Railway line monitoring alarm method and system

Abstract

The present invention provides a railway line monitoring alarm method and system, which relates to the field of computer technology. A dynamic monitoring data set including a track structure vibration waveform sequence, a track surface optical image sequence, and an environmental meteorological parameter sequence is continuously collected through multiple sensor nodes of a target railway line. Then, a multi-modal feature fusion process is performed on the dynamic monitoring data set to generate a comprehensive state evaluation matrix including a vibration energy distribution vector, etc., and a preset track safety evaluation model is used to classify the comprehensive state evaluation matrix into risk alarm levels, and an abnormal event type code and an abnormal area position code are output. Then, based on this, a multi-level linkage control instruction set including a train emergency braking trigger instruction is generated, and synchronously transmitted to a central dispatching system, so as to drive the system to execute track blocking, train speed control, and equipment maintenance operations according to instruction priority, so as to realize comprehensive and intelligent monitoring of railway lines and efficient alarm response.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a railway line monitoring alarm method and system. Background Art

[0002] In the field of railway transportation, safety is always the primary consideration. With the continuous expansion of the railway network and the continuous increase in train speed, real-time monitoring and timely alarm of railway line status have become particularly important. Traditionally, the monitoring of railway lines mainly relies on manual inspections and regular testing, which is not only time-consuming and labor-intensive, but also difficult to achieve real-time and comprehensive monitoring of track status.

[0003] In order to overcome the limitations of traditional monitoring methods, various sensor technologies have been gradually applied to the monitoring of railway lines. However, most of the current railway line monitoring systems use a single sensor type, or although multiple sensors are used, the data from different sensors are not effectively fused and processed, resulting in the monitoring system only being able to provide one-sided track status information, making it difficult to fully and accurately assess the safety status of the track. In addition, the existing monitoring system also has deficiencies in alarm and emergency response. Once an abnormality is detected, it can usually only send a simple alarm signal, and cannot take corresponding control measures based on the specific type and severity of the abnormality. Summary of the invention

[0004] In view of the above-mentioned problems, in combination with the first aspect of the present invention, an embodiment of the present invention provides a railway line monitoring alarm method, the method comprising:

[0005] A dynamic monitoring data set of the track area is continuously collected by multiple sensor nodes installed on the target railway line, wherein the dynamic monitoring data set includes a track structure vibration waveform sequence, a track surface optical image sequence, and an environmental meteorological parameter sequence;

[0006] Performing multimodal feature fusion processing on the dynamic monitoring data set to generate a comprehensive state evaluation matrix for each track area, wherein the comprehensive state evaluation matrix includes a vibration energy distribution vector, a surface defect probability vector, and an environmental stress influence vector;

[0007] Classify the risk alarm level of the comprehensive status assessment matrix according to the preset track safety assessment model, and output the abnormal event type code and abnormal area position code corresponding to the risk alarm level;

[0008] Generate a multi-level linkage control instruction set based on the abnormal event type code and the abnormal area position code, wherein the multi-level linkage control instruction set includes a train emergency brake trigger instruction, a signal light state switching instruction, and a maintenance robot dispatching instruction;

[0009] The multi-level linkage control instruction set is synchronously transmitted to the central dispatching system, driving the central dispatching system to execute track blocking operations, train speed control operations and equipment maintenance operations according to the instruction priority.

[0010] On the other hand, an embodiment of the present invention also provides a railway line monitoring and alarm system, including a processor and a machine-readable storage medium, wherein the machine-readable storage medium is connected to the processor, the machine-readable storage medium is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the machine-readable storage medium to implement the above method.

[0011] Based on the above aspects, the embodiment of the present application continuously collects various dynamic monitoring data sets of the track area through multiple sensor nodes installed on the target railway line, and performs multi-modal feature fusion processing on the collected dynamic monitoring data sets to generate a comprehensive state evaluation matrix for each track area, classifies the comprehensive state evaluation matrix according to the risk alarm level based on the preset track safety assessment model, and outputs the abnormal event type code and abnormal area location code corresponding to the risk alarm level, realizing the clear and precise risk classification and precise positioning from multi-source data, and can quickly and accurately identify safety risks of different types and levels, and clarify the specific location of the risk. The multi-level linkage control instruction set generated based on the abnormal event type code and the abnormal area location code includes the train emergency brake trigger instruction, the signal light state switching instruction and the maintenance robot dispatch instruction, etc., realizes the coordinated linkage control of various related systems in different risk scenarios, and can quickly and accurately allocate various resources according to the real-time risk status, take targeted measures, effectively avoid the occurrence of accidents or reduce the degree of harm of accidents, and the design of the multi-level linkage control instruction set fully considers the correlation and coordination requirements between different subsystems in the railway system, and significantly improves the ability of the railway system to respond to sudden safety incidents. Finally, the multi-level linkage control instruction set is synchronously transmitted to the central dispatching system, driving the central dispatching system to execute track blocking operations, train speed control operations and equipment maintenance operations according to the instruction priority, ensuring that when faced with complex security situations, the central dispatching system can coordinate various resources in an orderly manner, reasonably arrange response measures, and maximize the safety and efficiency of railway operations. Through the unified command and dispatch of the central dispatching system, various operating links work closely together to form an organic whole, realizing the integrated and coordinated operation of railway line monitoring, alarm and emergency response. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 It is a schematic diagram of the execution flow of the railway line monitoring and alarm method provided by an embodiment of the present invention.

[0013] Figure 2Schematic diagram of exemplary hardware and software components of a railway line monitoring and alarm system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0014] The present invention will be described in detail below with reference to the accompanying drawings. Figure 1 The figure is a flow chart of a railway line monitoring and alarm method provided by an embodiment of the present invention. The railway line monitoring and alarm method is introduced in detail below.

[0015] Step S110, continuously collecting a dynamic monitoring data set of the track area through multiple sensor nodes installed on the target railway line, wherein the dynamic monitoring data set includes a track structure vibration waveform sequence, a track surface optical image sequence, and an environmental meteorological parameter sequence.

[0016] In this embodiment, a mountain railway line is considered, which runs through mountains and has complex and changeable terrain. On this railway line, multiple sensor nodes can be installed at regular intervals, and these sensor nodes work together to continuously collect dynamic monitoring data sets of the track area.

[0017] For example, the distributed vibration sensor monitors the vibration of the track structure at a preset first sampling frequency, thereby obtaining a track structure vibration waveform sequence. Assume that at a certain moment, a heavy-loaded freight train is running on this railway line. When the wheels of the train pass a certain position on the track, the track structure vibrates. The distributed vibration sensor can capture this vibration and convert it into vibration waveform data and record it. As the train continues to run, the sensor continues to collect vibration waveforms at different positions to form a complete track structure vibration waveform sequence.

[0018] At the same time, the industrial camera shoots the track surface at a fixed shooting interval to obtain a sequence of optical images of the track surface. For example, when sunlight shines on the track surface, the industrial camera can clearly capture the track surface. During the shooting process, the industrial camera can capture possible defects such as fine cracks and rust spots on the track surface. For example, at a certain bend, due to the centrifugal force of the train and long-term natural erosion, some early tiny cracks may appear on the track surface. The industrial camera can record these track surface optical images containing crack information in an orderly manner at a fixed shooting interval.

[0019] In addition, the meteorological monitoring unit is also configured to record the environmental meteorological parameter sequence. For example, the meteorological conditions in mountainous areas are complex and changeable, and the meteorological monitoring unit can record air temperature gradient data, relative humidity change data, and wind speed fluctuation data. For example, in the early morning, the temperature in the mountainous area is low. As the sun rises, the temperature gradually rises. The meteorological monitoring unit can record the temperature gradient changes in this process. When the wind blows through the valley, the wind speed fluctuation data will also be accurately recorded. These track structure vibration waveform sequences, track surface optical image sequences, and environmental meteorological parameter sequences will be time-stamped and synchronized, and the data storage interval will be divided according to the mileage pile number of the track line. In this way, when you need to query the data near a certain mileage pile number later, you can easily obtain the corresponding complete dynamic monitoring data set.

[0020] Step S120, performing multimodal feature fusion processing on the dynamic monitoring data set to generate a comprehensive state evaluation matrix for each track area, wherein the comprehensive state evaluation matrix includes a vibration energy distribution vector, a surface defect probability vector, and an environmental stress influence vector.

[0021] Continuing with this mountain railway line as an example, after obtaining the dynamic monitoring data set, multimodal feature fusion processing can be performed.

[0022] For the vibration waveform sequence of the track structure, frequency domain energy analysis is performed to extract the proportion of vibration energy in each frequency band to form a vibration energy distribution vector. Assume that in the previously collected vibration waveform sequence, it is found that a certain section of track has a large low-frequency vibration energy component when a train passes through. This may be due to some potential problems in the track infrastructure, such as an unreliable roadbed. Through frequency domain energy analysis, the proportion of low-frequency vibration energy to total vibration energy is calculated, and the proportion of medium-frequency and high-frequency vibration energy is also obtained, thereby forming a vibration energy distribution vector.

[0023] For the track surface optical image sequence, pixel-level texture segmentation is performed to calculate the surface crack growth rate and the rust area coverage rate, and then generate a surface defect probability vector. For example, in the image of the track surface at the bend taken by the industrial camera before, the pixel range of the crack and the pixel range of the rust spot can be accurately identified through the pixel-level texture segmentation technology. According to the ratio of these pixel ranges to the total number of pixels of the entire track surface image, the surface crack growth rate and the rust area coverage rate are calculated. Assuming that within a certain period of time, the pixel range of the crack increases and the coverage rate of the rust spot also changes to a certain extent, the surface defect probability vector reflecting the track surface defect situation can be generated based on the above data.

[0024] For the environmental meteorological parameter sequence, cumulative effect modeling is performed to derive the offset of the thermal expansion coefficient of the rail and the risk index of bolt loosening, and to construct the environmental stress impact vector. Due to the large temperature difference between day and night in mountainous areas, the rail will expand due to the increase in temperature during the high temperature period during the day. The temperature data recorded by the meteorological monitoring unit is combined with the material properties of the rail, and the offset of the thermal expansion coefficient of the rail can be obtained through cumulative effect modeling. At the same time, wind speed fluctuations and temperature changes may affect the tightening state of the bolts, and the bolt loosening risk index is obtained through analysis and calculation. For example, after a long period of strong winds and drastic temperature changes, the bolt loosening risk index calculated based on meteorological data has increased, and these data constitute the environmental stress impact vector.

[0025] Finally, the vibration energy distribution vector, surface defect probability vector and environmental stress influence vector are horizontally spliced ​​according to the track area number to form a multi-dimensional comprehensive state assessment matrix. In this way, a comprehensive state assessment matrix is ​​generated for each track area, which can fully reflect the track state of the area.

[0026] Step S130, classifying the comprehensive status assessment matrix into risk alarm levels according to a preset track safety assessment model, and outputting abnormal event type codes and abnormal area position codes corresponding to the risk alarm levels.

[0027] Still taking the above mountain railway line as an example, the preset track safety assessment model is used to classify the risk alarm level.

