Intelligent train bogie defect detection and fault identification system and method

Abstract

The invention discloses an intelligent train bogie defect detection and fault identification system and method. The intelligent train bogie defect detection and fault identification system comprises a control module, a data acquisition module, a single-chip microcomputer module and a data sending module. The control module controls other modules to work and transmit data; the data acquisition module, the single chip microcomputer module and the data sending module are respectively connected with the control module; the data acquisition module comprises a high-definition imaging unit, a radar ranging unit and an infrared temperature measurement unit; the single-chip microcomputer module comprises an image processing unit, a defect detection unit and a fault identification unit. The data sending module comprises a coding unit and a wireless communication unit; and the fault judgment unit clusters the processed fault feature map and the fault information trained by the database to obtain the fault information, so that intelligent train bogie defect detection and fault identification are realized, the labor amount of workers is reduced, the bogie maintenance efficiency is improved, and the train operation safety is guaranteed.

Description

Technical Field

[0001] The present invention belongs to the technical field of defect and fault detection of railway vehicle bogies, and particularly relates to a system and method for intelligent defect detection and fault identification of train bogies. Background Art

[0002] In recent years, China has actively promoted scientific and technological innovation, and high-speed rail technology has leaped to the world's leading level. The development of this technology has provided passengers with a convenient and comfortable travel experience. As the core of passenger transportation, railway vehicles play a crucial role in the high-speed rail system. At the same time, as an important part of railway vehicles, the quality monitoring and maintenance of train bogies are crucial for train safety and comfortable passenger travel.

[0003] However, the traditional methods for detecting defects in train bogies are time-consuming and laborious, requiring the cooperation of multiple types of work and multiple departments, with poor real-time performance and easy to cause potential safety hazards. Moreover, when workers conduct "rush inspections" during the skylight period, the short-time and high-intensity manual operations are prone to misdetection and undetected cases.

[0004] Currently, a technology has been proposed for an inspection robot for the bogie under the train body, aiming to carry out maintenance work during the "skylight period" of the vehicle depot. However, this technology essentially cannot overcome the problem of insufficient real-time performance of maintenance caused by the "skylight period". It can be seen that designing a system for real-time and high-efficiency defect detection of bogies can reduce the workload of railway staff, improve the maintenance efficiency of bogies, and enhance the safety of the railway system. Summary of the Invention

[0005] The purpose of the present invention is to provide a system and method for intelligent defect detection and fault identification of train bogies to solve the above problems.

[0006] To achieve the above purpose, the present invention provides the following technical solutions: A system and method for intelligent defect detection and fault identification of train bogies include a control module, a data acquisition module, a single-chip microcomputer module, and a data transmission module; the control module controls the work of other modules and data transmission; the data acquisition module, the single-chip microcomputer module, and the data transmission module are respectively connected to the control module; the data acquisition module includes a high-definition imaging unit, a radar ranging unit, and an infrared temperature measurement unit; the single-chip microcomputer module includes an image processing unit, a defect detection unit, and a fault identification unit; the data transmission module includes an encoding unit and a wireless communication unit.

[0007] When the rail train enters and exits the station and enters and exits the yard line, the data acquisition module collects the data of the running gear of the rail train in real time, and transmits the collected data to the single-chip microcomputer module through the bus; further, after the data is processed by the single-chip microcomputer module, the fault result is input into the data sending module, and further, the data sending module sends the received processing result to the display screen of the terminal through wireless communication.

[0008] Specifically, the data acquisition module includes a high-definition imaging unit, a radar ranging unit, and an infrared temperature measurement unit;

[0009] The high-definition imaging unit is composed of a high-definition industrial camera and a fill light, and is used to scan the bottom of the train and extract the physical actual diagram of the bogie in real time;

[0010] The radar ranging unit is composed of a sound wave transmitting end and a sound wave receiving end; the sound wave transmitting end sends out sound waves, which are received by the sound wave receiving end, and the time interval between them is calculated for calculating the shooting distance.

[0011] The infrared temperature measurement unit is used to generate an infrared image to judge whether the temperature of the bottom of the train is too high.

