Gsm-r service quality analysis processing method and device based on neural network

Abstract

This invention proposes a neural network-based method and apparatus for GSM-R service quality analysis and processing, belonging to the field of communication technology analysis and processing technology. The method includes: acquiring GSM-R waveform data, inputting it into a service quality classification and identification model of the GSM-R system, and analyzing it to obtain the service quality problems corresponding to the GSM-R system under the input waveform data conditions; the service quality classification and identification model adopts a ResNet architecture, and the training method is as follows: labeling the categories of service quality problems and dividing the labeled sample set; training the artificial intelligence model using the training set, combining the categories of service quality problems, learning the mapping from GSM-R waveform data to service quality problem categories in the artificial intelligence model, and verifying and testing using a validation set and a test set to obtain the service quality classification and identification model; and displaying the GSM-R waveform data and service quality problems through a railway facility monitoring platform.

Description

Technical Field

[0001] This invention relates to the field of communication technology analysis and processing technology, and more particularly to a GSM-R service quality analysis and processing method and apparatus based on neural networks. Background Technology

[0002] This section is intended to provide background or context for embodiments of the present invention. The description herein is not intended to imply that it is prior art simply because it is included in this section.

[0003] GSM-R, a digital wireless communication system specifically designed for railway communication, not only boasts superior security and efficiency but also supports diverse data communication services. It has powerfully promoted the modernization of railway transportation production command methods, meeting the comprehensive needs of train dispatching, control, and railway informatization development, and providing solid communication support for the digital transformation and intelligent upgrading of the railway industry. As of the end of December 2020, GSM-R covered tens of thousands of kilometers of railway lines. In-depth research and analysis of GSM-R system service quality indicators can identify potential problems in the network, allowing for corresponding optimization measures. These optimization measures include, but are not limited to, adjusting network parameters, optimizing network layout, upgrading hardware equipment, and improving software algorithms. Through continuous practice and adjustments, the network service quality can be gradually improved to meet the ever-increasing business demands. The electromagnetic environment along railway lines is complex, and GSM-R is subject to significant interference from mobile operators (GSM). During communication testing, testers need to report four types of problems based on the waveform diagrams of GSM-R service quality testing: dropped calls, handover failures, handover anomalies, and poor voice quality. Despite continuous technological advancements, addressing traditional GSM-R service quality issues still relies on experienced and knowledgeable inspectors who meticulously identify and assess problems visually, enabling them to promptly detect and resolve complex situations that automated tools struggle to detect. To further improve the accuracy and efficiency of inspections, efforts are underway to explore combining advanced technologies with manual inspection, aiming to achieve higher-quality, more intelligent results.

[0004] With the continuous development of deep learning technology, the railway industry has adopted a large number of deep learning methods for defect identification, which has greatly optimized the efficiency and accuracy of defect identification. For example, a self-attention mechanism neural network is used to identify the string clamp bolts, an important component of the overhead contact line power supply line, improving identification accuracy while reducing computational load and parameter count. Simultaneously, image augmentation is performed to address the problem of insufficient defect samples in the dataset. An improved point cloud component-level segmentation method for railway bridges based on RandLA-Net solves a key problem in railway bridge construction progress management: how to more accurately extract component-level information from point cloud data. By employing a self-attention mechanism, the network can better capture long-range dependencies and semantic relationships between points. To innovate the method of finding lost items in railway passenger transport scenarios, based on the analysis of the needs and difficulties in finding lost items in railway passenger transport, and combining the cutting-edge technologies of face recognition and deep learning, an image search-based framework for finding lost items is established. Security check and non-security check lost item search schemes are designed for railway passenger transport scenarios. Research results show that this method can further improve the intelligence level of railway passenger transport and optimize the efficiency of finding lost items. Based on the data characteristics of waveform reporting for GSM-R Quality of Service (QoS) detection, this paper draws on deep learning theory, which has achieved significant performance improvements in computer vision tasks, to enhance detection efficiency. With the continuous development of deep learning technology, neural network structures have gradually evolved from shallow to deep layers. However, traditional neural networks experience performance degradation when the number of layers increases, mainly due to the vanishing and exploding gradient problems. To address these issues, the ResNet network was proposed in 2015. Besides achieving significant performance in image classification, ResNet also excels in object detection, image segmentation, natural language processing, and other tasks. This demonstrates that ResNet possesses strong generalization and adaptability, making it applicable to various deep learning tasks.

[0005] In summary, deep learning has been widely applied in multiple fields. However, there is currently no research on waveform reporting for GSM-R service quality detection, and a technical solution is urgently needed to overcome these shortcomings, improve GSM-R service quality detection methods, and enhance detection performance. Summary of the Invention

[0006] To address the problems existing in the prior art, this invention proposes a method and apparatus for GSM-R service quality analysis and processing based on neural networks. This invention classifies GSM-R service quality problems and proposes a GSM-R image recognition algorithm based on ResNet networks. It utilizes ResNet networks to construct the GSM-R image recognition algorithm, enriching the case library for various service quality problems and providing strong technical support for service quality analysis.