[0028] First, the vibration energy distribution vector is input into the first risk assessment sub-model. Assuming that the low-frequency vibration energy component accounts for too high a proportion in the vibration energy distribution vector of a certain track area before, after decomposing the vibration energy distribution vector into low-frequency vibration energy component, medium-frequency vibration energy component and high-frequency vibration energy component according to the preset frequency band division rule in the first risk assessment sub-model, it is found that the energy proportion data of the low-frequency vibration energy component exceeds the frequency band energy threshold interval stored in the first risk assessment sub-model. The starting mileage pile number and the ending mileage pile number of the abnormal vibration area are determined by the physical position of the sensor node corresponding to the abnormal vibration frequency band and the timestamp sequence of the vibration energy distribution vector, for example, the area between mileage pile number 100 and mileage pile number 120. Then, based on the energy proportion data of the abnormal vibration frequency band and the vibration duration calculation rule, the track structure fatigue score between this mileage pile number is calculated. The score is compared with the historical fatigue score of the same mileage pile number interval in the dynamic threshold curve by sliding window. If the score deviation is large, according to the rules set by the first risk assessment sub-model, it is determined to be a higher track fracture risk alarm level, and the corresponding coordinate interval is the area between mileage pile numbers 100-120.

[0029] Next, the surface defect probability vector is input into the second risk assessment sub-model. The pixel coordinate set of the surface crack growth rate and the rust area coverage is extracted from the surface defect probability vector. Using the image coordinate positioning system integrated in the second risk assessment sub-model, these pixel coordinates are converted into the track surface geographic coordinates and matched with the track area number of the corresponding mileage pile number. For example, in the surface defect probability vector of a certain track area, the crack growth rate has increased and the rust spot coverage has also increased. These spatiotemporal variation trends are input into the historical defect evolution pattern library, and pattern matching is performed with the crack growth rate template and rust diffusion template stored in the library to identify the closest historical defect evolution path. According to the diffusion direction and rate parameters of the path, the spatial diffusion boundary coordinates in the future time window are predicted. Assuming that the predicted spatial diffusion boundary coordinates have a large overlap ratio with the current pixel coordinate set, according to the definition of the second risk assessment sub-model, a higher defect diffusion alarm level and a pixel coordinate cluster set covering the diffusion boundary are generated.

[0030] Then, the environmental stress impact vector is input into the third risk assessment sub-model. The meteorological sensor node location data of the rail thermal expansion coefficient offset and the bolt loosening risk index are extracted from the environmental stress impact vector. Based on these data, a spatial distribution map of meteorological parameters is constructed and mapped to the three-dimensional geographic coordinate grid of the track line. Under extreme high temperature and extreme wind loading conditions, the track deformation variable distribution matrix is ​​simulated based on the temperature gradient data of each node in the three-dimensional geographic coordinate grid. The deformation hot zone nodes above the deformation threshold are extracted from the matrix, and the deformation accumulation of each node is calculated. For example, in a certain area, due to the sudden rise in temperature and strong winds, the rails have undergone large deformations, and the calculated deformation accumulation is large. According to the analysis of the third risk assessment sub-model, a higher environmental anomaly impact index and a set of geographic coordinates of the deformation hot zone nodes are generated.

[0031] Finally, according to the combined weight of the track fracture risk alarm level, defect diffusion alarm level and environmental anomaly impact index, the comprehensive risk alarm level is determined and mapped to the preset abnormal event type code. For example, when the track fracture risk alarm level is high, the defect diffusion alarm level is also high and the environmental anomaly impact index is large, the combined weight coefficient of the three indicates that the comprehensive risk is high. By matching with the preset abnormal event type code mapping table, the corresponding abnormal event type code is determined. At the same time, the coordinate interval, pixel coordinate cluster and deformation hot zone geographic coordinate set are spatially superimposed to generate the abnormal area location code containing mileage pile number and GPS positioning data. For example, the mileage pile number interval of the track fracture risk alarm level is converted into a GPS coordinate range, the pixel coordinate cluster set of the defect diffusion alarm level is converted into a polygon formed by geographic coordinates, and the deformation hot zone geographic coordinate set is merged into a polygon, and their spatial intersection area is calculated, and the vertex GPS coordinates of the intersection area and the covered mileage pile number interval are extracted to generate the abnormal area location code.

[0032] Step S140, generating a multi-level linkage control instruction set based on the abnormal event type code and the abnormal area position code, wherein the multi-level linkage control instruction set includes a train emergency brake trigger instruction, a signal light state switching instruction, and a maintenance robot dispatching instruction.

[0033] Taking this mountain railway line as an example, a multi-level linkage control instruction set is generated based on the abnormal event type code and abnormal area location code obtained previously.

[0034] When the comprehensive risk alarm level reaches the first critical value, for example, in the case of a serious track fracture risk, the GPS positioning data in the abnormal area position code is parsed. Assuming that the abnormal area is located between mileage pile numbers 100-120, the corresponding GPS positioning data covers a specific geographical range. Based on these data, a train emergency braking trigger instruction containing the blockade coordinates of the target track area and the contact network power-off time is generated. For example, the train emergency braking trigger instruction clearly stipulates that the emergency braking is triggered before the train travels to a certain distance from the target track area, and the contact network power-off time is determined to ensure that the train stops safely before reaching the dangerous area.

[0035] When the comprehensive risk alarm level is in the second critical range, for example, when the defect diffusion alarm level is high but has not yet reached the most serious level, the train dispatch schedule is matched according to the mileage post number encoded in the abnormal area position. Assuming that the abnormal area is near mileage post number 150, the train dispatch schedule is queried to find that multiple trains are about to pass through the area. According to the current position of the train, a signal light state switching instruction matching the current position of the train is generated, where the red light retention time is determined according to the estimated time for the train to arrive at the dangerous area, for example, the red light is maintained for 10 minutes. And according to the situation of the pixel coordinate cluster of the defect diffusion, the yellow flash warning mode parameters based on the pixel coordinate cluster are set, such as setting a yellow flash signal of a specific frequency near the dangerous area to remind the train driver to slow down.

[0036] When the comprehensive risk alarm level triggers the third response condition, for example, when the track is deformed to a certain extent due to abnormal environmental influences and there are defects on the surface, the spatial intersection of the geographical coordinate set of the deformation hotspot and the pixel coordinate cluster is extracted. Assuming that in a certain area, the deformation hotspot and the surface defect area partially overlap, a maintenance robot dispatch instruction carrying the three-dimensional model data of the track area is generated based on this intersection. The instruction contains a list of fault point coordinates that integrates mileage pile numbers and GPS data, for example, the mileage pile numbers and GPS coordinates of the specific locations that need to be repaired are listed. At the same time, the priority maintenance order identification is determined according to the area of ​​the deformation hotspot, and maintenance robots are given priority for maintenance in areas with larger areas or areas that have a greater impact on train operation.

[0037] Step S150, synchronously transmitting the multi-level linkage control instruction set to the central dispatching system, driving the central dispatching system to execute track blocking operations, train speed control operations and equipment maintenance operations according to the instruction priority.

[0038] Taking this mountain railway line as an example, after the multi-level linkage control instruction set is synchronously transmitted to the central dispatching system, the central dispatching system begins to perform corresponding operations according to the instruction priority.

[0039] First, the command priority, target track area blocking coordinates, red light holding time, and fault point coordinate list of each command in the multi-level linkage control command set are parsed. For example, the train emergency brake trigger command has the highest priority because it involves the safe parking of the train. Then, the train emergency brake trigger command, signal light state switching command, and maintenance robot dispatch command are sorted in execution order according to the command priority to generate a command distribution queue containing a time window identifier.

[0040] Next, the train emergency brake trigger command is verified by spatial overlap with the real-time train position trajectory data of the central dispatching system. Assuming that train A is about to enter the target track area, train A is determined to be the train number that needs to trigger emergency braking according to the command, and the corresponding contact network power-off time interval is determined, for example, the contact network power supply is cut off 2 minutes before train A reaches the danger zone to ensure the safe parking of the train.

[0041] For the signal light state switching instruction, the red light retention time is dynamically matched with the arrival time deviation of the train scheduling schedule. For example, Train B is originally expected to arrive at a station in 10 minutes, but due to the signal light state switching, it needs to stop in front of the abnormal area. Calculate the signal light switching delay compensation value, assuming it is 3 minutes, and update the yellow flash warning mode parameters, such as adjusting the yellow flash frequency, to better remind the train driver.

[0042] For the maintenance robot dispatch instruction, the fault point coordinate list is matched with the current position of the maintenance robot and the remaining operating capacity. Assuming that the maintenance robot C is close to the fault point and has sufficient remaining operating capacity, according to the task allocation path planning based on the priority maintenance sequence identification, the maintenance robot C is prioritized to go to the fault point for maintenance, and arrives at the fault point according to the planned path to perform equipment maintenance operations.

[0043] During the operation, the contact network power-off status code fed back by the train automatic driving subsystem, the signal light switching completion mark returned by the signal light control subsystem, and the fault point maintenance progress data reported by the maintenance robot dispatching subsystem are received in real time. For example, when the contact network of train A is successfully powered off, the train automatic driving subsystem feeds back the corresponding contact network power-off status code; when the signal light is switched, the signal light control subsystem returns the signal light switching completion mark; the maintenance robot C regularly reports the fault point maintenance progress data during the maintenance process. Based on these data, the track area blockade status update table, train speed control log and equipment maintenance real-time topology map are generated and synchronized to the global operation status interface of the central dispatching system, so that the dispatching personnel can grasp the operation status of the railway line in real time.

[0044] Based on the above steps, the embodiment of the present application continuously collects various dynamic monitoring data sets of the track area through multiple sensor nodes installed on the target railway line, and performs multimodal feature fusion processing on the collected dynamic monitoring data sets to generate a comprehensive state evaluation matrix for each track area, classifies the risk alarm level of the comprehensive state evaluation matrix based on the preset track safety assessment model, and outputs the abnormal event type code and abnormal area location code corresponding to the risk alarm level, realizing the clear and definite risk classification and precise positioning from multi-source data, and can quickly and accurately identify safety risks of different types and levels, and clarify the specific location of the risk. The multi-level linkage control instruction set generated based on the abnormal event type code and the abnormal area location code includes the train emergency brake trigger instruction, the signal light state switching instruction and the maintenance robot dispatch instruction, etc., realizes the coordinated linkage control of various related systems in different risk scenarios, and can quickly and accurately allocate various resources according to the real-time risk status, take targeted measures, effectively avoid the occurrence of accidents or reduce the degree of harm of accidents, and the design of the multi-level linkage control instruction set fully considers the correlation and coordination requirements between different subsystems in the railway system, and significantly improves the ability of the railway system to respond to sudden safety incidents. Finally, the multi-level linkage control instruction set is synchronously transmitted to the central dispatching system, driving the central dispatching system to execute track blocking operations, train speed control operations and equipment maintenance operations according to the instruction priority, ensuring that when faced with complex security situations, the central dispatching system can coordinate various resources in an orderly manner, reasonably arrange response measures, and maximize the safety and efficiency of railway operations. Through the unified command and dispatch of the central dispatching system, various operating links work closely together to form an organic whole, realizing the integrated and coordinated operation of railway line monitoring, alarm and emergency response.

