A bearing fault classification method, device, equipment and storage medium

Abstract

The present invention discloses a classification method, device, equipment and storage medium for bearing faults. The method comprises: determining bearing image data according to a vibration signal under a bearing fault state; processing the bearing image data by a residual block to obtain a bearing depth feature; the residual block is constructed by a causal hole convolution layer; extracting a bearing basic feature from the bearing image data by a basic feature extraction unit; fusing the bearing depth feature and the bearing basic feature, and determining the bearing fault type according to the fusion result. The embodiment of the present invention can simplify the target and difficulty of iterative learning and effectively improve the classification ability of the model.

Description

Technical Field

[0001] The present invention relates to the technical field of fault classification, and in particular to a bearing fault classification method, device, equipment and storage medium. Background Art

[0002] Bearing wear failure is very common in the use of motors. The main cause is the vibration and relative friction between the inner and outer rings of the bearing and the steel balls during transportation and use. For bearing failure, any local failure with subtle characteristics may be continuously extended and amplified to the entire operating system of the mechanical equipment, causing unnecessary equipment damage and economic losses. How to detect and determine the type of failure when a bearing fails has become an important issue in this field.

[0003] Currently, some researchers have begun to introduce deep learning models to learn bearing fault data, and to discover and classify faults based on deep learning models. However, due to the difficulty and high cost of collecting bearing fault data, most bearing fault data has a small amount of data and is subject to noise interference. The classification results of these models are generally poor and cannot meet the needs of modern equipment. Summary of the invention

[0004] The present invention provides a bearing fault classification method, device, equipment and storage medium to simplify the goal and difficulty of iterative learning and effectively improve the classification ability of the model.

[0005] According to one aspect of the present invention, a method for classifying bearing faults is provided, comprising:

[0006] Determine bearing image data according to the vibration signal under the bearing fault state;

[0007] The bearing image data is processed by a residual block to obtain a bearing depth feature; the residual block is constructed by a causal hole convolution layer;

[0008] Extracting basic bearing features from the bearing image data by a basic feature extraction unit;

[0009] The bearing depth feature and the bearing basic feature are fused, and the bearing fault type is determined according to the fusion result.

[0010] According to another aspect of the present invention, there is provided a bearing fault classification device, comprising:

[0011] An image data determination module, used to determine the bearing image data according to the vibration signal in the bearing fault state;

[0012] A depth feature determination module, used for processing the bearing image data through a residual block to obtain a bearing depth feature; the residual block is constructed through a causal hole convolution layer;

[0013] A basic feature determination module, used for extracting basic bearing features from the bearing image data through a basic feature extraction unit;

[0014] The fault type determination module is used to fuse the bearing depth feature and the bearing basic feature, and determine the bearing fault type according to the fusion result.

[0015] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0016] at least one processor; and

[0017] a memory communicatively connected to the at least one processor; wherein,

[0018] The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the method for classifying bearing faults described in any embodiment of the present invention.

[0019] According to another aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the method for classifying bearing faults described in any embodiment of the present invention when executed.

[0020] The embodiment of the present invention obtains the bearing deep features through the residual block constructed based on the causal hole convolution layer, and integrates it with the basic bearing features to determine the bearing fault type, which simplifies the goal and difficulty of iterative learning and effectively improves the classification ability under noisy conditions.

[0021] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0023] Figure 1A is a flow chart of a bearing fault classification method provided according to an embodiment of the present invention;

[0024] Figure 1B is a schematic diagram of obtaining bearing image data according to an embodiment of the present invention;

[0025] Figure 1C is a schematic diagram of a causal hole convolution layer provided according to an embodiment of the present invention;

[0026] Figure 2A is a flow chart of a bearing fault classification method provided according to another embodiment of the present invention;

[0027] Figure 2B is a schematic diagram of a residual block structure provided according to another embodiment of the present invention;

[0028] Figure 2C is a schematic diagram of a neural network model provided according to another embodiment of the present invention;

[0029] Figure 2D is a schematic diagram of a local binary pattern calculation process provided according to another embodiment of the present invention;

