Rail transit signal equipment health state intelligent monitoring terminal and system

Abstract

The application discloses a kind of track traffic signal equipment health state intelligent monitoring terminal and system, belong to the equipment state monitoring and security technology field of railway signal system, the signal equipment monitored is the track circuit, signal machine, switch machine, transponder and other signal equipment arranged along railway;The terminal includes state acquisition module and data processing module, and the voltage, current, temperature and other multidimensional operating parameters of signal equipment are collected.Data processing module is pre-set with the multi-parameter joint discrimination model specially for track traffic signal equipment, to establish benchmark eigenvector and covariance matrix with normal operation historical data, compare with threshold value to determine abnormal or failure risk, and identify fault type through deviation direction combination;The application solves the technical problems of high false alarm and missed alarm rate of single-parameter fixed threshold method, inability to identify coupled faults and lack of predictive ability, and realizes accurate discrimination and early warning of signal equipment health state.

Description

Technical Field

[0001] This invention relates to the field of railway signaling system and equipment status monitoring technology, and in particular to an intelligent monitoring terminal and system for the health status of rail transit signaling equipment applicable to railway sections, stations and entrances; the monitored signaling equipment includes railway signaling equipment such as track circuits, signals, switch machines, and transponders, and their operating status is directly related to train operation safety and traffic control efficiency. Background Technology

[0002] Traditional rail transit signaling equipment, including track circuits, signals, switch machines, and transponders, primarily employs a single-parameter fixed threshold comparison method for health status monitoring. This method sets fixed upper and lower limits for operating parameters such as voltage, current, and temperature. When a parameter collected in real-time exceeds the corresponding threshold range, an abnormality or malfunction is determined, triggering an alarm. This method is currently the most widely used and lowest-barrier-to-implement monitoring technique in the industry, and it is also the basic judgment method adopted in some current industry technical standards.

[0003] However, this method has fundamental flaws in engineering practice, mainly in the following aspects:

[0004] First, fixed thresholds cannot distinguish between normal dynamic fluctuations and true performance degradation. The operating parameters of rail transit signaling equipment are not constant values, but are affected by multiple factors, including ambient temperature changes (with seasonal temperature differences exceeding 60°C), power supply line voltage drop fluctuations, and changes in train operating conditions. Even under normal service conditions, significant dynamic changes occur. Applying a static, one-size-fits-all threshold to dynamically changing operating scenarios with fixed thresholds inevitably leads to the following: when the threshold is set too wide, obvious performance degradation may occur but be missed because the threshold has not been reached; when the threshold is set too narrow, frequent fluctuations in normal operating conditions trigger false alarms, overwhelming on-site maintenance personnel and making it easy to overlook true signs of accidents.

[0005] Secondly, single-parameter independent discrimination cannot identify fault modes coupled with multiple parameters. Many actual faults in signal equipment are not simple events of a single parameter "exceeding the standard," but rather complex processes involving the coordinated changes of multiple parameters. For example, when the filter capacitors in a power module age, the output voltage may slowly decrease, the ripple coefficient may increase significantly, and the equipment casing temperature may rise slightly. The individual changes in each of these three parameters may not exceed their respective fixed thresholds, but the coordinated changes of these three parameters clearly point to a fault mode caused by capacitor aging. Traditional single-parameter independent discrimination cannot capture this combined anomaly of multi-dimensional information, leading to the complete omission of early fault symptoms.

[0006] Third, there is a lack of ability to predict degradation trends based on historical data. Existing methods only perform a binary judgment of "whether the limit is exceeded" on the data collected at the current moment, which is a typical "post-event alarm" mode. It does not involve tracking the equipment performance degradation process or predicting the future state. When the alarm is triggered, the equipment is often already in a state of serious failure or near failure. The time window left for maintenance personnel to carry out preventive measures is extremely limited. Maintenance work is forced to remain in a passive mode of "failure occurs - post-event repair" for a long time, which not only increases the risk of safety accidents, but also causes a lot of unnecessary emergency repair costs during "window periods" and wasteful periodic blind replacement of equipment.

[0007] In summary, existing monitoring methods based on single-parameter fixed threshold comparisons are severely inadequate in terms of accuracy, early fault identification capability, and predictive maintenance support, and can no longer meet the urgent needs of current rail transit for equipment safety and intelligent operation and maintenance. Summary of the Invention

[0008] In view of this, the purpose of this invention is to provide an intelligent monitoring terminal for the health status of signaling equipment that can integrate multi-dimensional operating parameters for joint status discrimination and effectively predict equipment degradation trends. This intelligent monitoring terminal, as a condition monitoring device along the railway signaling system, serves to ensure the safe operation of signaling equipment such as track circuits, signals, switch machines, and transponders.

[0009] To achieve the above objectives, the present invention provides the following technical solution:

[0010] A smart monitoring terminal for the health status of rail transit signaling equipment, comprising:

[0011] A status acquisition module is connected to the railway signaling equipment to be monitored, the railway signaling equipment including track circuit equipment, signal lights, switch machines or transponders; the status acquisition module is used to acquire the operating status parameters of the signaling equipment, including at least voltage parameters, current parameters and temperature parameters;

[0012] The data processing module is electrically connected to the status acquisition module and is used to process and analyze the operating status parameters to generate equipment health status data.