[0012] In the above technical solution, the accuracy and frame rate of the high-definition industrial camera in the high-definition imaging unit can meet the relative conditions of the train entering and leaving the station and the yard line to scan and photograph the bottom of the train. The size of the photographed image is 1024 * × 2024; its fill light can provide sufficient lighting conditions for the high-definition industrial camera to shoot under poor lighting conditions, avoiding affecting the image quality of the shooting.

[0013] Specifically, the single-chip microcomputer module is composed of an image processing unit, a defect detection unit, and a fault identification unit;

[0014] Further, the image processing unit includes an image digitization model, and its process includes sampling and quantization.

[0015] The sampling is to discretize the continuous image in space to obtain a series of pixel points;

[0016] The quantization is to represent the color or gray value of each pixel point with a number, so as to convert the image into a digital matrix form for computer storage, processing and analysis.

[0017] The defect detection unit includes a convolutional layer, a pooling layer, and a fully connected layer;

[0018] The convolutional layer is used to extract the feature map of the input image;

[0019] The pooling layer is used to reduce the size of the feature map. After one operation, the new feature map obtained is the original feature map During the process, it needs to go through five convolutional layer processes and five pooling layer processes; the size of each processed feature image is 32 * 32;

[0020] The fully connected layer is used to reconnect and classify the feature images. After being processed by the n fully connected layers, a 1×m feature vector is finally obtained. Here, m represents m classification results, that is, m objects to be clustered

[0021] The fault recognition unit includes the K-means clustering algorithm, which is used to assign the obtained 1×m feature vector to one of the k clustering centers in the database for clustering; here, k represents the type of feature vectors in this database

[0022] Specifically, the fault type judgment unit selects k initialized samples as the initial clustering centers

[0023] a = a1, a2,... a k

[0024] Among them, a represents the initial selected value of the initial clustering center, k represents the number of selected values, that is, the samples are divided into k categories; the value of k needs to be based on the type of bogie faults that have been trained. Here, k = 6 is selected, representing that the algorithm can identify six types of faults such as oil leakage and wear

[0025] Furthermore, the fault recognition unit assigns each object to be clustered to the clustering center closest to it, and the assignment is based on the spatial Euclidean distance d(x, y) between the clustering center and the object to be clustered

[0026] Among them:

[0027]

[0028] n is the number of dimensions, (x1, x2,..., x m ) and (y1, y2,..., y m ) respectively represent the coordinates of two points in the feature space

[0029] Furthermore, the fault recognition unit assigns all objects to be clustered to the clustering center closest to it

[0030] Furthermore, for each category m j , the fault recognition unit recalculates its clustering center (i.e., the centroid of all samples belonging to this category) by taking the average value of the coordinates:

[0031]

[0032] Among them, k is the given number of clusters, c iRepresents the class in the i-th and k-th representative samples that is the closest in distance, c i The value of i is one from 1 to k; the centroid m j Represents our guess of the center point of the samples belonging to the same class.

[0033] Furthermore, the fault identification unit reassigns the object to be detected to the new clustering center, and the basis for the assignment is the spatial Euclidean distance between the new clustering center and the object to be detected.

[0034] Furthermore, this process will be continuously repeated by the fault identification unit until a certain termination condition is met; the termination condition can be that no (or the minimum number) of objects are reassigned to different clusters, no (or the minimum number) of clustering centers change anymore, and the sum of squared errors is locally minimized. That is, it ends when the coordinates of the new clustering center are the same as those of the previous clustering center; the points assigned to this clustering center can then be used to determine their fault types.

[0035] In the above technical solution, by processing the captured image into a digital model that can be processed by a computer, and through the processing of the convolutional layer, pooling layer, and fully connected layer, a feature map containing fault information can be quickly extracted; subsequently, the distance between the feature vector of the bogie feature map and the feature vector of the bogie fault type that has been trained in the database is calculated, and the type information of the fault can be quickly obtained to achieve intelligent bogie defect detection and fault identification, thereby helping to reduce the workload of railway workers and overcome the pain point that the current railway vehicle bogie can only be repaired during the skylight period.

[0036] Furthermore, the fault identification unit sends the processed result information to the data sending module through the bus;

[0037] The data sending module is composed of a wireless communication unit and an encoding unit.

[0038] The encoding unit can encode the result information input by the fault identification unit into a binary digital signal that can be sent;

[0039] The wireless communication unit can send the processed digital signal to the background terminal at a specific frequency.

