A method and apparatus for identifying bearing faults

Abstract

The embodiment of the application provides a kind of method and device for identifying bearing failure, the method comprises: obtaining the vibration signal and sound signal of target bearing operation;The vibration signal and sound signal are brought into convolution sparse characterization layer, and the output result of the convolution sparse characterization layer is obtained as feature data, the feature data includes first vibration feature and first sound feature;Second vibration feature and second sound feature of the feature data are adaptively extracted by classic convolution and hollow convolution respectively;The second vibration feature and second sound feature are fused convolution to obtain fusion feature, and the running state of the target bearing is identified according to the fusion feature, which improves the efficiency of identifying bearing failure.

Description

Technical Field

[0001] This invention relates to the field of mechanical equipment failure mode recognition, and specifically to a method and apparatus for identifying bearing failures. Background Technology

[0002] The procurement, operation, and maintenance costs of modern equipment are increasing, while advanced mechanical equipment condition monitoring and fault diagnosis technologies can significantly reduce maintenance costs and effectively improve production efficiency. Bearings, as key components of rotating machinery, are crucial to the normal operation of the equipment. Current technologies primarily use deep learning to acquire fault characteristics; however, due to the complexity of the operating environment, deep learning methods may learn many invalid features, thus affecting the final fault identification results. Summary of the Invention

[0003] The purpose of this invention is to provide a method and apparatus for identifying bearing faults, which improves the efficiency of bearing fault identification.

[0004] To achieve the above objectives, embodiments of the present invention provide a method for identifying bearing faults, the method comprising:

[0005] Acquire vibration and sound signals from the target bearing during operation;

[0006] The vibration signal and the sound signal are fed into a convolutional sparse representation layer to obtain the output of the convolutional sparse representation layer as feature data. The feature data includes a first vibration feature and a first sound feature.

[0007] The second vibration feature and the second sound feature of the feature data are adaptively extracted using classical convolution and dilated convolution, respectively.

[0008] The second vibration feature and the second sound feature are fused and convolved to obtain a fused feature, and the operating state of the target bearing is identified based on the fused feature.

[0009] Optionally, the step of feeding the vibration signal and sound signal into the convolutional sparse representation layer to obtain the output of the convolutional sparse representation layer as feature data includes:

[0010] One-dimensional convolution is performed on the vibration signal and sound signal to generate a convolution sparse graph;

[0011] A convolutional sparse representation layer is determined based on the convolutional sparse graph, and the convolutional sparse representation layer includes a fidelity loss function and a regularization loss function.

[0012] Feature data is obtained by processing the vibration signal and / or sound signal with a fidelity loss function and a regularization loss function.

[0013] Optionally, the step of processing the vibration signal and / or sound signal with a fidelity loss function and a regularization loss function to obtain feature data includes:

[0014] Fault features are extracted from the vibration signal and / or sound signal based on the fidelity loss function.

[0015] Feature data is obtained by suppressing redundant and invalid features in the fault features using the regularization loss function.

[0016] Optionally, the adaptive extraction of the second vibration feature and the second sound feature of the feature data through classical convolution and dilated convolution, respectively, includes:

[0017] The second vibration feature of the first vibration feature is extracted adaptively using classical convolution;

[0018] The second sound feature of the first sound feature is extracted adaptively by dilated convolution.

[0019] Optionally, the step of fusing and convolving the second vibration feature and the second sound feature to obtain the fused feature includes:

[0020] The second vibration feature and the second sound feature are combined in one dimension to obtain the fused feature.

[0021] Optionally, the method further includes:

[0022] If the target bearing malfunctions, the fault type of the target bearing is identified based on the fusion characteristics.

[0023] Optionally, the fault types include: bearing outer ring fault, bearing inner ring fault, bearing cage fault, and bearing rolling element fault.

[0024] On the other hand, the present invention also proposes a device for identifying bearing faults, the device comprising:

[0025] The acquisition module is used to acquire vibration and sound signals from the target bearing during operation.

[0026] The first processing module is used to input the vibration signal and the sound signal into the convolutional sparse representation layer to obtain the output result of the convolutional sparse representation layer as feature data, the feature data including a first vibration feature and a first sound feature.

[0027] The second processing module is used to adaptively extract the second vibration feature and the second sound feature of the feature data through classical convolution and dilated convolution, respectively.

[0028] The third processing module is used to perform fusion convolution on the second vibration feature and the second sound feature to obtain a fused feature, and to identify the operating state of the target bearing based on the fused feature.