[0045] In a possible implementation, step S110 includes:

[0046] Step S111, controlling the distributed vibration sensor to collect a vibration waveform sequence of the track structure at a first sampling frequency, and capturing a track surface optical image sequence at a fixed shooting interval through an industrial camera.

[0047] In the mountain railway line scenario mentioned above, the distributed vibration sensor is set to collect the vibration waveform sequence of the track structure at a specific first sampling frequency. On this mountain railway line, trains run frequently, for example, a large number of freight trains and passenger trains run through it every day. When the wheels of the train come into contact with the track, vibrations are generated, and this vibration propagates in the track structure in the form of waves. The distributed vibration sensor accurately captures the information of these vibration waves at a pre-set first sampling frequency and converts it into a digital signal to form a vibration waveform sequence of the track structure. For example, when a train loaded with cargo passes through a certain section of the track at a certain speed, the sensor records the vibration waveforms generated by the weight of the train, the speed of travel, and the structural characteristics of the track itself. As time goes by and trains continue to pass, the continuously accumulated vibration waveform data fully reflects the vibration of the track structure under different train loads and operating conditions.

[0048] At the same time, the industrial camera captures a sequence of optical images of the track surface at a fixed shooting interval. In a mountainous environment, the track surface is affected by many factors, such as climate change and geological conditions. The industrial camera takes pictures at a fixed shooting interval to ensure that the state of the track surface can be fully and orderly recorded. For example, in the case of sufficient light during the day, the industrial camera can clearly capture every detail of the track surface. The rust caused by rain erosion, the slight deformation caused by long-term rolling of the train, or the fine cracks caused by material aging can all be accurately reflected in the image. Over time, these sequences of optical images of the track surface acquired at fixed shooting intervals have become important records of changes in the state of the track surface.

[0049] Step S112, starting the meteorological monitoring unit to record the environmental meteorological parameter sequence in real time, wherein the environmental meteorological parameter sequence includes air temperature gradient data, relative humidity change data and wind speed fluctuation data.

[0050] For example, the meteorological conditions in mountainous areas are complex and changeable. The meteorological monitoring unit accurately records the air temperature gradient data, relative humidity change data, and wind speed fluctuation data in response to this situation. For example, in the morning, the temperature in the mountainous area is relatively low. As the sun rises, the air temperature gradually rises. The meteorological monitoring unit accurately records the temperature gradient changes in this process. When encountering strong winds, which are common in mountainous areas, the wind speed fluctuation data will be recorded in a timely manner. The change in relative humidity is also closely related to the state of the track. For example, when the humidity is high, the track is more likely to rust. These humidity change data are also fully recorded.

[0051] Step S113, performing time stamp synchronization marking on the track structure vibration waveform sequence, the track surface optical image sequence and the environmental meteorological parameter sequence, and dividing the data storage interval according to the mileage pile number of the track line.

[0052] In order to effectively associate and integrate the collected track structure vibration waveform sequence, track surface optical image sequence and environmental meteorological parameter sequence, they need to be time-stamped and synchronized. This is like putting an accurate time tag on each data to ensure that the acquisition time of each data can be accurately known in subsequent analysis. For example, when a distributed vibration sensor records a special vibration waveform at a certain moment, the track surface image captured by the industrial camera and the meteorological data recorded by the meteorological monitoring unit are marked with the same timestamp. Finally, the data storage interval is divided according to the mileage pile number of the track line. The mileage pile number of the mountain railway line is a clear geographical location identifier. The data storage interval is divided based on the mileage pile number, so that the data in a specific area can be quickly located when querying and analyzing the data. For example, the data between mileage pile numbers 100-200 are stored in a specific interval. When the track status data in this interval needs to be viewed, all relevant data such as the track structure vibration waveform sequence, the track surface optical image sequence and the environmental meteorological parameter sequence can be easily obtained from the corresponding storage interval.

[0053] In a possible implementation, step S120 includes:

[0054] Step S121, performing frequency domain energy analysis on the vibration waveform sequence of the track structure, extracting the proportion of vibration energy in each frequency band, and forming a vibration energy distribution vector.

[0055] For example, on mountain railway lines, the vibration waves generated by train operation contain rich information. Due to the different factors such as the type of train (such as passenger trains, freight trains), running speed, load, and the structural characteristics of the track itself (such as rail type, roadbed condition), the vibration waveform of the track structure presents diverse characteristics. When performing frequency domain energy analysis on the track structure vibration waveform sequence, the time domain vibration waveform is first converted into a frequency domain signal through a specific algorithm. For example, in the frequency domain, the energy corresponding to different frequency components can be clearly seen. In this mountain railway line, due to the possible local unevenness or structural damage of the track, the distribution of vibration energy in different frequency bands will change. It is assumed that low-frequency vibration may be related to the overall vibration of the track infrastructure (such as the roadbed), medium-frequency vibration may be related to the natural vibration frequency of the rail, and high-frequency vibration may be related to the microscopic contact vibration between the wheel and the rail. Through the frequency domain energy analysis algorithm, the proportion of low-frequency, medium-frequency and high-frequency vibration energy in the total vibration energy is calculated respectively. For example, after analysis, it was found that in a certain mileage range (such as mileage between 200 and 300), low-frequency vibration energy accounts for 30% of the total vibration energy, medium-frequency vibration energy accounts for 40%, and high-frequency vibration energy accounts for 30%. These percentage data constitute the vibration energy distribution vector of the area. This vibration energy distribution vector can reflect the vibration characteristics of the track structure from the perspective of energy distribution, providing an important basis for subsequent track status assessment.

[0056] Step S122, performing pixel-level texture segmentation on the track surface optical image sequence, calculating the surface crack growth rate and the corrosion area coverage rate, and generating a surface defect probability vector.

[0057] For example, in the operating environment of mountain railways, the track surface faces a variety of complex influencing factors, which may cause defects such as cracks and rust on the surface. The optical image sequence of the track surface acquired by the industrial camera at a fixed shooting interval has become an important data source for analyzing these defects. Through pixel-level texture segmentation technology, the areas corresponding to different texture features in the image can be accurately identified. For cracks on the track surface, they appear as specific texture changes in the image. Using the pixel-level texture segmentation algorithm, the pixel range of the crack in the image can be accurately defined. For example, in a certain track surface image, the pixel coordinate range where the crack is located is identified by the algorithm, and the ratio of the number of these pixels to the total number of pixels in the entire image is calculated to obtain the surface crack growth rate. Similarly, for the rusted area, it also has unique color and texture characteristics in the image. Through pixel-level texture segmentation technology, the rusted area can be accurately segmented from the image, and the ratio of the number of pixels in the rusted area to the total number of pixels in the image is calculated to obtain the coverage rate of the rusted area. Assuming that in the track area between mileage posts 350 and 400, after analyzing multiple track surface optical images, it is calculated that the surface crack growth rate is 5% and the rust area coverage rate is 8%. These data constitute the surface defect probability vector of the area, which can intuitively reflect the degree and development trend of track surface defects and provide key data for evaluating the track surface condition.

[0058] Step S123, performing cumulative effect modeling on the environmental meteorological parameter sequence, deriving the rail thermal expansion coefficient offset and bolt loosening risk index, and constructing an environmental stress influence vector.

[0059] For example, the meteorological conditions in mountainous areas are complex and changeable, which has a significant impact on the track structure. The air temperature gradient data, relative humidity change data, and wind speed fluctuation data in the environmental meteorological parameter sequence are all important factors affecting the track state. First, consider the temperature factor. Since the rail is a metal material, it has the characteristics of thermal expansion and contraction. In mountainous areas, the temperature difference between day and night is large, and the temperature difference between the high temperature period during the day and the low temperature period at night may reach a large value. Through cumulative effect modeling, the offset of the thermal expansion coefficient of the rail can be accurately derived based on the temperature data recorded for a long time and the material thermal properties of the rail. For example, during the high temperature period in summer, the rail will expand under the action of high temperature for a long time. If the temperature exceeds the normal working temperature range of the rail, the thermal expansion coefficient of the rail will shift. At the same time, wind speed fluctuations and relative humidity changes will also affect the connecting parts of the track (such as bolts). When the wind speed is high, lateral force may be generated on the track structure, and humidity changes may cause the bolts to rust or corrode. Through comprehensive analysis and modeling of wind speed fluctuation data, relative humidity change data, and mechanical performance parameters of bolts, the bolt loosening risk index can be calculated. For example, under certain meteorological conditions (such as strong winds and high humidity for several consecutive days), the model calculates that the bolt loosening risk index in the area is 0.2 (the index is a dimensionless value calculated based on the model, used to indicate the degree of loosening risk). These data such as rail thermal expansion coefficient offset and bolt loosening risk index construct the environmental stress impact vector of the area, which can reflect the stress impact of environmental factors on the track structure.

[0060] Step S124, the vibration energy distribution vector, the surface defect probability vector and the environmental stress influence vector are horizontally spliced ​​according to the track area number to form a multi-dimensional comprehensive state assessment matrix.

[0061] On this mountain railway line, each track area has its own unique number to identify and distinguish different location intervals. For example, the track area between mileage pile numbers 100-200 is numbered as area 1. For this area 1, the vibration energy distribution vector (such as low frequency 30%, medium frequency 40%, high frequency 30%), surface defect probability vector (such as surface crack growth rate 3%, rust area coverage rate 5%) and environmental stress influence vector (such as rail thermal expansion coefficient offset is 0.1, bolt loosening risk index is 0.15) have been calculated separately. According to the track area number, these three vectors are spliced ​​horizontally. In this way, a multi-dimensional comprehensive state assessment matrix is ​​formed, which can comprehensively and comprehensively reflect the track status of area 1. For each track area of ​​the entire mountain railway line, a comprehensive state assessment matrix is ​​constructed in this way. The comprehensive state assessment matrix can accurately reflect the track structure vibration characteristics, surface defects, environmental stress effects and other aspects of each track area, so as to achieve accurate assessment and monitoring of the track status of the entire mountain railway line.

[0062] In a possible implementation, step S130 includes:

[0063] Step S131, input the vibration energy distribution vector into the first risk assessment sub-model, determine the mileage pile number of the abnormal vibration area by comparing the frequency band energy threshold, calculate the track structure fatigue score and compare it with the dynamic threshold curve, and determine the track fracture risk alarm level and the corresponding coordinate interval.

[0064] Step S132, input the surface defect probability vector into the second risk assessment sub-model, match the historical defect evolution pattern library based on the image coordinate positioning system, predict the spatial diffusion range of the surface damage deterioration trend and generate the defect diffusion alarm level and pixel coordinate cluster.

[0065] Step S133, input the environmental stress impact vector into the third risk assessment sub-model, combine the topological position of the meteorological sensor to simulate the track deformation under preset extreme conditions, and output the environmental anomaly impact index and the geographical coordinate set of the deformation hot zone.

[0066] Step S134, according to the combined weight of the track fracture risk alarm level, defect diffusion alarm level and environmental anomaly impact index, the comprehensive risk alarm level is determined and mapped to the preset abnormal event type code, and the coordinate interval, pixel coordinate cluster and deformation hot zone geographic coordinate set are spatially superimposed to generate the abnormal area position code containing the mileage pile number and GPS positioning data.

