Railway bearing detection method and device, terminal and storage medium

Abstract

This invention provides a method, apparatus, terminal, and storage medium for detecting railway bearings. The method includes: acquiring an infrared image of the bearing to be detected from an infrared camera; adjusting the infrared image to a preset size to obtain a target infrared image; obtaining a diagnostic result for the bearing to be detected based on a preset classification list, the target infrared image, and a pre-trained bearing diagnostic model; the preset classification list stores the correspondence between numerical and textual diagnostic results; the diagnostic result for the bearing to be detected includes the fault type, fault severity, and rotational speed of the bearing. This invention can quickly diagnose railway bearings with high efficiency and accuracy, making bearing detection more intelligent and precise.

Description

Technical Field

[0001] This invention relates to the field of bearing testing technology, and in particular to a method, apparatus, terminal and storage medium for testing railway bearings. Background Technology

[0002] As a key component of the running gear, the health of rolling bearings directly affects the operational safety of trains, making health checks essential.

[0003] Existing intelligent fault diagnosis methods for rotor-bearing systems mainly focus on vibration analysis under stable operation. However, in practical applications, due to the long transmission path of vibration signals, variable working conditions, and strong noise, the processing of vibration signals is very complex. In addition, their adaptability to new scenarios is low, resulting in problems of poor accuracy and low efficiency in vibration analysis. Summary of the Invention

[0004] This invention provides a railway bearing testing method, device, terminal, and storage medium to address the problems of poor accuracy and low efficiency in current vibration analysis.

[0005] In a first aspect, embodiments of the present invention provide a method for detecting railway bearings, comprising:

[0006] Acquire infrared images of the bearing to be inspected from an infrared camera;

[0007] Adjust the infrared image to a preset size to obtain the target infrared image;

[0008] Based on a preset classification list, target infrared image, and pre-trained bearing diagnostic model, the diagnostic results of the bearing to be tested are obtained; the preset classification list stores the correspondence between numerical and textual diagnostic results; the diagnostic results of the bearing to be tested include the fault type, fault degree, and rotational speed of the bearing to be tested.

[0009] In one possible implementation, the bearing diagnostic model is a convolutional neural network model; the convolutional neural network model includes four convolutional layers and three fully connected layers connected in sequence.

[0010] In one possible implementation, in the convolutional neural network model, the image input of the first convolutional layer has a size of 100×100×3. The first convolutional layer is used to perform the first convolution processing, the first ReLU operation, and the first downsampling operation on the target infrared image to obtain a first image with a size of 55×55×32. The first convolution processing uses 32 filters with a size of 5×5×3 to multiply the corresponding regions of the 32 initial feature maps by the corresponding weights and add the corresponding biases to the regions obtained to obtain 32 convolutional feature maps. The width and height of the 32 convolutional feature maps are filled with pixels to obtain 32 new feature maps.

[0011] The second convolutional layer is used to perform a second convolution, a second ReLU operation, and a second downsampling operation on the first image to obtain a second image with a size of 27×27×64; the second convolution process uses 64 filters with a size of 5×5 to convolve the first image.

[0012] The third convolutional layer is used to perform the third convolution, the third ReLU operation, and the third downsampling operation on the second image to obtain a third image with a size of 13×13×128; the third convolution process uses 128 filters with a size of 3×3 to convolve the second image.

[0013] The fourth convolutional layer is used to perform the fourth convolution, the fourth ReLU operation, and the fourth downsampling operation on the third image to obtain a fourth image with a size of 6×6×128; the fourth convolution process uses 128 filters with a size of 3×3 to convolve the third image.

[0014] In one possible implementation, in the convolutional neural network model, the first fully connected layer uses 1024 neurons to perform the first fully connected processing and the first overfit suppression processing on the fourth image.

[0015] The second fully connected layer uses 512 neurons for the second fully connected processing and the second inhibition of overfitting.

[0016] The third fully connected layer uses 80 neurons for third fully connected processing and passes them through a Gaussian filter to obtain 80 classification values.

[0017] In one possible implementation, before obtaining the diagnostic result of the bearing to be detected based on a preset classification list, target infrared image, and pre-trained bearing diagnostic model, the railway bearing detection method further includes:

[0018] Acquire multiple infrared image samples of the bearing during normal and abnormal operation;

[0019] Multiple infrared image samples are segmented to obtain segmented infrared image samples; image segmentation is used to segment the specific location of bearing faults.