[0040] Technical effects and advantages of the present invention: The control module controls the data acquisition module to collect real-time data, and transmits the collected data to the single-chip microcomputer module via the bus, enabling real-time bogie detection and overcoming the problems of heavy labor and easy omission of detection by workers during the skylight period. Subsequently, the control module controls the single-chip microcomputer module to process images and judge faults. Its image processing unit can process the input image into a digital matrix form for subsequent operations; its defect detection unit uses convolutional layers, pooling layers, and connection layers to process the image potentially containing fault information into a feature map and reduces it. During this process, 5 convolutions and 5 poolings are used, and the original 1024 * 2024 image processing can be performed into a 32 * × 32 fault information feature map; its fault judgment unit clusters the processed fault feature map with the fault information trained in the database to obtain fault information, so as to realize intelligent defect detection and fault identification of the train bogie. Immediately afterwards, the data output module encodes and sends the information, which can visualize the train bogie data, provide intelligent feedback monitoring, pinpoint the fault location, and guide workers to the site for maintenance, thereby reducing the labor intensity of workers, improving the maintenance efficiency of the bogie, and ensuring the safe operation of the train. Description of the Drawings

[0041] Figure 1 It is a system structure block diagram for intelligent defect detection and fault identification of a train bogie;

[0042] Figure 2 It is a general execution step flowchart for intelligent defect detection and fault identification of a train bogie;

[0043] Figure 3 It is a flowchart for the execution of the single-chip microcomputer module;

[0044] Figure 4 It is a flowchart for the execution of the fault identification unit;

[0045] Figure 5 It is an execution diagram of the data sending model. Detailed Embodiments

[0046] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0047] Embodiment 1:

[0048] Please refer to Figure 1 , Figure 1It is a system structure block diagram for defect detection and fault identification of an intelligent train bogie shown in the embodiments of the present invention; specifically, it includes a control module, a data acquisition module, a single-chip microcomputer module, and a data transmission module; the data acquisition module, the single-chip microcomputer module, and the data transmission module are respectively connected to the control module. The control module controls other modules to work and data transmission. The data acquisition module includes a high-definition imaging unit, a radar ranging unit, and an infrared temperature measurement unit; the single-chip microcomputer module includes an image processing unit, a defect detection unit, and a fault identification unit; the data transmission module includes an encoding unit and a wireless communication unit;

[0049] Further, the high-definition imaging unit is composed of a high-definition industrial camera and a fill light, and is used to scan the bottom of the train and extract the physical actual image of the bogie in real time;

[0050] The radar ranging unit is composed of a sound wave sending end and a sound wave receiving end. The sound wave sending end sends sound waves, which are received by the sound wave receiving end, and the time interval between them is calculated for calculating the shooting distance;

[0051] The infrared temperature measurement unit is used to generate an infrared image to judge whether the temperature at the bottom of the train is too high;

[0052] Clearly, the accuracy and frame rate of the high-definition industrial camera in the high-definition imaging unit can meet the relative conditions of the train entering and leaving the station and the station yard line to scan and shoot the image of the bottom of the train, and the size of the captured image is 1024*2024. Its fill light can provide sufficient lighting conditions for the high-definition industrial camera to shoot in case of poor lighting conditions, avoiding affecting the quality of the captured image.

[0053] Further, the single-chip microcomputer module is composed of an image processing unit, a defect detection unit, and a fault identification unit;

[0054] Specifically, the image processing unit processes the image into a digital model;

[0055] Further, the defect detection model will perform convolutional layer, pooling layer, and fully connected layer processing to extract the feature map of potential faults and classify them;

[0056] Further, the fault identification module performs fault type clustering according to the K-means clustering algorithm.

[0057] Clearly, by processing the captured images into a computer - processable digital model, through the processing of convolutional layers, pooling layers, and fully - connected layers, a feature map containing fault information can be quickly extracted. Subsequently, by calculating the distance between the feature vectors of the bogie feature map and the feature vectors of the trained bogie fault types in the database, the type information of the fault can be obtained quickly, so as to realize intelligent bogie defect detection and fault recognition, which helps to reduce the workload of railway workers and overcome the pain point that the current railway vehicle bogie can only be repaired during the skylight period;

[0058] Further, the fault recognition unit transports the processed result information to the data sending module through the bus;

[0059] Specifically, the data sending module consists of a wireless communication unit and an encoding unit.