[0013] The data processing module is pre-configured with a multi-parameter joint discrimination model for the health status of signal equipment. This model establishes a baseline feature vector based on multi-dimensional parameters collected during normal equipment operation. In real-time monitoring, the current multidimensional feature vector X is calculated and compared with the reference feature vector. The Mahalanobis distance D between them is used to characterize the overall deviation of the equipment:

[0014] ;

[0015] in Let the covariance matrix be the covariance matrix between the parameters. Calculated based on at least 168 hours of historical equipment operation data; when the Mahalanobis distance D exceeds a preset first threshold. When a device is deemed to be malfunctioning, the Mahalanobis distance D exceeds a preset second threshold. The system determines when there is a risk of equipment failure and triggers an early warning. Simultaneously, the model analyzes the parameters of each dimension in the current feature vector X relative to the baseline feature vector. The deviation direction combination is matched with a preset deviation direction feature library of multiple typical fault modes to identify the fault type;

[0016] The data processing module also includes a preset algorithm for predicting equipment health degradation trends, defining equipment health H as:

[0017] ;

[0018] in The algorithm calculates the health status sequence based on a preset scaling factor according to the device type. Time series analysis is performed, and a least squares support vector regression model is used to predict the health status at future times. The training and prediction of the model are completed locally in the data processing module. When the predicted health status falls within a preset time window... The risk level drops to the set threshold. The following conditions trigger an early warning signal; the prediction model introduces a priori constraint that signal equipment degradation is monotonically non-increasing, and when the model's predicted value is higher than the current value, it is forcibly corrected to the current value, and prediction results that exceed a certain proportion of the known average service life of the equipment type are truncated.

[0019] The main control module is connected to the status acquisition module and the data processing module respectively, and is used to control the coordinated operation of each module.

[0020] Through the above technical solution, this invention constructs a multi-parameter joint discrimination model. By utilizing Mahalanobis distance to simultaneously consider the correlation between various parameters, it can accurately distinguish between fluctuations in normal operating conditions and actual performance degradation, overcoming the high false alarm and false negative rates of the single-parameter fixed threshold method. Furthermore, by combining health degradation trend prediction, it achieves a leap from "post-event alarm" to "pre-event warning." This solution is specifically designed for the fault mechanisms and operating characteristics of rail transit signaling equipment.

[0021] Preferably, the intelligent monitoring terminal further includes a CT induction power extraction module, which is installed near the power cable or contact network of the rail transit line. This module obtains electrical energy through electromagnetic induction and supplies power to various functional modules. The CT induction power extraction module includes an openable and closable power extraction core, an induction coil, a rectifier and filter unit, a voltage regulator output unit, and an energy storage unit. This design solves the on-site power supply problem, enabling the terminal to operate continuously for extended periods to accumulate sufficient historical data for model training and trend prediction.

[0022] Preferably, the intelligent monitoring terminal further includes a self-organizing network communication module, which adopts LoRa or LTE Mesh self-organizing network communication protocols to establish wireless self-organizing network communication connections with adjacent monitoring terminals or data aggregation gateways in environments without public network signals, thereby realizing multi-hop relay transmission of monitoring data. In the self-organizing network communication module, each terminal stores a neighbor node information table along the route, and routing prioritizes forwarding along the route extension direction. When a downstream node failure is detected, it automatically backtracks to the upstream node and triggers local route repair. This configuration ensures that monitoring data can be transmitted back to the operation and maintenance center in real time, providing data support for cloud verification and global data analysis.

[0023] Preferably, the intelligent monitoring terminal further includes a positioning module, which acquires the geographical location information of the intelligent monitoring terminal and associates it with the health status data of the device to achieve accurate location of the fault.

[0024] A smart health status monitoring system for rail transit signaling equipment, comprising:

[0025] Multiple intelligent monitoring terminals, as described above, serve as status monitoring devices along the railway signaling system. They are distributed and installed in sections, stations, and entrances of the rail transit line, and are connected to the railway signaling equipment at the corresponding locations.

[0026] The data aggregation gateway is communicatively connected to each of the aforementioned intelligent monitoring terminals, receives device health status data uploaded by each terminal, and uploads the data to the cloud management platform via wired or wireless means.

[0027] The cloud management platform communicates with the data aggregation gateway, receives, stores, and processes the device health status data, performs fault diagnosis and health assessment, and generates early warning information based on the assessment results; the cloud management platform is configured to retrieve historical data of any terminal for at least 30 days for early warning review to confirm whether the early warning triggered locally by the terminal is valid.

[0028] The early warning push module is connected to the cloud management platform and pushes the early warning information to the preset operation and maintenance terminal;

[0029] A handheld maintenance terminal is connected to the intelligent monitoring terminal via near-field communication. It is used to configure parameters, read data, and issue start commands for benchmark feature vector acquisition of the intelligent monitoring terminal on-site.

[0030] A method for intelligent monitoring of the health status of rail transit signaling equipment includes the following steps:

[0031] S1: The status acquisition module acquires the operating status parameters of the railway signaling equipment to be monitored at a set acquisition frequency. The railway signaling equipment includes track circuit equipment, signals, switch machines, or transponders. The operating status parameters include at least voltage, current, and temperature parameters. S2: The data processing module processes and analyzes the operating status parameters, and generates equipment health status data using a multi-parameter joint discrimination model. The model calculates the current multi-dimensional feature vector X and the baseline feature vector X. Mahalanobis distance between To characterize the overall deviation of the device, when the Mahalanobis distance D exceeds a preset first threshold. When the device is detected as malfunctioning and exceeds a preset second threshold, an anomaly is identified. The system will detect the presence of a fault risk and trigger an early warning. Simultaneously, fault type identification is performed through the combination of deviation directions of parameters in various dimensions; S3: The data processing module performs equipment health degradation trend prediction and defines equipment health. ,in To calculate the health status sequence based on the preset scaling coefficients according to the device type. The least squares support vector regression model is used to predict the health status at future times. The predicted health status will be within a preset time window. The risk level drops to the set threshold. The following steps trigger an early warning signal: S4: The device health status data and early warning signal are transmitted to the data aggregation gateway and / or cloud management platform; S5: The cloud management platform reviews and performs in-depth analysis on the device health status data, and generates early warning information and pushes it to the preset operation and maintenance terminal when an abnormality is confirmed.