[0030] Figure 3 is a structural schematic diagram of a bearing fault classification device provided according to another embodiment of the present invention;

[0031] Figure 4 It is a schematic diagram of the structure of an electronic device implementing an embodiment of the present invention. DETAILED DESCRIPTION

[0032] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0033] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0034] Figure 1A This is a flow chart of a bearing fault classification method provided by an embodiment of the present invention. This embodiment can be applied to classify bearing faults using signals in a bearing fault state. The method can be executed by a bearing fault classification device, which can be implemented in the form of hardware and / or software. The device can be configured in an electronic device with corresponding data processing capabilities. As shown in Figure 1, the method includes:

[0035] S110, determining bearing image data according to a vibration signal in a bearing fault state.

[0036] Specifically, when a bearing fails, the vibration signal generated during operation is collected by a collection device. However, the vibration signal at this time is mostly in the form of a waveform and cannot be directly processed by the neural network (model), so it needs to be converted into processable bearing image data. In the present invention, the bearing image data can be a grayscale image in the form of N×N to meet the input requirements of the neural network.

[0037] Optionally, acquiring bearing image data according to a vibration signal in a bearing fault state includes:

[0038] A preset number of continuous sampling points are collected from the vibration signal under the bearing fault state; the continuous sampling points are overlapped and sampled with a sampling step size less than the preset number to obtain multiple sampling data; the sampling data are converted into a time-frequency image through continuous wavelet transform; and the time-frequency image is converted into an N×N grayscale image as bearing image data.

[0039] Specifically, a preset number (e.g., 2048) of continuous sampling points are intercepted from the original bearing fault vibration signal, and the intercepted continuous sampling points are converted into a time-frequency image using a continuous wavelet transform (CWT). Finally, the time-frequency image is grayed and converted into a grayscale image of size N×N to be input into the neural network model. In the process of sampling from the original vibration signal, in order to expand the data set, continuous sampling points can be intercepted with a sampling step size (e.g., 512) that is less than the preset number, that is, at least two sampling data are obtained from a section of continuous sampling points in an overlapping interception manner, which expands the number and information richness of the bearing image data and improves the classification effect of the bearing fault type. The present invention uses CWT to preprocess the collected bearing fault signal as follows: For a one-dimensional vibration signal x(t), CWT on the signal can be expressed as the following equation (1): Wherein, a and b respectively control the two transformations of the wavelet, a is the time axis scale scaling parameter, and b is the time translation parameter. If |a|>1, the mother wavelet ψ(t) is stretched on the time axis and the amplitude is depressed, ψ a，b (t) contains the characteristics of low-frequency quantities; if |a| < 1, then ψ(t) is compressed on the time axis and the amplitude is increased, ψ a，b (t) contains the characteristics of high frequency. Different b values ​​indicate that the wavelet moves to different positions along the time axis.

[0040]

[0041] Since the Morlet wavelet matches the impact characteristics generated by the bearing defect, that is, the Morlet wavelet can be used to feedback the impact characteristics generated by the bearing fault. Therefore, in this study, the Morlet wavelet is applied to CWT. The definition of the Morlet wavelet function is shown in equation (2), where C represents the normalization constant.

[0042]

[0043] For example, Figure 1B This is a schematic diagram of obtaining bearing image data provided by an embodiment of the present invention. In which, 2048 sampling points are first intercepted from the original bearing fault signal, and then a plurality of sampling data are obtained with a sampling step of 512. Each segment of sampling data is subjected to CWT preprocessing, image graying and image size modification, and finally a grayscale image of size N×N is obtained as bearing image data, where N is preferably greater than or equal to 3.

[0044] S120, processing the bearing image data through a residual block to obtain a bearing depth feature; the residual block is constructed through a causal hole convolution layer.