[0067] Wherein, the track safety assessment model includes the first risk assessment sub-model, the second risk assessment sub-model and the third risk assessment sub-model.

[0068] In a possible implementation, step S131 includes:

[0069] Step S1311, decomposing the vibration energy distribution vector into a low-frequency vibration energy component, a medium-frequency vibration energy component and a high-frequency vibration energy component according to a preset frequency band division rule in the first risk assessment sub-model.

[0070] In the actual operation of mountain railways, the vibration of the track structure is complex and diverse. Take a specific track area as an example, such as the part between mileage pile numbers 700-800. First, assume that in this first risk assessment sub-model, the low-frequency band is set to 0-100Hz, the medium-frequency band is 100-500Hz, and the high-frequency band is 500-1000Hz. By analyzing the vibration energy distribution vector through a specific algorithm, it is concluded that the energy share of the low-frequency vibration energy component is 30%, the energy share of the medium-frequency vibration energy component is 40%, and the energy share of the high-frequency vibration energy component is 30%.

[0071] Step S1312, extracting the energy proportion data of the vibration energy components of each frequency band, and comparing them segment by segment with the frequency band energy threshold intervals stored in the first risk assessment sub-model, to identify the abnormal vibration frequency bands marked by the first risk assessment sub-model.

[0072] In this first risk assessment sub-model, the energy threshold interval of the low-frequency band is set to 10% - 20%, the energy threshold interval of the medium-frequency band is set to 30% - 40%, and the energy threshold interval of the high-frequency band is set to 20% - 30%. By comparison, it is found that the 30% proportion of the low-frequency vibration energy component exceeds its threshold interval, thereby identifying the low-frequency vibration band as the abnormal vibration band marked by the first risk assessment sub-model.

[0073] Step S1313, determining the starting mileage post number and the ending mileage post number of the abnormal vibration area output by the first risk assessment sub-model according to the physical position of the sensor node corresponding to the abnormal vibration frequency band and the timestamp sequence of the vibration energy distribution vector associated with the first risk assessment sub-model.

[0074] Assuming that the sensor node corresponding to the low-frequency vibration band is located near the mileage number 720, and according to the timestamp sequence analysis, low-frequency abnormal vibration continues to occur in a specific time period, combined with the structural characteristics of the track and the sensor layout, the starting mileage number of the abnormal vibration area output by the first risk assessment sub-model is determined to be 710, and the ending mileage number is 750.

[0075] Step S1314, based on the energy proportion data of the abnormal vibration frequency band and the vibration duration calculation rule in the first risk assessment sub-model, calculate the track structure fatigue score between the starting mileage pile number and the ending mileage pile number.

[0076] For example, the energy of the low-frequency vibration energy component accounts for 30%. According to the calculation formula in the model and the duration of the low-frequency vibration (assuming that the duration is 20% of the total monitoring time), the track structure fatigue score is calculated to be 0.4.

[0077] Step S1315, performing a sliding window comparison on the track structure fatigue score and the historical fatigue score of the same mileage pile number interval maintained by the first risk assessment sub-model in the dynamic threshold curve, and determining the track fracture risk alarm level and the corresponding coordinate interval boundary value according to the deviation amplitude set by the first risk assessment sub-model.

[0078] The dynamic threshold curve is established based on the long-term historical monitoring data of the area, and contains the fatigue score threshold ranges of different mileage pile number intervals at different times. In the same mileage pile number interval of 710-750, the fluctuation range of the historical fatigue score is between 0.2-0.3. The currently calculated score of 0.4 has a larger deviation compared with the historical fatigue score. According to the deviation amplitude determination rule set by the first risk assessment sub-model, it is determined to be a higher track fracture risk alarm level, and the corresponding coordinate interval boundary values ​​are the starting mileage pile number 710 and the ending mileage pile number 750.

[0079] And, step S132 includes:

[0080] Step S1321: extracting a set of pixel coordinates of the surface crack growth rate and the corrosion area coverage rate defined by the second risk assessment sub-model from the surface defect probability vector.

[0081] Consider the track surface conditions between mileage stakes 850 and 950 on a mountain railway. The pixel coordinate sets of the surface crack growth rate and the rust area coverage defined by the second risk assessment sub-model are extracted from the surface defect probability vector corresponding to this area. For example, the surface crack growth rate is 10% and the rust area coverage rate is 15%. The pixel coordinate sets related to these defect rates are determined through image analysis.

[0082] Step S1322: According to the image coordinate positioning system integrated with the second risk assessment sub-model, the pixel coordinate set is converted into the track surface geographic coordinates associated with the second risk assessment sub-model, and the track area number corresponding to the mileage pile number is matched.

[0083] The image coordinate positioning system uses a pre-established coordinate conversion algorithm to convert the pixel coordinates in the image into actual track surface geographic coordinates, thereby determining the specific locations of these defects on the track and clarifying the track area numbers corresponding to the mileage posts of 850-950.

[0084] Step S1323, input the spatiotemporal variation trends of the surface crack propagation rate and the corrosion area coverage into the historical defect evolution pattern library of the second risk assessment sub-model, perform pattern matching with the crack growth rate template and the corrosion diffusion template stored in the second risk assessment sub-model, and identify the closest historical defect evolution path output by the second risk assessment sub-model.

[0085] For example, the surface crack growth rate has shown a gradual increase over the past period of time, and the corrosion area coverage has a similar growth trend. In the historical defect evolution model library, crack growth rate templates and corrosion diffusion templates under different conditions are stored. By comparing the current spatiotemporal change trend with these templates, the closest historical defect evolution path output by the second risk assessment sub-model is identified.

[0086] Step S1324, predicting the spatial diffusion boundary coordinates within the future time window marked by the second risk assessment sub-model according to the diffusion direction of the historical defect evolution path and the rate parameter loaded by the second risk assessment sub-model.

[0087] Assuming that the diffusion direction of the historical defect evolution path is along the longitudinal direction of the track to both sides, the rate parameter indicates that cracks and rust will expand at a certain rate in the next month. Based on this information, the spatial diffusion boundary coordinates in the future time window are calculated and predicted, for example, four vertex coordinates are determined on the track surface geographic coordinates to define this diffusion range.

[0088] Step S1325 , based on the overlap ratio between the spatial diffusion boundary coordinates and the current pixel coordinate set in the second risk assessment sub-model, generate the defect diffusion alarm level defined by the second risk assessment sub-model and the pixel coordinate cluster set covering the diffusion boundary.

[0089] If the overlap ratio between the spatial diffusion boundary coordinates and the current pixel coordinate set is high, for example, reaching 60%, a higher defect diffusion alarm level is determined according to the alarm level division rule set in the second risk assessment sub-model. At the same time, a pixel coordinate cluster set covering the diffusion boundary is determined according to the diffusion boundary coordinates, and the pixel coordinate cluster set accurately depicts the spatial diffusion range of the surface damage deterioration trend.

[0090] And, step S133 includes:

[0091] Step S1331: extracting meteorological sensor node position data of rail thermal expansion coefficient offset and bolt loosening risk index configured by the third risk assessment sub-model from the environmental stress influence vector.

[0092] For the area between 900 and 1000 of mountain railway mileage, the meteorological sensor node location data of the rail thermal expansion coefficient offset and bolt loosening risk index configured by the third risk assessment sub-model are extracted from the environmental stress impact vector corresponding to the area. Assuming that the rail thermal expansion coefficient offset is 0.18 and the bolt loosening risk index is 0.22, the meteorological sensor node location data related to these data are obtained at the same time.

[0093] Step S1332: construct a spatial distribution map of meteorological parameters of the third risk assessment sub-model based on the meteorological sensor node location data associated with the third risk assessment sub-model, and map it to the three-dimensional geographic coordinate grid of the track line maintained by the third risk assessment sub-model.

[0094] Meteorological sensor nodes are distributed at different locations along the track. The collected meteorological data (such as temperature, humidity, wind speed, etc.) are used to construct a spatial distribution map of meteorological parameters, and then this distribution map is accurately mapped to the three-dimensional geographic coordinate grid of the track line, so that the meteorological data corresponds to the actual geographic location of the track.

[0095] Step S1333, based on the temperature gradient data of each node in the three-dimensional geographic coordinate grid and the extreme high temperature and extreme wind loading conditions set by the third risk assessment submodel, simulate the track deformation distribution matrix generated by the third risk assessment submodel.

[0096] In mountainous areas, the temperature gradient varies greatly, and weather conditions such as extreme high temperatures and strong winds have a significant impact on track deformation. Based on the temperature gradient data of different nodes, combined with extreme high temperature (for example, reaching above 40°C during the high temperature period in summer) and extreme wind (for example, instantaneous wind speed reaching above level 10) loading conditions, the deformation distribution matrix of the track under these extreme conditions is simulated through a pre-established physical model.

[0097] Step S1334: extracting deformation hotspot nodes above the deformation threshold preset by the third risk assessment sub-model from the track deformation distribution matrix, and calculating the deformation accumulation of each deformation hotspot node defined by the third risk assessment sub-model.

[0098] Assume that in the simulation results, the deformation threshold is set to 0.05, and it is found that the deformation variables of several nodes exceed this threshold, and these are determined to be deformation hotspot nodes. For each deformation hotspot node, according to the change of its deformation variable over time, according to the calculation method in the model, the deformation accumulation of each deformation hotspot node is calculated, for example, the deformation accumulation of a certain deformation hotspot node is 0.1.

[0099] Step S1335, generating an environmental anomaly impact index output by the third risk assessment sub-model and a set of geographic coordinates of deformation hotspot nodes according to the deformation accumulation and the spatial density distribution of deformation hotspot nodes analyzed by the third risk assessment sub-model.

[0100] If the deformation accumulation of the deformation hotspot node is large and the spatial density distribution is relatively concentrated, a higher environmental anomaly impact index, such as 0.5, is generated according to the calculation rules in the third risk assessment sub-model. At the same time, the geographical coordinate set of the deformation hotspot node is determined, which clarifies the geographical location of the track area with a large deformation amount under the influence of environmental stress, that is, the specific geographical coordinate set of the deformation hotspot between mileage pile numbers 900-1000.

[0101] In a possible implementation, step S134 includes:

[0102] Step S1341, spatially superimpose the coordinate interval boundary value corresponding to the track fracture risk alarm level and the pixel coordinate cluster set of the defect diffusion alarm level to calculate the overlapping risk area of ​​the track surface and the structural layer.

[0103] In the mountain railway line scenario described above, in the specific monitoring area of ​​the mountain railway, such as between mileage stakes 1000-1100, the coordinate interval boundary value corresponding to the track fracture risk alarm level is first spatially superimposed with the pixel coordinate cluster set of the defect diffusion alarm level to calculate the overlapping risk area of ​​the track surface and the structural layer. Assuming that the coordinate interval corresponding to the track fracture risk alarm level is mileage stakes 1020-1060, the actual track spatial range corresponding to this interval is spatially superimposed with the track surface defect area represented by the pixel coordinate cluster set of the defect diffusion alarm level (such as the surface cracks and rust areas determined by the previous image analysis). This process involves accurate spatial conversion and matching of the linear coordinates of the track with the pixel coordinates of the track surface, and calculating the actual area of ​​the overlapping part of the two on the track surface and the structural layer. For example, after complex calculations, the area of ​​the overlapping risk area is 10 square meters.

