Zero-speed out-of-gear protection system under network fault of turnout grinding wagon

Abstract

The invention relates to the technical field of turnout grinding wagon operation, in particular to a zero-speed out-of-gear protection system under a network fault of a turnout grinding wagon, which comprises a network state monitoring module, a speed monitoring module, a gear monitoring module, a central control module and an out-of-gear execution module, wherein the network state monitoring module detects faults through multi-modal fusion and deep learning; speed detection is combined with laser and traditional sensing, and Kalman filtering is used; gear monitoring depends on the Hall effect and image recognition; central control is carried out to reinforce learning of comprehensive multi-factor decisions; the out-of-gear execution is driven by magnetorheological fluid and is monitored by an MEMS (micro-electromechanical system); the AR technology is used for alarming and displaying, and voice interaction is achieved. The system has the advantages that the network, the speed and the gear are accurately monitored, and the out-of-gear is intelligently decided; off-gear execution is reliable, and alarm display is visual. The system has the functions of data secure storage, remote efficient monitoring and fault prediction, and the security under a network fault is comprehensively improved.

Description

Technical Field

[0001] This invention relates to the field of turnout grinding car operation technology, and in particular to a zero-speed disengagement protection system for a turnout grinding car under network fault. Background Technology

[0002] In railway maintenance, turnout grinding machines are key equipment for ensuring the safe and stable operation of turnouts. With the development of railway transportation towards high speed and heavy load, the intelligence and automation of turnout grinding machines are constantly improving. Their network systems bear the heavy responsibility of transmitting various control commands, sensor data, and coordinating the collaborative work of various components. However, the railway site environment is complex and changeable; electromagnetic interference, severe weather, and equipment aging can all easily lead to network failures in turnout grinding machines.

[0003] When a network failure occurs, the vehicle's control system may be unable to accurately obtain critical information such as speed and gear position, and may also struggle to achieve precise control over various actuators. In such a situation, if the vehicle is in operation, especially in the switch area, an operational error could lead to a serious safety accident. This could not only cause significant damage to equipment but also disrupt the normal operation of the railway, resulting in train delays or even cancellations and substantial economic losses to railway transportation.

[0004] Existing turnout grinding machines have significant shortcomings in handling network faults. Traditional protection mechanisms are often too simplistic, mostly only issuing alarms when network anomalies are detected, lacking a comprehensive assessment of the vehicle's actual operating status and effective countermeasures. For example, when a network fault causes the loss or error of speed signals, it cannot accurately determine whether the vehicle is truly at zero speed, potentially leading to erroneous disengagement while the vehicle still has some speed, damaging the transmission system; or, even when the vehicle is indeed at zero speed and requires disengagement protection, the disengagement may fail to be executed due to untimely or inaccurate judgment, increasing safety risks. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of the prior art by providing a zero-speed gear disengagement protection system for a turnout grinding car under network faults. This system accurately judges the vehicle status and promptly implements zero-speed gear disengagement protection, which is of urgent practical significance for improving the operational safety and reliability of turnout grinding cars and ensuring the efficient and stable operation of railway transportation.

[0006] The objective of this invention is achieved through the following technical solutions: A zero-speed gear disengagement protection system for a turnout grinding car under network fault, characterized in that it includes a network status monitoring module, a speed monitoring module, a gear monitoring module, a central control module, and a gear disengagement execution module; The network status monitoring module adopts multi-module fusion monitoring technology, collects network sensor monitoring signal parameters, introduces deep learning algorithms to analyze network traffic characteristics, constructs a network traffic model through convolutional neural networks, and transmits the results of the network traffic model to the central control module. The speed monitoring module combines the vehicle's speed relative to the ground and calculates the vehicle speed by measuring the wheel rotation speed. It then uses a Kalman filter algorithm to fuse and process the data before transmitting it to the central control module. The gear position monitoring module monitors the gear position based on the Hall effect principle, a gear position sensor, and intelligent image recognition technology. The gear position sensor detects changes in the magnetic field to determine the gear position, the image recognition algorithm identifies the gear position identifier, and when an abnormal change in the gear position is detected, it transmits an early warning information to the central control module. The central control module uses reinforcement learning algorithm to make a disengagement decision based on the monitoring data from the network status monitoring module, the speed monitoring module, and the gear position monitoring module. It constructs a disengagement decision model by inputting the comprehensive network fault index, speed, gear position information, vehicle load coefficient, and environmental factors. Through trial and error learning, it outputs a disengagement decision factor. When the factor exceeds the adaptive threshold, a disengagement command is issued. The gear disengagement execution module performs gear disengagement under the control of the central control module.