[0032] Through the above method, the present invention realizes a complete closed-loop process from data acquisition, multi-parameter joint discrimination, degradation trend prediction to cloud-based review and early warning, effectively solving the fundamental problems of low discrimination accuracy and lack of predictive ability of the single-parameter fixed threshold method.

[0033] Compared with the prior art, the present invention has the following advantages: The present invention constructs a multi-parameter joint discrimination model for rail transit signaling equipment, using Mahalanobis distance. This invention comprehensively assesses the overall equipment status, fully considering the correlation between various parameters. It effectively distinguishes between normal coordinated changes caused by factors such as ambient temperature and power supply fluctuations and abnormal changes caused by equipment performance degradation, fundamentally overcoming the high false alarm and false negative rates of the single-parameter fixed threshold method. By analyzing the deviation directions of parameters in various dimensions and matching them with a preset fault mode feature library, this invention can capture early fault signs indicated by the coordinated changes of multiple parameters, such as voltage slowdown, increased ripple, and slight temperature rise caused by capacitor aging, even when no single parameter exceeds its limit. This achieves early detection of coupled fault modes. This invention defines health status... The least squares support vector regression model is used to predict degradation trends. A prior constraint of monotonically non-increasing degradation of signal equipment is introduced, enabling maintenance personnel to identify risks and arrange preventative maintenance before equipment failure actually occurs, thus achieving a leap from "reactive maintenance" to "predictive maintenance." The covariance matrix of all model parameters in this invention... , baseline feature vector Scale coefficient Threshold , Risk threshold Prediction window Monotonic non-increasing prior constraints and lifetime proportional truncation are all specifically defined and constrained for the operating characteristics and degradation patterns of rail transit signaling equipment. Attached Figure Description

[0034] Figure 1 This is a structural block diagram of the intelligent health status monitoring terminal for rail transit signaling equipment provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the CT induction power harvesting module provided in an embodiment of the present invention; Figure 3 This is an overall architecture diagram of the intelligent health status monitoring system for rail transit signaling equipment provided in an embodiment of the present invention; Figure 4 This is a schematic diagram illustrating the spatial distance between the baseline feature vector and the real-time feature vector in the multi-parameter joint discrimination model provided in this embodiment of the invention. Figure 5 This is a schematic diagram of the device health degradation trend prediction curve provided in an embodiment of the present invention; Figure 6 This is a flowchart illustrating the intelligent health status monitoring method for rail transit signaling equipment provided in an embodiment of the present invention. Figure 7 A schematic diagram of node deployment and communication routing in a wireless ad hoc network under a railway strip topology; Figure 8 This is a block diagram of the functional modules of the cloud management platform. Detailed Implementation

[0035] 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.

[0036] Example 1: Overall Structure and Work Process of Intelligent Monitoring Terminal

[0037] like Figure 1 As shown in the figure, the intelligent monitoring terminal for the health status of rail transit signaling equipment provided in this embodiment includes a status acquisition module 100, a data processing module 200, a main control module 300, a CT induction power supply module 400, a self-organizing network communication module 500, and a positioning module 600.

[0038] The status acquisition module 100 is connected to the signal equipment to be monitored, taking the track circuit receiver as an example, via a signal cable. This module includes a DC voltage acquisition unit 110, an AC voltage acquisition unit 120, an AC / DC current acquisition unit 130, and a temperature acquisition unit. The DC voltage acquisition unit 110 is connected to the DC power terminal of the signal equipment via a resistor divider network and an isolation amplifier, with a measurement range of 0~300V and an accuracy of ±0.5%. The AC voltage acquisition unit 120 is connected to the AC power terminal of the signal equipment via a voltage transformer and a true RMS conversion circuit, with a measurement range of 0~250VAC and an accuracy of ±1%. The AC / DC current acquisition unit 130 acquires the operating current of the signal equipment using a Hall effect current sensor, with a measurement range of 0~20A and an accuracy of ±1%. The temperature acquisition unit uses an NTC thermistor mounted on the housing of the signal equipment, with a measurement range of -40℃ to +125℃ and an accuracy of ±0.5℃.

[0039] The data processing module 200 uses an STM32L4 series ultra-low-power ARM Cortex-M4 microprocessor with a main frequency of 80MHz, and has built-in 256KB SRAM and 1MB Flash. The data processing module performs 12-bit ADC sampling on the analog signal input from the status acquisition module at a sampling rate of 1kHz. After removing power frequency interference and random noise through a moving average filter (window length 16), it calculates the statistical characteristic values ​​of each parameter to form the current multidimensional feature vector. Where V is the effective value of the operating voltage, I is the effective value of the operating current, R is the voltage ripple coefficient, and T is the surface temperature of the equipment.

[0040] The data processing module 200 has a pre-stored reference feature vector for this device type. Covariance Matrix Benchmark eigenvectors The data was collected continuously for 168 hours after the equipment was installed and running normally. The mean value of each parameter was then obtained after filtering. Covariance matrix It is a 4×4 matrix, and its first... Line 1 Column elements It was also calculated based on the aforementioned 168-hour baseline data.

