Rolling bearing fault diagnosis method and system based on multi-source vibration signals and medium

Abstract

The invention discloses a rolling bearing fault diagnosis method and system based on multi-source vibration signals and a medium, and relates to the field of rolling bearing fault diagnosis, and the method comprises the steps: carrying out the fusion of vibration signals collected by all acceleration sensors through employing a PCA method; dividing the fused signal sample into a plurality of sub-samples by adopting a sliding window method; adopting time-frequency feature extraction to calculate a plurality of feature values of each sub-sample, judging the fault type of the rolling bearing based on a random forest model, and screening to obtain an important feature value of each sub-sample; generating a two-dimensional image of each sub-sample by using MTF, and dynamically adjusting the brightness and RGB color channel of the two-dimensional image corresponding to each sub-sample based on CBAM and the important feature value of each sub-sample; and inputting the dynamically adjusted two-dimensional images of all the sub-samples into the trained CBAM-CNN model, and outputting to obtain a rolling bearing fault type probability graph. According to the invention, the accuracy of rolling bearing fault diagnosis is improved.

Description

Technical Field

[0001] The present application relates to the technical field of rolling bearing fault diagnosis, and in particular to a rolling bearing fault diagnosis method, system and medium based on multi-source vibration signals. Background Art

[0002] Rolling bearings are widely used in modern machinery. Their main function is to support rotating parts and reduce friction, thereby achieving efficient and stable transmission. Rolling bearings are widely used in various equipment such as motors, fans, vehicles, aerospace equipment and precision instruments due to their simple structure, reliable operation and low maintenance cost.

[0003] However, since rolling bearings are subjected to high-speed rotation, heavy loads and even impact in mechanical systems, once a rolling bearing fails, it will cause severe vibration of the entire equipment, which may further damage other components and cause a decline in the overall performance of the equipment. Therefore, regular monitoring and diagnosis of the health status of rolling bearings is an important part of daily maintenance of rolling bearings.

[0004] In the prior art, rolling bearing faults are generally diagnosed through feature extraction analysis and image recognition analysis. Feature extraction analysis is to extract intrinsic mode functions, envelope spectra, singular value decomposition and other features from the vibration signal collected from a single measuring point using time-frequency analysis, wavelet transform, empirical mode decomposition and other technologies to characterize the fault. These fault features are used for rolling bearing diagnosis after data processing and classification; image recognition analysis is to generate images based on the vibration signal of the measuring point, and then combine convolutional neural networks for classification and discrimination. Although the above methods can realize rolling bearing fault diagnosis, feature extraction analysis is mostly based on a single signal source, which easily leads to incomplete extracted fault features, and the two-dimensional image generated by image recognition analysis contains a large amount of redundant information, which easily interferes with fault diagnosis. Therefore, the rolling bearing fault diagnosis results achieved based on the above methods are not accurate. Summary of the invention

[0005] The purpose of this application is to provide a rolling bearing fault diagnosis method, system and medium based on multi-source vibration signals, thereby improving the accuracy of rolling bearing fault diagnosis.

[0006] To achieve the above objectives, this application provides the following solutions:

[0007] In a first aspect, the present application provides a rolling bearing fault diagnosis method based on multi-source vibration signals, the rolling bearing fault diagnosis method based on multi-source vibration signals comprising:

[0008] Acquire vibration signals collected by multiple acceleration sensors; the multiple acceleration sensors are respectively arranged at different parts of the target device; the target device is a device containing a rolling bearing to be diagnosed;

[0009] The vibration signals collected by all acceleration sensors are fused using the PCA method to obtain fused signal samples;

[0010] The fused signal sample is divided into a plurality of sub-samples by using a sliding window method; each sub-sample contains a fused signal of a set sliding window length;

[0011] Time-frequency feature extraction is used to calculate multiple eigenvalues ​​of each subsample, and the rolling bearing fault type is identified based on the random forest model to screen out the important eigenvalues ​​of each subsample;

[0012] Generate a two-dimensional image of each sub-sample using MTF; and dynamically adjust the brightness and RGB color channels of the two-dimensional image corresponding to each sub-sample based on CBAM and the important eigenvalues ​​of each sub-sample;

[0013] The two-dimensional images after dynamic adjustment of all sub-samples are input into the trained CBAM-CNN model, and a rolling bearing fault type probability map is output; the rolling bearing fault type probability map is used to determine the fault type of the rolling bearing to be diagnosed in the target device.