[0007] It also includes an alarm and display module, which uses augmented reality (AR) technology to display alarm information. The AR glasses project vehicle network status, speed, and gear information, and a three-dimensional warning icon is used to alarm when a network failure or gear disengagement occurs.

[0008] It also includes a data storage module, which uses distributed blockchain storage technology to encrypt and store collected data, decision information and execution status, and compresses and stores data in real time according to timestamps and vehicle ID indexes using data compression algorithms.

[0009] It also includes a remote monitoring module, which is based on 5G communication technology combined with edge computing. It deploys edge computing devices on the vehicle to process and analyze raw data, obtains key vehicle information in real time through the 5G network, and uses digital twin technology to build a virtual vehicle model to simulate and predict vehicle status. The remote monitoring module is connected to a remote fault diagnosis expert system and provides remote fault troubleshooting guidance by interacting with vehicle data.

[0010] When determining network faults, the network status monitoring module introduces the network node importance index I and the fault propagation probability matrix P to calculate the weighted network fault comprehensive index N, using the following formula: Where w N0 As the initial weight, Δw N The adjustment coefficient is m, where m is the number of affected nodes.

[0011] The speed monitoring module includes a laser Doppler velocimeter and a speed sensor. The laser Doppler velocimeter measures the vehicle's speed relative to the ground, and the speed sensor is connected to the wheels to measure wheel rotation speed and convert it into vehicle speed. An adaptive weighted fusion algorithm is used to dynamically adjust the fusion weights based on vehicle operating conditions and sensor measurement accuracy. The algorithm is implemented using a formula... Calculate the data weight w of the laser Doppler velocimeter L , where σ T σ L These are the standard deviations of the measurement errors of the speed sensor and the laser Doppler velocimeter, respectively.

[0012] The image recognition algorithm of the gear monitoring module adopts transfer learning technology, uses a convolutional neural network model to fine-tune the gear image of the turnout grinding machine, and combines semantic segmentation technology to identify the position and status of the gear mark.

[0013] The formula for calculating and outputting the derailment decision factor F by the central control module is as follows: , where w i For the weights of each factor, f i The reward function is defined as the influence function of the corresponding factors. The reinforcement learning process introduces an adaptive adjustment mechanism for the reward function, dynamically adjusting its parameters according to different vehicle operating scenarios. The reward function is R = α × (1 + β × Td), where α and β are adjustment coefficients, and Td is the sum of the values ​​of α and β. d This is the ratio of the disengagement operation time to the preset optimal time.

[0014] The disengagement execution module adopts a magnetorheological fluid driven disengagement mechanism, which uses viscosity changes to control the current of the electromagnetic coil to adjust the disengagement force and speed. The magnetorheological fluid driven disengagement mechanism is equipped with a microelectromechanical system (MEMS) accelerometer to monitor mechanical vibration, and feeds back to the central control module when abnormalities occur. The magnetorheological fluid driven disengagement mechanism is equipped with an intelligent lubrication system and an intelligent cooling system.

[0015] It also includes an intelligent fault prediction module, which uses a deep neural network long short-term memory network (LSTM) model to learn and analyze historical monitoring data to predict network faults and equipment faults. When the predicted fault probability exceeds a threshold, it issues an early warning message.

[0016] The advantages of this invention are: accurate monitoring of network, speed, and gear position, intelligent decision-making for disengagement; reliable disengagement execution, intuitive alarm display; and the ability to securely store data, conduct remote and efficient monitoring, and predict faults, thus comprehensively improving security under network failure conditions. Attached Figure Description

[0017] Figure 1 This is a schematic block diagram illustrating the structure of the present invention; Figure 2 This is a schematic diagram comparing the network fault diagnosis accuracy of the present invention with that of existing technologies; Figure 3 This is a schematic diagram illustrating the change in speed measurement accuracy over time between the present invention and existing technologies; Figure 4 This is a schematic diagram comparing the response time of the disengagement operation of the present invention with that of the prior art. Detailed Implementation

[0018] The features and other related features of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments, so as to facilitate understanding by those skilled in the art: Example: Figures 1 to 4 As shown, the zero-speed disengagement protection system under network fault of the turnout grinding car in this embodiment is used to monitor the network, speed and gear of the turnout grinding car during operation and make intelligent decisions to disengage, so as to ensure the reliability of disengagement execution.