[0041] In real-time monitoring, after each acquisition of the current feature vector X, the data processing module 200 calculates the Mahalanobis distance:

[0042] ;

[0043] when Time (in this embodiment) Taking three times the standard deviation of the mean Mahalanobis distance in the baseline period (corresponding to approximately 99.7% confidence interval), the device is determined to be abnormal. The terminal records the abnormal event locally and increases the collection frequency for tracking; when Time (in this embodiment) If the standard deviation of the mean Mahalanobis distance in the baseline period is taken, and the equipment is determined to have a risk of failure, an early warning signal is immediately sent out through the self-organizing network communication module.

[0044] Meanwhile, the data processing module 200 calculates the parameters of each dimension in the current feature vector X relative to the baseline feature vector. The deviation direction combination is used to identify the fault type. For track circuit equipment, a deviation direction feature library of four typical fault modes is preset: (a) Power module aging: V is significantly lower ( (a) R is normal, I is normal, T is normal; (b) Load-side leakage current: I is significantly higher ( (c) Filter capacitor deterioration: R significantly increases ( (d) Heat dissipation failure: T significantly increased (V normal, I normal); V, I, and R are all normal. When the deviation direction combination matches one of the above patterns, the fault diagnosis unit outputs the corresponding fault type code.

[0045] The main control module 300 uses the same STM32L4 microprocessor as the data processing module (or a discrete STM32L0 series lower-power chip), and connects to each module via I²C, SPI, and UART buses. The main control module is responsible for task scheduling and power management of the terminal, employing an intermittent working mode: by default, it wakes up each module every 5 minutes, completes a full data acquisition, processing, and transmission cycle, and then enters sleep mode again, keeping the average power consumption of the terminal below 30mW.

[0046] CT inductive power supply module 400 Figure 2 As shown, the system includes an openable and closable power-harvesting core 410, an induction coil 420, a rectifier and filter unit 430, a voltage regulator output unit 440, an energy storage unit 450, and a maximum power point tracking control unit 460. The power-harvesting core 410 is made of nanocrystalline high-permeability magnetic material and is an openable and closable clamp structure composed of two semi-rings, fixed to the power cable. The induction coil 420 has 300 turns. When the primary current is greater than 50A, the power-harvesting module outputs ≥100mW. Besides meeting the immediate power consumption of the terminal, excess energy is stored in the energy storage unit 450. When no train passes and the primary current is zero, the energy storage unit supplies power to the terminal for a duration of ≥45 days.

[0047] The self-organizing network communication module 500 uses the SX1278 LoRa RF chip, operating in the 470MHz band with a transmit power of 20dBm. This module stores a table of upstream and downstream neighbor node information along the route. Routing prioritizes forwarding along the route's extension. When a downstream node failure is detected, it automatically backtracks upstream and triggers local route repair to ensure reliable data transmission.

[0048] The positioning module 600 uses the ATGM336H GPS / BeiDou dual-mode positioning chip to obtain the terminal's latitude and longitude coordinates and associate them with the device's health status data during the initial data report.

[0049] Example 2: Detailed Construction Method of Benchmark Eigenvectors and Covariance Matrix

[0050] This embodiment details the benchmark feature vector. Covariance Matrix The specific construction process is executed during the "benchmark learning phase" after the equipment installation and commissioning are completed.

[0051] For a specific model of track circuit receiving equipment, after the equipment is powered on and operating normally, maintenance personnel issue a "start benchmark learning" command to the monitoring terminal via a handheld maintenance terminal. Once the monitoring terminal enters benchmark learning mode, it continuously collects N=2016 sets of data (corresponding to 168 hours, or 7 days) at a frequency of once every 5 minutes. Each set of data includes voltage V, current I, and ripple coefficient. The four parameters are temperature T. Let the first... Group data .

[0052] The data processing module first performs outlier removal for each dimension: it calculates the mean and standard deviation of each dimension, removes data points that deviate from the mean by more than four times the standard deviation, recalculates the mean and standard deviation for the remaining data, and iterates three times until no new outliers are removed. Finally, the mean of each dimension is obtained.

[0053] ;

[0054] Where M is the number of valid data sets after removing outliers. This forms the baseline feature vector. .

[0055] covariance matrix It is a 4×4 symmetric positive definite matrix with the following elements:

[0056] ;

[0057] in Corresponding to the above In actual calculations, The typical numerical form is as follows (taking a certain type of track circuit equipment as an example):

[0058] ;

[0059] The diagonal elements represent the variance of each parameter, while the off-diagonal elements represent the covariance between the parameters. It can be seen that there is a certain positive correlation between voltage V and temperature T (voltage fluctuations occur due to changes in line voltage drop as ambient temperature increases). This correlation is expressed by the covariance matrix. Capture. In subsequent real-time monitoring, when calculating the Mahalanobis distance D, due to the existence of the covariance matrix, normal coordinated fluctuations of voltage and temperature will not cause D to increase significantly. Only abnormal changes that deviate from the normal coordinated relationship will cause D to increase, which can distinguish between normal dynamic fluctuations and actual performance degradation.

[0060] After benchmark learning is completed, the data processing module sets a threshold based on the Mahalanobis distance statistical distribution of the benchmark period: calculates the Mahalanobis distance of each data set relative to the benchmark eigenvector during the benchmark period. ,Pick ,in and These are the mean and standard deviation of the Mahalanobis distance for the baseline period, respectively.