[0029] Optionally, the adaptive extraction of the second vibration feature and the second sound feature of the feature data through classical convolution and dilated convolution, respectively, includes:

[0030] The second vibration feature of the first vibration feature is extracted adaptively using classical convolution;

[0031] The second sound feature of the first sound feature is extracted adaptively by dilated convolution.

[0032] Optionally, the third processing module is further configured to:

[0033] If the target bearing malfunctions, the fault type of the target bearing is identified based on the fusion characteristics;

[0034] The fault types include: bearing outer ring fault, bearing inner ring fault, bearing cage fault, and bearing rolling element fault.

[0035] This invention discloses a method for identifying bearing faults. The method includes: acquiring vibration and sound signals of a target bearing during operation; inputting the vibration and sound signals into a convolutional sparse representation layer to obtain the output of the convolutional sparse representation layer as feature data, wherein the feature data includes a first vibration feature and a first sound feature; adaptively extracting a second vibration feature and a second sound feature from the feature data using classical convolution and dilated convolution, respectively; performing a fusion convolution on the second vibration feature and the second sound feature to obtain a fused feature; and identifying the operating state of the target bearing based on the fused feature. This invention, by constructing a convolutional sparse representation layer, achieves the extraction of fault pulse components from the signal and suppresses invalid features therein, thereby avoiding interference from invalid features in the signal on the diagnostic results, and thus enhancing the model's feature extraction capability and diagnostic effect.

[0036] Other features and advantages of the embodiments of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0037] The accompanying drawings are provided to further illustrate embodiments of the present invention and form part of the specification. They are used together with the following detailed description to explain the embodiments of the present invention, but do not constitute a limitation thereof. In the drawings:

[0038] Figure 1 This is a schematic flowchart of a method for identifying bearing faults according to the present invention;

[0039] Figure 2This is a schematic diagram of the convolutional sparse representation layer of the present invention;

[0040] Figure 3 This is a diagram of the two-dimensional convolution kernel structure of the present invention;

[0041] Figure 4 This is a diagram comparing classic convolution and dilated convolution;

[0042] Figure 5 This is a schematic diagram of a fused convolutional layer network structure;

[0043] Figure 6 This is a schematic diagram of a bearing fault identification device according to the present invention;

[0044] Figure 7 This is a flowchart illustrating one embodiment of the present invention;

[0045] Figure 8 This is a schematic diagram of the confusion matrix of the fault mode recognition results of the present invention.

[0046] Explanation of reference numerals in the attached figures

[0047] 100 - A device for identifying bearing failures;

[0048] 200 - Acquisition Module;

[0049] 300 - First Processing Module;

[0050] 400 - Second Processing Module;

[0051] 500 - Third processing module. Detailed Implementation

[0052] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the scope of the present invention.

[0053] It should be noted that the acquisition, transmission, storage, use, and processing of data in the technical solution of this application all comply with the relevant provisions of national laws and regulations. In the embodiments of this application, certain existing industry solutions such as software, components, and models may be mentioned. These should be considered exemplary, intended only to illustrate the feasibility of implementing the technical solution of this application, and do not imply that the applicant has already used or necessarily used such solutions.

[0054] Example 1

[0055] Figure 1 This is a schematic flowchart of a method for identifying bearing faults according to the present invention, as shown below. Figure 1As shown, the present invention provides a method for identifying bearing faults, the method comprising: step S101 of acquiring vibration signals and sound signals of the target bearing during operation.

[0056] In one specific implementation, vibration and sound signals of the target bearing during operation are collected using vibration and noise sensors. Specifically, the vibration and sound signals are preprocessed, including operations such as filtering and noise reduction.

[0057] Step S102 involves feeding the vibration signal and sound signal into a convolutional sparse representation layer to obtain the output of the convolutional sparse representation layer as feature data. The feature data includes a first vibration feature and a first sound feature.

[0058] According to one specific implementation, the step of feeding the vibration signal and sound signal into a convolutional sparse representation layer to obtain the output of the convolutional sparse representation layer as feature data includes: performing one-dimensional convolution on the vibration signal and sound signal to generate a convolutional sparse map; determining a convolutional sparse representation layer based on the convolutional sparse map, wherein the convolutional sparse representation layer includes a fidelity loss function and a regularization loss function; and obtaining feature data by processing the vibration signal and / or sound signal with the fidelity loss function and the regularization loss function.

