Train bow net arc discharge anomaly detection method, device and system and storage medium
By constructing and training the residual neural network, a bow net arc drawing classification model is generated to detect whether there is arc drawing in the train bow net in real time, which solves the problems of slow response speed and low accuracy of traditional detection methods, and achieves efficient and accurate arc drawing abnormal detection of bow net arc drawing.
Patent Information
- Application Number
- CN202411964651.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-30
- Publication Date
- 2025-04-29
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The traditional train bow net arc drawing detection method has slow response speed and low accuracy, making it difficult to monitor and deal with bow net arc drawing abnormalities in real time.
The train bow net arc drawing anomaly detection method based on residual neural network is adopted. By constructing and training the residual neural network structure, the bow net arc drawing classification model is generated, and the real-time image is input into the model to detect whether the arc drawing exists in real time.
Real-time monitoring of the train bow net arc is realized, the accuracy and response speed of detection are improved, and the on-board mechanics can be notified in a timely manner to deal with abnormalities.
Smart Images
Figure CN119904806A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of rail transit safety technology, and in particular relates to a method and device, a system, and a storage medium for detecting abnormal arcing of a train bow catenary. Background Art
[0002] The bow-net arc detection system is one of the important systems for rail transit operation safety. Bow-net arc is usually caused by poor contact, mechanical vibration, environmental factors, etc., and is manifested as an instantaneous discharge of the arc between the bow and the net. Traditional detection methods often rely on manual inspections or simple electrical parameter monitoring, which have defects such as slow response speed and low accuracy. Summary of the invention
[0003] The technical problem to be solved by the present invention is to provide a method and device, a system and a storage medium for detecting abnormal arcing of a train bow catenary.
[0004] To achieve the above object, the present invention adopts the following technical solution:
[0005] A train pantograph arc abnormality detection method, comprising:
[0006] Step S1, obtaining a bow-net arc classification model according to a bow-net arc classification data set;
[0007] Step S2: input the real-time image of the train running bow-net into the bow-net arc classification model to obtain the result of whether there is arcing.
[0008] Preferably, step S1 comprises:
[0009] Step S11, constructing a residual neural network structure;
[0010] Step S12: training the residual neural network structure with the bow-net arc classification data set to obtain a bow-net arc classification model.
[0011] Preferably, the residual neural network structure includes: 18 convolutional layers, 1 pooling layer, 1 Flatten layer, and 1 fully connected layer; wherein the first layer is a convolutional layer with 64 convolution kernels of size 5×5 and a step size of 2; the second layer is a maximum pooling layer with a window size of 3×3 and a step size of 2; the third to sixth layers are convolutional layers with 64 convolution kernels of size 3×3 and a step size of 1; the seventh to tenth layers are convolutional layers with 128 convolution kernels of size 3×3, the seventh layer has a step size of 2, and the seventh layer has a step size of 2. The stride of the remaining layers is 1; the 11th to 14th layers are convolutional layers with 256 convolution kernels of size 3×3, the stride of the 11th layer is 2, and the stride of the remaining layers is 1; the 15th to 18th layers are convolutional layers with 512 convolution kernels of size 3×3, the stride of the 15th layer is 2, and the stride of the remaining layers is 1; the 19th layer is an average pooling layer; the 20th layer is a Flatten layer, which converts 8×16×16 data into a vector of length 2048; the 21st layer is a fully connected layer with an output vector length of 2.
[0012] The present invention also provides a train bow-cat arc abnormality detection device, comprising:
[0013] A training module is used to obtain a bow-net arc classification model based on a bow-net arc classification data set;
[0014] The detection module is used to input the real-time image of the train running bow-net into the bow-net arc classification model to obtain the result of whether arc exists.
[0015] Preferably, the training module includes:
[0016] Construction unit, used to construct the residual neural network structure;
[0017] The training unit is used to train the residual neural network structure with the bow-net arc classification data set to obtain the bow-net arc classification model.