[0020] The segmented infrared image samples are classified according to fault type, fault severity and rotation speed to obtain multiple sets of infrared image samples;

[0021] Multiple sets of infrared image samples are divided into training sample set and validation sample set according to a preset ratio;

[0022] The pre-built bearing diagnostic model is trained based on the training sample set, and the trained bearing diagnostic model is validated based on the validation sample set to obtain the pre-trained bearing diagnostic model.

[0023] In one possible implementation, the railway bearing inspection method also includes:

[0024] Import the pre-trained bearing diagnostic model into the Raspberry Pi and complete the system integration of hardware and software.

[0025] In one possible implementation, a pre-trained bearing diagnostic model is imported into the Raspberry Pi, and system integration of the hardware and software is completed, including:

[0026] Burn the Raspberry Pi and configure the wireless network and remote connection files on the Raspberry Pi;

[0027] Remote connection to Raspberry Pi via wireless network and remote file access;

[0028] Import the pre-trained bearing diagnostic model into the Raspberry Pi, run the pre-trained bearing diagnostic model, and save the valid data.

[0029] Secondly, embodiments of the present invention provide a railway bearing testing device, comprising:

[0030] The acquisition module is used to acquire infrared images of the bearing to be inspected captured by an infrared camera;

[0031] The adjustment module is used to adjust the infrared image to a preset size to obtain the target infrared image;

[0032] The diagnostic module is used to obtain the diagnostic results of the bearing to be tested based on a preset classification list, target infrared image and pre-trained bearing diagnostic model; the preset classification list stores the correspondence between numerical and textual diagnostic results; the diagnostic results of the bearing to be tested include the fault type, fault degree and speed of the bearing to be tested.

[0033] Thirdly, embodiments of the present invention provide a terminal, including a processor and a memory, wherein the memory is used to store a computer program, and the processor is used to call and run the computer program stored in the memory to execute the railway bearing detection method as described in the first aspect or any possible implementation thereof.

[0034] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the railway bearing detection method as described in the first aspect or any possible implementation thereof.

[0035] This invention provides a railway bearing detection method, device, terminal, and storage medium. It acquires infrared images of the bearing to be detected using an infrared camera, and based on these images, diagnoses the bearing using a preset classification list and a pre-trained bearing diagnostic model, obtaining the diagnostic results. This method enables rapid and efficient diagnosis of railway bearings, making bearing detection more intelligent and precise. Attached Figure Description

[0036] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0037] Figure 1 This is a schematic flowchart of the railway bearing testing method provided in an embodiment of the present invention;

[0038] Figure 2 This is a schematic diagram showing the diagnostic structure of the bearing to be tested provided in an embodiment of the present invention;

[0039] Figure 3 This is a schematic diagram of the structure of the railway bearing testing device provided in an embodiment of the present invention;

[0040] Figure 4 This is a schematic diagram of the terminal provided in an embodiment of the present invention. Detailed Implementation

[0041] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of the invention. However, those skilled in the art will understand that the invention can be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods are omitted so as not to obscure the description of the invention with unnecessary detail.

[0042] To make the objectives, technical solutions, and advantages of the present invention clearer, specific embodiments will be described below in conjunction with the accompanying drawings.

[0043] See Figure 1 The diagram illustrates a flowchart of the railway bearing testing method provided in this embodiment of the invention. The executing entity of the railway bearing testing method can be a terminal.

[0044] See Figure 1 The above-mentioned railway bearing testing methods include:

[0045] In step S101, an infrared image of the bearing to be tested is acquired from an infrared camera.

[0046] The infrared camera can be installed near the bearing to be inspected, in a location that captures infrared images of the bearing. The specific installation location can be determined based on the actual terrain, making it highly adaptable to various scenarios. The infrared camera can be set to a fixed angle to acquire infrared images of the bearing to be inspected.

[0047] In S102, the infrared image is adjusted to a preset size to obtain the target infrared image.

[0048] The size of the infrared image captured by the infrared camera does not match the size of the input image of the pre-trained bearing diagnostic model. Therefore, the size of the infrared image captured by the infrared camera needs to be adjusted to obtain the target infrared image.

[0049] The preset size refers to the size of the input image of the pre-trained bearing diagnostic model. For example, the size of the infrared image captured by the infrared camera can be 320×256×3, and the size of the input image of the pre-trained bearing diagnostic model can be 100×100×3 (RGB color image).

[0050] In S103, the diagnostic results of the bearing to be tested are obtained based on the preset classification list, the target infrared image and the pre-trained bearing diagnostic model; the preset classification list stores the correspondence between numerical diagnostic results and textual diagnostic results; the diagnostic results of the bearing to be tested include the fault type, fault degree and speed of the bearing to be tested.