[0007] In a first aspect of the present invention, a GSM-R service quality analysis and processing method based on a neural network is proposed, comprising:

[0008] Acquire GSM-R waveform data;

[0009] The GSM-R waveform data is input into the service quality classification and identification model of the GSM-R system. Analysis yields the service quality issues corresponding to the GSM-R system under the input waveform data conditions. The service quality classification and identification model employs a ResNet architecture and is pre-trained as follows:

[0010] For the GSM-R waveform data sample set, the service quality problem categories are labeled, and the labeled sample set is divided into training set, validation set and test set;

[0011] The artificial intelligence model is trained using the training set. In conjunction with the categories of service quality problems, the artificial intelligence model is mapped from GSM-R waveform data to service quality problem categories. The model is then validated and tested using the validation set and test set to obtain a service quality classification and recognition model. The service quality classification and recognition model takes GSM-R waveform data as input and the categories of service quality problems as output.

[0012] The railway facility monitoring platform displays the acquired GSM-R waveform data and the service quality issues identified through analysis.

[0013] In a second aspect of the present invention, a GSM-R service quality analysis and processing device based on a neural network is provided, comprising:

[0014] The data acquisition module is used to acquire GSM-R waveform data;

[0015] The service quality analysis module is used to input the GSM-R waveform data into the service quality classification and identification model of the GSM-R system, and to obtain the service quality problems corresponding to the GSM-R system under the input waveform data conditions through analysis; wherein, the service quality classification and identification model adopts the ResNet architecture and is pre-trained in the following manner:

[0016] For the GSM-R waveform data sample set, the service quality problem categories are labeled, and the labeled sample set is divided into training set, validation set and test set;

[0017] The artificial intelligence model is trained using the training set. In conjunction with the categories of service quality problems, the artificial intelligence model is mapped from GSM-R waveform data to service quality problem categories. The model is then validated and tested using the validation set and test set to obtain a service quality classification and recognition model. The service quality classification and recognition model takes GSM-R waveform data as input and the categories of service quality problems as output.

[0018] The display module is used to display the GSM-R waveform data obtained through the railway facility monitoring platform and the service quality issues identified after analysis.

[0019] In a third aspect of the present invention, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement a neural network-based GSM-R quality of service analysis and processing method.

[0020] In a fourth aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing a computer program that, when executed by a processor, implements a neural network-based GSM-R quality of service analysis and processing method.

[0021] In a fifth aspect of the present invention, a computer program product is provided, the computer program product comprising a computer program that, when executed by a processor, implements a neural network-based GSM-R quality of service analysis and processing method.

[0022] This invention proposes a neural network-based GSM-R service quality analysis and processing method and apparatus. Based on the analysis of a large number of GSM-R service quality problem images, it identifies the abnormal patterns in these images, determines their anomalous characteristics, and constructs an optimal algorithm based on these characteristics. The relevant parameters and thresholds in this invention were determined through extensive experimentation, selecting the best-performing results. Compared to existing technologies, it offers at least the following technical advantages: It uses a dataset generalization method based on transfer learning to extract features from GSM-R service quality problem images, enriching the multi-source heterogeneous data samples for data analysts; and it assists inspectors in achieving automated GSM-R service quality problem identification through a ResNet-based GSM-R image recognition algorithm, improving detection efficiency. By enriching the GSM-R service quality problem sample library, it provides valuable experimental materials for data analysts and developers, offering strong technical support for the identification of various communication problems. Attached Figure Description

[0023] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a schematic diagram of a GSM-R service quality analysis and processing method based on a neural network according to an embodiment of the present invention.

[0025] Figure 2 This is a schematic diagram illustrating the reporting of dropped call issues according to a specific embodiment of the present invention.

[0026] Figure 3 This is a schematic diagram illustrating the problem of dropped calls not being reported in a specific embodiment of the present invention.

[0027] Figure 4 This is a schematic diagram of a handover failure reporting according to a specific embodiment of the present invention.

[0028] Figure 5 This is a schematic diagram illustrating a specific embodiment of the present invention where a switching failure is not reported.

[0029] Figure 6 This is a schematic diagram of a switching anomaly reporting according to a specific embodiment of the present invention.

[0030] Figure 7 This is a schematic diagram of a switching anomaly reporting according to another specific embodiment of the present invention.

[0031] Figure 8 This is a schematic diagram of poor voice quality reporting according to a specific embodiment of the present invention.

[0032] Figure 9 This is a schematic diagram of the model training process according to a specific embodiment of the present invention.

[0033] Figure 10 This is a schematic diagram of the ResNet network structure according to a specific embodiment of the present invention.

[0034] Figure 11 This is a schematic diagram of the ResNet18 network structure according to a specific embodiment of the present invention.

[0035] Figure 12 This is a schematic diagram of the jump connection relationship in a specific embodiment of the present invention.

[0036] Figure 13A This is a schematic diagram of the loss value for each iteration of a specific embodiment of the present invention.

[0037] Figure 13B This is a schematic diagram illustrating the accurate values ​​for each iteration of a specific embodiment of the present invention.

[0038] Figure 14 This is a schematic diagram of the processing flow of a disease case library according to a specific embodiment of the present invention.

[0039] Figure 15 This is a schematic diagram of a communication-related disease database query interface according to a specific embodiment of the present invention.

[0040] Figure 16 This is a schematic diagram of the interface for importing a communication-related disease database according to a specific embodiment of the present invention.

