Intelligent monitoring methods for urban rail transit industrial control systems

Abstract

This invention relates to the field of urban rail transit technology, specifically to an intelligent monitoring method for an industrial control system of urban rail transit. The method includes the following steps: Step 1: Based on the control requirements of urban rail transit, corresponding sensors are matched to each key device of the control system to collect data parameters in real time and transmit them via network communication technology to generate raw data acquisition reports; Step 2: Based on the load indicators of each channel in network communication, load thresholds for each channel are preset according to historical load and flow, the channel status is checked, and the flow allocation is dynamically adjusted through channel status feedback to generate a data transmission scheme. In this invention, the performance change trend of simulated equipment can be compared with the threshold to provide early warning, allowing for more sufficient processing time, thereby minimizing the threat posed by potential hazards. Furthermore, by marking and analyzing frequently malfunctioning equipment, a more timely response can be achieved.

Description

Technical Field

[0001] This invention relates to the field of urban rail transit technology, and in particular to an intelligent monitoring method for urban rail transit industrial control systems. Background Technology

[0002] The urban rail transit industrial control system refers to the comprehensive management and control of various aspects such as tracks, trains, signals, power supply, and stations during the operation of urban rail transit. With the rapid development of intelligent technology, the urban rail transit industrial control system is now increasingly being monitored in real time through digitalization and automation to ensure its long-term, efficient, and stable operation.

[0003] However, due to the relatively large scale of urban rail transit and the involvement of multiple safety factors such as speed, location, and pressure, there is a relatively large amount of real-time monitoring data to ensure the normal operation of rail transit. This may cause the transmission channels for the relevant monitoring data to operate under high load, resulting in some abnormal data not being transmitted to the data terminal in a timely manner. This may further have a certain impact on urban rail transit. In addition, some areas of urban rail transit may frequently experience malfunctions due to pressure, heat dissipation, and compression. However, in the early stages of a malfunction, maintenance personnel may fail to detect it in time because the malfunction is not obvious, which may delay the resolution of the malfunction and increase unnecessary trouble and costs. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of existing technologies and propose an intelligent monitoring method for urban rail transit industrial control systems.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: an intelligent monitoring method for an urban rail transit industrial control system, comprising the following steps:

[0006] Step 1: Based on the control requirements of urban rail transit, match the corresponding sensors to each key device of the control system, collect data parameters in real time, and transmit them through network communication technology to generate raw data collection reports.

[0007] Step 2: Based on the load indicators of each channel in network communication, preset the load threshold of each channel according to historical load and traffic, check the channel status, dynamically adjust the traffic allocation through channel status feedback, complete the transmission of the original data acquisition report, and generate a data transmission scheme.

[0008] Step 3: Based on the data transmission scheme, the raw data is stored in the cloud and preprocessed, including data standardization and normalization, extraction of key features, and combined with historical data. A fault diagnosis algorithm is constructed using machine learning to screen abnormal data, identify fault types, and generate a diagnostic prediction report by analyzing and predicting the equipment performance degradation trend through time series analysis.

[0009] Step 4: Based on the diagnostic prediction report, match the corresponding time predicted by the equipment performance degradation, combine the equipment performance redundancy and degradation hazards, calculate the emergency weight ratio and prioritize them, establish a multi-level early warning mechanism, and generate a fault warning report.

[0010] Step 5: Based on the fault warning report, integrate the known information of each device to be processed one by one according to priority, retrieve historical data, count the number of times the device is processed according to the set time period, mark the frequently processed devices, adjust the allocation of monitoring and emergency resources, and generate a dispatch report for devices to be processed.

[0011] Step Six: Based on the pending dispatch report, use a drone carrying maintenance assistance tools to perform multi-dimensional scanning and verification, and simultaneously display the status of the corresponding equipment in real time through AR glasses to generate a visual assistance report;

[0012] Step 7: Based on the aforementioned visualization auxiliary report, conduct periodic feedback to evaluate monitoring efficiency and accuracy, optimize the monitoring algorithm and model based on the feedback information, and generate monitoring management record reports.