[0060] Further, the encoding unit can encode the result information input by the fault recognition unit into a binary digital signal that can be sent;

[0061] Further, the wireless communication unit can send the processed digital signal to the background terminal at a specific frequency.

[0062] Embodiment 2:

[0063] This embodiment is a further supplementary description based on Embodiment 1, as follows:

[0064] Please refer to Figure 2 , Figure 2 which is the overall execution step flowchart for intelligent train bogie defect detection and fault recognition.

[0065] S101, the control module controls the data acquisition module to collect multi - dimensional data;

[0066] S102, the data collected by the control module controlling the data acquisition module is input into the single - chip microcomputer module through the bus;

[0067] S103, the control module controls the single - chip microcomputer module to perform image processing;

[0068] S104, the data processed by the control module controlling the single - chip microcomputer module is input into the data sending module through the bus;

[0069] S105, the control module controls the data sending module to send the received data to the background terminal;

[0070] Embodiment 3:

[0071] This embodiment is a supplementary description based on Embodiment 1 and Embodiment 2, as follows:

[0072] Please refer to Figure 3 ,Figure 4 , Figure 5 , Figure 3 is the execution flow chart of the single-chip microcomputer module, Figure 4 is the execution flow chart of the defect detection unit, Figure 4 is the execution flow chart of the fault identification unit, Figure 5 is the execution diagram of the data sending model.

[0073] S201, the control module controls the image processing unit to perform sampling and quantization. In this process, the sampling is to discretize the continuous image in space to obtain a series of pixel points; quantization is to represent the color or gray value of each pixel point with a number, so as to convert the image into a digital matrix form for computer storage, processing, and analysis.

[0074] S202, the control module performs convolution, pooling, and fully connected processing on the defect detection unit. The defect detection unit includes a convolutional layer, a pooling layer, and a fully connected layer. In this process, the convolutional layer is used to extract the feature map of the input image; the pooling layer is used to reduce the size of the feature map. After one operation, the new feature map obtained is the original feature Figure 1 / 4. In this process, five convolutional layer processes and five pooling layer processes are required. The size of each processed feature image is 32*32; the fully connected layer, whose function is to reconnect and classify the feature images. After n times of fully connected layer processing, a 1×m feature vector is finally obtained. Where m represents m classification results, that is, m objects to be clustered.

[0075] S203, the control module controls the fault identification unit to work;

[0076] S204, the fault type judgment unit selects k initialized samples as the initial clustering centers. The selection formula is:

[0077] a = a1, a2,... a k

[0078] Among them, a represents the initial selected value of the initial clustering center, k represents the number of selected values, that is, the samples are divided into k categories; the value of k needs to be based on the types of bogie faults that have been trained. Here, k = 6 is selected, representing that the algorithm can identify six types of faults such as oil leakage and wear;

[0079] S205, the fault identification unit assigns each object to be clustered to the clustering center closest to it. The basis for the assignment is based on the spatial Euclidean distance d(x, y) between the clustering center and the object to be clustered; among them:

[0080]

[0081] n is the number of dimensions, (x1, x2,..., x m), and (y1, y2, …, y m ) represent the coordinates of two points in the feature space respectively;

[0082] S206, the fault recognition unit assigns all objects to be clustered to the cluster center closest to it;

[0083] S207, for each category, the fault recognition unit recalculates its cluster center by taking the coordinate average, and the basis formula is as follows:

[0084]

[0085] where k is the given number of clusters, and c i represents the class closest to the sample i among the k classes, and the value of c i is one from 1 to k. The centroid m j represents our guess of the center point of the samples belonging to the same class.