[0014] In a second aspect, the present application also provides a computer system comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the rolling bearing fault diagnosis method based on multi-source vibration signals described in the first aspect.

[0015] In a third aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the rolling bearing fault diagnosis method based on multi-source vibration signals described in the first aspect is implemented.

[0016] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0017] The present application first obtains the vibration signals collected by different acceleration sensors, realizes the collection of multi-source vibration signals, and at the same time uses the principal component analysis (PCA) method to realize the fusion of multi-source vibration signals, thereby improving the comprehensiveness of rolling bearing fault information acquisition. In order to facilitate the subsequent feature extraction, the present application also uses a sliding window to divide the fused signal sample into multiple sub-samples, and through the time-frequency feature extraction and the screening of the random forest model, the important eigenvalue with the strongest correlation with the rolling bearing fault type discrimination is determined from the multiple eigenvalues ​​of each sub-sample. After the Markov Transition Field (MTF) is used to generate a two-dimensional image of each sub-sample, the two-dimensional image of each sub-sample is dynamically adjusted by the above-mentioned important eigenvalues, ensuring that the dynamically adjusted two-dimensional image contains the most representative feature information for rolling bearing fault diagnosis. In addition, in order to realize the accurate recognition of two-dimensional images, the present application introduces a convolutional block attention mechanism module (CBAM) in the convolutional neural network (CNN) model, and focuses on the dynamically adjusted area in the MTF image through CBAM, effectively reducing the interference of redundant information. Therefore, the present application improves the accuracy of rolling bearing fault diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0019] Figure 1 An overall flow chart of a rolling bearing fault diagnosis method based on multi-source vibration signals provided in an embodiment of the present application;

[0020] Figure 2 An execution module diagram of a rolling bearing fault diagnosis method based on multi-source vibration signals provided in an embodiment of the present application;

[0021] Figure 3 A decision tree structure diagram provided for an embodiment of the present application;

[0022] Figure 4 A statistical diagram of the contribution of various characteristic values ​​provided in the embodiment of the present application to rolling bearing fault identification;

[0023] Figure 5The MTF image provided in the embodiment of the present application is not generated through dynamic adjustment;

[0024] Figure 6 The MTF image generated through dynamic adjustment provided in the embodiment of the present application;

[0025] Figure 7 A rolling bearing fault type probability diagram provided in an embodiment of the present application;

[0026] Figure 8 This is a diagram of the internal structure of a computer system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0027] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0028] The purpose of this application is to provide a rolling bearing fault diagnosis method, system and medium based on multi-source vibration signals, thereby improving the accuracy of rolling bearing fault diagnosis.

[0029] The development of intelligent production has put forward higher requirements for early rolling bearing fault diagnosis. Therefore, collecting information from multiple acceleration sensors to more completely describe the system status has become an important research direction for rolling bearing fault diagnosis. Compared with a single acceleration sensor, the fusion of multiple acceleration sensors can greatly improve the detection accuracy, robustness and reliability.

[0030] Feature extraction is the process of simplifying complex data into key numerical indicators (such as mean, standard deviation, etc.). It effectively identifies key information that helps fault diagnosis and is widely used in machine learning and signal processing. Commonly used feature extraction methods include time-frequency analysis, wavelet transform coefficient features, intrinsic mode function features after empirical mode decomposition, envelope spectrum features, and singular value decomposition features.

[0031] MTF is a technique for converting time series data into images, aiming to capture the temporal dependencies and patterns in the sequence. The core idea is to use the state transition matrix of the Markov chain to represent the transition probabilities between each moment in the time series, and then represent these probability information in the form of a two-dimensional image.