[0041] Figure 17 This is a schematic diagram of the architecture of a neural network-based GSM-R service quality analysis and processing device according to an embodiment of the present invention.

[0042] Figure 18 This is a schematic diagram of the architecture of a GSM-R service quality analysis and processing device based on a neural network, according to another embodiment of the present invention.

[0043] Figure 19 This is a schematic diagram of a computer device structure according to an embodiment of the present invention. Detailed Implementation

[0044] The principles and spirit of the invention will now be described with reference to several exemplary embodiments. It should be understood that these embodiments are given merely to enable those skilled in the art to better understand and implement the invention, and are not intended to limit the scope of the invention in any way. Rather, these embodiments are provided to make this disclosure more thorough and complete, and to fully convey the scope of this disclosure to those skilled in the art.

[0045] Those skilled in the art will recognize that embodiments of the present invention can be implemented as a system, apparatus, device, method, or computer program product. Therefore, this disclosure can be specifically implemented in the following forms: entirely hardware, entirely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

[0046] According to an embodiment of the present invention, a method and apparatus for GSM-R service quality analysis and processing based on neural networks are proposed. This invention relates to the application of intelligent classification and decision-making in railway inspection and monitoring within the field of railway communication technology. Since GSM-R (GSM-R Dedicated Mobile Communication System) is a digital wireless communication system for railways, it is susceptible to electromagnetic interference. Inspection personnel report four types of problems—dropped calls, handover failures, handover anomalies, and poor voice quality—based on the waveform diagrams of GSM-R service quality detection. The intelligent classification and processing of GSM-R image recognition and service quality problems based on ResNet networks is of great significance for further improving the accuracy and efficiency of inspection.

[0047] The principles and spirit of the present invention will be explained in detail below with reference to several representative embodiments.

[0048] Figure 1 This is a schematic flowchart of a GSM-R service quality analysis and processing method based on a neural network according to an embodiment of the present invention. Figure 1 As shown, the method includes:

[0049] S101, acquire GSM-R waveform data;

[0050] S102, the GSM-R waveform data is input into the service quality classification and identification model of the GSM-R system, and the service quality problem corresponding to the GSM-R system under the input waveform data conditions is obtained through analysis; wherein, the service quality classification and identification model adopts the ResNet architecture and is pre-trained in the following manner:

[0051] For the GSM-R waveform data sample set, the service quality problem categories are labeled, and the labeled sample set is divided into training set, validation set and test set;

[0052] The artificial intelligence model is trained using the training set. In conjunction with the categories of service quality problems, the artificial intelligence model is mapped from GSM-R waveform data to service quality problem categories. The model is then validated and tested using the validation set and test set to obtain a service quality classification and recognition model. The service quality classification and recognition model takes GSM-R waveform data as input and the categories of service quality problems as output.

[0053] S103 displays the acquired GSM-R waveform data and the service quality issues identified through analysis via the railway facility monitoring platform.

[0054] This invention proposes a neural network-based GSM-R service quality analysis and processing method and apparatus. Based on the analysis of a large number of GSM-R service quality problem images, it identifies the abnormal patterns in these images, determines their anomalous characteristics, and constructs an optimal algorithm based on these characteristics. The relevant parameters and thresholds in this invention were determined through extensive experimentation, selecting the best-performing results. Compared to existing technologies, it offers at least the following technical advantages: It uses a dataset generalization method based on transfer learning to extract features from GSM-R service quality problem images, enriching the multi-source heterogeneous data samples for data analysts; and it assists inspectors in achieving automated GSM-R service quality problem identification through a ResNet-based GSM-R image recognition algorithm, improving detection efficiency. By enriching the GSM-R service quality problem sample library, it provides valuable experimental materials for data analysts and developers, offering strong technical support for the identification of various communication problems.

[0055] To provide a clearer explanation of the above-mentioned neural network-based GSM-R service quality analysis and processing method, each step will be described in detail below.

[0056] In one embodiment, the categories of the quality of service issues include at least: dropped calls, handover failures, handover anomalies, and poor voice quality.

[0057] Specifically, regarding service quality issues related to dropped calls, if the test interface indicates a connection loss and there is continuous voice quality reaching the set level, then the call is determined to have been dropped at the corresponding location. If the test interface indicates a connection loss and there is no voice quality issue, the historical data of the line is replayed. If the same phenomenon occurs at the same location, then the call is determined to have been dropped at the corresponding location. If the same phenomenon does not occur, then the test system is determined to be abnormal.

[0058] For service quality issues related to failed switching, when the test interface indicates a failed switching, determine the location where the switching failed.

[0059] For service quality issues caused by handover anomalies, a handover anomaly is defined as one that occurs when no handover takes place between two base stations or when multiple handovers occur.

[0060] For service quality issues with poor voice quality, if the voice quality on the test interface is within the set level range and no dropped calls occur, it is judged as poor voice quality.

[0061] The following describes different service quality issues using a specific example.

[0062] 1. Call dropped:

[0063] When the test interface indicates a connection loss and there are consecutive voice quality levels of 7, it is determined that the call was dropped at that location. Figure 2 The diagram shown illustrates a specific embodiment of the present invention regarding call drop reporting. Device 1001 corresponds to Lushan, device 1012 corresponds to JJK-LS01, and device 1002 corresponds to JJN-JJX01.