[0045] Specifically, in order to extract the deeper fault features of the bearing fault vibration signal, a residual block is added to realize the cross-layer connection of the model, so that the deeper layers in the model can directly learn the shallow feature vectors. The residual block can protect the integrity of the feature vector by transmitting the input bearing fault feature vector to the output in the form of branches, so that the entire model only needs to learn the parts where the input and output of each module are different, which greatly simplifies the goal and difficulty of iterative learning. Since the bearing fault signal data is presented in the form of a sequence, the time dimension of the fault sequence needs to be considered when constructing the residual block. Based on the above considerations, the present invention uses a causal hole convolution layer instead of a traditional two-dimensional convolution layer when constructing a residual block. Similar to the classic two-dimensional convolutional network, causal convolution still has the problem that the modeling length of time is limited by the size of the convolution kernel. In order to solve this problem, the present invention uses a causal hole convolution instead of a causal convolution to construct a residual block, and uses the residual block structure to obtain the deep features of the bearing, thereby obtaining a longer dependency when filtering the data, and effectively improving the classification results under noise conditions.

[0046] Optionally, the equivalent convolution kernel size of the causal atrous convolution layer is:

[0047] k′=k+(k+1)×(d-1)

[0048] Among them, k′ is the equivalent convolution size, k is the convolution kernel size of causal dilated convolution, and d is the dilation coefficient.

[0049] Specifically, the calculation of the receptive field of the current causal hole convolution layer is shown in equation (3). i+1 Represents the receptive field of the current layer, RF i represents the receptive field of the previous layer, and k′ represents the size of the convolution kernel. i represents the product of the step lengths in all previous layers except this layer. For S i The definition of is shown in equation (4):

[0050] RF i+1 =RF i +(k′-1)×S i (3)

[0051]

[0052] For example, Figure 1C is a schematic diagram of a causal atrous convolutional layer provided by an embodiment of the present invention. According to the input x1, ..., x1 at the previous time t, t-1 To predict y t , so that y t Closer to the true value in the label, the specific process is shown in equation (5):

[0053]

[0054] S130, extracting basic bearing features from the bearing image data by a basic feature extraction unit;

[0055] Among them, the basic feature extraction unit in the present invention can be a 1×1 convolution layer (kernel). Compared with various two-dimensional convolutional networks, the 1×1 convolution layer can retain the shallow (basic) features of the data as much as possible.

[0056] Specifically, the shallow features of the bearing fault data are obtained through the basic feature extraction unit, avoiding the loss of the shallow features of the bearing fault data due to the deepening of the convolutional network layers in the process of obtaining the deep features of the bearing, thereby improving the classification effect of the model on the fault type.

[0057] S140, fusing the bearing depth feature and the bearing basic feature, and determining the bearing fault type according to the fusion result

[0058] Specifically, the obtained bearing depth features and bearing basic features are summed to obtain the output feature vector of the residual block - the fusion result, and the fusion result is processed in sequence by the flattening layer, the fully connected layer, and the Softmax layer, and finally the classification result of the bearing fault is obtained by using the neural network.

[0059] The embodiment of the present invention obtains the bearing deep features through the residual block constructed based on the causal hole convolution layer, and integrates it with the basic bearing features to determine the bearing fault type, which simplifies the goal and difficulty of iterative learning and effectively improves the classification results under noisy conditions.

[0060] Figure 2A This is a flow chart of a bearing fault classification method provided by another embodiment of the present invention. This embodiment is optimized and improved on the basis of the above embodiment. Figure 2A As shown, the method includes:

[0061] S210. Determine bearing image data according to a vibration signal in a bearing fault state.

[0062] S220, extracting the bearing causal hole features of the bearing image data through the causal hole convolution layer; batch normalizing and activating the bearing causal hole features to obtain bearing depth features.

[0063] Specifically, in addition to using causal hole convolution instead of causal convolution to construct the residual block, batch normalization (BN) and ReLU activation function are also introduced in the residual block. After extracting the bearing causal hole feature of the bearing image data through the causal hole convolution layer, batch normalization operation is performed and the feature is activated using the ReLU activation function, and then the processed feature vector is input into the Dropout layer to finally obtain the bearing depth feature.

[0064] For example, Figure 2B : is a schematic diagram of a residual block structure provided by another embodiment of the present invention. The residual block uses the causal hole convolution layer and other residual block structures to extract features twice to obtain the bearing depth feature, and sums the bearing depth feature with the bearing basic feature obtained by the basic basic feature extraction unit (1×1 convolution layer) to obtain an output feature vector, which is the fusion result.