[0059] The step of processing the vibration signal and / or sound signal with a fidelity loss function and a regularization loss function to obtain feature data includes: extracting fault features from the vibration signal and / or sound signal according to the fidelity loss function; and suppressing redundant and invalid features in the fault features according to the regularization loss function to obtain feature data.

[0060] Due to the harsh operating environment of mechanical equipment, the observed signals are subject to numerous interferences such as noise, making feature extraction by the network quite challenging. To suppress the influence of these interferences, convolutional sparse layers are constructed in the deep network to extract the fault pulse component of the signal and suppress invalid features, thereby greatly preserving the pulse features beneficial for fault identification. Specifically, for example... Figure 2 As shown, this application uses one-dimensional convolution to generate a convolutional sparse graph D, and its specific calculation is as follows:

[0061]

[0062] In the formula: is the ReLU function (an activation function for neural networks); x is the acquired signal; w k and b k These represent the weights and kernel bias of the convolution kernel; c k is the k-th channel of the convolutional sparse graph.

[0063] The convolutional sparse map D obtained after the above one-dimensional convolution operation exhibits features hidden in the signal in each channel. These features are considered as atoms in the convolutional sparse layer. The sparse representation of the signal is achieved through linear combinations of atoms in the convolutional sparse map.

[0064] For the resulting convolutional sparse graph D = {c1, c2, ..., c} consisting of m atoms, m} and the initialized sparse vector v∈R m×1 The sparse representation of the original signal can be expressed as:

[0065] x i =D i v

[0066] In the formula: x i Let represent a single raw input signal, where i is the i-th sample. Updates to the convolutional sparse graph D and the sparse vector v are performed during training.

[0067] For a convolutional sparse representation layer, the training loss consists of two parts: a fidelity term and a sparse regularization term composed of the l1 norm. The fidelity term aims to preserve useful impulse components in the original signal as effectively as possible, and its specific calculation formula is as follows:

[0068]

[0069] The main purpose of the sparse regularization term composed of the l1 norm is to suppress redundant and invalid features in the signal, making the signal sparser. The formula for calculating the l1 norm sparse regularization term is:

[0070]

[0071] In the formula: λ is the coefficient representing the trade-off between the fidelity term and the sparsity regularization term. Therefore, the total loss function of the convolutional sparse representation layer is:

[0072] Loss = Loss1 + loss2

[0073] The overall loss function of the convolutional sparse representation layer is a combination of the loss functions of the two methods mentioned above, thus ensuring that important components of the original signal are preserved while suppressing invalid components. Therefore, compared with conventional convolutional layers that directly extract features from the original noisy signal, the convolutional sparse representation layer can more effectively improve feature extraction performance.

[0074] This application constructs a convolutional sparse representation layer to extract fault pulse components from signals and suppress invalid features, thereby avoiding interference from invalid features in the signal on the diagnostic results and enhancing the model's feature extraction capability and diagnostic effect.

[0075] Step S103 involves adaptively extracting the second vibration feature and the second sound feature of the feature data using classical convolution and dilated convolution, respectively.

[0076] According to one specific implementation, the adaptive extraction of the second vibration feature and the second sound feature of the feature data by classical convolution and dilated convolution respectively includes: adaptively extracting the second vibration feature of the first vibration feature by classical convolution; and adaptively extracting the second sound feature of the first sound feature by dilated convolution.

[0077] Classical convolution is one of the fundamental components of convolutional neural networks, and its primary function is the adaptive extraction of data features. Classical convolution consists of multiple convolutional kernels of fixed size with learnable parameters.

[0078] Specifically, taking the classic two-dimensional convolutional layer as an example, its structure is as follows: Figure 3 As shown, the process of performing convolution on the input 2D data involves sliding the convolution kernel from top to bottom and from left to right, sequentially traversing the input data map. Each traversal results in the sum of the product of the value in the convolution kernel and the corresponding value in the input data map. After each convolution operation, a corresponding feature map is obtained. Since each convolutional layer consists of multiple convolutional kernels, the final output of the convolutional layer is the result of stacking multiple feature maps together.

[0079] The formula for calculating the size of each feature map in the output of a convolutional layer is:

[0080]

[0081] D out =K

[0082] In the above formula: W in W represents the width of the input data. out H is the width of the output feature map. in and H out These represent the heights of the input and output data, respectively; D out The output feature map is denoted by F; the filter size is denoted by P; the boundary padding size is denoted by S; the step size is denoted by K; and the number of filters is denoted by K.