[0019] Specifically, the zero-speed gear slippage protection system in this embodiment includes: Network Status Monitoring Module: Employing multimodal fusion monitoring technology, this module not only utilizes traditional network sensors to detect parameters such as signal strength, packet loss rate, and latency, but also incorporates deep learning algorithms to analyze network traffic characteristics. It constructs a network traffic model using a convolutional neural network (CNN) and compares the actual traffic with the normal traffic model in real time. When the signal strength falls below an adaptive threshold S... min (t)(S min (t) dynamically adjusted based on historical data and real-time operating conditions), or the packet loss rate is higher than the threshold P derived from Bayesian inference. max (x) (x is a set of parameters related to network topology and service load), or the latency exceeds the threshold T calculated based on the queuing theory model. max (y) (where y is a set of parameters such as the number of network nodes and link bandwidth), and the CNN model determines that a network failure has occurred when it detects abnormal traffic. This module collects and analyzes data at a frequency of 5 times per second, and transmits the results to the central control module in real time via a high-speed data bus.

[0020] In this embodiment, the network status monitoring module considers the importance of network nodes and the fault propagation model when judging network faults. It introduces a network node importance index I (calculated using algorithms such as betweenness centrality in complex network analysis) and a fault propagation probability matrix P. When a fault occurs at a critical node and, according to the fault propagation model, may trigger a large-scale network paralysis, the central control module significantly increases the weight of the network fault comprehensive index N when calculating the failure decision factor F. , where w N0 As the initial weight, Δw NThe adjustment coefficient is m, where m is the number of affected nodes.

[0021] Speed ​​Detection Module: Utilizes a redundant detection method combining laser Doppler velocimetry technology and speed sensors. The laser Doppler velocimeter directly measures the vehicle's speed relative to the ground, unaffected by factors such as wheel slippage, achieving an accuracy of ±0.01 km / h. Speed ​​sensors are mounted on the wheel axles, calculating vehicle speed by measuring wheel rotation speed. A Kalman filter algorithm is used to fuse the data from both speed measurement methods, updating vehicle speed data in real time to obtain highly accurate vehicle speed data. The speed detection module updates the speed data every 0.05 seconds and sends it to the central control module.

[0022] In this embodiment, the speed detection module employs an adaptive weighted fusion algorithm during the data fusion process between the laser Doppler velocimeter and the traditional speed sensor. The fusion weights of the two sensor data are dynamically adjusted based on the vehicle's operating conditions (acceleration, deceleration, constant speed) and the real-time measurement accuracy of the sensors. For example, during acceleration or deceleration, when the laser Doppler velocimeter's measurement accuracy is higher than that of the traditional speed sensor, the weight of the laser speed measurement data is increased. L The calculation formula is: , where σ T σ L These represent the standard deviations of measurement errors for traditional speed sensors and laser Doppler velocimeters, respectively.

[0023] Gear Position Monitoring Module: Utilizing a high-precision gear position sensor based on the Hall effect principle, combined with intelligent image recognition technology, the module monitors gear position. The sensor detects real-time changes in the magnetic field of the transmission's gear control mechanism to determine the gear. Simultaneously, a camera is installed in the transmission gear position display area, using image recognition algorithms to identify the gear indicator, providing double assurance of gear position information accuracy. This module also features a sudden gear position change detection function; when an abnormal gear position change is detected within a very short time, a warning message is immediately transmitted to the central control module.