[0018] Preferably, the residual neural network structure includes: 18 convolutional layers, 1 pooling layer, 1 Flatten layer, and 1 fully connected layer; wherein the first layer is a convolutional layer with 64 convolution kernels of size 5×5 and a step size of 2; the second layer is a maximum pooling layer with a window size of 3×3 and a step size of 2; the third to sixth layers are convolutional layers with 64 convolution kernels of size 3×3 and a step size of 1; the seventh to tenth layers are convolutional layers with 128 convolution kernels of size 3×3, the seventh layer has a step size of 2, and the seventh layer has a step size of 2. The stride of the remaining layers is 1; the 11th to 14th layers are convolutional layers with 256 convolution kernels of size 3×3, the stride of the 11th layer is 2, and the stride of the remaining layers is 1; the 15th to 18th layers are convolutional layers with 512 convolution kernels of size 3×3, the stride of the 15th layer is 2, and the stride of the remaining layers is 1; the 19th layer is an average pooling layer; the 20th layer is a Flatten layer, which converts 8×16×16 data into a vector of length 2048; the 21st layer is a fully connected layer with an output vector length of 2.
[0019] An embodiment of the present invention also provides a train bow-cat arcing anomaly detection system, comprising: a memory and a processor, wherein the memory stores a computer program executed by the processor, and the computer program executes a train bow-cat arcing anomaly detection method when executed by the processor.
[0020] An embodiment of the present invention further provides a storage medium having a computer program stored thereon, and the computer program executes a train bow-cat arcing anomaly detection method when the computer program is running.
[0021] The present invention obtains a bow-net arc classification model based on a bow-net arc classification data set; a real-time image of a train running bow-net is input into the bow-net arc classification model to obtain a result of whether arcing exists; by adopting the technical solution of the present invention, whether arcing occurs can be monitored in real time during the operation of the train, and once it occurs, the on-board mechanic needs to be notified immediately. BRIEF DESCRIPTION OF THE DRAWINGS
[0022] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.
[0023] Figure 1 This is a flow chart of a train bow-net arc abnormality detection method according to an embodiment of the present invention;
[0024] Figure 2 It is a data processing diagram for detecting abnormal arcing of train bow catenary according to the present invention. DETAILED DESCRIPTION
[0025] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions 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 are within the scope of protection of the present invention.
[0026] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.
[0027] Embodiment 1:
[0028] like Figure 1 , 2 As shown, an embodiment of the present invention provides a method for detecting abnormal arcing of a train bow and catenary, comprising:
[0029] Step S1, obtaining a bow-net arc classification model according to a bow-net arc classification data set;
[0030] Step S2: input the real-time image of the train running bow-net into the bow-net arc classification model to obtain the result of whether there is arcing.
[0031] As an implementation of the embodiment of the present invention, step S1 includes:
[0032] Step S11, constructing a residual neural network structure;
[0033] Step S12: training the residual neural network structure with the bow-net arc classification data set to obtain a bow-net arc classification model.
[0034] Further, in step S11, the residual neural network structure includes: a ResNet18 backbone network, 1 pooling layer, 1 Flatten layer, and 1 fully-connected layer. The ResNet18 backbone network contains 18 convolutional layers; the convolutional layers are used to extract image features; the pooling layer uses the average pooling method; the Flatten layer converts the output data of the previous layer into a one-dimensional array; the fully-connected layer maps the feature representations of the convolutional layer and the pooling layer to the labels of the data samples; among them, the first layer is a convolutional layer with 64 convolutional kernels of size 5×5, and the stride is 2; the second layer is a max pooling layer with a window size of 3×3, and the stride is 2; the third to sixth layers are convolutional layers with 64 convolutional kernels of size 3×3, and the stride is 1; the seventh to tenth layers are convolutional layers with 128 convolutional kernels of size 3×3, the stride of the seventh layer is 2, and the strides of the remaining layers are 1; the eleventh to fourteenth layers are convolutional layers with 256 convolutional kernels of size 3×3, the stride of the eleventh layer is 2, and the strides of the remaining layers are 1; the fifteenth to eighteenth layers are convolutional layers with 512 convolutional kernels of size 3×3, the stride of the fifteenth layer is 2, and the strides of the remaining layers are 1; the nineteenth layer is an average pooling layer; the twentieth layer is a Flatten layer that converts the data of 8×16×16 into a vector with a length of 2048; the twenty-first layer is a fully-connected layer, and the output vector length is 2.