[0061] Example 3: Detailed Implementation of the Health Degradation Trend Prediction Algorithm

[0062] This embodiment details the definition of equipment health HH and the specific implementation process of the degradation trend prediction algorithm.

[0063] Define the device health H as:

[0064] ;

[0065] Where D is the Mahalanobis distance at the current moment. These are pre-defined scaling factors based on the equipment type. For track circuit equipment, Take 0.15; for switch machine equipment, Take 0.20; for signal equipment, Set the value to 0.12. This mapping ensures that H≈100 when the device is in its initial state (D≈0), and H monotonically decreases as D slowly increases. Set a risk threshold. This means that when the health level is below 60, the equipment is considered to need maintenance.

[0066] After each data collection and calculation, the monitoring terminal saves the most recent N=100 historical health data points (corresponding to approximately 8.3 hours of data, assuming a collection interval of 5 minutes) in the local memory of the data processing module. Based on these 100 data points, a least squares support vector regression (LSSVR) model is used to predict the degradation trend.

[0067] Specifically, constructing a training sample set ; where the input vector Output (Input dimension, corresponding to the 10 most recent historical points). Minutes (sampling interval), P=90 (number of training samples, generated by rolling over the most recent 90 data sets). Radial basis function kernel is used:

[0068] ;

[0069] Among them, the kernel width parameter The value is taken as the mean of the variances of each dimension of the input vector. The LSSVR model solves the following optimization problem:

[0070] ;

[0071] The constraints are ,in This is the implicit mapping corresponding to the kernel function. Let C be the error variable and C be the regularization parameter (50 in this embodiment). The model is obtained by solving a system of linear equations using the Lagrange multiplier method.

[0072] After obtaining the prediction model, the latest health status sequence is used as input to predict the health status values ​​at the next q time points. , , where q satisfies , The preset prediction time window is 90 days (90 days for track circuit equipment, corresponding to q=25920 steps), but in actual prediction, a coarse-grained prediction is performed with a step size of 1440 minutes to reduce the amount of calculation.

[0073] In the prediction process, a prior constraint of monotonically non-increasing signal equipment degradation is introduced: if the predicted value at a certain moment... Then a mandatory order This is because signal equipment does not spontaneously recover its performance during its normal service life. Furthermore, the prediction period is set to no more than 10% of the known average lifespan of this equipment type (the average lifespan of track circuit equipment is 15 years, 10% is 1.5 years). Predictions exceeding this period are truncated and not accepted. When the prediction curve deviates from the risk threshold... When they intersect, record the prediction time corresponding to the intersection point. ,like If this occurs, a warning signal will be triggered.

[0074] In this embodiment, when the LSSVR model runs on the STM32L4 processor, the model size is approximately 48KB, the training time is approximately 180ms, and the prediction time is approximately 8ms, which is fully adapted to the computing power and power consumption limitations of the terminal. Figure 5 This is a schematic diagram of the health degradation trend prediction curve of a certain track circuit equipment in this embodiment from the time it was put into use to the 8th year. It can be seen that the health gradually decreased from the initial 98 to about 65. The prediction curve reached the risk threshold of 60 in the 7th and 8th year, and the warning was issued about 3 months in advance.

[0075] Example 4: Adaptive Working Mode of Terminal under Different Power Supply Conditions

[0076] This embodiment illustrates the adaptive working strategy of the monitoring terminal under two conditions: sufficient and insufficient power supply from the CT sensor.

[0077] When the primary current in the power cable exceeds 50A (typically due to train operation or high traction power supply), the output power of the CT induction power module 400 exceeds 100mW. The main control module 300 determines that the current mode is "sufficient power supply," and the control terminal operates at the normal acquisition frequency (once every 5 minutes). In this mode, the data processing module 200 performs complete multi-parameter joint discrimination and LSSVR model training and prediction after each acquisition, resulting in the highest real-time early warning performance.

[0078] When the primary current in the power cable is below 50A (usually when there are no trains running or during off-peak hours at night), the output power of the power extraction module drops below 50mW. The main control module determines that the current mode is "power supply limited" and automatically executes the following energy-saving strategies: (1) reducing the data acquisition frequency from once every 5 minutes to once every 30 minutes; (2) stopping the periodic heartbeat data reporting of the self-organizing network communication module 500, and only reporting abnormal data when abnormal data is collected. (2) When the number of data packets stored locally reaches the cache limit (50 packets), a batch data transmission is initiated proactively; (3) The training frequency of the LSSVR model in the data processing module is reduced from once every 5 data points to once every 50 data points; (4) The calculation frequency of health trend prediction is reduced from execution after each collection to execution once a day. In power-constrained mode, the average power consumption of the terminal can be reduced to below 8mW, and the energy storage unit (450) can support the terminal to work continuously for more than 60 days.

[0079] When the primary current recovers to above 50A, the main control module automatically switches back to the full power supply mode and uses the power drawn to quickly charge the energy storage unit, storing energy for the next low-power period.

[0080] Example 5: Data Verification and Early Warning Confirmation Mechanism of Cloud Management Platform

[0081] This embodiment illustrates the specific process by which the cloud management platform 30 performs cloud-based review of alerts triggered locally on the terminal.