[0065] S230, obtaining initial features of the bearing image data through a local binary convolution layer; batch normalizing and activating the initial features to obtain initial activated features of the bearing image, and using the initial activated features as input of the residual block.

[0066] Specifically, the bearing image data is input into the local binary convolution layer (LBC) layer for feature extraction to obtain the initial features. The feature vector is then normalized and activated by the ReLU activation function. The processed feature vector is further input into the two-dimensional convolution layer and the BN+ReLU layer for further deep feature extraction, and the deep feature extraction based on the two-dimensional convolution layer and the BN+ReLU layer is repeated at least once or multiple times. In order to avoid the loss of shallow features of bearing fault data due to the deepening of the network layers, a 1x1 convolution layer is added in the present invention to extract shallow features of the data, and the obtained (initial activation features) deep features are used together with the shallow features extracted by the 1x1 convolution layer as inputs of the residual block.

[0067] For example, Figure 2CIt is a schematic diagram of a neural network model provided according to another embodiment of the present invention. Among them, the preprocessed fault data (bearing image data) is input into the LBC layer, and the bearing image data is converted into a feature vector by LBC, and then normalized and activated by the activation function. Two branches are processed at the same time. One is to repeatedly use the two-dimensional convolution layer and the BN+Relu layer to obtain the deep features of the bearing fault data, and the other is to use the 1×1 convolution layer to retain the shallow features of the bearing fault data. The deep features are further processed by the residual block and then fused with the shallow features. The fusion process of the residual block can be repeated once or multiple times. Finally, the fusion result is processed in turn by inputting the flattening layer, the fully connected layer, and the Softmax layer to obtain the final output classification result.

[0068] Optionally, the obtaining of initial features of the bearing image data through a local binary convolution layer includes:

[0069] A target difference map after filtering of the bearing image data is obtained through at least two convolution kernels in the local binary convolution layer; the target difference map is activated by a nonlinear activation function, and the initial features of the bearing image data are obtained based on the corresponding learning weights of the target difference map.

[0070] Specifically, LBP is used for the center pixel (x n ,y n The feature extraction process of ) is shown in equation (6), where N represents the size of the original matrix, i m Represents the value of the mth adjacent element, i n It represents the value of the center element of the matrix. m >i n , then s(i m ,i n )=1, otherwise s(i m ,i n )=0. N represents the size of the original matrix, i m represents the value of the mth adjacent element, and in represents the value of the center element of the matrix. m >i n , then s(i m ,i n )=1, otherwise s(i m ,i n )=0.

[0071]

[0072] For the LBC layer, the present invention designs an LBC convolution kernel based on the local binary pattern. Similar to the traditional convolution kernel, the LBC convolution kernel also uses the convolution kernel to scan the input matrix. The convolution process of the LBC layer is shown in equation (7), where s represents the input channel and t represents the output channel. The LBC layer in the model consists of m predefined convolution filters, where b i , i∈[m]. The convolution kernel in the LBC layer is applied to the input image x l Filtering is performed to generate m difference maps, and then the generated difference maps are activated by a nonlinear activation function to generate m activated feature maps. The present invention uses the ReLU activation function to replace the Heaviside step function used in the traditional local binary mode. Then the m activated feature maps are combined with m learnable weights v l，i Combined to generate a channel of LBC layer response. The feature map output by the LBC layer in the model is used as the input x1+1 of the next layer of the model.

[0073]

[0074] For example, Figure 2D is a schematic diagram of a local binary pattern calculation process provided according to another embodiment of the present invention. Assuming that the matrix size obtained by the filter is 3×3, each neighborhood pixel value of the central pixel is quantized with the gray value of the central pixel as the threshold, and the pixel value greater than or equal to the central pixel is encoded as 1, otherwise it is 0, thereby forming a local binary pattern. Subsequently, these binary numbers are read in sequence and multiplied by a predefined decimal number composed of base 2, such as with a predefined weight vector [2 0 , 2 1 , 2 2 , 2 3 , 2 4 , 2 5 , 2 6 , 2 7 ] are multiplied. The final result can be obtained by summing the products of the corresponding elements of these matrices. Each pixel in the image can be calculated by local binary value to obtain a local binary pattern result.