[0024] In this embodiment, the image recognition algorithm of the gear monitoring module adopts an advanced image recognition algorithm. This algorithm is based on transfer learning technology and fully utilizes convolutional neural network models pre-trained on massive industrial equipment image datasets. These pre-trained models have undergone training on a large number of diverse industrial images and possess powerful general image feature extraction capabilities, such as keen capture of basic features like edges, textures, and shapes. Considering the specific characteristics of the gear images of the turnout grinding machine, fine-tuning of the pre-trained model is necessary. During the fine-tuning process, researchers carefully collected a large amount of gear image data from the turnout grinding machine under different working conditions, lighting conditions, and angles to construct a specialized dataset. Using the backpropagation algorithm, the parameters of the pre-trained model are iteratively updated based on this dataset, enabling it to better adapt to the feature distribution of the turnout grinding machine gear images. Simultaneously, semantic segmentation technology is deeply integrated. This technology, through pixel-level classification, can accurately divide the region of the gear indicator in the image and clarify its specific location. Through detailed classification of the pixels of each part of the gear indicator, not only can its location be determined, but its state can also be accurately identified, such as gear shifting and wear conditions. For example, semantic segmentation technology can identify subtle wear marks on gear position markings based on changes in pixel features. By organically combining transfer learning and semantic segmentation technology, the gear position monitoring module significantly improves the accuracy and robustness of gear position recognition, effectively avoiding misjudgments caused by factors such as image interference and lighting changes, and providing a solid foundation for reliable operation of the zero-speed gear disengagement protection system of the turnout grinding car.

[0025] The central control module receives data from various monitoring modules and uses reinforcement learning algorithms to make gear disengagement decisions. A gear disengagement decision model is constructed, with inputs including the network fault comprehensive index N (derived from multiple parameters of the network status monitoring module via analytic hierarchy process), the fused vehicle speed V, gear information G, vehicle load coefficient L (calculated based on grinding motor current, vehicle load, etc.), and environmental factors E. Through continuous trial and error learning, the model outputs a gear disengagement decision factor F, calculated using the following formula: , where w i For the weights of each factor, f i The influence function of the corresponding factors is continuously optimized through reinforcement learning. When F exceeds the adaptive threshold F... max When (z) (z is a set of parameters such as vehicle historical operation data and current task type), the central control module issues a disengagement command.

[0026] In this embodiment, the central control module introduces an adaptive adjustment mechanism for the reward function during reinforcement learning. The parameters of the reward function are dynamically adjusted based on different vehicle operating scenarios (such as different line types and operational tasks). For example, during operations on busy railway mainlines, a higher reward is given for quick and accurate gear shifting, with the reward function R = α × (1 + β × T).d ), where α and β are adjustment coefficients, and T d This is the ratio of the disengagement operation time to the preset optimal time.

[0027] The gear disengagement execution module employs a rapid gear disengagement mechanism driven by magnetorheological fluid. This mechanism utilizes the rapid viscosity change of the magnetorheological fluid under a magnetic field to precisely adjust the disengagement force and speed by controlling the current of the electromagnetic coil. Simultaneously, it is equipped with a microelectromechanical system (MEMS)-based accelerometer to monitor mechanical vibrations during the disengagement process in real time. When abnormal vibrations are detected, feedback is immediately sent to the central control module for timely adjustment of the disengagement strategy. The gear disengagement execution module also features automatic lubrication, using an intelligent lubrication system to automatically add lubricant based on the disengagement frequency and operating time, reducing mechanical wear.

[0028] In this embodiment, the magnetorheological fluid drive mechanism in the disengagement execution module is equipped with an intelligent cooling system. This intelligent cooling system possesses real-time monitoring and control capabilities. The system incorporates high-precision sensors that can accurately capture the magnitude of the electromagnetic coil current and the real-time temperature of the magnetorheological fluid. Changes in the electromagnetic coil current directly affect the working state of the magnetorheological fluid; excessively high or low temperatures will adversely affect its performance. Therefore, accurately acquiring these two key parameters is crucial. Based on the acquired real-time data, the system employs an advanced fuzzy control algorithm for intelligent control. This algorithm uses current deviation, temperature deviation, and the rate of change of these deviations as input quantities. Specifically, the current deviation reflects the difference between the current and the set current value, the temperature deviation reflects the difference between the current temperature of the magnetorheological fluid and the optimal operating temperature, and the rate of change of deviation represents the trend of deviation change over time. By fuzzifying these input quantities and using pre-set fuzzy rules for inference calculations, a precise control quantity is finally output. This control quantity is used to adjust the cooling fan speed and coolant flow rate in real time. When the electromagnetic coil current increases or the magnetorheological fluid temperature rises, the fuzzy control algorithm adjusts the cooling fan speed and coolant flow rate accordingly to enhance heat dissipation. Conversely, when the current and temperature decrease, the cooling fan speed and coolant flow rate are reduced to prevent over-cooling and energy waste. This intelligent control method ensures the magnetorheological fluid operates within a suitable temperature range, effectively reducing performance degradation and component wear caused by temperature anomalies. This significantly improves the reliability and service life of the disengagement mechanism and guarantees the stable operation of the zero-speed disengagement protection system on the turnout grinding car.