[0082] After receiving a warning signal reported by a terminal, the data receiving unit 31 of the cloud management platform does not immediately push the final warning to the operation and maintenance personnel, but first enters the "review and confirmation" process. Specifically, the fault diagnosis unit 33 retrieves all historical data of the terminal for the past 30 days from the data storage unit 36 ​​(including daily voltage V, current I, ripple coefficient R, temperature T, Mahalanobis distance D, and health H sequence), re-runs the multi-parameter joint discrimination model in the cloud (using double-precision floating-point arithmetic, which has higher precision than the terminal's single-precision arithmetic), and compares the discrimination results reported by the terminal with the results recalculated in the cloud.

[0083] If the cloud computing result is consistent with the result reported by the terminal (i.e.) If the fault type matches, the warning is confirmed to be valid. The warning generation unit (35) generates structured warning information according to the warning level (general / important / urgent) and pushes it to the preset maintenance personnel through three channels: WeChat, SMS and platform in-site message within 30 seconds via the warning push module (40).

[0084] If the cloud-based calculation results differ from the terminal-reported results (for example, the D value reported by the terminal is artificially high due to communication interference or transient pulses, but the cloud recalculates the 30-day sliding window data), If the event occurs, it is determined to be in the "to be observed" state. The cloud management platform will not push an alert, but will only record the event and mark the terminal as "requiring manual attention" on the system interface. At the same time, it will issue an instruction to the terminal to increase the collection frequency in the following 48 hours (from once every 5 minutes to once every 1 minute) to closely track the event.

[0085] If the warning is confirmed to be valid after three consecutive verifications, the cloud management platform will automatically upgrade the warning level of the terminal by one level (generally to important, important to emergency), and include the device model and fault mode in the network fault statistics database for subsequent optimization of the health assessment model parameters for similar devices.

[0086] Example 6: Field Application and Benchmark Learning Startup of Handheld Maintenance Terminal

[0087] This embodiment describes the specific operation process of starting benchmark learning on the handheld maintenance terminal 50 during the equipment installation and commissioning phase.

[0088] After the intelligent monitoring terminal (10) has completed hardware installation and is powered on and running normally, the maintenance personnel arrive at the site with a handheld maintenance terminal (an industrial-grade rugged tablet with built-in NFC and Bluetooth 5.0). The maintenance personnel open the dedicated configuration APP on the handheld terminal, scan for nearby intelligent monitoring terminals 10 via Bluetooth, and establish a connection. The APP interface displays the current operating status of the terminal (power consumption, storage capacity, firmware version, etc.). After the maintenance personnel verify that everything is correct, they select the type of signal device to be monitored on the APP (in this embodiment, "track circuit device"), and then click the "Start Reference Learning" button.

[0089] The APP sends a baseline learning start command to the monitoring terminal via Bluetooth. The command includes the device type code and the estimated learning duration (168 hours). After receiving the command, the main control module (300) of the monitoring terminal confirms that the energy storage unit has sufficient power (>80%), and then begins to execute the baseline learning process described in Example 2. The APP interface displays the learning progress (number of data sets collected / 2016 sets) and intermediate statistical results (real-time mean and standard deviation of each parameter) in real time, and maintenance personnel can remotely monitor the learning process.

[0090] After 168 hours of benchmark learning, the data processing module 200 of the monitoring terminal automatically calculates the benchmark feature vector. Covariance matrix and threshold and The results are then transmitted via Bluetooth to a handheld maintenance terminal for confirmation by maintenance personnel. After reviewing the calculation results and confirming that the mean values ​​of each parameter are within the normal range specified in the equipment specifications, the maintenance personnel click the "Confirm Fixed Line" button. The monitoring terminal then writes the baseline eigenvector and covariance matrix into the protected area of ​​its local Flash memory, after which the terminal enters normal operation monitoring mode. If the maintenance personnel have doubts about the baseline learning results (e.g., due to extreme weather conditions causing data anomalies during the learning process), they can select "Relearn" to reset the entire baseline learning process.

[0091] Example 7: Configuration Scheme and Actual Effects of System in Large-Scale Deployment Across the Entire Network

[0092] This embodiment illustrates the large-scale deployment scheme and practical application effect of the system of the present invention on a typical railway line (200 km long, with 4000 track circuit devices along the line).

[0093] Following the principle of deploying one intelligent monitoring terminal per 100 meters, a total of 2,000 monitoring terminals will be deployed along the 200-kilometer line (covering all track circuit equipment and some key signal machines). The hardware cost of each terminal is approximately 2,000 yuan, with a total hardware investment of 4 million yuan. Data aggregation gateways (20) will be set up along the line, one every 4 kilometers, for a total of 50 gateways. The unit price of each gateway is 15,000 yuan, with a gateway investment of 750,000 yuan. The development and deployment cost of the cloud management platform is 1.5 million yuan (including 3 years of operation and maintenance). The construction and installation cost is 1.2 million yuan. The total investment of the project is approximately 7.45 million yuan.

[0094] After the line deployment was completed, all 2,000 terminals entered normal operation and monitoring mode. In the first month after the system was put into operation, the 2,000 terminals collected a total of approximately 172,800 sets of data (288 sets per day per terminal, calculated at once every 5 minutes, totaling 8,640 sets in 30 days, and 17.28 million sets of data for 2,000 terminals; the actual average collection frequency per terminal under power-constrained mode was slightly lower, approximately 140,000 sets / month). All terminals successfully completed baseline learning and entered a stable monitoring state.