[0013] As a further aspect of the present invention, the original data acquisition report includes temperature, vibration, current, voltage, pressure, speed, position, smoke, and acceleration. By monitoring the equipment operating status and environmental conditions in real time, it provides basic reference data, wherein temperature and pressure are used to detect the thermodynamic state of the system, vibration and acceleration are used to reflect the health status of the mechanical system, current and voltage are used to analyze the working efficiency of electrical equipment, speed and position are used to dynamically track the movement of the equipment, and smoke concentration is used for fire safety judgment.

[0014] As a further aspect of the present invention, the load metrics include channel throughput, latency, and queue length. Channel throughput measures the amount of data successfully transmitted per unit time to determine network transmission capacity. Latency reflects the time required for data to travel from the source to the destination to determine user experience and real-time application response speed. Queue length reflects the number of data packets waiting to be processed in the network device's buffer to determine the probability of network congestion and packet loss.

[0015] As a further aspect of the present invention, the specific steps for generating the data transmission scheme are as follows:

[0016] Based on the load index, the load threshold of each channel is determined one by one, and channels that are congested, have packet loss, or have increased latency are identified and marked to generate a channel health table.

[0017] Based on the channel health table, the health status of each channel is compared, low-load channels are filtered in real time, and a candidate channel table is generated.

[0018] Based on the alternative channel table, the data to be transmitted in the original data acquisition report is dynamically split to generate a data transmission scheme.

[0019] As a further aspect of the present invention, the specific steps of the preprocessing are as follows:

[0020] Based on the original data, data with excessive missing values ​​are deleted, and missing data are filled with the mean, median, and most common values. Simultaneously, interpolation is used to fill in continuous data, and duplicate data is removed to generate a cleaned data set.

[0021] Based on the data cleanup set, different types of data columns are logically checked, and data that does not conform to the format is converted to generate a standard dataset.

[0022] Based on a standard dataset, different data of the same type are integrated into a unified framework through inner and outer joins to complete data preprocessing.

[0023] As a further aspect of the present invention, the specific steps for generating the diagnostic prediction report are as follows:

[0024] Based on the historical data, a neural network is used for simulation to analyze the correlation between various historical monitoring data in the event and to establish a fault algorithm.

[0025] Based on the fault identification algorithm, the known feature information and known fault types are matched one by one, and the unknown fault type information is marked to generate a fault identification report.

[0026] Based on the fault identification report, an LSTM neural network is used in conjunction with fault feature information to predict the performance change trend and cross-interference direction of the corresponding equipment without remedial measures, and a diagnostic prediction report is generated.

[0027] As a further aspect of the present invention, the early warning mechanism includes pop-ups, emails, text messages, telephone calls, and alarm bells.

[0028] As a further aspect of the present invention, the monitoring and emergency resources include monitoring sensors and emergency repair personnel, and relevant periodic inspection and maintenance plans are formulated according to the risk causes of the faulty equipment, while emergency resources of corresponding levels are allocated according to the risk level of the faulty equipment.

[0029] As a further aspect of the present invention, the specific steps for the AR glasses to display the corresponding device status in real time are as follows:

[0030] Real-time data transmission is performed based on the MQTT protocol, and known information is displayed through virtual labels, charts, and icons.

[0031] Information is linked to corresponding devices based on spatial anchor points, and the current data of the devices is displayed through floating information boxes;

[0032] Based on drone scanning and sensor detection, real-time updated data is pushed to AR glasses via WebSocket technology to update the displayed content and verify the processing results of maintenance personnel.

[0033] As a further aspect of the present invention, the optimized monitoring algorithm and model evaluate the performance of the model through evaluation indicators, perform difference analysis between the model prediction results and the actual results, and periodically train the optimal model through hyperparameter optimization and apply it.