[0032] The emergence of CNN has greatly promoted the development of image recognition, especially its unique convolutional layer structure, which can automatically extract and learn features in images and no longer rely on manually designed features, which makes it perform well in various image classification tasks. Compared with traditional manual feature extraction methods, CNN not only reduces the dependence on expert experience, but also can extract deeper fault features from raw data, thereby improving the accuracy and robustness of diagnosis. Although CNN performs well in the fields of image processing and fault diagnosis, there may be a lot of redundant information in the features extracted by CNN, which not only increases the computational burden, but also may reduce the efficiency and accuracy of the model.

[0033] CBAM is a module that combines spatial attention and channel attention, designed to enhance the convolutional neural network's attention to important features. CBAM first calculates channel attention to identify which channels contain more critical information, and then calculates spatial attention to determine which spatial regions in the feature map are more important. This dual attention mechanism can effectively guide the model to focus on the more diagnostically significant parts during feature extraction, avoid the interference of redundant information, and help improve the performance of the model.

[0034] The above methods are widely used in the classification of other fault fields, but few of them are combined and applied to the research of rolling bearing fault diagnosis. This application considers combining and applying these advanced technologies to the fault diagnosis of rolling bearings. By introducing CBAM and dynamically adjusting the MTF image according to the key features in the vibration signal to highlight its key features, the model can be effectively guided to focus on the more diagnostically significant parts during the feature extraction process, avoiding the interference of redundant information, and helping to improve the performance of the model. This fault diagnosis method, which combines the optimization of time-frequency domain features and utilizes the fusion of vibration signals with a convolutional neural network containing an attention mechanism, significantly improves the accuracy and reliability of model classification. This combined method has broad application prospects and is expected to play an important role in the field of fault diagnosis of other equipment.

[0035] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0036] Example 1

[0037] This embodiment provides a rolling bearing fault diagnosis method based on multi-source vibration signals. Figure 1 and Figure 2 As shown, the rolling bearing fault diagnosis method based on multi-source vibration signals includes:

[0038] Step S1: Acquire vibration signals collected by multiple acceleration sensors.

[0039] In this embodiment, a plurality of acceleration sensors are respectively arranged at different positions of the target device; the target device is a device containing a rolling bearing to be diagnosed; and the sampling frequencies and transmission rates of all acceleration sensors are the same.

[0040] As a preferred implementation mode, three acceleration sensors of the same model are selected and respectively set at the base end, fan end and drive end of the target device, ensuring that the sampling frequency of the three acceleration sensors is 12KHz, so that they can simultaneously collect vibration signals of the base end, fan end and drive end.

[0041] Step S2: The vibration signals collected by all acceleration sensors are fused using the PCA method to obtain fused signal samples.

[0042] In this embodiment, step S2 specifically includes:

[0043] The first step is to construct the initial signal matrix X based on the vibration signals collected by all acceleration sensors.

[0044] Wherein, X is an n×m matrix, n is the number of acceleration sensors, m is the number of data points collected by each acceleration sensor, and each row element in X corresponds to a vibration signal collected by an acceleration sensor.

[0045] The second step is to normalize the initial signal matrix to obtain the standardized signal matrix X std .

[0046] X std =(X-μ) / σ;

[0047] Where μ is the mean of each column of data in X, and σ is the standard deviation of each column of data in X.

[0048] The third step is to calculate the covariance matrix of the standardized signal matrix and perform eigendecomposition on the covariance matrix to obtain multiple eigenvalues ​​and multiple eigenvectors.

[0049]

[0050] Cv i =λ i v i ;

[0051] Where C is the covariance matrix, λ i is the i-th eigenvalue, v i is the same as λ i The corresponding i-th eigenvector.

[0052] The fourth step is to select the first k eigenvalues ​​from all eigenvalues ​​as the principal component components, and take the largest eigenvalue among the first k eigenvalues ​​as the principal component eigenvalue.