[0029] Alarm and Display Module: Utilizes Augmented Reality (AR) technology for alarm and information display. In the operator's work area, AR glasses project information such as the vehicle's network status, speed, and gear position. When a network failure or gear slippage occurs, a prominent 3D warning icon will trigger an alarm. The alarm and display module also supports voice interaction, allowing operators to query detailed fault information and operating instructions via voice commands.

[0030] In this embodiment, the AR display content of the alarm and display module has a personalized customization function. Operators can fine-tune the AR display content based on their unique operating habits and actual work needs. The AR settings interface provides a series of intuitive and easy-to-use parameter adjustment options. Regarding information display position, operators can place important fault prompts, vehicle status parameters, and other information in the most easily observable area of ​​their field of vision through precise coordinate settings or visual drag-and-drop operations. For example, operators who prefer to browse information from left to right can fix key warning information on the left side of the AR field of vision. Regarding information display size, adjustable scaling sliders or inputting specific values ​​allow for flexible adjustments from tiny prompt icons to large data panels. For personnel with poor eyesight or working under specific lighting conditions, the displayed content can be appropriately enlarged to ensure clear readability. Color parameter adjustment provides a rich color wheel selection and custom RGB value input function. Operators can assign striking colors according to different information categories and urgency levels. For example, a high-contrast red can be set for critical fault alarms, and yellow for general prompts, facilitating quick differentiation and response. The alarm and display module fully considers the usage needs of operators in different regions around the world, supporting multilingual switching. The system has built-in character libraries and language packs for multiple mainstream languages, including but not limited to English, Chinese, Spanish, and French. Operators simply need to click the corresponding language option in the settings interface, and the text information in the AR display, such as fault descriptions and operation guides, will instantly switch to the selected language, effectively eliminating language barriers and improving the system's international versatility and ease of use. This ensures that operators in different regions can efficiently and accurately obtain and understand the information provided by the system.

[0031] In this embodiment, in addition to the functions described above, the following modules are also included. By using these modules, the functionality and performance of the zero-speed gear slippage protection system in this embodiment can be further improved: Data storage module: Utilizing distributed blockchain storage technology based on a P2P network architecture, all nodes are equal in status. Real-time data collected by various monitoring modules, such as speed, pressure, and temperature, as well as precise decision-making information from the central control module based on complex algorithms, and execution status such as the action status and response time of the off-grid execution module, are all encrypted using asymmetric encryption algorithms and then distributed across multiple nodes. During data storage, an indexing mechanism is established based on precise timestamps and unique vehicle IDs. The timestamps are accurate to the millisecond level, and the vehicle IDs use globally unique identification codes. This not only ensures that the data is difficult to illegally tamper with, but also allows for precise tracing of the data source and flow process in case of problems. Furthermore, the module integrates a highly efficient data compression algorithm, using a combination of the LZ77 algorithm and Huffman coding to compress stored data in real time. Without affecting data integrity and accuracy, this significantly reduces storage space usage, improves storage efficiency, and enables the efficient and secure storage and management of massive amounts of data, providing a solid guarantee for the stable operation of the turnout grinding machine and subsequent data analysis.