[0034] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0035] 1. In this invention, by combining historical data with real-time collected monitoring data, the performance change trend of each monitoring device can be simulated. At the same time, the simulation results are compared with the threshold to provide early warning when the fault is not obvious, allowing more sufficient processing time, thereby minimizing the threat posed by hidden dangers. Furthermore, by marking and analyzing equipment with frequent faults, the response speed can be enhanced, reducing the probability of rail transit being affected.

[0036] 2. In this invention, low-load channels can be screened out by real-time monitoring and analysis of wireless network channels, while high-load channels can be marked. This avoids further increasing the transmission pressure on high-load channels and allows for the real-time transmission of raw data monitored by various sensors in the urban rail transit industrial control system by flexibly selecting appropriate channels. Furthermore, by continuously refreshing the current data, the efficiency of discovering potential faults can be improved, preventing them from continuing to develop and causing dangerous events.

[0037] 3. In this invention, drones can be used to further scan and verify some faulty equipment before maintenance personnel go to the equipment to be processed, so as to verify the prediction results of the algorithm and model and avoid wasting maintenance resources. At the same time, AR glasses can guide maintenance personnel and display various known information. Then, after the maintenance is completed, the maintenance results can be verified by refreshing the data information collected by the sensors in real time, forming an efficient and accurate intelligent maintenance closed loop. Attached Figure Description

[0038] Figure 1 This is a schematic diagram of the main steps of the present invention;

[0039] Figure 2 This is a schematic diagram of the data transmission scheme steps of the present invention;

[0040] Figure 3 This is a schematic diagram illustrating the steps of the visualization-assisted reporting method of the present invention. Detailed Implementation

[0041] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0042] Please see Figure 1 - Appendix Figure 3 This invention provides a technical solution: an intelligent monitoring method for an industrial control system of urban rail transit, comprising the following steps:

[0043] Step 1: Based on the control requirements of urban rail transit, match corresponding sensors to each key device of the control system, collect data parameters in real time, and transmit them through network communication technology to generate raw data collection reports. Use various types of sensors to monitor each key device on the urban rail line in real time to continuously collect the raw parameters of the monitored objects, thereby providing a reference for subsequent judgment on whether there are any hidden dangers.

[0044] Step 2: Based on the load indicators of each channel in network communication, preset the load threshold for each channel according to historical load and traffic, check the channel status, dynamically adjust traffic allocation through channel status feedback, complete the transmission of raw data collection reports, and generate a data transmission plan. Based on historical load and traffic, collect time series data through protocol analysis tools to predict future traffic and load. Then, calculate the typical load level of each channel by analyzing traffic fluctuations. Furthermore, by comprehensively considering channel bandwidth, channel quality, transmission rate, channel capacity, and network load and traffic changes, the load threshold of each channel can be set to improve resource utilization and transmission performance. At the same time, by comparing the load indicators of each channel when transmitting raw data, each channel can be divided into high load and low load. When a certain load exceeds the set threshold, the average load of each channel is calculated using a rolling time window, and then the low load channel is given priority for data transmission to actively adjust traffic allocation. This ensures that the large amount of raw data collected by each sensor can be sent to the monitoring center for processing and analysis in a timely manner, reducing the probability that some sudden abnormal data will not be detected in time due to channel congestion.

[0045] Step 3: Based on the data transmission scheme, the raw data is stored in the cloud and preprocessed, including data standardization and normalization, extraction of key features, and the use of machine learning to construct a fault diagnosis algorithm by combining historical data. This process filters abnormal data, identifies fault types, and generates a diagnostic prediction report by analyzing and predicting equipment performance degradation trends through time series analysis. Cloud storage of raw data improves data security and ease of subsequent retrieval. Furthermore, preprocessing the raw data reduces the error rate and ensures consistency of similar data. This facilitates the extraction of key abnormal features such as chassis temperature rise, increased track amplitude, or increased current from the raw data through Fourier transform or time-domain analysis. Further, machine learning can be used to build various types of fault algorithms from historical data. After incorporating the currently filtered key feature information, the fault type is determined. Then, an LSTM neural network is used to predict the performance change trend of each device with potential faults in the future, assessing the direction of performance degradation. This achieves the goal of preventing serious harm from further development through early prediction.