[0064] When the test interface indicates a connection loss but there is no voice quality issue, to avoid false alarms, it is necessary to review the historical data of this line from the previous month. If the same phenomenon occurred at the same location last month, then it is determined that the call was dropped at that location; if there were no abnormalities last month, then it is determined that there is a sudden abnormality in the test system, and the problem is not reported for the time being, but continued observation is maintained. Figure 3 The diagram shown illustrates the failure to report dropped calls according to a specific embodiment of the present invention. Device 1012 corresponds to ZX-BJ03.

[0065] 2. Switching failed:

[0066] When the test interface indicates that the switch failed, determine if the switch at that location failed. Figure 4 As shown. Device 1004 corresponds to Luoyuan 02, device 1002 corresponds to Luoyuan 01, and device 1007 corresponds to Luoyuan.

[0067] When the test module operates for an extended period, occasional anomalies may occur, manifesting as the test interface displaying a "switch failure" message with the "switch command" and "switch failure" signals having the same timestamp. In such cases, do not report the issue immediately; continue observation. Figure 5 As shown.

[0068] 3. Switching error:

[0069] There should be at least one handover between two base stations. If no handover occurs or multiple handovers occur between the two base stations, it is considered a handover anomaly (co-line or hub areas are considered separately). For this type of problem, the following two aspects should be considered when reporting the issue:

[0070] If the ping-pong handover is caused by continuous cross-coverage of field strength, for C3 lines, no notification is required if the overall interference rate is within acceptable limits, but the railway bureau maintenance manager must still be informed; for C2 lines and regular speed lines, no notification is required if the voice quality is not affected, but the railway bureau maintenance manager must still be informed.

[0071] If the currently serving cell is forced to switch from a high-level signal to a low-level signal, this is considered to be caused by equipment or interference and should be reported. Figure 6 , 7 As shown. In Figure 6 In this context, device 1001 corresponds to DXD-YZ04, device 1003 corresponds to DXD-YZ03, and device 1009 corresponds to DXB-YZ02. Figure 7 In this context, device 1008 corresponds to TJN-CZX24, device 1004 corresponds to TJN-CZX23, device 1000 corresponds to TJN-CZX22, and device 1013 corresponds to TJN-CZX21.

[0072] 4. Poor voice quality:

[0073] When the test interface shows a voice quality level of 6-7, but no dropped calls occur, it is judged as poor voice quality. This issue is handled differently for different lines. For C3 lines, if the overall interference rate is within acceptable limits, no report is issued, but the railway bureau maintenance manager must still be notified. For C2 lines and regular speed lines, if the voice quality is consistently between 5-7 for more than 500 meters, a report is required. Figure 8 As shown. Device 1003 corresponds to HeFeiXiA, device 1010 corresponds to HFN-CAJ02-BU, and device 1004 corresponds to HFN-CAJ01.

[0074] In one embodiment, for S101, GSM-R waveform data is acquired.

[0075] Furthermore, image preprocessing is performed on the GSM-R waveform data; specifically, random image cropping is performed on the GSM-R waveform data, and the size of the cropped image is adjusted to a set pixel specification; during the training process, the image is horizontally flipped with a set probability; and the image is normalized to convert the pixel values ​​of the image to a distribution with a mean of 0 and a standard deviation of 1.

[0076] In one embodiment, for S102, the GSM-R waveform data is input to the service quality classification and identification model of the GSM-R system, and the service quality problem corresponding to the GSM-R system under the input waveform data conditions is obtained through analysis; wherein, the service quality classification and identification model adopts the ResNet architecture, referring to... Figure 9 It was obtained by training in advance as follows:

[0077] S901, for the GSM-R waveform data sample set, label the categories of service quality problems, divide the labeled sample set to obtain the training set, validation set and test set;

[0078] S902, the artificial intelligence model is trained using the training set, and the mapping from GSM-R waveform data to service quality problem categories is learned by combining the categories of service quality problems with the training set. The model is then validated and tested using the validation set and test set to obtain a service quality classification and recognition model. The service quality classification and recognition model takes GSM-R waveform data as input and the categories of service quality problems as output.

[0079] The service quality classification and identification model adopts the ResNet architecture as ResNet18, which includes 16 convolutional layers and 2 fully connected layers.

[0080] Convolutional layers are used to extract multi-level feature representations from the input image, while fully connected layers are used for classification or regression based on the extracted features.

[0081] ResNet18 incorporates residual blocks, each consisting of two convolutional layers and one skip connection. The skip connection allows information to be passed directly from one layer to the next in the network.

[0082] In one embodiment, this invention proposes a deep learning-based image recognition method to address GSM-R quality of service issues. To enhance the model's generalization ability and robustness, this invention employs a series of image preprocessing techniques. First, the input image is randomly cropped, which is an effective data generalization method, enabling the model to learn image features at different locations and sizes. After random cropping, the size of the cropped image is further adjusted to 224×224 pixels, ensuring that the model has a consistent input format when processing images of different sizes.

[0083] To further enhance the model's robustness, a random horizontal flipping strategy was introduced. During training, images are horizontally flipped with a certain probability, which helps the model learn orientation-independent features in the images. Through this data augmentation technique, the model exhibits greater adaptability when handling various complex and variable scenes.