[0051] By inputting the target infrared image into a pre-trained bearing diagnostic model, a numerical diagnostic result for the bearing under test can be obtained. Since the bearing diagnostic model outputs numerical values, and the numerical value alone cannot quickly determine its specific meaning, a textual diagnostic result corresponding to the numerical diagnostic result of the bearing under test can be obtained based on a preset classification list, serving as the final diagnostic result for the bearing under test. This diagnostic result includes the bearing's fault type, fault severity, and rotational speed. The fault type can include inner ring fault (FN), outer ring fault (FW), and rolling element fault (FQ) of rolling bearings; the fault severity ranges from 0.1 to 1, and the rotational speed is between 10 and 50 rpm.

[0052] The target infrared image can be placed on a specified path. When performing fault diagnosis, the target infrared image is obtained from the specified path to perform the fault diagnosis.

[0053] In some possible implementations, the diagnostic results of the bearing under test may also include information such as the probability of the diagnosis being correct, for staff reference. The diagnostic results of the bearing under test can also be saved as historical data.

[0054] This embodiment uses an infrared camera to acquire infrared images of the bearing to be tested. Based on these infrared images, the bearing to be tested is diagnosed using a preset classification list and a pre-trained bearing diagnostic model. This results in a diagnosis of the bearing to be tested, enabling rapid diagnosis of railway bearings with high efficiency and accuracy, making bearing testing more intelligent and precise.

[0055] In some embodiments, the bearing diagnostic model is a convolutional neural network model; the convolutional neural network model includes four convolutional layers and three fully connected layers connected in sequence.

[0056] This embodiment uses a convolutional neural network to establish a bearing diagnostic model, which can quickly determine the operating status of railway bearings and locate faults, thus preventing accidents and having great engineering application value. Furthermore, the model based on the convolutional neural network can eliminate spatiotemporal obstacles and detect vehicles, which can greatly reduce misjudgments caused by human subjective factors.

[0057] In some embodiments, in the convolutional neural network model, the image input size of the first convolutional layer is 100×100×3. The first convolutional layer is used to perform a first convolutional processing, a first ReLU operation, and a first downsampling operation on the target infrared image to obtain a first image with a size of 55×55×32. The first convolutional processing uses 32 filters with a size of 5×5×3 to multiply the corresponding regions of the 32 initial feature maps by the corresponding weights and add the corresponding biases to the regions obtained to obtain 32 convolutional feature maps. The width and height of the 32 convolutional feature maps are filled with pixels to obtain 32 new feature maps.

[0058] The second convolutional layer is used to perform a second convolution, a second ReLU operation, and a second downsampling operation on the first image to obtain a second image with a size of 27×27×64; the second convolution process uses 64 filters with a size of 5×5 to convolve the first image.

[0059] The third convolutional layer is used to perform the third convolution, the third ReLU operation, and the third downsampling operation on the second image to obtain a third image with a size of 13×13×128; the third convolution process uses 128 filters with a size of 3×3 to convolve the second image.

[0060] The fourth convolutional layer is used to perform the fourth convolution, the fourth ReLU operation, and the fourth downsampling operation on the third image to obtain a fourth image with a size of 6×6×128; the fourth convolution process uses 128 filters with a size of 3×3 to convolve the third image.

[0061] In some embodiments, in the convolutional neural network model, the first fully connected layer uses 1024 neurons to perform first fully connected processing and first overfit suppression processing on the fourth image.

[0062] The second fully connected layer uses 512 neurons for the second fully connected processing and the second inhibition of overfitting.

[0063] The third fully connected layer uses 80 neurons for third fully connected processing and passes them through a Gaussian filter to obtain 80 classification values.

[0064] In this embodiment, the convolutional neural network model includes four convolutional layers and three fully connected layers connected in sequence. The four convolutional layers are a first convolutional layer, a second convolutional layer, a third convolutional layer, and a fourth convolutional layer connected in sequence. The three fully connected layers are a first fully connected layer, a second fully connected layer, and a third fully connected layer connected in sequence.

[0065] In the first convolutional layer:

[0066] The first convolutional layer (conv1) uses 32 filters of size 5×5×3 (i.e., convolutional kernels with a stride of 1). The difference in this network is that this layer's filters multiply the corresponding regions in the 32 feature maps by their respective weights, convolve the resulting regions after adding a bias, and then pad both the width and height with 2 pixels to obtain 32 new feature maps. The new feature map size is 100×100×32 [(100-5+2×2) / 1+1=100]. The same method is used for convolution in subsequent convolutional layers.