[0053]

[0054] λ′=max{λ1,λ2,…,λ k};

[0055] Where η is the cumulative contribution rate and λ′ is the eigenvalue of the principal component.

[0056] The fifth step is to determine the eigenvector corresponding to the eigenvalue of the principal component, and fuse the standardized signal matrix with the eigenvector corresponding to the eigenvalue of the principal component to obtain a fused signal sample.

[0057] Z=X std v′;

[0058] Where Z is the fusion signal sample, and v′ is the feature vector corresponding to λ′.

[0059] Step S3: Divide the fused signal sample into multiple sub-samples using a sliding window method.

[0060] In this embodiment, in order to facilitate the subsequent feature extraction of the fused signal, the fused signal sample is divided into multiple sub-samples for direct processing through a sliding window, and each sub-sample contains a fused signal with a set sliding window length, so as to ensure that the complete vibration characteristics of the rolling bearing in one operating cycle can be captured, which is convenient for subsequent feature extraction and fault identification. The sliding window length is set as:

[0061] k min ;

[0062] L min =F s *[1 / (N / 60)];

[0063] Where k is an integer greater than or equal to 1, L min is the length of data collected by each acceleration sensor after the rolling bearing to be diagnosed runs for one cycle (the length of data collected by each acceleration sensor is the same), F s is the sampling frequency of the acceleration sensor, and N is the rotational speed of the rolling bearing to be diagnosed.

[0064] As a preferred implementation, in order to further enrich the fault characteristics and improve the accuracy of rolling bearing fault diagnosis, L min Set to 401, and the sliding window length is set to 2L min(ie 802), the sliding step size is set to 802 to ensure that the data in each window is completely independent, reduce repeated calculations of the same information, save computing resources, and avoid redundancy.

[0065] Step S4: Calculate multiple eigenvalues ​​of each subsample using time-frequency feature extraction, and identify the rolling bearing fault type based on the random forest model to screen out important eigenvalues ​​of each subsample (the important eigenvalues ​​of each subsample are the same).

[0066] In this embodiment, step S4 specifically includes:

[0067] In the first step, time-frequency feature extraction is used to calculate the multi-class feature values ​​under each sliding window.

[0068] Among them, one sliding window corresponds to one sub-sample, and the number of eigenvalue categories under each sliding window is the same. For example, 10 categories of eigenvalues ​​are extracted under each sliding window.

[0069] In the second step, based on the multi-class eigenvalues ​​under all sliding windows and the out-of-bag error increment in the random forest model (the out-of-bag error increment measures the importance of features by evaluating the contribution of each feature to the prediction performance in the random forest model), the contribution value of each type of eigenvalue to the identification of rolling bearing fault types is determined.

[0070] like Figure 3 The random forest structure shown in the figure includes 100 decision trees. Each leaf node includes at least one sample. All features are selected at each split, and the maximum split depth is not limited. This method uses out-of-bag data that is not seen during the decision tree construction process as the test set. For each feature, its value in the out-of-bag data is randomly shuffled, and the out-of-bag error of the model is recalculated. The difference between the out-of-bag error of a feature after shuffling and the original error is the increment of the feature, indicating its importance. The larger the error increment, the more significant the impact of the feature on the model effect. Figure 4 As shown, s1, s2, s3, ..., s15 are 15 feature labels (corresponding to 15 types of feature values), and the ordinate is the contribution value of each type of feature value to the identification of the rolling bearing fault type.

[0071] In the third step, multiple categories of eigenvalues ​​whose contribution values ​​are greater than the set threshold are taken as important eigenvalues, and the important eigenvalues ​​of each sub-sample are screened.

[0072] As a preferred implementation, Figure 4 The top four eigenvalues ​​in contribution value are taken as the important eigenvalues ​​of each subsample and used in subsequent analysis; multiple eigenvalues ​​with contribution value greater than 2.5 can also be taken as the important eigenvalues ​​of each subsample and used in subsequent analysis.