[0046] Step 4: Based on the diagnostic prediction report, match the predicted time of equipment performance degradation, combine equipment performance redundancy and degradation hazards, calculate the urgency weight ratio and prioritize them, establish a multi-level early warning mechanism, generate fault warning reports, and match the predicted development trend with the corresponding time points according to the statistically identified potential faulty equipment. This allows maintenance personnel to handle the equipment before the danger develops further. At the same time, based on the remaining performance redundancy value of the equipment with potential hazards and the degree of harm that may be caused by its degradation, different types of equipment can be standardized or coded. Then, through weighted comprehensive calculation, the actual score of each equipment to be handled can be obtained. At this time, by ranking the scores, all equipment with potential hazards can be divided into different urgency levels, ensuring that maintenance personnel can prioritize the handling of relatively critical equipment with potential hazards.

[0047] Step 5: Based on the fault warning report, integrate the known information of each device to be processed one by one according to priority, retrieve historical data, count the number of times the device is processed according to the set time period, mark the frequently processed devices, adjust the allocation of monitoring and emergency resources, and generate a dispatch report for the devices to be processed. By integrating the spatial coordinates, development curves, remaining development time and abnormal data of each device to be processed, it is easier for subsequent maintenance personnel to go to the corresponding location, understand the remaining resolution time and the problems to be solved. By statistically analyzing the devices that frequently fail, maintenance personnel or dedicated monitoring sensors can be arranged near each device to improve the response speed when a fault occurs and avoid maintenance personnel wasting time by frequently going to the location of the faulty device.

[0048] Step Six: Based on the pending dispatch report, a drone carrying maintenance auxiliary tools is used to perform multi-dimensional scanning and verification. Simultaneously, AR glasses display the corresponding equipment status in real time, generating a visual auxiliary report. Depending on the specific dispatch task, after communication with the patrol maintenance personnel, the on-duty maintenance personnel operate the drone carrying necessary maintenance tools to the area where the faulty equipment is located. The drone's high-definition camera, thermal imager, and LiDAR scanner then perform a comprehensive, three-dimensional scan, capturing potential problems such as abnormal equipment temperature, vibration, and mechanical wear. This verifies some faults and, to a certain extent, avoids delays caused by diagnostic algorithm errors. Simultaneously, the AR glasses worn by the patrol personnel overlay drone scan data, sensor data, fault causes, coordinate distances, and equipment shape information onto their field of vision, providing precise maintenance guidance and improving work efficiency while reducing human error. After maintenance, sensors and the drone collect information related to the results, which is then verified by intelligent algorithms. The AR glasses provide timely feedback on the maintenance effect, forming an efficient and accurate intelligent maintenance closed loop.

[0049] Step 7: Based on the visualization auxiliary report, conduct periodic feedback to evaluate monitoring efficiency and accuracy. Optimize monitoring algorithms and models based on the feedback information, generate monitoring management record reports, and analyze the reliability of various algorithms and models used for monitoring based on the feedback content of AR glasses during maintenance and the actual problems encountered by inspection personnel. Then, based on the analysis results, algorithms or models with defects can be screened out and optimized to ensure the long-term effectiveness of the algorithms and models.

[0050] Please see Figure 1 The raw data acquisition report includes temperature, vibration, current, voltage, pressure, speed, position, smoke, and acceleration. By monitoring the equipment's operating status and environmental conditions in real time, it provides basic reference data. Temperature and pressure are used to detect the thermodynamic state of the system, vibration and acceleration are used to reflect the health status of the mechanical system, current and voltage are used to analyze the working efficiency of electrical equipment, speed and position are used to dynamically track the movement of the equipment, and smoke concentration is used to make fire safety judgments.