[0084] In the final stage of image preprocessing, this invention performs image normalization. This step transforms the image's pixel values ​​to a distribution with a mean of 0 and a standard deviation of 1, which helps the model converge faster during training. Normalization not only improves the model's training efficiency but also helps the model learn more robust feature representations. This lays the foundation for automated and intelligent fault detection in railway communication systems.

[0085] ResNet18 is a deep convolutional neural network architecture, such as Figure 10 As shown, it is suitable for processing image recognition and other visual tasks. Its core design concept lies in the introduction of residual blocks. Through this innovative structure, the ResNet series of networks successfully solved the gradient vanishing and performance degradation problems that occur during the training of deep networks. Each residual block consists of two convolutional layers and a skip connection. Skip connections allow information to be directly passed from one layer to the next in the network, thus enabling rapid information flow.

[0086] exist Figure 10In this context, "layer name" refers to the name of each layer in the network, typically used to identify and distinguish different layers, such as "conv1" and "conv2". "output size" refers to the size of the feature map output by the network layer, usually including width and height. "conv" refers to a convolutional layer, the basic unit in deep learning used for feature extraction. It generates new feature maps by performing weighted summations on the input data through a sliding window (convolutional kernel). "stride" refers to the size of the convolutional kernel during the convolution operation. The stride determines the size of the output feature map; a larger stride leads to a smaller output feature map size. "max pool" refers to a max pooling layer, which reduces the spatial dimension of the feature map by selecting the maximum value from a local region in the input feature map, helping to extract important features and reduce computational cost. "average pool" refers to an average pooling layer, similar to max pooling, but it reduces the spatial dimension of the feature map by calculating the average value of local regions in the input feature map. 1000-d FC: Refers to a 1000-dimensional fully connected layer, used in convolutional neural networks to map extracted features to the final output class. For example, in the ImageNet classification task, the number of output classes is 1000. Softmax is an activation function used in multi-class classification problems to transform the raw scores of the neural network output into a probability distribution. The softmax function ensures that the sum of the probabilities of all output classes is 1. FLOPs stands for Floating Point Operations, a metric for measuring the computational complexity of a model, representing the total number of floating-point operations the model needs to perform during forward propagation. Higher FLOPs generally indicate a higher computational cost.

[0087] ResNet18, such as Figure 11 As shown, it contains a total of 18 layers, of which 16 are convolutional layers responsible for extracting multi-level feature representations from the input image; and 2 are fully connected layers used for classification or regression based on the extracted features. The residual block type of ResNet18 is BasicBlock.

[0088] exist Figure 11In the exemplary ResNet variant network architecture, the input image for input layer x is 3×64×64 pixels with 3 channels, each channel being 64×64 pixels. Convolutional layers (Conv) consist of multiple 3×3 convolutional layers, with the number of kernels (C) and stride (S) labeled for each layer. For example, the first layer is a 3×3 convolution with 64 kernels and a stride of 1. Max pooling layers follow the convolutional layers with a 3×3 stride of 2, typically used to reduce the size of feature maps. Residual blocks, each containing two 3×3 convolutional layers, have the number of kernels labeled for each block. Residual blocks are the core of the ResNet network, allowing the network to learn residual maps, thus addressing the vanishing gradient problem in deep network training. Skip connections, indicated by solid arrows, directly pass feature maps; this is part of skip connections, allowing information in the network to bypass some layers and pass directly. The Global Average Pooling (GAP) layer, located at the end of the network, compresses each channel of the feature map into a single value. Following the GAP layer is a Fully Connected layer with an output size of 512, typically used to map the features to the final classification result. Different stages, distinguished by different colors, each contain a certain number of residual blocks. These stages usually correspond to different depths of the network. The output layer y represents the number of final output classes. Arrows indicate direct passing of the feature map, while dashed arrows indicate downsampling of the feature map through a 1×1 convolutional layer.

[0089] In deep neural networks, the vanishing gradient problem is a common issue. This problem refers to the phenomenon where, during backpropagation, the gradient diminishes and eventually disappears as it is multiplied by the weight matrix layer by layer. Skip connections can avoid this problem because each residual block contains two convolutional layers and one skip connection. Skip connections allow information to be transferred more quickly within the network. Specifically, after the input passes through the first convolutional layer and the first pooling layer, it undergoes feature mapping through a fully connected layer. Then, after passing through the second convolutional layer and the second pooling layer, it undergoes feature mapping again through a fully connected layer. The outputs of these two fully connected layers are added to the original input, forming two new outputs. This approach increases the model's expressive power and stability. The cross-layer connections in ResNet directly pass the input gradient to subsequent layers, thus avoiding the vanishing gradient problem. This reflects the fundamental design philosophy of the ResNet series of networks: incorporating skip connections to alleviate the vanishing gradient problem during deep network training. ResNet18 not only improves model performance while maintaining network depth but also accelerates the training process and increases the model's convergence speed. like Figure 12 The diagram shown illustrates the relationship of skip connections in a specific embodiment of the present invention. Skip connections allow information in the network to bypass some layers and be directly passed through feature maps. Residual blocks are core components in the ResNet network. The purpose of residual blocks is to simplify the flow of information in the network, making the training of deep networks easier. Figure 12 In this diagram, 'x' represents the input feature. 'Weight layer' represents a layer with weights, typically a convolutional layer. Within the residual block, there are usually two such layers, each followed by a ReLU activation function. The first ReLU stands for Rectified Linear Unit, a commonly used activation function that sets all negative values ​​to 0 and retains positive values. 'F(x)' represents the residual function, a combination of the two weighted layers and the ReLU activation function; this function aims to learn the residual of the input 'x'. 'Identity' represents the identity mapping, directly passing the input 'x' to the output without any transformation. 'F(x)+x' represents adding the output of the residual function to the input 'x'; a crucial operation in the residual block, it allows the network to learn the difference (residual) between the input and output, rather than directly learning the output. The second ReLU indicates that after the residuals are added, the network passes the output through another ReLU activation function to obtain the final output.