[0035] Further, the pantograph-catenary arcing classification data set includes: arcing pictures and non-arcing pictures, with 189,786 pictures each in the embodiments of the present invention. The ratio of arcing to non-arcing is 4:6, the training set ratio is 80%, and the test set ratio is 20%. The residual neural network structure is trained with the training set to obtain the pantograph-catenary arcing classification model.
[0036] Further, in step S12, the Adam optimizer is used to train the residual neural network structure. The predetermined number of iterations is 500 steps, and the learning rate is 0.0001. The stochastic optimization method with adaptive momentum is used to continuously update the network parameters and recalculate the loss function value until the loss function value converges or reaches the predetermined number of iterations.
[0037] As an implementation manner of the embodiment of the present invention, the step S2 includes:
[0038] Step S21: Input the real-time image of the pantograph-catenary during train operation into the pantograph-catenary arcing classification model. After passing through 18 convolutional layers, 1 pooling layer, 1 Flatten layer, and 1 fully-connected layer, the detection result is obtained;
[0039] Step S22: The detection result is a vector [x0, x1] with a length of 2. The first value x0 is the probability of no arcing, and the second value x1 is the probability of arcing; when x0 > x1, the output result is no arcing, and when x0 < x1, the output result is arcing.
[0040] Embodiment 2:
[0041] The embodiment of the present invention further provides a train bow-cat arc abnormality detection device, comprising:
[0042] A training module is used to obtain a bow-net arc classification model based on a bow-net arc classification data set;
[0043] The detection module is used to input the real-time image of the train running bow-net into the bow-net arc classification model to obtain the result of whether arc exists.
[0044] As an implementation of an embodiment of the present invention, the training module includes:
[0045] Construction unit, used to construct the residual neural network structure;
[0046] The training unit is used to train the residual neural network structure with the bow-net arc classification data set to obtain the bow-net arc classification model.
[0047] As an implementation of an embodiment of the present invention, the residual neural network structure includes: 18 convolutional layers, 1 pooling layer, 1 Flatten layer, and 1 fully connected layer; wherein the first layer is a convolutional layer with 64 convolution kernels of size 5×5 and a step size of 2; the second layer is a maximum pooling layer with a window size of 3×3 and a step size of 2; the third to sixth layers are convolutional layers with 64 convolution kernels of size 3×3 and a step size of 1; the seventh to tenth layers are convolutional layers with 128 convolution kernels of size 3×3, and the step size of the seventh layer is 2. The stride length is 2, and the stride length of the remaining layers is 1; the 11th to 14th layers are 256 convolution layers with a convolution kernel size of 3×3, the stride length of the 11th layer is 2, and the stride length of the remaining layers is 1; the 15th to 18th layers are 512 convolution layers with a convolution kernel size of 3×3, the stride length of the 15th layer is 2, and the stride length of the remaining layers is 1; the 19th layer is an average pooling layer; the 20th layer is a Flatten layer, which converts 8×16×16 data into a vector of length 2048; the 21st layer is a fully connected layer, and the output vector length is 2.
[0048] Embodiment 3:
[0049] An embodiment of the present invention also provides a train bow-cat arcing anomaly detection system, comprising: a memory and a processor, wherein the memory stores a computer program executed by the processor, and the computer program executes a train bow-cat arcing anomaly detection method when executed by the processor.
[0050] Embodiment 4:
[0051] An embodiment of the present invention further provides a storage medium having a computer program stored thereon, and the computer program executes a train bow-cat arcing anomaly detection method when the computer program is running.