[0051] Various types of sensors can be used to monitor and collect data on temperature, vibration, current, voltage, pressure, speed, position, smoke, and acceleration in real time, so as to understand the raw data information related to urban rail transit from multiple dimensions and improve the reliability of subsequent identification of potential hazards in various equipment on the track.

[0052] Please see Figure 1Load metrics include channel throughput, latency, and queue length. Channel throughput measures the amount of data successfully transmitted per unit time, which helps determine network transmission capacity. Latency reflects the time required for data to travel from the source to the destination, which helps determine user experience and real-time application response speed. Queue length reflects the number of data packets waiting to be processed in the network device's buffer, which helps determine the probability of network congestion and packet loss.

[0053] By using Prometheus, Grafana, and other similar tools, one can query actual information such as channel throughput, latency, and queue length. This information can then be used as a reference to determine the actual load of network communication channels, preventing some channels from being overloaded and going unnoticed. Conversely, it can also be used to filter out low-load channels.

[0054] Please see Figure 2 The specific steps for generating a data transmission scheme are as follows:

[0055] Based on load metrics, load thresholds are determined for each channel one by one. Channels experiencing congestion, packet loss, or increased latency are identified and marked, and a channel health table is generated. Based on load metrics, load thresholds are set for each channel one by one, and channels with congestion, packet loss, or increased latency are marked, thus initially identifying them as high-load channels and preventing them from continuing to transmit raw data collected by sensors, which would cause continuous congestion.

[0056] Based on the channel health table, the health status of each channel is compared, low-load channels are filtered in real time, and a candidate channel table is generated. Low-load channels that can be used for traffic offloading are determined through real-time filtering.

[0057] Based on the alternative channel table, the data to be transmitted in the original data acquisition report is dynamically split to generate a data transmission scheme. By actively splitting the data, it can be ensured that each channel can be fully utilized, reducing the probability of data congestion or loss.

[0058] Please see Figure 1 The specific steps of preprocessing are as follows:

[0059] Based on the original data, data with too many missing values ​​are deleted, and missing data are filled with the mean, median, and most common values. Interpolation is used to fill in continuous data, and duplicate data is removed to generate a cleaned data set. By appropriately deleting and filling data, the integrity of the data can be improved.

[0060] Based on the data cleaning set, different types of data columns are logically checked, and data that does not conform to the format is converted to generate a standard dataset. Data conversion can improve data reliability.

[0061] Based on standard datasets, different data of the same type are integrated into a unified framework through inner and outer joins to complete data preprocessing. Data integration can improve the convenience of subsequent data use, and preprocessing can provide a cleaner and more standardized foundation for machine learning and data analysis.

[0062] Please see Figure 1 The specific steps for generating a diagnostic prediction report are as follows:

[0063] Based on historical data, a neural network is used for simulation to analyze the correlation between various historical monitoring data in events, and a fault algorithm is established. By retrieving historical data, the corresponding sensor monitoring data when different faults occur can be obtained. Then, the neural network can be used to analyze each event to obtain the corresponding development curve.

[0064] Based on the fault algorithm, the known feature information and known fault types are matched one by one, and the unknown fault type information is marked to generate a fault identification report. By substituting the known feature information, the fault type can be initially identified, so that areas with possible problems can be screened out from a large amount of data. The information of unknown fault types can be stored and retrieved for future analysis.

[0065] Based on the fault identification report, the system uses an LSTM neural network and combines fault feature information to predict the performance change trend and cross-interference direction of the corresponding equipment without remedial measures, and generates a diagnostic prediction report. By simulating the development trend of the faulty equipment, we can understand the potential harm it may cause and its impact on other equipment, providing a reference for determining its treatment level.

[0066] Please see Figure 1 The early warning mechanism includes pop-ups, emails, text messages, phone calls, and alert bells;

[0067] Pop-ups, emails, text messages, phone calls, and alerts can be used to address various situations ranging from low to high hazard levels, making it easier for maintenance personnel to prioritize and handle potentially faulty equipment with greater risks.