[0067] In the first ReLU operation, ReLU is used as the activation function to ensure that the value range of the feature map is reasonable. The data format after ReLU1 is 100×100×32.

[0068] In the first downsampling operation pool1, the kernel size is 2×2, the step size is 2, and the data after downsampling in pool1 (pooling layer) is 55×55×32[(100-2) / 2+1=55].

[0069] In the second convolutional layer:

[0070] The input data size is 55×55×32.

[0071] The second convolutional processing, conv2, uses 64 filters of size 5×5 (with a stride of 1) to further extract features from the 55×55×32 feature maps. The size after convolution is 55×55×64[(55-5+2×2) / 1+1=55], which is 64 feature maps of size 55×55.

[0072] The data after the second ReLU operation, ReLU2, is 55×55×64.

[0073] The kernel of the second downsampling operation pool2 is 2×2, the step size is 2, and the data after pool2 (pooling layer) downsampling is 27×27×64[(55-2) / 2+1=27].

[0074] In the third convolutional layer:

[0075] The input data size is 27×27×64.

[0076] The third convolution process, conv3, uses 128 filters of size 3×3 (with a stride of 1). The data after convolution is: 27×27×128[(27+1×2-3) / 1+1=27].

[0077] A third ReLU operation, ReLU3, is then performed to ensure that the feature map values ​​are within a reasonable range. The data format after ReLU3 is 27×27×128.

[0078] The kernel of the third downsampling operation pool3 is 2×2, the step size is 2, and the downsampled data is 13×13×128[(27-2) / 2+1=13].

[0079] In the fourth convolutional layer:

[0080] The input data is a feature map of 13×13×128.

[0081] The fourth convolution process, conv4, uses 128 filters of size 3×3 (with a stride of 1). The data after convolution is: 13×13×128[(13+1×2-3) / 1+1=13].

[0082] The data format after the fourth ReLU operation, ReLU4, is 27×27×128.

[0083] The kernel of the fourth downsampling operation pool4 is 2×2, the step size is 2, and the downsampled data is 6×6×128[(13-3) / 2+1=6].

[0084] In the first fully connected layer: 1024 neurons are used to process 128 6×6 feature maps in the first fully connected process, and then convolution is performed to transform them into a feature point. For a point in the 1024 neurons, it is obtained by multiplying the feature point obtained by convolution of some feature maps in the 128 feature maps by the corresponding weight, and then adding a bias. Then, the first overfitting suppression is performed (the dropout rate of 0.5 has the best effect and generates the most random network structures). Some node information is randomly dropped from the 1024 nodes (values ​​are cleared to 0) to obtain a new 1024 neurons.

[0085] In the second fully connected layer: similar to the first fully connected layer, the 1024 data points output by the first fully connected layer are fully connected with the 512 neurons in this layer, and then convolution is performed, multiplied by weights, bias is added, and the second suppression of overfitting is applied to obtain 512 neurons.

[0086] In the third fully connected layer: 80 neurons are used, and then the 512 neurons in the second fully connected layer are processed by the third fully connected layer. Then, they are passed through a Gaussian filter to obtain 80 float values, which are all the possible predicted faults.

[0087] In some embodiments, before obtaining the diagnostic result of the bearing to be detected based on a preset classification list, a target infrared image, and a pre-trained bearing diagnostic model, the railway bearing detection method further includes:

[0088] Acquire multiple infrared image samples of the bearing during normal and abnormal operation;

[0089] Multiple infrared image samples are segmented to obtain segmented infrared image samples; image segmentation is used to segment the specific location of bearing faults.

[0090] The segmented infrared image samples are classified according to fault type, fault severity and rotation speed to obtain multiple sets of infrared image samples;

[0091] Multiple sets of infrared image samples are divided into training sample set and validation sample set according to a preset ratio;

[0092] The pre-built bearing diagnostic model is trained based on the training sample set, and the trained bearing diagnostic model is validated based on the validation sample set to obtain the pre-trained bearing diagnostic model.

[0093] This embodiment can use an infrared camera to collect infrared images of railway bearings at different speeds during normal operation and abnormal operation (various fault types and fault degrees) as infrared image samples, and save them in the dataset folder.