[0090] The training data for this invention consists of four folders named dh, qhsb, qhyc, and yyzz, representing dropped calls, handover failures, handover anomalies, and poor voice quality, respectively, each containing 170 samples. The test data contains 75 images for each class. Intelligent classification and prediction of GSM-R service quality issues is performed based on ResNet. The loss and accuracy values ​​corresponding to each epoch are as follows: Figure 13A and Figure 13B As shown. The highest accuracy rate is 84%.

[0091] In one embodiment, for S103, the acquired GSM-R waveform data and the service quality issues obtained after analysis are displayed through the railway facility monitoring platform.

[0092] Further reference Figure 14 In order to collect and analyze disease cases and provide strong technical support to maintenance personnel, this invention also proposes the following processing methods:

[0093] S1401, Create a service quality case library, which stores different categories of service quality problems, as well as descriptions, causes, and solutions for the service quality problems;

[0094] S1402, After the service quality problem is identified through analysis, the description, cause, and solution corresponding to the service quality problem are queried through the disease case database and provided to the operation and maintenance personnel;

[0095] S1403, the service quality problem database is updated according to the service quality problems output by the service quality classification and identification model, and used as sample data for iterative training of the service quality classification and identification model to continuously learn and train the model.

[0096] Disease database and case database (such as) Figure 15 , 16 As a crucial component of the railway infrastructure inspection and monitoring data platform, the richness and accuracy of its content directly impact the safety and efficiency of railway operations. The introduction of the GSM-R service quality issue intelligent classification algorithm provides abundant and accurate sample case input materials, greatly enriching the sample case resources in the case library and making it more targeted and practical. These cases not only cover various service quality issues that may occur in the GSM-R system, but also include detailed descriptions of the problems, their causes, and solutions, providing valuable reference and guidance for data analysts and inspection personnel.

[0097] Furthermore, the continuous optimization and upgrading of the GSM-R service quality issue intelligent classification algorithm has enabled faster updates to the case database. With the emergence of new technologies and the ongoing development of railway communication networks, new service quality issues are constantly arising. The intelligent classification algorithm can track and identify these issues and incorporate them into the case database, ensuring its timeliness and completeness.

[0098] In practical applications, the GSM-R system primarily provides scheduling communication services such as railway train dispatching, achieving a combination of wired and wireless scheduling communication, and providing safe and reliable train-to-ground information transmission services for high-speed train operation control systems. When GSM-R service quality issues arise, train operation safety is affected. Inspection personnel report problems based on waveform images from GSM-R service quality detection. The algorithm proposed in this invention extracts features from different problem waveform images, identifies the problem image types, and establishes a sample library of GSM-R service quality problems. Addressing the characteristics of the aforementioned problem image types, the technical solution adopted in this invention is to use transfer learning for GSM-R service quality problem images to better generalize the problem image dataset, increase the model's robustness, and normalize the images to facilitate convergence during model training. A ResNet model is built, with training parameters used as initialization for full fine-tuning, and pre-trained parameters used as fixed feature extraction for classifier fine-tuning. At the beginning of each training iteration, the gradients of the parameters are cleared to zero to ensure that gradient updates for each batch are independent. Gradient calculation is controlled, with backpropagation and gradient updates only performed in training mode. A learning rate scheduler is used to dynamically adjust the learning rate during training to improve training effectiveness. The accuracy of each epoch is calculated and stored by comparing the model's predictions with the true labels to evaluate model performance. A deep neural network is used to perform four-class classification on the GSM-R service quality problem dataset, labeling the problem images to establish a GSM-R service quality problem sample library.

[0099] It should be noted that although the operation of the method of the present invention has been described in a specific order in the above embodiments and figures, this does not require or imply that the operations must be performed in that specific order, or that all the operations shown must be performed to achieve the desired result. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step, and / or one step may be broken down into multiple steps.

[0100] After introducing the method of exemplary embodiments of the present invention, the following references are made. Figure 17 An exemplary embodiment of the present invention, a neural network-based GSM-R service quality analysis and processing apparatus, is described.