[0032] The remote monitoring module deeply integrates 5G communication and edge computing technologies. The edge computing devices deployed on the vehicle utilize high-performance chips with powerful computing capabilities. Their built-in advanced algorithms can quickly and accurately preprocess raw data collected from various sensors, such as temperature, pressure, and vibration frequency, across multiple dimensions. Through feature extraction and data filtering, redundant information is eliminated, retaining only key data, significantly reducing data transmission volume and making efficient use of limited network bandwidth. The 5G network, with its peak speeds of tens of Gbps and ultra-low latency down to milliseconds, establishes a high-speed information channel between the remote monitoring center and the vehicle. The remote monitoring center can obtain key vehicle operating information in real time through this channel. Using digital twin technology, a highly realistic virtual model of the physical turnout grinding machine is constructed through detailed 3D modeling and real-time data mapping. This model not only reflects the vehicle's external appearance in real time but also accurately simulates the operating conditions of various internal components, achieving comprehensive, real-time simulation and forward-looking prediction of the vehicle's status. Furthermore, the remote monitoring module is equipped with a remote fault diagnosis expert system, which is an intelligent fault handling system. It possesses a large and constantly updated database of fault cases, covering fault scenarios under various complex operating conditions. When a vehicle malfunctions, the system quickly interacts with the vehicle, using deep learning algorithms and intelligent reasoning mechanisms to rapidly locate the root cause of the fault and provide operators with detailed and targeted remote troubleshooting guidance, effectively ensuring the stable operation of the turnout grinding machine.

[0033] In this embodiment, the system possesses an intelligent fault prediction function. This function relies on the Long Short-Term Memory (LSTM) network model in deep neural networks, which has unique advantages in processing time-series data. The system first comprehensively collects and integrates historical monitoring data from the turnout grinding machine. This data encompasses rich information obtained from various sensors, such as network-related data like bandwidth, latency, and packet loss rate, as well as equipment status data like temperature, vibration frequency, and pressure during equipment operation. This data is carefully organized into a standardized time-series dataset in chronological order. During training, the LSTM model, through its unique memory unit structure, can automatically learn long-term dependencies and short-term trends in the data. The input gate, forget gate, and output gate in the memory unit work together to precisely control the inflow, retention, and outflow of information. For example, when analyzing network faults, the model remembers the network bandwidth change patterns over a period of time. When it detects a recent continuous decline in bandwidth approaching a historical fault threshold, the forget gate selectively discards some irrelevant early information, the input gate introduces current real-time data for comprehensive analysis, and the output gate outputs a prediction of the future network state based on this information. For equipment fault prediction, the model deeply analyzes the key operating parameters of the equipment. Taking equipment temperature as an example, the LSTM model not only focuses on the current temperature value but also learns features such as the rate of temperature change over time and the fluctuation period. When the model detects an abnormal trend in temperature changes by learning from historical data and predicts that the probability of a fault exceeds a pre-set threshold, it quickly triggers an early warning mechanism, sending detailed warning information to the operator, including the possible location of the fault, the type of fault, and the approximate probability of the fault occurring. Through this intelligent fault prediction method, the zero-speed derailment protection system under network faults of the turnout grinding car can anticipate potential risks in advance, giving operators sufficient time to take preventative maintenance measures, effectively reducing the possibility of faults and ensuring the reliable operation of the turnout grinding car.

[0034] Although the above embodiments have described the concept and embodiments of the present invention in detail with reference to the accompanying drawings, those skilled in the art will recognize that various improvements and modifications can still be made to the present invention without departing from the scope of the claims, and therefore will not be elaborated here.

Claims

1. A zero-speed gear slippage protection system under network fault of a turnout grinding car, characterized in that: It includes a network status monitoring module, a speed monitoring module, a gear position monitoring module, a central control module, and a gear disengagement execution module; The network status monitoring module adopts multi-module fusion monitoring technology, collects network sensor monitoring signal parameters, introduces deep learning algorithms to analyze network traffic characteristics, constructs a network traffic model through convolutional neural networks, and transmits the results of the network traffic model to the central control module. The speed monitoring module combines the vehicle's speed relative to the ground and calculates the vehicle speed by measuring the wheel rotation speed. It then uses a Kalman filter algorithm to fuse and process the data before transmitting it to the central control module. The gear position monitoring module monitors the gear position based on the Hall effect principle, a gear position sensor, and intelligent image recognition technology. The gear position sensor detects changes in the magnetic field to determine the gear position, the image recognition algorithm identifies the gear position identifier, and when an abnormal change in the gear position is detected, it transmits an early warning information to the central control module. The central control module uses reinforcement learning algorithm to make a disengagement decision based on the monitoring data from the network status monitoring module, the speed monitoring module, and the gear position monitoring module. It constructs a disengagement decision model by inputting the comprehensive network fault index, speed, gear position information, vehicle load coefficient, and environmental factors. Through trial and error learning, it outputs a disengagement decision factor. When the factor exceeds the adaptive threshold, a disengagement command is issued. The gear disengagement execution module performs gear disengagement under the control of the central control module.