[0095] When the system had been running for 6 months, the data analysis unit (32) of the cloud management platform found that the health H of terminal #37 in a certain section gradually decreased from 92 in the 4th month to 68 in the 6th month, and the LSSVR model predicted that it would continue to decrease to below the risk threshold of 60 within the next 45 days. The cloud management platform retrieved the complete data (voltage V, current I, ripple coefficient R, temperature T) of the terminal for the most recent 30 days for verification and found that V had decreased by 3.2% compared with the baseline value and R had increased by 41%. The deviation direction combination was completely matched with the "filter capacitor deterioration" fault mode. The cloud management platform pushed an "important warning" to the maintenance work area of ​​the section through the warning push module. The warning information clearly pointed out the predicted fault type, the suggested handling time limit (within 45 days), and the handling suggestion (replacing the power module filter capacitor).

[0096] After receiving the warning, the maintenance team scheduled a preventative maintenance check on the equipment during the nearest maintenance window (day 5). On-site inspection confirmed that the filter capacitor had slightly bulged, and the measured capacitance value was 28% lower than the nominal value, consistent with the warning diagnosis. After replacing the capacitor, the equipment voltage returned to normal, the ripple coefficient decreased to the baseline range, and the health index (H) rose to 91. From the discovery of the symptoms to the completion of the repair, only 5 days were needed, avoiding a potential sudden failure on day 45 that could have caused a signal malfunction. In a traditional approach, the slow aging process of the filter capacitor would not trigger a single fixed threshold alarm (a voltage drop of 3.2% does not exceed the ±10% conventional voltage threshold). When the capacitor completely fails, it would cause abnormal track circuit signals, potentially leading to train speed limits or even accidents, with losses far exceeding the cost of this preventative maintenance.

[0097] After deploying the system of this invention on this line, 37 potential early equipment degradation issues were identified in the first year, and all were addressed before any failures occurred, achieving the maintenance goal of "zero sudden failures" for the line's signal equipment. Compared with the traditional manual periodic inspection plan, the number of inspection personnel was reduced from 200 to 45, and the annual labor cost was reduced from 30 million yuan to 6.75 million yuan, resulting in annual labor cost savings of 23.25 million yuan. The project investment payback period is approximately 4 months, demonstrating significant economic and social benefits.

[0098] The above description is only a preferred embodiment of the present invention. All equivalent changes and modifications made within the scope of the claims of the present invention should be included in the scope of the present invention.

Claims

1. A smart monitoring terminal for the health status of rail transit signaling equipment, characterized in that, include: A status acquisition module is connected to the railway signaling equipment to be monitored, the railway signaling equipment including track circuit equipment, signal lights, switch machines or transponders; The status acquisition module is used to acquire the operating status parameters of the signal equipment, including at least voltage parameters, current parameters, and temperature parameters; The data processing module, electrically connected to the status acquisition module, is used to process and analyze the operating status parameters to generate equipment health status data. The data processing module contains a pre-set multi-parameter joint discrimination model for signal equipment health status. This model establishes a baseline feature vector based on multi-dimensional parameters collected during normal equipment operation. In real-time monitoring, the current multidimensional feature vector X is calculated and compared with the reference feature vector. The Mahalanobis distance D between them is used to characterize the overall deviation of the equipment: ； in Let the covariance matrix be the covariance matrix between the parameters. Calculated based on at least 168 hours of historical equipment operation data; when the Mahalanobis distance D exceeds a preset first threshold. When a device is deemed to be malfunctioning, the Mahalanobis distance D exceeds a preset second threshold. The system determines when there is a risk of equipment failure and triggers an early warning. Simultaneously, the model analyzes the parameters of each dimension in the current feature vector X relative to the baseline feature vector. The deviation direction combination is matched with a preset feature library of deviation directions for multiple typical fault modes to identify the fault type; the data processing module also has a preset equipment health degradation trend prediction algorithm, defining the equipment health H as: ； in The algorithm calculates the health status sequence based on a preset scaling factor according to the device type. Time series analysis is performed, and a least squares support vector regression model is used to predict the health status at future times. The training and prediction of the model are completed locally in the data processing module. When the predicted health status falls within a preset time window... The risk level drops to the set threshold. The following conditions trigger an early warning signal; the prediction model incorporates a priori constraint that signal equipment degradation is monotonically non-increasing, and when the model's predicted value is higher than the current value, it is forcibly corrected to the current value, and prediction results that exceed a certain proportion of the known average service life of the equipment type are truncated; the main control module is connected to the status acquisition module and the data processing module respectively, and is used to control the collaborative work of each module.

2. The intelligent health status monitoring terminal for rail transit signaling equipment according to claim 1, characterized in that, Also includes: The CT induction power harvesting module is installed near the power cable or contact network of the rail transit line. It obtains electrical energy through electromagnetic induction and supplies power to various functional modules. The CT induction power harvesting module includes an openable and closable power harvesting core, an induction coil, a rectifier and filter unit, a voltage regulator output unit, and an energy storage unit. The CT induction power harvesting module also includes a maximum power point tracking control unit, which dynamically adjusts the load impedance according to the output characteristics of the induction coil, so that the power harvesting module operates at the maximum power output point under different primary side current conditions. The energy storage unit stores electrical energy when the power cable current is greater than a set threshold and supplies power to the terminal when the current is lower than the set threshold. The self-organizing network communication module adopts LoRa or LTE. The Mesh self-organizing network communication protocol establishes wireless self-organizing network communication connections with adjacent monitoring terminals or data aggregation gateways in environments without public network signals, enabling multi-hop relay transmission of monitoring data. In the self-organizing network communication module, each terminal stores a neighbor node information table along the route. Routing selection prioritizes forwarding along the route extension direction. When a downstream node failure is detected, it automatically backtracks to the upstream node and triggers local route repair. The positioning module acquires the geographical location information of the intelligent monitoring terminal and associates it with the device health status data.