[0101] The implementation of the GSM-R service quality analysis and processing device based on neural networks can refer to the implementation of the above method, and repeated details will not be elaborated further. The terms "module" or "unit" used below can refer to a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0102] Based on the same inventive concept, this invention also proposes a GSM-R service quality analysis and processing device based on neural networks, such as... Figure 17 As shown, the device includes:

[0103] Data acquisition module 1710 is used to acquire GSM-R waveform data;

[0104] The service quality analysis module 1720 is used to input the GSM-R waveform data into the service quality classification and identification model of the GSM-R system, and to obtain the service quality problems of the GSM-R system under the input waveform data conditions through analysis; wherein, the service quality classification and identification model adopts the ResNet architecture and is pre-trained in the following manner:

[0105] For the GSM-R waveform data sample set, the service quality problem categories are labeled, and the labeled sample set is divided into training set, validation set and test set;

[0106] The artificial intelligence model is trained using the training set. In conjunction with the categories of service quality problems, the artificial intelligence model is mapped from GSM-R waveform data to service quality problem categories. The model is then validated and tested using the validation set and test set to obtain a service quality classification and recognition model. The service quality classification and recognition model takes GSM-R waveform data as input and the categories of service quality problems as output.

[0107] The display module 1730 is used to display the GSM-R waveform data obtained through the railway facility monitoring platform and the service quality issues obtained after analysis.

[0108] In one embodiment, the categories of service quality issues include at least: dropped calls, handover failures, handover anomalies, and poor voice quality.

[0109] In one embodiment, for service quality issues related to dropped calls, if the test interface indicates a connection loss and there is continuous voice quality reaching a set level, then the corresponding location is determined to be a dropped call; if the test interface indicates a connection loss and there is no voice quality issue, the historical data of the line is replayed. If the same phenomenon occurs at the same location, then the corresponding location is determined to be a dropped call; if the same phenomenon does not occur, then the test system is determined to be abnormal.

[0110] For service quality issues related to failed switching, when the test interface indicates a failed switching, determine the location where the switching failed.

[0111] For service quality issues caused by handover anomalies, a handover anomaly is defined as one that occurs when no handover takes place between two base stations or when multiple handovers occur.

[0112] For service quality issues with poor voice quality, if the voice quality on the test interface is within the set level range and no dropped calls occur, it is judged as poor voice quality.

[0113] In one embodiment, reference Figure 18 The device also includes: a preprocessing module 1740;

[0114] The preprocessing module 1740 is used to perform image preprocessing on GSM-R waveform data. Specifically, it performs random image cropping on the GSM-R waveform data and adjusts the size of the cropped image to a set pixel specification. During training, it horizontally flips the image with a set probability. It also performs image normalization processing, converting the image's pixel values ​​to a distribution with a mean of 0 and a standard deviation of 1.

[0115] In one embodiment, the service quality classification and identification model adopts the ResNet architecture as ResNet18, which includes 16 convolutional layers and 2 fully connected layers;

[0116] Convolutional layers are used to extract multi-level feature representations from the input image, while fully connected layers are used for classification or regression based on the extracted features.

[0117] ResNet18 incorporates residual blocks, each consisting of two convolutional layers and one skip connection. The skip connection allows information to be passed directly from one layer to the next in the network.

[0118] In one embodiment, reference Figure 18 The device also includes: a case library module 1750;

[0119] The case library module 1750 is used to create a case library of service quality issues, which stores different categories of service quality issues, as well as descriptions, causes, and solutions for the service quality issues.

[0120] Once service quality issues are identified through analysis, the description, cause, and solution for the service quality issues are queried from the aforementioned problem case database and provided to the operations and maintenance personnel.

[0121] The service quality analysis module 1720 is also used to update the defect case library based on the service quality problems output by the service quality classification and identification model, and to use it as sample data for iterative training of the service quality classification and identification model to continuously learn and train the model.

[0122] It should be noted that although several modules of the GSM-R service quality analysis and processing device based on neural networks have been mentioned in the detailed description above, this division is merely exemplary and not mandatory. In fact, according to embodiments of the present invention, the features and functions of two or more modules described above can be embodied in one module. Conversely, the features and functions of one module described above can be further divided and embodied by multiple modules.

[0123] Based on the aforementioned inventive concept, such as Figure 19 As shown, the present invention also proposes a computer device 1900, including a memory 1910, a processor 1920, and a computer program 1930 stored in the memory 1910 and executable on the processor 1920. When the processor 1920 executes the computer program 1930, it implements the aforementioned neural network-based GSM-R service quality analysis and processing method.

[0124] Based on the aforementioned inventive concept, this invention proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned neural network-based GSM-R quality of service analysis and processing method.

[0125] Based on the aforementioned inventive concept, this invention proposes a computer program product, which includes a computer program that, when executed by a processor, implements a neural network-based GSM-R quality of service analysis and processing method.

[0126] This invention proposes a neural network-based GSM-R service quality analysis and processing method and apparatus. Based on the analysis of a large number of GSM-R service quality problem images, it identifies the abnormal patterns in these images, determines their anomalous characteristics, and constructs an optimal algorithm based on these characteristics. The relevant parameters and thresholds in this invention were determined through extensive experimentation, selecting the best-performing results. Compared to existing technologies, it offers at least the following technical advantages: It uses a dataset generalization method based on transfer learning to extract features from GSM-R service quality problem images, enriching the multi-source heterogeneous data samples for data analysts; and it assists inspectors in achieving automated GSM-R service quality problem identification through a ResNet-based GSM-R image recognition algorithm, improving detection efficiency. By enriching the GSM-R service quality problem sample library, it provides valuable experimental materials for data analysts and developers, offering strong technical support for the identification of various communication problems.