[0075] S240. Extracting basic bearing features from the bearing image data through a basic feature extraction unit.

[0076] S250: Fusing the bearing depth feature and the bearing basic feature, and determining the bearing fault type according to the fusion result.

[0077] The embodiment of the present invention improves the performance of the model for fault feature extraction by introducing a local binary convolution layer, and on this basis, a new type of residual block is designed to construct a time domain network, thereby classifying the extracted feature vectors. The embodiment of the present invention combines the designed LBC layer with the constructed time domain network to realize bearing fault diagnosis, which has important theoretical significance and practical value for improving the accuracy of rolling bearing fault diagnosis, and improving the accuracy in the absence of fault data samples and in a noisy environment, and ensuring the safety and stable operation of rolling bearings.

[0078] Figure 3 A structural schematic diagram of a bearing fault classification device provided by another embodiment of the present invention. Figure 3 As shown, the device comprises:

[0079] An image data determination module 310 is used to determine bearing image data according to a vibration signal in a bearing fault state;

[0080] A depth feature determination module 320, configured to process the bearing image data through a residual block to obtain a bearing depth feature; the residual block is constructed through a causal hole convolution layer;

[0081] A basic feature determination module 330 is used to extract the basic features of the bearing from the bearing image data through a basic feature extraction unit;

[0082] The fault type determination module 340 is used to fuse the bearing depth feature and the bearing basic feature, and determine the bearing fault type according to the fusion result.

[0083] The bearing fault classification device provided in the embodiment of the present invention can execute the bearing fault classification method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0084] Optionally, the depth feature determination module is specifically used to:

[0085] The bearing causal hole features of the bearing image data are extracted through the causal hole convolution layer; the bearing causal hole features are batch normalized and activated to obtain bearing depth features.

[0086] Optionally, the equivalent convolution kernel size of the causal atrous convolution layer is:

[0087] k′=k+(k+1)×(d-1)

[0088] Among them, k′ is the equivalent convolution size, k is the convolution kernel size of causal dilated convolution, and d is the dilation coefficient.

[0089] Optionally, the image data determination module 310 is specifically configured to:

[0090] A preset number of continuous sampling points are collected from the vibration signal under the bearing fault state; the continuous sampling points are overlapped and sampled with a sampling step size less than the preset number to obtain multiple sampling data; the sampling data are converted into a time-frequency image; and the time-frequency image is converted into an N×N grayscale image as bearing image data.

[0091] Optionally, the device further comprises:

[0092] An initial feature acquisition module, used to acquire the initial features of the bearing image data through a local binary convolution layer;

[0093] The activation feature determination module is used to batch normalize and activate the initial features to obtain the initial activation features of the bearing image, and to use the initial activation features as input of the residual block.

[0094] Optionally, the initial feature acquisition module is specifically used to:

[0095] A target difference map after filtering of the bearing image data is obtained through at least two convolution kernels in the local binary convolution layer; the target difference map is activated by a nonlinear activation function, and the initial features of the bearing image data are obtained based on the corresponding learning weights of the target difference map.

[0096] The bearing fault classification device further described can also execute the bearing fault classification method provided by any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0097] Figure 4 A schematic diagram of the structure of an electronic device 40 that can be used to implement an embodiment of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0098] like Figure 4As shown, the electronic device 40 includes at least one processor 41, and a memory connected to the at least one processor 41, such as a read-only memory (ROM) 42, a random access memory (RAM) 43, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 41 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 42 or the computer program loaded from the storage unit 48 to the random access memory (RAM) 43. In the RAM 43, various programs and data required for the operation of the electronic device 40 can also be stored. The processor 41, the ROM 42, and the RAM 43 are connected to each other through a bus 44. An input / output (I / O) interface 45 is also connected to the bus 44.