[0094] The acquired infrared image samples are segmented to extract bearing fault location images suitable for training. The original infrared image samples can be segmented into images of a preset size. Then, based on fault type, fault severity, and rotational speed, the infrared image samples are divided into 80 groups. Each group of infrared images can be divided into a training sample set and a validation sample set according to a preset ratio, for example, a 4:1 ratio. When training the pre-built bearing diagnostic model, the batch size can be set to 64, and the model can be iterated 20 times to obtain the pre-trained bearing diagnostic model.

[0095] During training, a convolutional neural network model architecture can be used, and existing data can be utilized to analyze the impact of hyperparameters such as kernel size and learning rate on feature extraction.

[0096] Specifically, this embodiment can use Python or other languages ​​to write application software, within which a bearing diagnosis model based on a convolutional neural network is built. Infrared image samples are categorized and placed into corresponding folders. The model's training program is then run to obtain the trained model parameters, which are saved to the model coefficient file within the model. When bearing fault diagnosis is required, the target infrared image is placed in the specified path, the trained bearing diagnosis model is run, and the processed infrared image and diagnosis results are output. (See [link to relevant documentation]). Figure 2 .

[0097] In some possible implementations, the model training process can be as follows:

[0098] A. Determine the dataset path;

[0099] Determine the path to the dataset to ensure that the dataset can be accurately located during model training.

[0100] B. Randomly divide the database (dataset) into a training sample set and a validation sample set in a 4:1 ratio;

[0101] The database is mainly used for training and validating models. Since the structural parameters of these two types of data are the same, the database can be randomly partitioned.

[0102] C. Model training requires certain rules, so we set the initial training parameters of the model to batch size of 64 and iteration count of 20.

[0103] D. Add data labels;

[0104] The training sample set in the database was divided into 80 label groups according to fault type, fault severity, and rotational speed, namely "FN / W / Q, 0.X, 10-50".

[0105] E. Compile the initial model;

[0106] F. Iteratively train the parameters.

[0107] The procedure for detecting rolling bearing faults is as follows:

[0108] a. Open the folder containing the infrared images of the rolling bearings that need to be inspected;

[0109] b. Scale the infrared image of the rolling bearing to be inspected to a preset size;

[0110] Because the input to the model network is 100*100*3, the target infrared image needs to be scaled down to 100*100 for the model to work.

[0111] c. Load model parameters;

[0112] First, import the initial model you've built, then load the model parameters obtained through training. Only then can the model run normally.

[0113] d. Load the preset category list;

[0114] A pre-defined category list needs to be loaded; otherwise, the model will only save the numerical values ​​and will not be able to output the meaning of the values.

[0115] e. Detect and diagnose, and output results;

[0116] The final output includes the current input target infrared image, the diagnostic structure (fault type, fault severity, rotation speed), and the probability of correctness, which can be used as a reference for testing personnel.

[0117] f. Save the results.

[0118] This step is optional; when saving the results, in addition to the output results, the detection time, etc., will also be included.

[0119] In some embodiments, the railway bearing testing method further includes:

[0120] Import the pre-trained bearing diagnostic model into the Raspberry Pi and complete the system integration of hardware and software.

[0121] The Raspberry Pi is a miniature computer. By importing a pre-trained bearing diagnostic model into the Raspberry Pi and completing the system integration of hardware and software, fault diagnosis can be performed quickly and easily using the portable Raspberry Pi, improving the efficiency of fault diagnosis. Some related experiments can also be conducted on the Raspberry Pi.

[0122] In some embodiments, a pre-trained bearing diagnostic model is imported into the Raspberry Pi, and system integration of the hardware and software is completed, including:

[0123] Burn the Raspberry Pi and configure the wireless network and remote connection files on the Raspberry Pi;

[0124] Remote connection to Raspberry Pi via wireless network and remote file access;

[0125] Import the pre-trained bearing diagnostic model into the Raspberry Pi, run the pre-trained bearing diagnostic model, and save the valid data.

[0126] In this embodiment, the device's IP address can be obtained through the DNS domain name resolution protocol; the Raspberry Pi is burned with a wireless network and remote connection files; the Raspberry Pi is connected by entering its IP address in PuTTY, opening it, and entering the username and password to establish a remote connection; the Windows system's built-in "Remote Desktop Connection" is opened, the Raspberry Pi's IP address is entered, and the connection is clicked; the Raspberry Pi is opened, the account and password are entered, and the Raspberry Pi's main page is accessed; the model is run, and valid data is saved; the Raspberry Pi is shut down, and the experiment ends.