[0104] Step S1342, position matching is performed on the area of ​​the overlapping risk region and the geographical coordinate set of the deformation hot zone node to determine a multi-modal risk intersection area that includes vibration anomalies, surface defects and deformation hot zones.

[0105] Between the previously determined mileage posts 1000 - 1100, the geographic coordinates of the deformation hotspot nodes represent areas where track deformation is greater due to environmental stresses (such as thermal expansion of rails and loose bolts). By matching the overlapping risk area with the geographic coordinates of these deformation hotspot nodes, the intersection of track breakage risk, surface defect risk, and deformation risk due to environmental factors can be accurately found. For example, at a specific location, it is found that the overlapping risk area partially overlaps with the deformation hotspot, and the overlapping part is the multimodal risk intersection area.

[0106] Step S1343, assigning a combined weight coefficient according to the fatigue score of the vibration energy distribution vector in the multi-modal risk intersection area, the diffusion alarm level of the surface defect probability vector, and the deformation accumulation of the environmental stress impact index.

[0107] In this multimodal risk intersection area, the fatigue score of the vibration energy distribution vector reflects the degree of fatigue that may be caused to the track structure due to vibration, assuming that the score is 0.5. The diffusion alarm level of the surface defect probability vector indicates the degree of diffusion risk of defects such as surface cracks and rust, for example, a medium alarm level. The cumulative deformation of the environmental stress impact index reflects the cumulative impact of environmental factors on the track deformation, assuming that the cumulative deformation is 0.3. According to the pre-set weight allocation rules, considering the different contributions of track structure fatigue, surface defect diffusion risk and environmental deformation impact to the overall risk, the three factors are assigned combined weight coefficients. For example, the weight coefficient of track structure fatigue is 0.3, the weight coefficient of surface defect diffusion risk is 0.4, and the weight coefficient of environmental deformation impact is 0.3.

[0108] Step S1344, based on the combined weight coefficient and a preset abnormal event type code mapping table, matching the abnormal event type code consistent with the risk characteristics of the multimodal risk intersection area.

[0109] For example, the preset abnormal event type code mapping table contains abnormal event type codes corresponding to different combination weight coefficient ranges, and these abnormal event type codes represent different types of track risk events. According to the combined weight coefficient calculated previously, for example, the weight result after comprehensive calculation is a certain value, by searching the mapping table, the abnormal event type code matching it is found, and the abnormal event type code can accurately identify the risk type corresponding to the multi-modal risk intersection area, such as the risk type of coexistence of local track structural weakening and surface damage.

[0110] Step S1345, generating an abnormal area location code including a start-end stake identifier and GPS boundary coordinates according to the mileage stake range and geographic coordinate set of the multi-modal risk intersection area.

[0111] In the multimodal risk intersection area, the mileage pile number range is 1030 - 1050. By integrating this mileage pile number range with the corresponding geographic coordinates, the linear identification of the track is combined with the actual geographic location coordinates. For example, the track position corresponding to the mileage pile number is converted into accurate GPS coordinates through the coordinate conversion algorithm to determine the GPS boundary coordinates of this area, such as (x1, y1), (x2, y2), etc. Then, based on this information, an abnormal area position code containing a polygon vertex sequence and a pile number interval identifier is generated. The abnormal area position code can uniquely determine the track area with multimodal risks and provide accurate positioning information for subsequent risk processing and maintenance operations.

[0112] Step S1346: Convert the coordinate interval boundary value of the track fracture risk alarm level into a GPS coordinate range to generate a first risk area polygon.

[0113] In a specific area of ​​mountain railways, such as between mileage stakes 1100 and 1200, the coordinate interval boundary value of the track fracture risk alarm level is first converted into a GPS coordinate range to generate the first risk area polygon. Assuming that the coordinate interval corresponding to the track fracture risk alarm level is mileage stakes 1120-1160, the start and end mileage stakes of this interval are converted into corresponding GPS coordinates through the conversion algorithm of track mileage stakes and GPS coordinates, for example, the start coordinates are (x3, y3) and the end coordinates are (x4, y4). Based on these two coordinates and the geometry of the track, a polygon representing the track fracture risk area is generated, and the vertex coordinates of the polygon are (x3, y3) and (x4, y4), etc., thereby forming the first risk area polygon.

[0114] Step S1347: convert the pixel coordinate cluster set of the defect diffusion alarm level into a second risk area polygon through a geographic coordinate mapping rule.

[0115] The pixel coordinate clusters of the defect diffusion alarm level obtained by analyzing the optical image of the track surface represent the defect area on the track surface. These pixel coordinates are converted into actual geographic coordinates using the geographic coordinate mapping rules. For example, the coordinate points (p1, q1), (p2, q2), etc. in the pixel coordinate cluster are converted into geographic coordinates (x5, y5), (x6, y6), etc. Based on these converted geographic coordinates, the second risk area polygon is constructed according to the shape of the track surface and the defect distribution.

[0116] Step S1348: Merge the geographical coordinate set of the deformation hotspot into a third risk area polygon according to a spatial density clustering algorithm.

[0117] Between mileage pile numbers 1100 and 1200, the geographic coordinate set of the deformation hotspot contains multiple node coordinates that have large deformation variables due to environmental stress. Through the spatial density clustering algorithm, these geographic coordinates are clustered according to the density of their spatial distribution. For example, coordinate points with close distances are merged into a cluster area, and then the third risk area polygon is constructed based on the boundary coordinates of these cluster areas. Assume that the polygon vertex coordinates obtained after clustering are (x7, y7), (x8, y8), etc.

[0118] Step S1349, calculating the spatial intersection area of ​​the first risk area polygon, the second risk area polygon and the third risk area polygon, and extracting the GPS coordinates of the vertices of the intersection area and the covered mileage pile number interval.

[0119] For example, through spatial geometric calculation, the overlapping parts of the first risk area polygon, the second risk area polygon and the third risk area polygon, i.e., the spatial intersection area, can be found. For example, after calculation, it is found that the three polygons intersect in a certain area, and the GPS coordinates of the vertices of the intersection area are (x9, y9), (x10, y10), etc., and the mileage pile number interval covered by this intersection area on the track is 1130 - 1150.

[0120] Step S13410, generating an abnormal area position code including a polygon vertex sequence and a mileage stake number interval identifier according to the vertex GPS coordinates and mileage stake number interval of the intersection area.

[0121] For example, the GPS coordinates of the vertices (x9, y9), (x10, y10), etc. of the intersection area are arranged in a certain order to form a polygon vertex sequence, and the stake number interval identifier 1130 - 1150 is added to generate a complete abnormal area location code. The abnormal area location code can accurately locate the area on the track where there is a risk of track fracture, surface defect risk and environmental deformation risk, providing accurate location information for subsequent track maintenance, train scheduling and other operations.

[0122] In a possible implementation, step S140 includes:

[0123] Step S141, when the comprehensive risk alarm level reaches the first critical value, the GPS positioning data in the abnormal area position code is parsed to generate a train emergency braking trigger instruction including the target track area blocking coordinates and the contact network power outage time.

[0124] In the mountain railway line scenario described above, during the operation of the mountain railway line, when the comprehensive risk alarm level reaches the first critical value, it indicates that the track is facing a serious risk situation and emergency measures need to be taken immediately to ensure the safety of train operation. At this time, the GPS positioning data in the abnormal area position code is parsed to generate a train emergency braking trigger instruction containing the target track area blocking coordinates and the contact network power-off time. For example, in a certain mountain railway section, the GPS positioning data in the abnormal area position code shows that the target track area is located within a specific longitude and latitude range, corresponding to the track section between mileage pile numbers 1200-1250. According to this location information, the target track area blocking coordinates are determined, such as a coordinate range with a certain distance (the specific width is determined according to railway safety regulations) extending on both sides based on the center line of the track. At the same time, the contact network power-off time is determined by considering factors such as the train's running speed, braking performance, and distance to the target area. Assuming that a passenger train is heading towards this dangerous area, according to the current speed of the train, the distance from the target area, and the braking characteristics of the train, it is calculated that the contact network power-off should be triggered 30 seconds before the train reaches the target track area to ensure that the train can stop safely before entering the dangerous area. This generates a train emergency braking trigger instruction containing the target track area blocking coordinates (precise geographical coordinate range) and the contact network power outage time (30 seconds).

[0125] Step S142, when the comprehensive risk alarm level is in the second critical range, the train dispatch timetable is matched according to the mileage post number encoded in the abnormal area position, and a signal light state switching instruction matching the current position of the train is generated, and the signal light state switching instruction includes the red light holding time and the yellow flash warning mode parameters based on the pixel coordinate cluster.

[0126] When the comprehensive risk alarm level is in the second critical range, although the risk level is slightly lower than the first critical value, the train operation still needs to be adjusted. According to the mileage pile number encoded in the abnormal area position, the train dispatch schedule is matched to generate a signal light state switching instruction that matches the current position of the train. The signal light state switching instruction includes the red light holding time and the yellow flash warning mode parameters based on the pixel coordinate cluster. For example, the mileage pile number in the abnormal area position code is between 1300-1350. By querying the train dispatch schedule, it is found that multiple trains are about to pass through this area. For trains closer to the area, the signal light state switching instruction is determined according to the distance relationship between the current position of the train and the dangerous area. Assuming that a freight train is currently a certain distance away from the dangerous area, according to the train's speed and the remaining distance to the dangerous area, the red light should be kept for 5 minutes. At the same time, since the previous monitoring of track surface defects has obtained track surface defect information based on pixel coordinate clusters, the yellow flash warning mode parameters are set according to the distribution of these defects (pixel coordinate clusters). For example, if the defects are mainly concentrated on one side of the track, a yellow flashing signal with a specific frequency and flashing pattern is set on that side to remind the train driver to pay special attention when passing through this area, thereby generating a signal light state switching instruction including the red light holding time (5 minutes) and the yellow flashing warning mode parameters (specific flashing frequency and pattern) based on the pixel coordinate cluster.

[0127] Step S143, when the comprehensive risk alarm level triggers the third response condition, the spatial intersection of the deformation hot zone geographic coordinate set and the pixel coordinate cluster is extracted, and a maintenance robot dispatching instruction carrying the three-dimensional model data of the track area is generated. The maintenance robot dispatching instruction includes a list of fault point coordinates that integrates the mileage pile number and GPS data and a priority maintenance order identifier based on the area of ​​the deformation hot zone.