[0073] Step S5: Generate the MTF image of each sub-sample using CBAM, and dynamically adjust the brightness and RGB color channels of the corresponding MTF image according to the important feature values ​​of each sub-sample during the MTF image generation process.

[0074] In this embodiment, first, on each sub-sample, the fused signal is converted into a two-dimensional image through the Markov transition field. Specifically, for the fused signals Z1, Z2, ..., Z n , discretize it into Q bins according to its amplitude, and obtain a Q×Q Markov transition matrix. In order to make up for the shortcomings of the Markov transition field that is not sensitive to time series and step size, and considering the time position information, the matrix containing the amplitude is expanded to the Markov transition field containing the time axis, retaining the time correlation of the original signal, calculating the transition probability between the states to which each time series data point belongs, and obtaining an n×n transfer matrix, which is encoded into image pixels, thereby retaining the dynamic characteristics and local dependencies of the time series data.

[0075] Furthermore, the introduction of CBAM (different from the CBAM in the CBAM-CNN model, which exists independently) is mainly to enhance the feature extraction ability of the subsequent model through channel attention and spatial attention. If the four types of feature parameters, namely margin factor, centroid frequency, amplitude mean and spectral entropy, are determined according to the feature importance output by the random forest, then different scaling factors are applied to each color channel, and the response of each channel is adjusted using various eigenvalues ​​to enhance specific features; the contrast of each channel is adjusted through the imadjust function so that the model pays attention to important features on different channels; a spatial attention map is generated according to the spatial position of the image, and the feature expression of a specific area is enhanced by adjusting the spatial structure of the image. The before and after comparison of the MTF image generated by the above adjustments can be seen in Figure 5 and Figure 6 .

[0076] Step S6: The MTF images of all sub-samples are input into the trained CBAM-CNN model, and the rolling bearing fault type probability map is output.

[0077] In this embodiment, in order to obtain a trained CBAM-CNN model, the following steps need to be performed:

[0078] The first step is to build a CBAM-CNN model. The CBAM-CNN model is a convolutional neural network model that introduces a convolutional block attention mechanism module. The convolutional block attention mechanism module in the CBAM-CNN model is used to strengthen the attention to the brightness and RGB color channels in the MTF image of each sub-sample, so that it can focus on the feature areas that are critical to rolling bearing fault diagnosis through channel attention mechanism and spatial attention mechanism, thereby improving the accuracy and efficiency of feature extraction.

[0079] Specifically, the CBAM-CNN model includes: one CBAM layer, two convolutional layers, a maximum pooling layer, a tiling layer, and a fully connected layer. The CBAM-CNN model uses the Adam optimizer, and uses the LeakyReLU activation function and Dropout to prevent overfitting. Among them, the first convolutional layer consists of 64 3×3 convolutional kernels; the activation layer is LeakyReLU; the second convolutional layer consists of 128 3×3 convolutional kernels; the maximum pooling layer uses a 2×2 pooling window; the fully connected layer contains 512 neurons, and the activation function is ReLU. In order to prevent the CBAM-CNN model from overfitting, Dropout is also added to randomly discard 30% of the neurons. The output layer activation function is softmax, and the fault classification of rolling bearings is realized through the softmax activation function. The output dimension is the number of categories.

[0080] The second step is to obtain and process the training data. The data source is the Case Western Reserve University Bearing Data (CWRU). The bearing speed under CWRU is 1796RPM, the sampling frequency of the acceleration sensor is 12KHz, and the sources of the vibration signal include the base end, fan end and drive end.

[0081] Based on the CWRU data, the above steps S2 to S5 are executed to obtain the MTF images of multiple sub-samples, and the generated MTF images are divided into a training set and a validation set in a ratio of 8:2. Through cropping, 151 sub-samples are intercepted for each type of fault, and a total of 1963 MTF images are generated, including 1573 training sets and 390 validation sets.