2. The zero-speed disengagement protection system under network fault of a turnout grinding car according to claim 1, characterized in that: It also includes an alarm and display module, which uses augmented reality (AR) technology to display alarm information. The AR glasses project vehicle network status, speed, and gear information, and a three-dimensional warning icon is used to alarm when a network failure or gear disengagement occurs.

3. The zero-speed disengagement protection system under network fault of a turnout grinding car according to claim 1, characterized in that: It also includes a data storage module, which uses distributed blockchain storage technology to encrypt and store collected data, decision information and execution status, and compresses and stores data in real time according to timestamps and vehicle ID indexes using data compression algorithms.

4. The zero-speed disengagement protection system under network fault of a turnout grinding car according to claim 1, characterized in that: It also includes a remote monitoring module, which is based on 5G communication technology combined with edge computing. It deploys edge computing devices on the vehicle to process and analyze raw data, obtains key vehicle information in real time through the 5G network, and uses digital twin technology to build a virtual vehicle model to simulate and predict vehicle status. The remote monitoring module is connected to a remote fault diagnosis expert system and provides remote fault troubleshooting guidance by interacting with vehicle data.

5. The zero-speed disengagement protection system under network fault of a turnout grinding car according to claim 1, characterized in that: When determining network faults, the network status monitoring module introduces the network node importance index I and the fault propagation probability matrix P to calculate the weighted network fault comprehensive index N, using the following formula: Where w N0 As the initial weight, Δw N The adjustment coefficient is m, where m is the number of affected nodes.

6. The zero-speed disengagement protection system under network fault of a turnout grinding car according to claim 1, characterized in that: The speed monitoring module includes a laser Doppler velocimeter and a speed sensor. The laser Doppler velocimeter measures the vehicle's speed relative to the ground, and the speed sensor is connected to the wheels to measure the wheel rotation speed and convert it into vehicle speed. An adaptive weighted fusion algorithm is used to dynamically adjust the fusion weights according to the vehicle's operating conditions and the sensor's measurement accuracy. Through formula Calculate the data weight w of the laser Doppler velocimeter L , where σ T σ L These are the standard deviations of the measurement errors of the speed sensor and the laser Doppler velocimeter, respectively.

7. The zero-speed disengagement protection system under network fault of a turnout grinding car according to claim 1, characterized in that: The image recognition algorithm of the gear monitoring module adopts transfer learning technology, uses a convolutional neural network model to fine-tune the gear image of the turnout grinding machine, and combines semantic segmentation technology to identify the position and status of the gear mark.

8. The zero-speed disengagement protection system under network fault of a turnout grinding car according to claim 1, characterized in that: The formula for calculating and outputting the derailment decision factor F by the central control module is as follows: , where w i For the weights of each factor, f i The reward function is defined as the influence function of the corresponding factors. The reinforcement learning process introduces an adaptive adjustment mechanism for the reward function, dynamically adjusting its parameters according to different vehicle operating scenarios. The reward function is R = α × (1 + β × Td), where α and β are adjustment coefficients, and Td is the sum of the values ​​of α and β. d This is the ratio of the disengagement operation time to the preset optimal time.

9. The zero-speed disengagement protection system under network fault of a turnout grinding car according to claim 1, characterized in that: The disengagement execution module adopts a magnetorheological fluid driven disengagement mechanism, which uses viscosity changes to control the current of the electromagnetic coil to adjust the disengagement force and speed. The magnetorheological fluid driven disengagement mechanism is equipped with a microelectromechanical system (MEMS) accelerometer to monitor mechanical vibration, and feeds back to the central control module when abnormalities occur. The magnetorheological fluid driven disengagement mechanism is equipped with an intelligent lubrication system and an intelligent cooling system.

10. The zero-speed disengagement protection system under network fault of a turnout grinding car according to claim 1, characterized in that: It also includes an intelligent fault prediction module, which uses a deep neural network long short-term memory network (LSTM) model to learn and analyze historical monitoring data to predict network faults and equipment faults. When the predicted fault probability exceeds a threshold, it issues an early warning message.

Images