[0068] Please see Figure 1 Monitoring and emergency resources include sensors for monitoring and emergency repair personnel. Regular inspection and maintenance plans are developed based on the risk causes of equipment failure, and emergency resources of corresponding levels are allocated according to the risk level of the equipment failure.

[0069] By developing separate handling plans and inspection cycles for equipment that frequently malfunctions, and assigning dedicated personnel for maintenance, it can be ensured that timely responses and handling are available when malfunctions occur frequently. This prevents maintenance personnel from wasting time by frequently visiting the track inspection area due to frequent malfunctions. Furthermore, adding additional specialized monitoring sensors can further extend the service life of the equipment.

[0070] Please see Figure 3 The specific steps for AR glasses to display the real-time status of the corresponding device are as follows:

[0071] Real-time data transmission is performed based on the MQTT protocol, and known information is displayed through virtual labels, charts, and icons. Different display methods can be selected to clearly present various types of known information.

[0072] Based on spatial anchor points, information is connected to the corresponding equipment, and the current data of the equipment is displayed through floating information boxes. Spatial anchor points make it easier for maintenance personnel to find faulty equipment, and floating information boxes can display data such as temperature and pressure on the surface of the corresponding equipment.

[0073] Based on drone scanning and sensor detection, real-time refreshed data is pushed to AR glasses via WebSocket technology to update the displayed content and verify the processing results of maintenance personnel. By refreshing the displayed content in real time, a closed loop can be formed after the maintenance is completed, so that maintenance personnel can determine the maintenance effect.

[0074] Please see Figure 1 The monitoring algorithm and model are optimized by evaluating the performance of the model through evaluation metrics and analyzing the difference between the model's prediction results and the actual results. The optimal model is trained regularly through hyperparameter optimization and then applied. Since the monitoring environment is dynamic, the algorithm and model need to be updated regularly through hyperparameter optimization training to avoid performance degradation caused by environmental changes.

[0075] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. An intelligent monitoring method for an industrial control system of urban rail transit, characterized in that, Includes the following steps: Step 1: Based on the control requirements of urban rail transit, match the corresponding sensors to each key device of the control system, collect data parameters in real time, and transmit them through network communication technology to generate raw data collection reports. Step Two: Based on the load indicators of each channel in network communication, preset the load threshold for each channel according to historical load and traffic, check the channel status, dynamically adjust traffic allocation through channel status feedback, complete the transmission of the original data acquisition report, and generate a data transmission scheme. The specific steps for generating the data transmission scheme are as follows: Based on the load index, the load threshold of each channel is determined one by one, and channels that are congested, have packet loss, or have increased latency are identified and marked to generate a channel health table. Based on the channel health table, the health status of each channel is compared, low-load channels are filtered in real time, and a candidate channel table is generated. Based on the alternative channel table, the data to be transmitted in the original data acquisition report is dynamically split to generate a data transmission scheme. Step 3: Based on the data transmission scheme, the raw data is stored in the cloud and preprocessed, including data standardization and normalization, extraction of key features, and the use of machine learning to build a fault diagnosis algorithm in combination with historical data. Abnormal data is screened, fault types are identified, and a diagnostic prediction report is generated by time series analysis and prediction of equipment performance degradation trends. Step 4: Based on the diagnostic prediction report, match the corresponding time predicted by the equipment performance degradation, combine the equipment performance redundancy and degradation hazards, calculate the emergency weight ratio and prioritize them, establish a multi-level early warning mechanism, and generate a fault warning report. Step 5: Based on the fault warning report, integrate the known information of each device to be processed one by one according to priority, retrieve historical data, count the number of times the device is processed according to the set time period, mark the frequently processed devices, adjust the allocation of monitoring and emergency resources, and generate a dispatch report for devices to be processed. Step Six: Based on the pending dispatch report, use a drone carrying maintenance assistance tools to perform multi-dimensional scanning and verification, and simultaneously display the status of the corresponding equipment in real time through AR glasses to generate a visual assistance report; Step 7: Based on the aforementioned visualization auxiliary report, conduct periodic feedback to evaluate monitoring efficiency and accuracy, optimize the monitoring algorithm and model based on the feedback information, and generate monitoring management record reports.