[0128] When the comprehensive risk alarm level triggers the third response condition, it means that the condition of the track requires maintenance operations, but the situation is not urgent enough to stop immediately. At this time, the spatial intersection of the deformation hot zone geographic coordinate set and the pixel coordinate cluster is extracted to generate a maintenance robot dispatch instruction carrying the track area three-dimensional model data. The maintenance robot dispatch instruction contains a list of fault point coordinates that integrate the mileage pile number and GPS data and a priority maintenance order identifier based on the area of ​​the deformation hot zone. For example, in the area between mileage pile numbers 1400-1450, the deformation hot zone geographic coordinate set shows the area where the track has a large deformation due to environmental stress and other factors, while the pixel coordinate cluster represents the defective area that may exist on the track surface. By calculating the spatial intersection of the two, an area with both deformation and surface defects is obtained. According to this area, the track area three-dimensional model data is constructed, and the track area three-dimensional model contains the geometric shape, structural characteristics, and detailed information of defects and deformation of the track. Then, a list of fault point coordinates that integrate the mileage pile number and GPS data is determined, for example, the coordinate information of multiple fault points such as (x, y) corresponding to the GPS coordinates at the mileage pile number 1420 is listed. At the same time, the priority maintenance sequence identification is determined according to the size of the deformation hot zone. For example, areas with larger deformation hot zones or areas that have a greater impact on train operation (such as areas near curves or bridges) are marked as having a higher priority maintenance level, thereby generating a maintenance robot dispatch instruction that includes a fault point coordinate list (detailed coordinate list) that integrates mileage pile numbers and GPS data and a priority maintenance sequence identification based on the area of ​​the deformation hot zone (different levels of maintenance priority).

[0129] In a possible implementation, step S150 includes:

[0130] Step S151, parsing the instruction priority, target track area blocking coordinates, red light holding time and fault point coordinate list of each instruction in the multi-level linkage control instruction set.

[0131] For example, in the received multi-level linkage control command set, the train emergency brake trigger command has the highest priority because it involves the immediate stop of the train to avoid a major accident. The target track area blocking coordinates clearly define the specific location of the track to be blocked, such as the precise geographic coordinate range corresponding to the mileage post number 1200 - 1250 mentioned earlier. Information such as the length of time the red light is maintained (such as 5 minutes in the previous signal light state switching command) and the list of fault point coordinates (such as the coordinates at the mileage post number 1420) are also accurately parsed out.

[0132] Step S152, sorting the execution order of the train emergency brake triggering instruction, the signal light state switching instruction and the maintenance robot scheduling instruction according to the instruction priority, and generating an instruction distribution queue including a time window identifier.

[0133] Since the train emergency brake trigger instruction has the highest priority, it is placed at the front of the instruction distribution queue, followed by the signal light state switching instruction, and finally the maintenance robot dispatch instruction. And each instruction is assigned a time window identifier, for example, the time window identifier of the train emergency brake trigger instruction is immediate execution, the time window identifier of the signal light state switching instruction is executed within a certain time after the train emergency brake trigger instruction is executed (the specific time is determined according to the train operation status and signal light switching logic), and the time window identifier of the maintenance robot dispatch instruction is executed at an appropriate time after the train operation adjustment is completed.

[0134] Step S153, verifying the spatial overlap of the train emergency brake triggering instruction in the instruction distribution queue with the real-time train position trajectory data of the central dispatching system to determine the train number to be triggered and the corresponding contact network power outage time interval.

[0135] The central dispatching system has real-time position and trajectory data of all trains. For example, at a certain moment, there are multiple trains running on mountain railway lines. By comparing the blocking coordinates of the target track area in the train emergency brake trigger command with the real-time position and trajectory data of each train, it is found that train A is about to enter the target track area. According to the train number of train A and the remaining distance and speed of the train to the target area, the corresponding contact network power-off time interval is determined. Assuming that train A is still a certain distance away from the target area, according to the running speed and braking performance of the train, it is determined that the contact network power-off starts 20 seconds before train A arrives at the target track area, thereby determining the train number to be triggered (train A) and the corresponding contact network power-off time interval (20 seconds before train A arrives at the target area).

[0136] Step S154, dynamically matching the red light holding time in the signal light state switching instruction with the arrival time deviation of the train scheduling schedule, calculating the signal light switching delay compensation value and updating the yellow flash warning mode parameters.

[0137] For example, according to the train dispatch schedule, train B is expected to arrive at a station in 10 minutes, but because the red light in the signal light state switching instruction is kept for 5 minutes, this will cause the arrival time of train B to be delayed. By calculating the impact of train B's original speed, remaining distance, and the red light holding time on the train operation, the signal light switching delay compensation value is obtained to be 3 minutes. According to this compensation value, the yellow flash warning mode parameters are updated, such as adjusting the frequency of yellow flashes or the flashing time interval, to better adapt to the adjustment of train operation.

[0138] Step S155, matching the fault point coordinate list in the maintenance robot scheduling instruction with the current position and remaining operation capacity of the maintenance robot, and generating a task allocation path plan based on a priority maintenance sequence identifier.

[0139] Assume that there are multiple maintenance robots distributed along the mountain railway, and each maintenance robot has its current location and remaining operating capacity information. For the list of fault point coordinates in the maintenance robot scheduling instruction (such as the fault point at mileage pile number 1420), each maintenance robot is assigned a task based on factors such as the distance from the current position of the maintenance robot to the fault point, the moving speed of the maintenance robot, and the remaining operating capacity. For example, if maintenance robot C is close to the fault point and has sufficient remaining operating capacity, and the fault point is in an area with a higher priority maintenance level, then maintenance robot C will be given priority to go to the fault point for maintenance. Based on factors such as the current position of maintenance robot C, the location of the fault point, and the terrain and track layout of the mountain railway, a task allocation path planning is generated to ensure that the maintenance robot can reach the fault point in the shortest path and the fastest time.

[0140] Step S156, sending the updated train emergency brake trigger instruction, signal light state switching instruction and maintenance robot scheduling instruction to the train automatic driving subsystem, signal light control subsystem and maintenance robot scheduling subsystem in the order of the instruction distribution queue.

[0141] After receiving the train emergency braking trigger command, the train automatic driving subsystem executes the train emergency braking operation according to the parameters in the command; after receiving the signal light state switching command, the signal light control subsystem switches the signal light state according to the red light holding time and yellow flash warning mode parameters in the command; after receiving the maintenance robot scheduling command, the maintenance robot dispatches the maintenance robot to the fault point for maintenance according to the task allocation path planning.

[0142] Step S157, receiving in real time the contact network power-off status code fed back by the train automatic driving subsystem, the signal light switching completion mark returned by the signal light control subsystem, and the fault point maintenance progress data reported by the maintenance robot scheduling subsystem.

[0143] For example, after executing the contact network power-off operation, the train automatic driving subsystem will feedback a contact network power-off status code, indicating whether the contact network is successfully powered off and related status information of the power-off; after completing the signal light switching, the signal light control subsystem returns a signal light switching completion flag, indicating that the signal light has completed the state switching according to the instruction; during the maintenance process of the maintenance robot, the maintenance robot scheduling subsystem regularly reports the fault point maintenance progress data, such as the proportion of completed maintenance workload, the estimated remaining maintenance time, etc.

[0144] Step S158, generates a track area blocking status update table, a train speed control log and a real-time equipment maintenance topology diagram according to the contact network power-off status code, the signal light switching completion mark and the fault point maintenance progress data, and synchronizes them to the global operation status interface of the central dispatching system.

[0145] For example, the blocking status of the track area is determined according to the contact network power-off status code, such as whether it is completely blocked, whether there are still risks in some areas, etc., and this information is updated to the track area blocking status update table. According to the train speed control situation (such as the train slowing down or stopping due to the switching of the signal light status), the train speed control log is recorded, including the train number, speed change time, speed change value and other information. According to the maintenance progress data of the fault point of the maintenance robot, the real-time topology diagram of equipment maintenance is drawn to display the location of the maintenance robot, the maintenance status of the fault point and other information. Finally, these track area blocking status update tables, train speed control logs and equipment maintenance real-time topology diagrams are synchronized to the global operation status interface of the central dispatching system, so that the dispatching personnel can fully grasp the operation status, maintenance status and risk control of the mountain railway line in real time.

[0146] For example, in a possible implementation, after step S150, the method further includes:

[0147] Step S160, receiving in real time the instruction execution status data fed back by the central dispatching system, wherein the instruction execution status data includes track blockade area update information, train real-time speed curve and maintenance work completion report.

[0148] During the operation of mountain railway lines, the central dispatching system will continuously feedback these important data information after executing various operations. For example, the track blockade area update information includes the track blockade coordinate list and the blockade start time series. The track blockade coordinate list accurately identifies the geographical location of the blocked track area. For example, part of the track between mileage pile numbers 1500-1550 is blocked. These coordinate information is recorded in detail through a specific geographic coordinate system. The blockade start time series records a series of time data from the time when the blockade of this area begins. For example, the blockade starts at 10 am, and this time point and subsequent related time records (such as the blockade duration, etc.) are included. The real-time train speed curve reflects the speed change of the train during operation, which is closely related to the track blockade operation and the train speed control operation. It records in detail the speed values ​​of the train at different times. For example, the speed of train A gradually decreases when approaching the track blockade area. The speed is 50 kilometers per hour at a certain moment and 30 kilometers per hour at another moment. These speed values ​​are arranged in chronological order to form the real-time train speed curve. The repair work completion report includes the coordinates of the repaired track area and the repair operation timestamp. The coordinates of the repaired track area clearly indicate which track areas have completed the repair work, such as the track between mileage pile numbers 1420 and 1430 has been repaired after the operation of the maintenance robot. The repair operation timestamp records the specific time of the maintenance operation, such as the maintenance work started at 11 am and ended at 1 pm. These time information corresponds to the coordinates of the repaired track area.

[0149] Step S170: reconstructing a dynamic threshold curve of a track safety assessment model according to the instruction execution status data.

[0150] For example, in a possible implementation, step S170 includes:

[0151] Step S171, receiving the track area blocked coordinate list and the blocking start time sequence in the track blocked area update information, extracting the speed drop gradient data associated with the track area blocked coordinate list in the real-time speed curve of the train, and parsing the repaired track area coordinates and the repair operation timestamp in the maintenance work completion report.

[0152] For example, for the track blockade area between mileage posts 1500 - 1550, the speed data of the train when approaching and passing this area is found from the real-time train speed curve, and the speed drop gradient data is calculated. Assuming that the speed of the train begins to drop 1 km before entering the blockade area, from 80 km / h to 30 km / h at the edge of the blockade area, the speed drop gradient is calculated to be 50 km / h per kilometer based on the distance and speed changes. At the same time, the coordinates of the repaired track area (such as mileage posts 1420 - 1430) and the repair operation timestamp (11 am - 1 pm) are parsed from the maintenance work completion report.

[0153] Step S172, based on the time correspondence between the blocking start time sequence and the speed drop gradient data, calculate the vibration energy attenuation rate corresponding to the track area blocking coordinate list, and generate the first adjustment coefficient of the dynamic threshold sub-curve of the vibration energy distribution vector.

[0154] Since the decrease in train speed will cause changes in track vibration energy, the vibration energy attenuation rate corresponding to the track area blockade coordinate list (mileage pile number 1500 - 1550) is calculated based on the blockade start time series (blockade starts at 10 am) and the speed reduction gradient data (50 km / h reduction per kilometer), combined with the vibration characteristics of the track structure and the train-track interaction model. For example, after complex calculations, it is found that the vibration energy decays at a rate of 10% per hour for a period of time after the blockade. The first adjustment coefficient of the dynamic threshold sub-curve of the vibration energy distribution vector is generated based on this attenuation rate. The first adjustment coefficient reflects the impact of track blockade and train speed control on the vibration energy distribution. Assuming that this first adjustment coefficient is 0.8, it means that the vibration energy threshold should be adjusted accordingly according to this coefficient.