[0083] For convolutional neural networks used for one-dimensional signals such as vibration signals, the convolutional layer operation is the same as the above process, except that the convolutional kernel is one-dimensional and the resulting feature map is also one-dimensional.

[0084] Because the shrinking and enlarging of feature maps can cause accuracy loss, the concept of dilated convolution was proposed to address this issue. Unlike classic convolution, dilated convolution introduces a hyperparameter α called the dilation rate. This hyperparameter α defines the distance between convolution kernels during data processing, allowing for different detection ranges to be obtained by changing its value, thus capturing more information. Dilated convolution uses a dilation rate α to expand the convolution kernel. Figure 4 This demonstrates the classic two-dimensional convolution ( Figure 4 (Left figure) and expansion rates of 1 and 2 respectively (respectively) Figure 4 China Library and Figure 4 The receptive field of the convolution kernel corresponding to the right figure is from Figure 4 As can be seen, with the increase of the dilation rate, the size of the convolution kernel will increase, thereby expanding the receptive field of the convolution kernel.

[0085] Therefore, let q be the size of the convolution kernel in classical convolution, then the mathematical expression for the kernel size of dilated convolution is as follows:

[0086] Q = q + (q-1)(α-1)

[0087] In the formula: Q is the actual kernel size of the dilated convolution.

[0088] According to another specific implementation, the step of adaptively extracting the second vibration feature and the second sound feature of the feature data by classical convolution and dilated convolution respectively includes: adaptively extracting the second sound feature of the first sound feature by classical convolution; and adaptively extracting the second vibration feature of the first vibration feature by dilated convolution.

[0089] Both types of signals contain fault features. A classic convolutional layer can extract local features of one signal, while a dilated convolutional layer can extract global features of the other signal. By fusing these features, we can ensure that fault features at multiple scales can be extracted.

[0090] Classical convolutional layers excel at extracting local features but have a limited receptive field. Dilated convolutional layers, by adjusting their dilation rate, can expand the receptive field and capture information from farther distances. This application combines both to extract features at different scales, improving model performance. Specifically, the complementarity in computation and parameters includes: classical convolutional layers have relatively low computational cost and fewer parameters; dilated convolutional layers, while maintaining relatively low computational cost and fewer parameters, can capture information from farther distances. Combining the two can improve model performance while maintaining model efficiency. The complementarity in application scenarios includes: in some tasks (such as object detection and image super-resolution), dilated convolutional layers may be more suitable for capturing long-distance dependencies. In other tasks (such as image classification and edge detection), classical convolutional layers may be better at extracting local features. Depending on the specific task requirements, appropriate convolutional layer types can be selected or both can be used in combination.

[0091] Step S104 involves fusing and convolving the second vibration feature and the second sound feature to obtain a fused feature, and then identifying the operating state of the target bearing based on the fused feature.

[0092] According to one specific implementation, the step of fusing and convolving the second vibration feature and the second sound feature to obtain the fused feature includes: performing one-dimensional splicing of the second vibration feature and the second sound feature to obtain the fused feature.

[0093] Specifically, dilated convolution adds a dilation rate α to the classic convolution to expand the receptive field of the convolution kernel, thereby obtaining more global and multi-scale information from the input features and improving model performance. However, if the parameter is too large, the convolutional layer will lose a lot of original information during the operation, resulting in a net loss. Therefore, fusion convolutional layers use both classic convolution and dilated convolution to complement each other's advantages, satisfying the need to obtain more information from the original signal while overcoming the problem of some information loss. The specific network structure of the fusion convolutional layer is as follows: Figure 5 As shown.

[0094] This application constructs a fusion convolutional layer to fuse multi-source heterogeneous features, making full use of multi-source heterogeneous signals and improving the final diagnostic effect.

[0095] The method further includes: if the target bearing malfunctions, identifying the fault type of the target bearing based on the fusion features. The fault types include: bearing outer ring fault, bearing inner ring fault, bearing cage fault, and bearing rolling element fault.