[0099] A number of components in the electronic device 40 are connected to the I / O interface 45, including: an input unit 46, such as a keyboard, a mouse, etc.; an output unit 47, such as various types of displays, speakers, etc.; a storage unit 48, such as a disk, an optical disk, etc.; and a communication unit 49, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 49 allows the electronic device 40 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0100] The processor 41 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 41 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 41 executes the various methods and processes described above, such as a classification method for bearing faults.

[0101] In some embodiments, the classification method of bearing faults can be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 48. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 40 via the ROM 42 and / or the communication unit 49. When the computer program is loaded into the RAM 43 and executed by the processor 41, one or more steps of the classification method of bearing faults described above can be performed. Alternatively, in other embodiments, the processor 41 can be configured to execute the classification method of bearing faults in any other appropriate manner (for example, by means of firmware).

[0102] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0103] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0104] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in combination with an instruction execution system, device or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0105] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0106] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0107] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0108] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0109] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for classifying bearing faults, comprising: Determine bearing image data according to the vibration signal under the bearing fault state; The bearing image data is processed by a residual block to obtain a bearing depth feature; the residual block is constructed by a causal hole convolution layer; Extracting basic bearing features from the bearing image data by a basic feature extraction unit; fusing the bearing depth feature and the bearing basic feature, and determining the bearing fault type according to the fusion result; Before the bearing image data is processed by the residual block to obtain the bearing depth feature, the method further includes: Acquire the initial features of the bearing image data through a local binary convolution layer; Batch normalize and activate the initial features to obtain initial activated features of the bearing image, and use the initial activated features as inputs of the residual block. When extracting deep features, the input of the residual block also includes basic features extracted by a 1x 1 convolutional layer. The step of processing the bearing image data by using a residual block to obtain a bearing depth feature includes: Extracting bearing causal hole features of the bearing image data through the causal hole convolution layer; The bearing causal hole features are batch normalized and activated to obtain bearing depth features.

2. The method according to claim 1, characterized in that The equivalent convolution kernel size of the causal hole convolution layer is: k ′ =k+(k+1)×(d-1) Among them, k ′ is the equivalent convolution size, k is the convolution kernel size of the causal dilated convolution, and d is the dilation coefficient.

3. The method according to claim 1, characterized in that The step of obtaining bearing image data according to the vibration signal in the bearing fault state includes: Collecting a preset number of continuous sampling points from the vibration signal of the bearing under fault condition; Overlappingly sampling the continuous sampling points with a sampling step length less than a preset number to obtain a plurality of sampling data; By converting the sampled data into a time-frequency image; The time-frequency image is converted into an N×N grayscale image as bearing image data.

4. The method according to claim 1, characterized in that: The initial features of the bearing image data obtained by using a local binary convolution layer include: Obtaining a target difference image after filtering the bearing image data through at least two convolution kernels in the local binary convolution layer; The target difference map is activated by using a nonlinear activation function, and the initial features of the bearing image data are obtained based on the corresponding learning weights of the target difference map.

5. A bearing fault classification device, comprising: An image data determination module, used to determine the bearing image data according to the vibration signal in the bearing fault state; A depth feature determination module, used for processing the bearing image data through a residual block to obtain a bearing depth feature; the residual block is constructed through a causal hole convolution layer; A basic feature determination module, used for extracting basic bearing features from the bearing image data through a basic feature extraction unit; A fault type determination module is used to fuse the bearing depth feature and the bearing basic feature, and determine the bearing fault type according to the fusion result; An initial feature acquisition module, used to acquire the initial features of the bearing image data through a local binary convolution layer; An activation feature determination module, used for batch normalization and activation of the initial features to obtain the initial activation features of the bearing image, and used for using the initial activation features as inputs of the residual block. When extracting deep features, the input of the residual block also includes basic features extracted by a 1x 1 convolutional layer; Wherein, the depth feature determination module is specifically used for: Extracting bearing causal hole features of the bearing image data through the causal hole convolution layer; The bearing causal hole features are batch normalized and activated to obtain bearing depth features.

6. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can perform the bearing fault classification method according to any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the bearing fault classification method according to any one of claims 1 to 4 when executed.

Images