[0155] Step S173, matching the repair area pixel coordinate cluster in the track surface optical image sequence according to the repaired track area coordinates and the repair operation timestamp, extracting the surface texture feature change rate after repair, and generating a second adjustment coefficient of the dynamic threshold sub-curve of the surface defect probability vector.

[0156] For the repaired track area between mileage posts 1420 and 1430, the corresponding pixel coordinate cluster of the repaired area is found in the track surface optical image sequence according to the repair operation timestamp (11 a.m. to 1 p.m.). By using image analysis technology, the track surface texture features before and after repair are compared, and the surface texture feature change rate after repair is calculated. For example, there are cracks and rusted areas on the track surface before repair. After repair, these defects are improved and the surface texture becomes smoother. The surface texture feature change rate is calculated to be 30%. The second adjustment coefficient of the dynamic threshold sub-curve of the surface defect probability vector is generated based on this change rate. Assuming that this coefficient is 1.2, it means that the threshold of the surface defect probability should be adjusted according to this coefficient due to the repair operation, that is, the tolerance for surface defects can be appropriately increased.

[0157] Step S174, according to the geographical location distribution of the blocked coordinate list of the track area, associate the temperature fluctuation amplitude and humidity accumulation of the same geographical location in the environmental meteorological parameter sequence, calculate the rail deformation stress relaxation offset, and generate the third adjustment coefficient of the dynamic threshold sub-curve of the environmental stress influence vector.

[0158] For the track blockade area between mileage pile numbers 1500 and 1550, according to its geographical location distribution, the temperature fluctuation amplitude and humidity accumulation data of the same geographical location are found in the environmental meteorological parameter sequence. For example, during the blockade period, the temperature fluctuation amplitude in the area was 5°C and the humidity accumulation was 20% (the humidity accumulation here refers to the comprehensive change in humidity over a period of time). Combined with the material characteristics and thermal and mechanical properties of the rail, the deformation stress relaxation offset of the rail is calculated. Assuming that the deformation stress relaxation offset is calculated to be 0.05, the third adjustment coefficient of the dynamic threshold sub-curve of the environmental stress influence vector is generated based on this offset. For example, this coefficient is 0.9, indicating that the threshold related to environmental stress should be adjusted according to this coefficient.

[0159] Step S175, input the first adjustment coefficient of the dynamic threshold sub-curve of the vibration energy distribution vector, the second adjustment coefficient of the dynamic threshold sub-curve of the surface defect probability vector, and the third adjustment coefficient of the dynamic threshold sub-curve of the environmental stress influence vector into the multi-dimensional threshold fusion device, linearly superimpose them according to the priority weights of the track area blocking coordinate list, and output the amplitude attenuation gradient, surface defect suppression factor and environmental stress tolerance threshold of the reconstructed dynamic threshold curve.

[0160] Assuming that the priority weight of the track area blockade coordinate list (mileage pile number 1500 - 1550) is 0.5, a linear superposition calculation is performed in the multi-dimensional threshold fuser based on this weight and the first adjustment coefficient (0.8), second adjustment coefficient (1.2) and third adjustment coefficient (0.9) calculated previously. After calculation, it is found that the amplitude attenuation gradient of the reconstructed dynamic threshold curve is a specific value (for example, 0.03, indicating the attenuation rate of the vibration energy amplitude), the surface defect suppression factor is a specific value (for example, 1.1, indicating the ability to suppress surface defects), and the environmental stress tolerance threshold is a certain value (for example, 0.04, indicating the environmental stress range that the rail can withstand).

[0161] Step S176: updating the maximum allowable vibration energy peak value in the frequency band energy threshold comparison rule of the first risk assessment sub-model according to the amplitude attenuation gradient of the reconstructed dynamic threshold curve.

[0162] Since the change of the amplitude attenuation gradient reflects the change of the overall state of the track, the maximum allowable vibration energy peak in the frequency band energy threshold comparison rule of the first risk assessment sub-model is updated according to the amplitude attenuation gradient (0.03) of the reconstructed dynamic threshold curve. For example, the original maximum allowable vibration energy peak in a certain frequency band is 10 joules. According to the new relationship between the amplitude attenuation gradient and the vibration characteristics of the track structure, the maximum allowable vibration energy peak is adjusted to 8 joules to more accurately reflect the actual vibration energy tolerance of the track after a series of operations.

[0163] Step S177, adjusting the tolerance upper limit of the crack growth rate in the historical defect evolution pattern library of the second risk assessment sub-model according to the surface defect suppression factor of the reconstructed dynamic threshold curve.

[0164] Because the surface defect suppression factor (1.1) reflects the change in the ability to control surface defects, the upper limit of the crack growth rate tolerance in the historical defect evolution model library of the second risk assessment sub-model is adjusted according to this factor. For example, the original upper limit of the crack growth rate tolerance is 5% per year. According to the new surface defect suppression factor and the repair status of the track surface, the upper limit of the crack growth rate tolerance is adjusted to 4% per year, so that the model is more in line with the actual maintenance and operation status of the track when evaluating the evolution of surface defects.

[0165] Step S178, resetting the thermal expansion coefficient safety interval in the track deformation simulation algorithm of the third risk assessment sub-model according to the environmental stress tolerance threshold of the reconstructed dynamic threshold curve.

[0166] Since the environmental stress tolerance threshold (0.04) reflects the change in the rail's ability to withstand environmental stress, the thermal expansion coefficient safety interval in the track deformation simulation algorithm of the third risk assessment sub-model is reset according to this threshold. For example, the original thermal expansion coefficient safety interval is [10×10⁻ 6 - 12×10⁻ 6 ] / ℃, according to the new environmental stress tolerance threshold and the material properties of the rail and the influence of meteorological conditions, the thermal expansion coefficient safety interval is adjusted to [9×10⁻ 6 - 11×10⁻ 6 ] / ℃, so as to more accurately consider the influence of environmental factors when simulating track deformation.

[0167] Step S179, synchronizing the updated maximum allowable vibration energy peak, crack growth rate tolerance upper limit and thermal expansion coefficient safety interval to the risk alarm level classification strategy of the rail safety assessment model, overwriting the parameter configuration of the original dynamic threshold curve.

[0168] In this embodiment, the track safety assessment model can more accurately assess the safety status of the track according to the new parameter configuration, adapt to the actual operation of mountain railway lines after track closures, train speed control and equipment maintenance operations, and improve the accuracy and effectiveness of track safety monitoring and risk assessment.

[0169] Figure 2 The schematic diagram shows exemplary hardware and software components of the railway line monitoring alarm system 100 provided by some embodiments of the present application that can implement the concept of the present application. For example, the processor 120 can be used in the railway line monitoring alarm system 100 and used to perform the functions in the present application.

[0170] The railway line monitoring alarm system 100 can be a general server or a special-purpose server, both of which can be used to implement the railway line monitoring alarm method of the present application. Although only one server is shown in the present application, for convenience, the functions described in the present application can be implemented in a distributed manner on multiple similar platforms to balance the processing load.

[0171] For example, the railway line monitoring alarm system 100 may include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and storage media 140 in different forms, such as a disk, ROM, or RAM, or any combination thereof. Exemplarily, the railway line monitoring alarm system 100 may also include program instructions stored in ROM, RAM, or other types of non-transitory storage media, or any combination thereof. The method of the present application can be implemented according to these program instructions. The railway line monitoring alarm system 100 also includes an input / output (I / O) interface 150 between the computer and other input / output devices.

[0172] For ease of explanation, only one processor is described in the railway line monitoring and alarm system 100. However, it should be noted that the railway line monitoring and alarm system 100 in the present application may also include multiple processors, so the steps performed by one processor described in the present application may also be performed jointly or individually by multiple processors. For example, if the processor of the railway line monitoring and alarm system 100 performs step A and step B, it should be understood that step A and step B may also be performed jointly by two different processors or individually in one processor. For example, the first processor performs step A, the second processor performs step B, or the first processor and the second processor perform steps A and B together.

[0173] In addition, an embodiment of the present invention further provides a readable storage medium, in which computer executable instructions are preset. When a processor executes the computer executable instructions, the above railway line monitoring and alarm method is implemented.

[0174] It should be noted that in order to simplify the description of the present invention and thus help understand one or more embodiments of the invention, in the foregoing description of the embodiments of the present invention, various features are sometimes combined into one embodiment, drawing or description thereof.

Claims

1. A railway line monitoring and alarm method, characterized in that: The method comprises: A dynamic monitoring data set of the track area is continuously collected by multiple sensor nodes installed on the target railway line, wherein the dynamic monitoring data set includes a track structure vibration waveform sequence, a track surface optical image sequence, and an environmental meteorological parameter sequence; Performing multimodal feature fusion processing on the dynamic monitoring data set to generate a comprehensive state evaluation matrix for each track area, wherein the comprehensive state evaluation matrix includes a vibration energy distribution vector, a surface defect probability vector, and an environmental stress influence vector; Classifying the risk alarm level of the comprehensive status assessment matrix according to a preset track safety assessment model, and outputting the abnormal event type code and abnormal area position code corresponding to the risk alarm level; Generate a multi-level linkage control instruction set based on the abnormal event type code and the abnormal area position code, wherein the multi-level linkage control instruction set includes a train emergency brake trigger instruction, a signal light state switching instruction, and a maintenance robot dispatching instruction; Synchronously transmitting the multi-level linkage control instruction set to the central dispatching system, driving the central dispatching system to execute track blocking operations, train speed control operations and equipment maintenance operations according to instruction priorities; The multi-modal feature fusion processing is performed on the dynamic monitoring data set to generate a comprehensive status evaluation matrix for each track area, including: Performing frequency domain energy analysis on the vibration waveform sequence of the track structure, extracting the proportion of vibration energy in each frequency band, and forming a vibration energy distribution vector; Performing pixel-level texture segmentation on the track surface optical image sequence, calculating the surface crack growth rate and the corrosion area coverage rate, and generating a surface defect probability vector; The cumulative effect modeling is performed on the environmental meteorological parameter sequence, the rail thermal expansion coefficient offset and the bolt loosening risk index are derived, and the environmental stress influence vector is constructed; The vibration energy distribution vector, surface defect probability vector and environmental stress influence vector are horizontally spliced ​​according to the track area number to form a multi-dimensional comprehensive state assessment matrix.

2. The railway line monitoring and alarm method according to claim 1, characterized in that: The method of continuously collecting a dynamic monitoring data set of the track area by installing multiple sensor nodes on the target railway line includes: Controlling the distributed vibration sensor to collect a vibration waveform sequence of the rail structure at a first sampling frequency, and capturing a track surface optical image sequence at a fixed shooting interval through an industrial camera; Start the meteorological monitoring unit to record the environmental meteorological parameter sequence in real time, wherein the environmental meteorological parameter sequence includes air temperature gradient data, relative humidity change data and wind speed fluctuation data; The track structure vibration waveform sequence, track surface optical image sequence and environmental meteorological parameter sequence are synchronously marked with time stamps, and the data storage intervals are divided according to the mileage pile numbers of the track line.