[0127] This embodiment combines infrared imaging technology with deep learning to make vehicle fault detection more intelligent and accurate. With the continuous development of information technology, models based on convolutional neural networks can eliminate spatiotemporal obstacles, monitor vehicles, and significantly reduce misjudgments caused by subjective human factors. Establishing a deep learning model reduces model training time, allowing for better detection methods with less data, thus improving the efficiency and accuracy of locomotive fault detection. This embodiment also solves the problems of vibration analysis affecting equipment structure and the difficulty of sensor installation. Furthermore, the method provided in this application has strong scalability; once the model has been trained on infrared images of corresponding samples, the number of diagnostic categories can be quickly increased, resulting in a larger number of diagnostic classifications. Moreover, its training time is short, and its diagnostic detection speed is fast. No manual monitoring is required throughout the process; diagnostic results can be monitored and output within seconds, greatly improving the detection efficiency of railway bearings and reducing misjudgments caused by subjective human factors to a certain extent.

[0128] In practical applications, the method provided in this application can be written in Python, which has excellent cross-platform running characteristics. It can realize model building, training and prediction functions by calling TensorFlow modules. The software is divided into source file classification and model parameters. The source files include model training, bearing detection and graphical interface. The classification includes image segmentation. The main parameters of the model are checkpoint, model.ckpt, model.ckpt.index and model.ckpt.meta.

[0129] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0130] The following are device embodiments of the present invention. For details not described in detail, please refer to the corresponding method embodiments described above.

[0131] Figure 3 A schematic diagram of the railway bearing testing device provided in an embodiment of the present invention is shown. For ease of explanation, only the parts related to the embodiment of the present invention are shown, and are described in detail below:

[0132] like Figure 3 As shown, the railway bearing testing device 30 may include: an acquisition module 31, an adjustment module 32, and a diagnostic module 33.

[0133] The acquisition module 31 is used to acquire the infrared image of the bearing to be tested captured by the infrared camera;

[0134] Adjustment module 32 is used to adjust the infrared image to a preset size to obtain the target infrared image;

[0135] The diagnostic module 33 is used to obtain the diagnostic results of the bearing to be tested based on the preset classification list, the target infrared image and the pre-trained bearing diagnostic model. The preset classification list stores the correspondence between numerical diagnostic results and textual diagnostic results. The diagnostic results of the bearing to be tested include the fault type, fault degree and speed of the bearing to be tested.

[0136] In one possible implementation, the bearing diagnostic model is a convolutional neural network model; the convolutional neural network model includes four convolutional layers and three fully connected layers connected in sequence.

[0137] In one possible implementation, in the convolutional neural network model, the image input of the first convolutional layer has a size of 100×100×3. The first convolutional layer is used to perform the first convolution processing, the first ReLU operation, and the first downsampling operation on the target infrared image to obtain a first image with a size of 55×55×32. The first convolution processing uses 32 filters with a size of 5×5×3 to multiply the corresponding regions of the 32 initial feature maps by the corresponding weights and add the corresponding biases to the regions obtained to obtain 32 convolutional feature maps. The width and height of the 32 convolutional feature maps are filled with pixels to obtain 32 new feature maps.

[0138] The second convolutional layer is used to perform a second convolution, a second ReLU operation, and a second downsampling operation on the first image to obtain a second image with a size of 27×27×64; the second convolution process uses 64 filters with a size of 5×5 to convolve the first image.

[0139] The third convolutional layer is used to perform the third convolution, the third ReLU operation, and the third downsampling operation on the second image to obtain a third image with a size of 13×13×128; the third convolution process uses 128 filters with a size of 3×3 to convolve the second image.

[0140] The fourth convolutional layer is used to perform the fourth convolution, the fourth ReLU operation, and the fourth downsampling operation on the third image to obtain a fourth image with a size of 6×6×128; the fourth convolution process uses 128 filters with a size of 3×3 to convolve the third image.

[0141] In one possible implementation, in the convolutional neural network model, the first fully connected layer uses 1024 neurons to perform the first fully connected processing and the first overfit suppression processing on the fourth image.

[0142] The second fully connected layer uses 512 neurons for the second fully connected processing and the second inhibition of overfitting.

[0143] The third fully connected layer uses 80 neurons for third fully connected processing and passes them through a Gaussian filter to obtain 80 classification values.

[0144] In one possible implementation, the railway bearing testing device 30 also includes a training module.

[0145] The training module is used for:

[0146] Acquire multiple infrared image samples of the bearing during normal and abnormal operation;

[0147] Multiple infrared image samples are segmented to obtain segmented infrared image samples; image segmentation is used to segment the specific location of bearing faults.