[0096] Example 2

[0097] This invention also proposes a device for identifying bearing faults, such as... Figure 6As shown, the bearing fault identification device 100 includes: an acquisition module 200 for acquiring vibration and sound signals of the target bearing during operation; a first processing module 300 for inputting the vibration and sound signals into a convolutional sparse representation layer to obtain the output of the convolutional sparse representation layer as feature data, the feature data including a first vibration feature and a first sound feature; a second processing module 400 for adaptively extracting the second vibration feature and the second sound feature of the feature data through classical convolution and dilated convolution, respectively; and a third processing module 500 for performing a fusion convolution on the second vibration feature and the second sound feature to obtain a fused feature, and identifying the operating state of the target bearing based on the fused feature.

[0098] According to one specific implementation, the step of adaptively extracting the second vibration feature and the second sound feature of the feature data by classical convolution and dilated convolution respectively includes: adaptively extracting the second vibration feature of the first vibration feature by classical convolution; and adaptively extracting the second sound feature of the first sound feature by dilated convolution.

[0099] The third processing module 500 is further configured to: if the target bearing malfunctions, identify the fault type of the target bearing based on the fusion features; the fault type includes: bearing outer ring fault, bearing inner ring fault, bearing cage fault, and bearing rolling element fault. The bearing fault identification device 100 performs fault mode identification based on the input signal to determine the bearing fault type.

[0100] The bearing fault identification device 100 extracts fault pulse components from signals and suppresses invalid features by constructing a convolutional sparse representation layer, thereby avoiding interference from invalid features in the signal on the diagnostic results and enhancing the model's feature extraction capability and diagnostic effect. Furthermore, in order to make full use of the fault information contained in multi-source heterogeneous signals, a deep sparse representation network for heterogeneous information fusion is proposed. This device constructs a fusion convolutional layer to fuse multi-source heterogeneous features, enabling the model to make full use of multi-source heterogeneous signals and improve its final diagnostic effect.

[0101] Example 3

[0102] This embodiment is a multi-source dataset consisting of 6207 deep groove ball bearings in both inner and outer ring faults, rolling element faults, cage faults, and normal conditions. Figure 7 As shown, the specific steps are as follows:

[0103] Step 1: Collect vibration and sound signals of the experimental bearing using an accelerometer and a noise sensor, such as... Figure 8 As shown. The signal sampling frequency is 20480Hz, the bearing rotation speed is 3000r / min, and the radial load is 2KN.

[0104] Step 2: Organize the above dataset to have 400 samples of each type, including 200 vibration signal samples and 200 sound signal samples, with each sample containing 1024 sampling points. This results in a new dataset of 2000 samples. 80% of this dataset will be used as the training dataset, and the remaining 20% ​​as the test dataset.

[0105] Step 3: Input the training dataset into the bearing fault identification device to learn the corresponding parameters.

[0106] Step 4: Input the test dataset into the bearing fault identification device to perform fault mode identification on the bearing.

[0107] Step 5: Display the recognition results obtained in Step 4 using a confusion matrix, such as... Figure 7 As shown, its recognition accuracy is 99.5%. This verifies that the method can accurately identify bearing failure modes.

[0108] The convolutional sparse representation layer of this invention extracts fault pulse components from the signal and suppresses invalid features therein, thereby avoiding interference from invalid features in the signal on the diagnostic results; while the fusion convolutional layer is used to fuse multi-source heterogeneous features, so that the model can make full use of multi-source heterogeneous signals, thereby enhancing the model's feature extraction capability and improving its final fault model recognition effect.

[0109] This invention discloses a method for identifying bearing faults. The method includes: acquiring vibration and sound signals of a target bearing during operation; inputting the vibration and sound signals into a convolutional sparse representation layer to obtain the output of the convolutional sparse representation layer as feature data, wherein the feature data includes a first vibration feature and a first sound feature; adaptively extracting a second vibration feature and a second sound feature from the feature data using classical convolution and dilated convolution, respectively; performing a fusion convolution on the second vibration feature and the second sound feature to obtain a fused feature; and identifying the operating state of the target bearing based on the fused feature. This invention, by constructing a convolutional sparse representation layer, extracts fault pulse components from the signal and suppresses invalid features, thereby avoiding interference from invalid features in the signal on the diagnostic results, and thus enhancing the model's feature extraction capability and diagnostic effect.

[0110] On the other hand, embodiments of the present invention provide a storage medium on which a program is stored, which, when executed by a processor, implements the method for identifying bearing faults.

[0111] This invention provides a processor for running a program, wherein the program executes the method for identifying bearing faults.

[0112] This invention provides a device including a processor, a memory, and a program stored in the memory and executable on the processor, which is implemented when the processor executes the program. The device described herein can be a server, PC, PAD, mobile phone, etc.

