch mean at time step t, α is the smoothing factor of EMA, is the exponential sliding average at time step t-1, is the exponential moving average of the variance at time step t, is the current batch variance at time step t, is the exponential moving average of the variance at time step t-1.

7. The method according to claim 1, characterized in that The selective fault attention module comprises: A concatenation layer, used for combining the three third fault signal features to obtain a fusion feature; A global average pooling layer, used for processing the fused features to obtain a global feature vector; An adaptive maximum pooling layer, used for processing the fused features to obtain an adaptive feature vector; A fusion layer, used to fuse the global feature vector and the adaptive feature vector using a Sigmoid function to obtain a combined feature; The selection layer is used to select a feature vector from the combined features by adopting a soft attention mechanism to obtain the fault feature.

8. The method according to claim 1, characterized in that The classification module includes a third convolutional layer, a second batch normalization layer, a maximum pooling layer and a fully connected layer which are connected in sequence. The fully connected layer is used to perform a classification operation using a Softmax function to obtain the fault type.

9. A rolling bearing fault diagnosis system based on an improved SK-MSCNN model, characterized in that: The system comprises a processor configured to execute the method according to any one of claims 1 to 8.

Images