[0086] S208, the fault recognition unit reassigns the object to be detected to the new cluster center. The basis for its assignment is the spatial Euclidean distance between the new cluster center and the object to be detected;

[0087] S209, after reaching the termination condition, the fault recognition unit stops working; the termination condition can be that no (or the minimum number) of objects are reassigned to different clusters, no (or the minimum number) of cluster centers change anymore, and the sum of squared errors is locally minimized. That is, it ends when the coordinates of the new cluster center are the same as those of the previous cluster center; the points assigned to this cluster center can then be used to determine their fault types;

[0088] Obviously, in the above process, by processing the captured image into a digital model that can be processed by a computer, and through the processing of the convolutional layer, pooling layer, and fully connected layer, the feature map containing fault information can be quickly extracted. Subsequently, by calculating the distance between the feature vector of the bogie feature map and the feature vectors of the bogie fault types that have been trained in the database, the type information of the fault can be quickly obtained to achieve intelligent bogie defect detection and fault recognition, thereby helping to reduce the workload of railway workers and overcome the pain point that the current railway vehicle bogies can only be repaired during the skylight period;

[0089] S210, the fault recognition unit sends the processed result information to the data sending module through the bus;

[0090] S211, the control module controls the encoding unit to work. The encoding unit can encode the result information input by the fault recognition unit into a binary digital signal that can be sent;

[0091] S212, the control module controls the wireless communication module to send information. The wireless communication unit can send the processed digital signal to the background terminal at a specific frequency.

[0092] In the embodiment of the present invention, the control module controls the data acquisition module to collect real-time data, and transmits the collected data to the single-chip microcomputer module through the bus, so as to detect the bogie in real time, overcoming the problems of large workload of workers and easy omission of inspection brought by the skylight period. Subsequently, the control module controls the single-chip microcomputer module to process the image and judge the fault. Its image processing unit can process the input image into a digital matrix form for subsequent operations; its defect detection unit uses convolutional layers, pooling layers, and connection layers to process the image potentially containing fault information into a feature map and shrink it. In this process, 5 times of convolution and 5 times of pooling are used, and the original 1024*2024 image can be processed into a 32*32 fault information feature map; its fault judgment unit clusters the processed fault feature map with the fault information trained in the database to obtain the fault information, so as to realize intelligent defect detection and fault identification of the train bogie. Immediately afterwards, the data output module encodes and sends the information, which can visualize the train bogie data, provide intelligent feedback monitoring, pinpoint the fault location, guide the workers to the site for maintenance, thereby reducing the workload of the workers, improving the maintenance efficiency of the bogie, and ensuring the safe operation of the train.

[0093] Compared with the previous manual bogie maintenance method, the system and method for train bogie defect detection and fault identification disclosed in the present invention can quickly and intelligently judge the fault type and guide the workers to perform maintenance; at the same time, compared with the existing deep learning-based technical solutions, the train bogie defect detection and fault identification method disclosed in the present invention has lower requirements for the memory, and the existing single-chip microcomputer configuration can meet the operation requirements, which is beneficial to reducing the manufacturing cost of related equipment and facilitating the lightweight and miniaturization of related equipment.

[0094] Finally, it should be noted that the above are only the preferred embodiments of the present invention and are not used to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A system for intelligent train bogie defect detection and fault identification, characterized by: It includes control module, data acquisition module, single chip computer module and data transmission module; The control module controls other modules to work and transmit data; The data acquisition module, the single-chip computer module, and the data transmission module are respectively connected to the control module, and the data acquisition module includes a high-definition imaging unit, a radar ranging unit, and an infrared temperature measuring unit; The single-chip microcomputer module includes an image processing unit, a defect detection unit, and a fault identification unit; The data sending module includes an encoding unit and a wireless communication unit.

2. The intelligent train bogie defect detection and fault identification system according to claim 1 is characterized by: The high-definition imaging unit is composed of a high-definition industrial camera and a fill light, which is used to scan the bottom of the train and capture and extract the actual physical image of the bogie in real time. The radar ranging unit is composed of a sound wave transmitting end and a sound wave receiving end; the sound wave transmitting end sends sound waves, which are received by the sound wave receiving end, and the time interval between them is calculated to calculate the shooting distance. The infrared temperature measurement unit is used to generate an infrared image to determine whether the temperature under the train is too high.

3. The intelligent train bogie defect detection and fault identification system according to claim 1 is characterized in that: The image processing unit includes an image digitization model, and its process includes sampling and quantization; The sampling is to discretize the continuous image in space to obtain a series of pixel points; The quantization is to represent the color or gray value of each pixel with numbers, thereby converting the image into a digital matrix form for computer storage, processing and analysis.