[0052] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.
Claims
1. A train bow-cat arc abnormality detection method, characterized in that: include: Step S1, obtaining a bow-net arc classification model according to a bow-net arc classification data set; Step S2: input the real-time image of the train running bow-net into the bow-net arc classification model to obtain the result of whether there is arcing.
2. The train bow-cat arcing abnormality detection method according to claim 1, characterized in that: Step S1 includes: Step S11, constructing a residual neural network structure; Step S12: training the residual neural network structure with the bow-net arc classification data set to obtain a bow-net arc classification model.
3. The train bow-cat arc abnormality detection method according to claim 2, characterized in that: The residual neural network structure includes: 18 convolutional layers, 1 pooling layer, 1 Flatten layer, and 1 fully connected layer; the first layer is a convolutional layer with 64 convolution kernels of size 5×5 and a step size of 2; the second layer is a maximum pooling layer with a window size of 3×3 and a step size of 2; the third to sixth layers are convolutional layers with 64 convolution kernels of size 3×3 and a step size of 1; the seventh to tenth layers are convolutional layers with 128 convolution kernels of size 3×3, the seventh layer has a step size of 2, and the remaining layers have a step size of 2. The length is 1; the 11th to 14th layers are 256 convolution layers with a convolution kernel size of 3×3, the 11th layer has a stride of 2, and the rest of the layers have a stride of 1; the 15th to 18th layers are 512 convolution layers with a convolution kernel size of 3×3, the 15th layer has a stride of 2, and the rest of the layers have a stride of 1; the 19th layer is an average pooling layer; the 20th layer is a Flatten layer, which converts 8×16×16 data into a vector of length 2048; the 21st layer is a fully connected layer, and the output vector length is 2.
4. A train bow-cat arc abnormality detection device, characterized in that: include: A training module is used to obtain a bow-net arc classification model based on a bow-net arc classification data set; The detection module is used to input the real-time image of the train running bow-net into the bow-net arc classification model to obtain the result of whether arc exists.
5. The train bow-cat arc abnormality detection device according to claim 4, characterized in that: The training modules include: Construction unit, used to construct the residual neural network structure; The training unit is used to train the residual neural network structure with the bow-net arc classification data set to obtain the bow-net arc classification model.
6. The train bow-cat arc abnormality detection device according to claim 5, characterized in that: The residual neural network structure includes: 18 convolutional layers, 1 pooling layer, 1 Flatten layer, and 1 fully connected layer; the first layer is a convolutional layer with 64 convolution kernels of size 5×5 and a step size of 2; the second layer is a maximum pooling layer with a window size of 3×3 and a step size of 2; the third to sixth layers are convolutional layers with 64 convolution kernels of size 3×3 and a step size of 1; the seventh to tenth layers are convolutional layers with 128 convolution kernels of size 3×3, the seventh layer has a step size of 2, and the remaining layers have a step size of 2. The length is 1; the 11th to 14th layers are 256 convolution layers with a convolution kernel size of 3×3, the 11th layer has a stride of 2, and the rest of the layers have a stride of 1; the 15th to 18th layers are 512 convolution layers with a convolution kernel size of 3×3, the 15th layer has a stride of 2, and the rest of the layers have a stride of 1; the 19th layer is an average pooling layer; the 20th layer is a Flatten layer, which converts 8×16×16 data into a vector of length 2048; the 21st layer is a fully connected layer, and the output vector length is 2.
7. A train bow-cat arc abnormality detection system, characterized in that: include: A memory and a processor, wherein the memory stores a computer program executed by the processor, and when the computer program is executed by the processor, the train bow-cat arc abnormality detection method as described in any one of claims 1 to 3 is executed.
8. A storage medium, characterized in that: The storage medium stores a computer program, which, when running, executes the train bow-cat arc abnormality detection method as described in any one of claims 1 to 3.
Citation Information
Patent Citations
Visual detection method for pantograph-catenary arcing of electrified railway
CN112766195A