3. The railway line monitoring and alarm method according to claim 1, characterized in that: The risk alarm level classification of the comprehensive status assessment matrix according to the preset track safety assessment model, and outputting the abnormal event type code and abnormal area position code corresponding to the risk alarm level, include: The vibration energy distribution vector is input into the first risk assessment sub-model, the mileage pile number of the abnormal vibration area is determined by comparing the frequency band energy threshold, the track structure fatigue score is calculated and compared with the dynamic threshold curve, and the track fracture risk alarm level and the corresponding coordinate interval are determined; The surface defect probability vector is input into the second risk assessment sub-model, and the image coordinate positioning system is used to match the historical defect evolution pattern library, predict the spatial diffusion range of the surface damage deterioration trend, and generate the defect diffusion alarm level and pixel coordinate cluster; The environmental stress impact vector is input into the third risk assessment sub-model, and the track deformation under preset extreme conditions is simulated in combination with the topological position of the meteorological sensor, and the environmental anomaly impact index and the geographical coordinate set of the deformation hot zone are output; According to the combined weight of the track fracture risk alarm level, defect diffusion alarm level and environmental anomaly impact index, the comprehensive risk alarm level is determined and mapped to the preset abnormal event type code, and the coordinate interval, pixel coordinate cluster and deformation hot zone geographic coordinate set are spatially superimposed to generate the abnormal area location code containing mileage pile number and GPS positioning data; Wherein, the track safety assessment model includes the first risk assessment sub-model, the second risk assessment sub-model and the third risk assessment sub-model.

4. The railway line monitoring and alarm method according to claim 3 is characterized in that: The vibration energy distribution vector is input into the first risk assessment sub-model, the mileage pile number of the abnormal vibration area is determined by comparing the frequency band energy threshold, the track structure fatigue score is calculated and compared with the dynamic threshold curve, and the track fracture risk alarm level and the corresponding coordinate interval are determined, including: Decomposing the vibration energy distribution vector into a low-frequency vibration energy component, a medium-frequency vibration energy component and a high-frequency vibration energy component according to a preset frequency band division rule in the first risk assessment sub-model; Extracting energy proportion data of vibration energy components in each frequency band, and comparing them segment by segment with frequency band energy threshold intervals stored in the first risk assessment sub-model, to identify abnormal vibration frequency bands marked by the first risk assessment sub-model; Determine the starting mileage and ending mileage of the abnormal vibration area output by the first risk assessment sub-model according to the physical position of the sensor node corresponding to the abnormal vibration frequency band and the timestamp sequence of the vibration energy distribution vector associated with the first risk assessment sub-model; Calculate the track structure fatigue score between the starting mileage pile number and the ending mileage pile number based on the energy proportion data of the abnormal vibration frequency band and the vibration duration calculation rule in the first risk assessment sub-model; Perform a sliding window comparison on the track structure fatigue score and the historical fatigue score of the same mileage pile number interval maintained by the first risk assessment sub-model in the dynamic threshold curve, and determine the track fracture risk alarm level and the corresponding coordinate interval boundary value according to the deviation amplitude set by the first risk assessment sub-model; And, the surface defect probability vector is input into the second risk assessment sub-model, and the image coordinate positioning system is used to match the historical defect evolution pattern library, predict the spatial diffusion range of the surface damage deterioration trend, and generate the defect diffusion alarm level and pixel coordinate cluster, including: Extracting a pixel coordinate set of the surface crack growth rate and the rust area coverage rate defined by the second risk assessment sub-model from the surface defect probability vector; According to the image coordinate positioning system integrated with the second risk assessment sub-model, the pixel coordinate set is converted into the track surface geographic coordinates associated with the second risk assessment sub-model, and the track area number corresponding to the mileage pile number is matched; Input the spatiotemporal variation trends of the surface crack growth rate and the corrosion area coverage rate into the historical defect evolution pattern library of the second risk assessment sub-model, perform pattern matching with the crack growth rate template and the corrosion diffusion template stored in the second risk assessment sub-model, and identify the closest historical defect evolution path output by the second risk assessment sub-model; Predicting the spatial diffusion boundary coordinates within the future time window marked by the second risk assessment submodel according to the diffusion direction of the historical defect evolution path and the rate parameter loaded by the second risk assessment submodel; Generate a defect diffusion alarm level defined by the second risk assessment sub-model and a pixel coordinate cluster set covering the diffusion boundary based on an overlap ratio between the spatial diffusion boundary coordinates and the current pixel coordinate set in the second risk assessment sub-model; And, the environmental stress impact vector is input into the third risk assessment sub-model, and the track deformation under extreme conditions is simulated in combination with the topological position of the meteorological sensor, and the environmental anomaly impact index and the geographical coordinate set of the deformation hot zone are output, including: Extracting meteorological sensor node location data of rail thermal expansion coefficient offset and bolt loosening risk index configured by the third risk assessment sub-model from the environmental stress influence vector; According to the meteorological sensor node location data associated with the third risk assessment sub-model, a meteorological parameter spatial distribution map of the third risk assessment sub-model is constructed, and mapped to the three-dimensional geographic coordinate grid of the track line maintained by the third risk assessment sub-model; Simulating the track deformation distribution matrix generated by the third risk assessment submodel based on the temperature gradient data of each node in the three-dimensional geographic coordinate grid and the extreme high temperature and extreme wind loading conditions set by the third risk assessment submodel; Extracting deformation hotspot nodes above the deformation threshold preset by the third risk assessment sub-model from the track deformation distribution matrix, and calculating the deformation accumulation of each deformation hotspot node defined by the third risk assessment sub-model; According to the deformation accumulation and the spatial density distribution of the deformation hot zone nodes analyzed by the third risk assessment sub-model, the environmental anomaly impact index output by the third risk assessment sub-model and the geographical coordinate set of the deformation hot zone nodes are generated.

5. The railway line monitoring and alarm method according to claim 4, characterized in that: The comprehensive risk alarm level is determined according to the combined weight of the track fracture risk alarm level, the defect diffusion alarm level and the environmental anomaly impact index and mapped to a preset abnormal event type code, including: The coordinate interval boundary value corresponding to the rail fracture risk alarm level is spatially superimposed with the pixel coordinate cluster set of the defect diffusion alarm level to calculate the overlapping risk area of ​​the rail surface and the structural layer; Positionally match the area of ​​the overlapping risk region with the geographical coordinate set of the deformation hot zone node to determine a multi-modal risk intersection area that includes vibration anomalies, surface defects and deformation hot zones; Allocating a combined weight coefficient according to the fatigue score of the vibration energy distribution vector in the multi-modal risk intersection area, the diffusion alarm level of the surface defect probability vector, and the deformation accumulation of the environmental stress impact index; Based on the combined weight coefficient and a preset abnormal event type code mapping table, matching the abnormal event type code consistent with the risk characteristics of the multimodal risk intersection area; According to the mileage pile number range and geographic coordinate set of the multimodal risk intersection area, an abnormal area location code including the start-end pile number identification and GPS boundary coordinates is generated.

6. The railway line monitoring and alarm method according to claim 3, characterized in that: The coordinate interval, pixel coordinate cluster and deformation hot zone geographic coordinate set are spatially superimposed to generate an abnormal area location code containing mileage pile number and GPS positioning data, including: Converting the coordinate interval boundary value of the track fracture risk alarm level into a GPS coordinate range to generate a first risk area polygon; Converting the pixel coordinate cluster set of the defect diffusion alarm level into a second risk area polygon through a geographic coordinate mapping rule; Merging the deformation hot zone geographic coordinate set into a third risk area polygon according to a spatial density clustering algorithm; Calculate the spatial intersection area of ​​the first risk area polygon, the second risk area polygon and the third risk area polygon, and extract the GPS coordinates of the vertices of the intersection area and the mileage pile number interval covered; According to the GPS coordinates of the vertices and the mileage and pile number intervals of the intersection area, an abnormal area position code including a polygon vertex sequence and a pile number interval identifier is generated.

7. The railway line monitoring and alarm method according to claim 3, characterized in that: The generating of a multi-level linkage control instruction set based on the abnormal event type code and the abnormal area position code includes: When the comprehensive risk alarm level reaches a first critical value, the GPS positioning data in the abnormal area position code is parsed to generate a train emergency brake trigger instruction including the target track area blocking coordinates and the contact network power-off time; When the comprehensive risk alarm level is in the second critical range, the train dispatching timetable is matched according to the mileage pile number encoded in the abnormal area position, and a signal light state switching instruction matching the current position of the train is generated, wherein the signal light state switching instruction includes the red light holding time and the yellow flash warning mode parameters based on the pixel coordinate cluster; When the comprehensive risk alarm level triggers the third response condition, the spatial intersection of the deformation hot zone geographic coordinate set and the pixel coordinate cluster is extracted, and a maintenance robot dispatching instruction carrying the three-dimensional model data of the track area is generated. The maintenance robot dispatching instruction includes a list of fault point coordinates that integrates the mileage pile number and GPS data and a priority maintenance order identifier based on the area of ​​the deformation hot zone.

8. The railway line monitoring and alarm method according to claim 7, characterized in that: The synchronous transmission of the multi-level linkage control instruction set to the central dispatching system, driving the central dispatching system to perform track blocking operations, train speed control operations and equipment maintenance operations according to instruction priorities, includes: Analyze the instruction priority, target track area blocking coordinates, red light holding time and fault point coordinate list of each instruction in the multi-level linkage control instruction set; The train emergency brake trigger instruction, the signal light state switching instruction and the maintenance robot dispatch instruction are sorted in execution order according to the instruction priority, and an instruction distribution queue including a time window identifier is generated; The train emergency brake triggering instruction in the instruction distribution queue is verified by spatial overlap with the real-time train position trajectory data of the central dispatching system to determine the train number to be triggered and the corresponding contact network power-off time interval; Dynamically match the red light holding time in the signal light state switching instruction with the arrival time deviation of the train scheduling timetable, calculate the signal light switching delay compensation value and update the yellow flash warning mode parameters; Matching the fault point coordinate list in the maintenance robot dispatch instruction with the current position and remaining operation capacity of the maintenance robot to generate a task allocation path plan based on a priority maintenance sequence identifier; Send the updated train emergency brake triggering instruction, signal light state switching instruction and maintenance robot dispatching instruction to the train automatic driving subsystem, signal light control subsystem and maintenance robot dispatching subsystem in the order of instruction distribution queue; Receive in real time the contact network power-off status code fed back by the train automatic driving subsystem, the signal light switching completion mark returned by the signal light control subsystem, and the fault point maintenance progress data reported by the maintenance robot scheduling subsystem; According to the contact network power-off status code, signal light switching completion mark and fault point maintenance progress data, a track area blockade status update table, a train speed control log and a real-time topology diagram of equipment maintenance are generated and synchronized to the global operation status interface of the central dispatching system.

9. A railway line monitoring and alarm system, characterized in that: The railway line monitoring and alarm system includes a processor and a memory, the memory is connected to the processor, the memory is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the memory to implement the railway line monitoring and alarm method described in any one of claims 1 to 8 above.

Images