[0082] The third step is to train the CBAM-CNN model. The MTF images in the training set are input into the CBAM-CNN model, and CNN will calculate the spatial attention and channel attention through CBAM. Among them, the channel attention ensures that the color adjustment part of the image is focused on. The reduction ratio of the channel attention is set to 9 to control the compression degree of each channel feature and reduce the number of neurons proportionally. Two Dense layers, the first one compresses the channel dimension, and the second one restores it to the original number of channels. The ReLU activation function is used as the activation of the first layer, and the final weight is generated through the Sigmoid activation function.

[0083] Spatial attention enhances the regions adjusted by spatial position. A 2D convolution layer with a kernel size of 4×4 is used to extract spatial information, and a spatial attention map is generated through a Sigmoid activation function. These modules focus on and weight the channel and spatial features of the image through the above parameters and structures, ensuring that the CBAM-CNN model pays more attention to the key parts of the image, that is, the parts adjusted by important features, during the classification process.

[0084] Finally, the performance of the CBAM-CNN model is evaluated by accuracy and classification cross entropy loss function, and the trained CBAM-CNN model is tested using the test set. Finally, the visualization result is output, i.e., the rolling bearing fault type probability map, which is used to determine the fault type of the rolling bearing to be diagnosed in the target equipment. Figure 7 As shown, F1, F2, F3...F13 correspond to 13 types of rolling bearing faults respectively. By comparing the sizes of the vertical coordinate probability values, the fault type of the rolling bearing to be diagnosed in the target equipment can be determined.

[0085] Example 2

[0086] This embodiment provides a computer system, which may be a server or a terminal, and its internal structure diagram may be as follows: Figure 8 As shown. The computer system includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer system is used to provide computing and control capabilities. The memory of the computer system includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer system is used to store vibration signals collected by multiple acceleration sensors. The input / output interface of the computer system is used to exchange information between the processor and an external device. The communication interface of the computer system is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the above-mentioned rolling bearing fault diagnosis method based on multi-source vibration signals is implemented.

[0087] Those skilled in the art will understand that Figure 8 The structure shown in the figure is only a block diagram of a part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer system to which the scheme of the present application is applied. The specific computer system may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.

[0088] Example 3

[0089] This embodiment provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0090] In summary, this application mainly has the following advantages:

[0091] (1) Multi-source signal fusion improves detection accuracy: The main components are extracted from multi-source signals through fusion algorithms, shielding signals with more noise but less contribution, highlighting key fault features, and significantly improving the accuracy and robustness of diagnosis. This method can integrate information from multi-dimensional signals, make up for the shortcomings of a single signal source in terms of information comprehensiveness and noise resistance, and provide more comprehensive diagnostic support.

[0092] (2) Dynamic feature image generation and CBAM to enhance CNN accuracy: In the process of image generation and recognition, in order to reduce the interference of redundant information on the model recognition ability, this application introduces a dynamic image adjustment technology based on feature screening. By screening out the feature values ​​with the strongest correlation with the rolling bearing fault type, the image is dynamically adjusted to ensure that the image contains the most representative feature information for fault diagnosis. Furthermore, CBAM is introduced into the CNN used for rolling bearing fault diagnosis, so that the CNN focuses on the dynamic adjustment area during the learning process, thereby effectively reducing the interference of redundant information. This method is equivalent to artificially prompting important features to the neural network, which helps the network to accurately capture key information, thereby significantly improving classification accuracy.

[0093] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0094] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0095] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0096] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0097] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A rolling bearing fault diagnosis method based on multi-source vibration signals, characterized in that: The rolling bearing fault diagnosis method based on multi-source vibration signals includes: Acquire vibration signals collected by multiple acceleration sensors; the multiple acceleration sensors are respectively arranged at different parts of the target device; the target device is a device containing a rolling bearing to be diagnosed; The vibration signals collected by all acceleration sensors are fused using the PCA method to obtain fused signal samples; The fused signal sample is divided into a plurality of sub-samples by using a sliding window method; each sub-sample contains a fused signal of a set sliding window length; Time-frequency feature extraction is used to calculate multiple eigenvalues ​​of each subsample, and the rolling bearing fault type is identified based on the random forest model to screen out the important eigenvalues ​​of each subsample; The MTF is used to generate a two-dimensional image of each sub-sample, and the brightness and RGB color channels of the two-dimensional image corresponding to each sub-sample are dynamically adjusted based on the CBAM and the important eigenvalues ​​of each sub-sample. The two-dimensional images after dynamic adjustment of all sub-samples are input into the trained CBAM-CNN model, and a rolling bearing fault type probability map is output; the rolling bearing fault type probability map is used to determine the fault type of the rolling bearing to be diagnosed in the target device.