[0127] The acquisition, storage, use, and processing of data in this application comply with relevant laws and regulations.

[0128] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, apparatus, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0129] This invention is described with reference to flowchart illustrations and / or block diagrams of methods and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0130] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0131] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0132] Finally, it should be noted that the above-described embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit it. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the technical scope disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, 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 covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A neural network-based GSM-R service quality analysis processing method, characterized in that, include: Acquire GSM-R waveform data; The GSM-R waveform data is input into the service quality classification and identification model of the GSM-R system. Analysis yields the service quality issues corresponding to the GSM-R system under the input waveform data conditions. The service quality classification and identification model employs a ResNet architecture and is pre-trained as follows: For the GSM-R waveform data sample set, the service quality problem categories are labeled, and the labeled sample set is divided into training set, validation set and test set; The artificial intelligence model is trained using the training set. In conjunction with the categories of service quality problems, the artificial intelligence model is mapped from GSM-R waveform data to service quality problem categories. The model is then validated and tested using the validation set and test set to obtain a service quality classification and recognition model. The service quality classification and recognition model takes GSM-R waveform data as input and the categories of service quality problems as output. The service quality issues are categorized into at least the following: dropped calls, handover failures, handover anomalies, and poor voice quality. For dropped calls, if the test interface indicates connection loss and there is continuous voice quality reaching a set level, then the corresponding location is considered a dropped call. If the test interface indicates connection loss but there is no poor voice quality, historical line data is replayed. If the same phenomenon occurs at the same location, then the corresponding location is considered a dropped call. If the same phenomenon does not occur, then the test system is considered abnormal. For handover failures, if the test interface indicates handover failure, then the corresponding location is considered a handover failure. For handover anomalies, if no handover occurs between two base stations or multiple handovers occur, then it is considered a handover anomaly. For poor voice quality, if the test interface shows voice quality within a set level range and no dropped calls occur, then it is considered poor voice quality. The railway facility monitoring platform displays the acquired GSM-R waveform data and the service quality issues identified through analysis.

2. The GSM-R service quality analysis and processing method according to claim 1, characterized in that, The method also includes: Image preprocessing is performed on GSM-R waveform data; specifically, random image cropping is performed on the GSM-R waveform data, and the size of the cropped image is adjusted to a set pixel specification; during training, the image is horizontally flipped with a set probability; and the image is normalized to convert the pixel values ​​of the image to a distribution with a mean of 0 and a standard deviation of 1.

3. The GSM-R service quality analysis and processing method according to claim 1, characterized in that, The service quality classification and identification model adopts the ResNet architecture as ResNet18, which includes 16 convolutional layers and 2 fully connected layers. Convolutional layers are used to extract multi-level feature representations from the input image, while fully connected layers are used for classification or regression based on the extracted features. ResNet18 incorporates residual blocks, each consisting of two convolutional layers and one skip connection. The skip connection allows information to be passed directly from one layer to the next in the network.

4. The GSM-R service quality analysis and processing method according to claim 1, characterized in that, The method also includes: Create a service quality case library, which stores different categories of service quality issues, along with their descriptions, causes, and solutions. Once service quality issues are identified through analysis, the description, cause, and solution for the service quality issues are queried from the aforementioned problem case database and provided to the operations and maintenance personnel. The service quality issues output by the service quality classification and identification model are used to update the disease case database and serve as sample data for iterative training of the service quality classification and identification model, enabling continuous learning and training of the model.

5. A GSM-R service quality analysis and processing device based on neural networks, characterized in that, include: The data acquisition module is used to acquire GSM-R waveform data; The service quality analysis module is used to input the GSM-R waveform data into the service quality classification and identification model of the GSM-R system, and to obtain the service quality problems corresponding to the GSM-R system under the input waveform data conditions through analysis; wherein, the service quality classification and identification model adopts the ResNet architecture and is pre-trained in the following manner: For the GSM-R waveform data sample set, the service quality problem categories are labeled, and the labeled sample set is divided into training set, validation set and test set; The artificial intelligence model is trained using the training set. In conjunction with the categories of service quality problems, the artificial intelligence model is mapped from GSM-R waveform data to service quality problem categories. The model is then validated and tested using the validation set and test set to obtain a service quality classification and recognition model. The service quality classification and recognition model takes GSM-R waveform data as input and the categories of service quality problems as output. The service quality issues are categorized into at least the following: dropped calls, handover failures, handover anomalies, and poor voice quality. For dropped calls, if the test interface indicates connection loss and there is continuous voice quality reaching a set level, then the corresponding location is considered a dropped call. If the test interface indicates connection loss but there is no poor voice quality, historical line data is replayed. If the same phenomenon occurs at the same location, then the corresponding location is considered a dropped call. If the same phenomenon does not occur, then the test system is considered abnormal. For handover failures, if the test interface indicates handover failure, then the corresponding location is considered a handover failure. For handover anomalies, if no handover occurs between two base stations or multiple handovers occur, then it is considered a handover anomaly. For poor voice quality, if the test interface shows voice quality within a set level range and no dropped calls occur, then it is considered poor voice quality. The display module is used to display the GSM-R waveform data obtained through the railway facility monitoring platform and the service quality issues identified after analysis.

6. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method of any one of claims 1 to 4.

8. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the method of any one of claims 1 to 4.

Images