3. The intelligent health status monitoring terminal for rail transit signaling equipment according to claim 2, characterized in that, The acquisition frequency of the status acquisition module is dynamically adjusted by the main control module according to the current power supply: when the power supply is higher than the set threshold, it operates at the first acquisition frequency; when the power supply is lower than the set threshold, it operates at the second acquisition frequency, which is lower than the first acquisition frequency. The self-organizing network communication module stops periodic heartbeats when the power supply is lower than the set threshold, and only initiates data transmission when abnormal data is acquired or the stored data reaches the cache limit.

4. An intelligent health status monitoring system for rail transit signaling equipment, characterized in that, include: Multiple intelligent monitoring terminals as described in any one of claims 1 to 3 serve as railway signal system status monitoring devices, distributedly installed in sections, stations, and entrances of the rail transit line, and connected to the railway signal equipment at the corresponding locations; a data aggregation gateway is communicatively connected to each of the intelligent monitoring terminals, receives equipment health status data uploaded by each of the intelligent monitoring terminals, and uploads the data to the cloud management platform via wired or wireless means; The cloud management platform communicates with the data aggregation gateway, receives, stores and processes the device health status data, performs fault diagnosis and health assessment, and generates early warning information based on the assessment results. The cloud management platform is configured to retrieve at least 30 days of historical data from any terminal for early warning review; The early warning push module is connected to the cloud management platform and pushes the early warning information to the preset operation and maintenance terminal; A handheld maintenance terminal is connected to the intelligent monitoring terminal via near-field communication. It is used to configure parameters, read data, and issue start commands for benchmark feature vector acquisition of the intelligent monitoring terminal on-site.

5. The intelligent health status monitoring system for rail transit signaling equipment according to claim 4, characterized in that, The cloud management platform includes: a data receiving unit, which receives device health status data from each intelligent monitoring terminal uploaded by the data aggregation gateway; a data analysis unit, which processes and analyzes the device health status data, extracts device operating characteristic parameters, and compares them with preset health status benchmark values; a fault diagnosis unit, which determines whether the signal equipment has any abnormalities or fault risks based on the analysis results of the data analysis unit, and generates diagnostic results; a health assessment unit, which estimates the remaining service life of each device based on the time series change trend of the device health status data and combined with the statistical lifespan data of the same model of equipment across the entire network; and an early warning generation unit, which generates early warning information including device location, abnormality type, and risk level when the fault diagnosis unit confirms an abnormality or the health assessment unit assesses a health level lower than a preset threshold.

6. A method for intelligent monitoring of the health status of rail transit signaling equipment, characterized in that, Includes the following steps: S1: The status acquisition module acquires the operating status parameters of the railway signaling equipment to be monitored at a set acquisition frequency. The railway signaling equipment includes track circuit equipment, signals, switch machines, or transponders. The operating status parameters include at least voltage, current, and temperature parameters. S2: The data processing module processes and analyzes the operating status parameters, and generates equipment health status data using a multi-parameter joint discrimination model. The model calculates the current multi-dimensional feature vector X and the baseline feature vector X. Mahalanobis distance between To characterize the overall deviation of the device, when the Mahalanobis distance D exceeds a preset first threshold. When the device is detected as malfunctioning and exceeds a preset second threshold, an anomaly is identified. The system will detect the presence of a fault risk and trigger an early warning. Simultaneously, fault type identification is performed through the combination of deviation directions of parameters in various dimensions; S3: The data processing module performs equipment health degradation trend prediction and defines equipment health. ,in To calculate the health status sequence based on the preset scaling coefficients according to the device type. The least squares support vector regression model is used to predict the health status at future times. The predicted health status will be within a preset time window. The risk level drops to the set threshold. The following conditions will trigger a warning signal; S4: Transmit the device health status data and early warning signals to the data aggregation gateway and / or cloud management platform; S5: The cloud management platform reviews and performs in-depth analysis on the device health status data, and generates early warning information and pushes it to the preset operation and maintenance terminal when an abnormality is confirmed.

7. The intelligent monitoring method for the health status of rail transit signaling equipment according to claim 6, characterized in that, In step S2, the reference feature vector Covariance Matrix The first threshold is calculated based on at least 168 hours of historical equipment operation data; The second threshold is calculated as three times the standard deviation of the mean Mahalanobis distance in the baseline period. Take 6 times the standard deviation of the mean Mahalanobis distance in the base period.

8. The intelligent monitoring method for the health status of rail transit signaling equipment according to claim 6, characterized in that, In step S3, the prediction model incorporates a priori constraint that signal equipment degradation is monotonically non-increasing. When the model's predicted value is higher than the current value, it is forcibly corrected to the current value, and predictions exceeding a certain proportion of the known average lifespan of the equipment type are truncated. The risk threshold... Set to 60.

9. The intelligent monitoring method for the health status of rail transit signaling equipment according to claim 6, characterized in that, In step S1, when the power supply of the monitoring terminal is higher than a set threshold, it operates at a first acquisition frequency; when the power supply is lower than the set threshold, it operates at a second acquisition frequency lower than the first acquisition frequency. The data aggregation gateway uploads the device health status data to the cloud management platform via wired or wireless means.

10. The intelligent monitoring method for the health status of rail transit signaling equipment according to claim 6, characterized in that, It also includes step S6: The maintenance personnel communicate with the intelligent monitoring terminal through a handheld maintenance terminal, and set up parameters, read data and issue start commands for the intelligent monitoring terminal to collect the reference feature vector on site.

Images