[0148] The segmented infrared image samples are classified according to fault type, fault severity and rotation speed to obtain multiple sets of infrared image samples;

[0149] Multiple sets of infrared image samples are divided into training sample set and validation sample set according to a preset ratio;

[0150] The pre-built bearing diagnostic model is trained based on the training sample set, and the trained bearing diagnostic model is validated based on the validation sample set to obtain the pre-trained bearing diagnostic model.

[0151] In one possible implementation, the railway bearing testing device 30 also includes a Raspberry Pi integrated module.

[0152] The Raspberry Pi Integration Module is used to import pre-trained bearing diagnostic models into the Raspberry Pi and complete the system integration of hardware and software.

[0153] In one possible implementation, the Raspberry Pi integration module is specifically used for:

[0154] Burn the Raspberry Pi and configure the wireless network and remote connection files on the Raspberry Pi;

[0155] Remote connection to Raspberry Pi via wireless network and remote file access;

[0156] Import the pre-trained bearing diagnostic model into the Raspberry Pi, run the pre-trained bearing diagnostic model, and save the valid data.

[0157] Figure 4 This is a schematic diagram of a terminal provided in an embodiment of the present invention. Figure 4As shown, the terminal 4 in this embodiment includes a processor 40 and a memory 41. The memory 41 stores a computer program 42, and the processor 40 calls and runs the computer program 42 stored in the memory 41 to execute the steps described in the various railway bearing detection method embodiments above, for example... Figure 1 S101 to S103 are shown. Alternatively, the processor 40 is used to call and run the computer program 42 stored in the memory 41 to implement the functions of each module / unit in the above-described device embodiments, for example... Figure 3 The functions of modules / units 31 to 33 shown.

[0158] For example, the computer program 42 can be divided into one or more modules / units, which are stored in the memory 41 and executed by the processor 40 to complete the present invention. The one or more modules / units can be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program 42 in the terminal 4. For example, the computer program 42 can be divided into... Figure 3 Modules / units 31 to 33 are shown.

[0159] The terminal 4 may be a computing device such as a computer or server. The terminal 4 may include, but is not limited to, a processor 40 and a memory 41. Those skilled in the art will understand that... Figure 4 This is merely an example of terminal 4 and does not constitute a limitation on terminal 4. It may include more or fewer components than shown, or combine certain components, or different components. For example, the terminal may also include input / output devices, network access devices, buses, etc.

[0160] The processor 40 may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0161] The memory 41 can be an internal storage unit of the terminal 4, such as a hard disk or memory of the terminal 4. The memory 41 can also be an external storage device of the terminal 4, such as a plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the terminal 4. Furthermore, the memory 41 can include both internal storage units and external storage devices of the terminal 4. The memory 41 is used to store the computer program and other programs and data required by the terminal. The memory 41 can also be used to temporarily store data that has been output or will be output.

[0162] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0163] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0164] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0165] In the embodiments provided by this invention, it should be understood that the disclosed devices / terminals and methods can be implemented in other ways. For example, the device / terminal embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0166] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0167] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0168] If the integrated module / unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various railway bearing detection method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals.

[0169] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for testing railway bearings, characterized in that, include: Acquire infrared images of the bearing to be inspected from an infrared camera; The infrared image is adjusted to a preset size to obtain the target infrared image; Based on a preset classification list, the target infrared image, and a pre-trained bearing diagnostic model, the diagnostic result of the bearing to be detected is obtained. The preset classification list stores the correspondence between numerical and textual diagnostic results. The diagnostic result of the bearing to be detected includes the fault type, fault severity, and rotational speed of the bearing. The bearing diagnostic model is a convolutional neural network model. The convolutional neural network model includes four convolutional layers and three fully connected layers connected in sequence. The image input specification of the first convolutional layer is 100×100×3. The first convolutional layer is used to perform a first convolutional processing, a first ReLU operation, and a first downsampling operation on the target infrared image to obtain a first image with a specification of 55×55×32. The second convolutional layer performs a second convolution, a second ReLU operation, and a second downsampling operation on the first image to obtain a second image with dimensions of 27×27×64. The third convolutional layer performs a third convolution, a third ReLU operation, and a third downsampling operation on the second image to obtain a third image with dimensions of 13×13×128. The fourth convolutional layer performs a fourth convolution, a fourth ReLU operation, and a fourth downsampling operation on the third image to obtain a fourth image with dimensions of 6×6×128. The third fully connected layer uses 80 neurons for the third fully connected processing and passes them through a Gaussian filter to obtain 80 classification values. These 80 classification values ​​represent all predicted possible faults.