[0113] This application also provides a computer program product, which, when executed on a data processing device, is suitable for executing an initialization program with the following method steps: acquiring vibration and sound signals of a target bearing in operation; inputting the vibration and sound signals into a convolutional sparse representation layer to obtain the output of the convolutional sparse representation layer as feature data, the feature data including a first vibration feature and a first sound feature; adaptively extracting a second vibration feature and a second sound feature from the feature data using classical convolution and dilated convolution, respectively; performing a fusion convolution on the second vibration feature and the second sound feature to obtain a fused feature; and identifying the operating state of the target bearing based on the fused feature.

[0114] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0115] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0116] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0117] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0118] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0119] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0120] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, 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 technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0121] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0122] The above are merely embodiments of this application and are not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application. Those skilled in the art should understand that the embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0123] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0124] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0125] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0126] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0127] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0128] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, 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 technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0129] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0130] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for identifying bearing faults, characterized in that, The method includes: Acquire vibration and sound signals from the target bearing during operation; The vibration signal and the sound signal are fed into a convolutional sparse representation layer to obtain the output of the convolutional sparse representation layer as feature data. The feature data includes a first vibration feature and a first sound feature. The second vibration feature and the second sound feature of the feature data are adaptively extracted using classical convolution and dilated convolution, respectively. The second vibration feature and the second sound feature are fused and convolved to obtain a fused feature, and the operating state of the target bearing is identified based on the fused feature.

2. The method according to claim 1, characterized in that, The step of feeding the vibration signal and sound signal into the convolutional sparse representation layer to obtain the output of the convolutional sparse representation layer as feature data includes: One-dimensional convolution is performed on the vibration signal and the sound signal to generate a convolutional sparse graph; A convolutional sparse representation layer is determined based on the convolutional sparse graph, and the convolutional sparse representation layer includes a fidelity loss function and a regularization loss function. Feature data is obtained by processing the vibration signal and / or sound signal with a fidelity loss function and a regularization loss function.

3. The method according to claim 2, characterized in that, The feature data obtained after processing the vibration signal and / or sound signal with a fidelity loss function and a regularization loss function includes: Fault features are extracted from the vibration signal and / or sound signal based on the fidelity loss function. Feature data is obtained by suppressing redundant and invalid features in the fault features using the regularization loss function.

4. The method according to claim 1, characterized in that, The method of adaptively extracting the second vibration feature and the second sound feature of the feature data through classical convolution and dilated convolution respectively includes: The second vibration feature of the first vibration feature is extracted adaptively using classical convolution; The second sound feature of the first sound feature is extracted adaptively by dilated convolution.

5. The method according to claim 1, characterized in that, The process of fusing and convolving the second vibration feature and the second sound feature to obtain the fused feature includes: The second vibration feature and the second sound feature are combined in one dimension to obtain the fused feature.

6. The method according to claim 1, characterized in that, The method also includes: If the target bearing malfunctions, the fault type of the target bearing is identified based on the fusion characteristics.

7. The method according to claim 6, characterized in that, The fault types include: bearing outer ring fault, bearing inner ring fault, bearing cage fault, and bearing rolling element fault.

8. A device for identifying bearing faults, characterized in that, The device includes: The acquisition module is used to acquire vibration and sound signals from the target bearing during operation. The first processing module is used to input the vibration signal and the sound signal into the convolutional sparse representation layer to obtain the output result of the convolutional sparse representation layer as feature data, the feature data including a first vibration feature and a first sound feature. The second processing module is used to adaptively extract the second vibration feature and the second sound feature of the feature data through classical convolution and dilated convolution, respectively. The third processing module is used to perform fusion convolution on the second vibration feature and the second sound feature to obtain a fused feature, and to identify the operating state of the target bearing based on the fused feature.

9. The apparatus according to claim 8, characterized in that, The method of adaptively extracting the second vibration feature and the second sound feature of the feature data through classical convolution and dilated convolution respectively includes: The second vibration feature of the first vibration feature is extracted adaptively using classical convolution; The second sound feature of the first sound feature is extracted adaptively by dilated convolution.

10. The apparatus according to claim 8, characterized in that, The third processing module is also used for: If the target bearing malfunctions, the fault type of the target bearing is identified based on the fusion characteristics; The fault types include: bearing outer ring fault, bearing inner ring fault, bearing cage fault, and bearing rolling element fault.

Images