2. The intelligent monitoring method for urban rail transit industrial control systems according to claim 1, characterized in that, The raw data acquisition report includes temperature, vibration, current, voltage, pressure, speed, position, smoke, and acceleration. By monitoring the equipment's operating status and environmental conditions in real time, it provides basic reference data. Temperature and pressure are used to detect the system's thermodynamic state, vibration and acceleration are used to reflect the health status of the mechanical system, current and voltage are used to analyze the working efficiency of electrical equipment, speed and position are used to dynamically track equipment movement, and smoke concentration is used for fire safety assessment.

3. The intelligent monitoring method for urban rail transit industrial control systems according to claim 1, characterized in that, The load metrics include channel throughput, latency, and queue length. Channel throughput measures the amount of data successfully transmitted per unit time, which helps determine network transmission capacity. Latency reflects the time required for data to travel from the source to the destination, which helps determine user experience and real-time application response speed. Queue length reflects the number of data packets waiting to be processed in the network device's buffer, which helps determine the probability of network congestion and packet loss.

4. The intelligent monitoring method for urban rail transit industrial control systems according to claim 1, characterized in that, The specific steps of the preprocessing are as follows: Based on the original data, data with excessive missing values ​​are deleted, and missing data are filled with the mean, median, and most common values. Simultaneously, interpolation is used to fill in continuous data, and duplicate data is removed to generate a cleaned data set. Based on the data cleanup set, different types of data columns are logically checked, and data that does not conform to the format is converted to generate a standard dataset. Based on a standard dataset, different data of the same type are integrated into a unified framework through inner and outer joins to complete data preprocessing.

5. The intelligent monitoring method for urban rail transit industrial control systems according to claim 1, characterized in that, The specific steps for generating the diagnostic prediction report are as follows: Based on the historical data, a neural network is used for simulation to analyze the correlation between various historical monitoring data in the event and to establish a fault algorithm. Based on the fault identification algorithm, the known feature information and known fault types are matched one by one, and the unknown fault type information is marked to generate a fault identification report. Based on the fault identification report, an LSTM neural network is used in conjunction with fault feature information to predict the performance change trend and cross-interference direction of the corresponding equipment without remedial measures, and a diagnostic prediction report is generated.

6. The intelligent monitoring method for urban rail transit industrial control systems according to claim 1, characterized in that, The early warning mechanism includes pop-ups, emails, text messages, phone calls, and alert bells.

7. The intelligent monitoring method for urban rail transit industrial control systems according to claim 1, characterized in that, The monitoring and emergency resources include monitoring sensors and emergency repair personnel. Relevant regular inspection and maintenance plans are developed based on the risk causes of the faulty equipment, and emergency resources of corresponding levels are allocated according to the risk level of the faulty equipment.

8. The intelligent monitoring method for urban rail transit industrial control systems according to claim 1, characterized in that, The specific steps for the AR glasses to display the real-time status of the corresponding device are as follows: Real-time data transmission is performed based on the MQTT protocol, and known information is displayed through virtual labels, charts, and icons. Information is linked to corresponding devices based on spatial anchor points, and the current data of the devices is displayed through floating information boxes; Based on drone scanning and sensor detection, real-time updated data is pushed to AR glasses via WebSocket technology to update the displayed content and verify the processing results of maintenance personnel.

9. The intelligent monitoring method for urban rail transit industrial control systems according to claim 1, characterized in that, The optimized monitoring algorithm and model evaluate the model's performance using evaluation metrics, analyze the difference between the model's predictions and actual results, and periodically train the optimal model through hyperparameter optimization and apply it.

Images