2. The railway bearing testing method according to claim 1, characterized in that, The first convolutional processing uses 32 filters with a specification of 5×5×3. The corresponding regions of the 32 initial feature maps are multiplied by the corresponding weights and then convolved with the corresponding biases to obtain 32 convolutional feature maps. The width and height of the 32 convolutional feature maps are filled with pixels to obtain 32 new feature maps. The second convolution process uses 64 filters of size 5×5 to convolve the first image; The third convolution process uses 128 filters of size 3×3 to convolve the second image; The fourth convolution process uses 128 filters of size 3×3 to convolve the third image.

3. The railway bearing testing method according to claim 2, characterized in that, In the convolutional neural network model, the first fully connected layer uses 1024 neurons to perform the first fully connected processing and the first overfit suppression processing on the fourth image; The second fully connected layer uses 512 neurons for the second fully connected processing and the second suppression of overfitting.

4. The railway bearing testing method according to claim 1, characterized in that, Before obtaining the diagnostic result of the bearing to be detected from the preset classification list, the target infrared image, and the pre-trained bearing diagnostic model, the railway bearing detection method further includes: Acquire multiple infrared image samples of the bearing during normal and abnormal operation; The multiple infrared image samples are segmented to obtain segmented infrared image samples; the image segmentation is used to segment the specific fault location of the bearing. The segmented infrared image samples are classified according to fault type, fault severity and rotation speed to obtain multiple sets of infrared image samples; Multiple sets of infrared image samples are divided into training sample set and validation sample set according to a preset ratio; The pre-built bearing diagnostic model is trained based on the training sample set, and the trained bearing diagnostic model is verified based on the verification sample set to obtain the pre-trained bearing diagnostic model.

5. The railway bearing testing method according to any one of claims 1 to 4, characterized in that, The railway bearing testing method also includes: Import the pre-trained bearing diagnostic model into the Raspberry Pi and complete the system integration of hardware and software.

6. The railway bearing testing method according to claim 5, characterized in that, The step of importing the pre-trained bearing diagnostic model into the Raspberry Pi and completing the system integration of hardware and software includes: Burn the Raspberry Pi and configure the wireless network and remote connection files on the Raspberry Pi; Based on the wireless network and the remote connection file, a remote connection is established with the Raspberry Pi; Import the pre-trained bearing diagnostic model into the Raspberry Pi, run the pre-trained bearing diagnostic model, and save the valid data.

7. A railway bearing testing device, characterized in that, include: The acquisition module is used to acquire infrared images of the bearing to be inspected captured by an infrared camera; The adjustment module is used to adjust the infrared image to a preset size to obtain the target infrared image; The diagnostic module is used to obtain the diagnostic result of the bearing to be tested based on a preset classification list, the target infrared image, and a pre-trained bearing diagnostic model. The preset classification list stores the correspondence between numerical and textual diagnostic results. The diagnostic result of the bearing to be tested includes the fault type, fault degree, and rotational speed of the bearing. The bearing diagnostic model is a convolutional neural network model. The convolutional neural network model includes four convolutional layers and three fully connected layers connected in sequence. The image input specification of the first convolutional layer is 100×100×3. The diagnostic module is also used to perform a first convolution processing, a first ReLU operation, and a first downsampling operation on the target infrared image based on the first convolutional layer to obtain a first image with a specification of 55×55×32. The diagnostic module is also used to perform a second convolution, a second ReLU operation, and a second downsampling operation on the first image based on the second convolution layer to obtain a second image with a size of 27×27×64. The diagnostic module is further configured to perform a third convolution, a third ReLU operation, and a third downsampling operation on the second image based on the third convolutional layer to obtain a third image with a size of 13×13×128; the diagnostic module is further configured to perform a fourth convolution, a fourth ReLU operation, and a fourth downsampling operation on the third image based on the fourth convolutional layer to obtain a fourth image with a size of 6×6×128; the third fully connected layer uses 80 neurons for the third fully connected processing and passes them through a Gaussian filter to obtain 80 classification values; the 80 classification values ​​are all predicted possible faults.

8. A terminal, characterized in that, It includes a processor and a memory, the memory being used to store a computer program, and the processor being used to call and run the computer program stored in the memory to perform the railway bearing testing method as described in any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the railway bearing testing method as described in any one of claims 1 to 6.

Images