2. The rolling bearing fault diagnosis method based on multi-source vibration signals according to claim 1 is characterized in that: The PCA method is used to fuse the vibration signals collected by all acceleration sensors to obtain fused signal samples, including: Based on the vibration signals collected by all acceleration sensors, an initial signal matrix is ​​constructed; each row element in the initial signal matrix corresponds to a vibration signal collected by an acceleration sensor; Performing standardization processing on the initial signal matrix to obtain a standardized signal matrix; Calculating a covariance matrix of the standardized signal matrix, and performing eigendecomposition on the covariance matrix to obtain a plurality of eigenvalues ​​and a plurality of eigenvectors; Select the first k eigenvalues ​​from all eigenvalues ​​as the principal component components, and take the largest eigenvalue among the first k eigenvalues ​​as the principal component eigenvalue; Determine the eigenvector corresponding to the principal component eigenvalue, and fuse the standardized signal matrix with the eigenvector corresponding to the principal component eigenvalue to obtain a fused signal sample.

3. The rolling bearing fault diagnosis method based on multi-source vibration signals according to claim 1 is characterized in that: The sliding window length is set to k*L min ; k is an integer greater than or equal to 1, L min It is the length of data collected by each acceleration sensor after the rolling bearing to be diagnosed runs for one cycle.

4. The rolling bearing fault diagnosis method based on multi-source vibration signals according to claim 3 is characterized in that: The formula for calculating the length of data collected by each acceleration sensor after the rolling bearing to be diagnosed runs for one cycle is: L min =F s *[1 / (N / 60)]; In the formula, F s is the sampling frequency of the acceleration sensor, and N is the rotational speed of the rolling bearing to be diagnosed.

5. The rolling bearing fault diagnosis method based on multi-source vibration signals according to claim 1 is characterized in that: Time-frequency feature extraction is used to calculate multiple eigenvalues ​​of each subsample, and the rolling bearing fault type is identified based on the random forest model to screen out the important eigenvalues ​​of each subsample, including: The multi-class feature values ​​under each sliding window are calculated by using time-frequency feature extraction; one sliding window corresponds to one subsample; the number of feature value categories under each sliding window is the same; Based on the multi-class eigenvalues ​​under all sliding windows, the contribution value of each eigenvalue to the identification of rolling bearing fault types is determined according to the out-of-bag error increment in the random forest model. The multi-category eigenvalues ​​whose contribution values ​​are greater than the set threshold are taken as important eigenvalues, and the important eigenvalues ​​of each sub-sample are screened.

6. The rolling bearing fault diagnosis method based on multi-source vibration signals according to claim 1 is characterized in that: All accelerometers have the same sampling frequency and transmission rate.

7. The rolling bearing fault diagnosis method based on multi-source vibration signals according to claim 1 is characterized in that: The CBAM-CNN model is a convolutional neural network model that introduces a convolutional block attention mechanism module.

8. The rolling bearing fault diagnosis method based on multi-source vibration signals according to claim 7 is characterized in that: The convolutional block attention mechanism module in the CBAM-CNN model is used to strengthen the attention to the brightness and RGB color channels in the MTF image of each sub-sample.

9. A computer system comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the rolling bearing fault diagnosis method based on multi-source vibration signals as described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the rolling bearing fault diagnosis method based on multi-source vibration signals described in any one of claims 1 to 8 is implemented.

Images