4. The intelligent train bogie defect detection and fault identification system according to claim 1 is characterized by: The defect detection unit includes a convolutional layer, a pooling layer, and a fully connected layer; The convolutional layer is used to extract a feature map of the input image; The pooling layer is used to reduce the size of the feature map. After one operation, the new feature map obtained is the original feature map. In the process, it needs to be processed by five convolution layers and five pooling layers. The size of each feature image after processing is 32 * 32; The fully connected layer is used to reconnect and classify the feature image, and finally obtain the feature vector after being processed by the secondary fully connected layer; It represents a classification result, that is, an object to be clustered.

5. The intelligent train bogie defect detection and fault identification system according to claim 1 is characterized by: The fault identification unit includes a K-means clustering algorithm, which is used to assign the 1×m feature vectors obtained above to k cluster centers in the database for clustering; wherein k represents the type of feature vector in the database; Specifically, the fault type judgment unit selects the initialized k samples as the initial clustering centers <h2 style=";text-align:left;direction:ltr">a = a1, a2,...a<h2 style=";text-align:left;direction:ltr"> k Among them, a represents the initial selected value of the initial cluster center, and k represents the number of selected values, that is, the samples are divided into k categories; the value of k needs to be selected according to the types of bogie faults that have been trained. Here, k=6 is selected, which means that the algorithm can identify six types of faults such as oil leakage and wear; The fault identification unit assigns each object to be clustered to the cluster center closest to it, and the basis for the assignment is the spatial Euclidean distance d9x,y) between the cluster center and the object to be clustered; wherein: n is the number of dimensions, (x1, x2, ..., x m ) and (y1, y2, …, y m ) represent the coordinates of two points in the feature space respectively; Furthermore, the fault identification unit assigns all the objects to be clustered to the cluster center closest to it; The fault identification unit is for each category m j , take the average of the coordinates and recalculate its cluster center (that is, the centroid of all samples belonging to this class): Among them, k is the given number of clusters, c i represents the class that is closest to sample i and class k, c i The value is from 1 to k; the centroid m j Represents our guess about the center point of samples belonging to the same class; The fault identification unit reallocates the object to be detected to the new cluster center, and the allocation is based on the spatial Euclidean distance between the new cluster center and the object to be detected; Furthermore, the fault identification unit will continuously repeat this process until a certain termination condition is met; the termination condition may be that no (or a minimum number) objects are reallocated to different clusters, no (or a minimum number) cluster centers change again, and the sum of squared errors is locally minimized; that is, it ends when the new cluster center coordinates are consistent with the previous cluster center coordinates; the point assigned to this cluster center can determine its fault type.

6. The intelligent train bogie defect detection and fault identification system according to claim 1 is characterized by: The fault identification unit transmits the processed result information to the data sending module through the bus.

7. The intelligent train bogie defect detection and fault identification system according to claim 1 is characterized by: The data sending module is composed of a wireless communication unit and a coding unit; The encoding unit can encode the result information input by the fault identification unit into a transmittable binary digital signal; The wireless communication unit can send the processed digital signal to the background terminal at a specific frequency.

8. The method of a system for intelligent train bogie defect detection and fault identification according to claim 1, characterized in that: Step 1: When the rail train enters or leaves the station, or enters or leaves the station yard, the data acquisition module collects the rail train running gear data in real time, and transmits the collected data to the single-chip microcomputer module via the bus; Step 2: After the data is processed by the single-chip microcomputer module, the fault result is input into the data transmission module, and the data transmission module sends the received processing result to the display screen of the terminal via wireless communication; Step 3: Processing the image captured by the high-definition imaging unit in the data acquisition module into a digital model that can be processed by a computer through an image processing unit; Step 4: The defect detection unit includes a convolution layer, a pooling layer, and a fully connected layer. The digital model obtained in step 3 is processed by the convolution layer, the pooling layer, and the fully connected layer in the defect detection unit to quickly extract a feature map containing fault information; Then, the fault identification unit calculates the distance between the feature vector of the bogie characteristic graph and the feature vector of the bogie fault type trained in the database, and can quickly obtain the type of fault information to achieve intelligent bogie defect detection and fault identification; Step 5: The encoding unit can encode the result information input by the fault identification unit into a transmittable binary digital signal, and can send the processed digital signal to the background terminal at a specific frequency through the wireless communication unit.

Images